The role of neuromorphic principles in the future of biomedicine and healthcare

1 The role of neuro m orphic p rin c i pl es in the f utu re of bio med ic in e an d he althcar e Gr ace M. H wan g 1 , Jess ica D. F alco ne 2 , Jo s eph D . Mo na co 1 , Court ney R. P i nar d 3 , Je s si ca A. Mollic k 4 , Ro ger L. M iller 5 , S tep ha nie L. G age 6 , And re y V. Kan ae v 6 , Mar gar e t Ki m 7 , R. Ale L u k as zew 7 , Ste ve n M. Zehnde r 7 , Da vid Rampulla 2 1 N ati on al Inst itute of Neurolog ic a l Dis orders a nd St roke (N IND S) , N atio n al Ins titute s of He alth ( NI H) , The B rai n Research T hr ou gh A dvancing Inno va t ive Neu rotech nol ogies ® (BRA I N) I nit i a tiv e, B ethesda, MD 20892 , USA 2 Na tion a l Inst it u t e of Bio med ica l Imag ing a nd Bioen gin eering ( NIB IB) , N I H, Bethes da , MD 20892 , US A 3 Na tion a l Inst it u t e of M enta l H ealth (NI MH), N I H, B eth e sd a , M D 20892 , U S A 4 Na tion a l Inst it u t e of D rug Abus e (NI DA ), N I H, Bethe s d a , MD 20892 , USA 5 Nati onal In sti tu te o n De a fne ss and Other C o mmuni cation Disorde rs (NI D CD ), N I H, Bethes da, M D 20892 , USA 6 U nit ed S t ate s N ati on al Scie nc e Found at io n (N SF ), Direc torat e for C om puter a nd Inform a tion Scien ce an d E ng ineering ( C IS E ) , A lexa nd ria , VA , 22314 , US A 7 U nit ed S t ate s N a tion a l Scienc e Fou ndat ion (NS F ) , Direc to rate for E ng ineerin g ( EN G ) , Al ex andr ia , V A, 22314 , US A A bstrac t Neurom orph ic eng ineering has ma tured over t he p ast fou r deca de s a nd i s c urrent l y experien cin g explos iv e grow t h w ith t h e pot ent ia l to t r a nsfo rm b iomed ical en gin eering a nd n eur ot ec hn ol o gies . Par ticip ants at the Neu romo rphic Prin cip les in Biom edic ine a n d H ea lthca re (NPBH ) W orks hop (Oct ober 2024) — repres ent ing a b roa d c ross - sec t ion of th e co mm uni ty , includ ing early - ca reer an d esta blis hed scho l ar s, engineers , scien tist s, c lin icia ns, indus t r y, and fun ders — c onvened t o disc u s s the s tat e of t he ﬁeld, curre nt and fut ure challenge s , a nd strate gies for adv a n cing neuromor phi c re search a nd dev elopm ent for biom edic a l app licat ions . Pub licly a pp roved reco rdings wi th tr ans cr ipts ( https ://2024.ne uro - me d. or g/ pro gr am / se ssio n - v ideo - an d - tran scr ipts ) a nd s lid es ( h t tps ://2024.neur o - med.org/pr ogram/ses s ion - s lides ) c a n be fo und at the works hop web s ite. Ke ywo rds : neuro m or phic, workshop , b rai n - ins pired , Neu roAI , neur omod ulat io n, bioin terfa c es , br ai n - mac hin e int erfa c e Int r o duct io n Neurom orph ic e ng ineerin g wa s orig ina lly desc ribed in the 1980s a s s imp le a n a log circu its t h a t emula te d th e tem pora l p roc e s sing o f s igna ls in th e bra in (e. g., s ilic on retin a (1 – 5) ). Th e ﬁeld h as s inc e evo lved f rom stri ctly sm al l - sc a le har dw ar e - spec iﬁc im plemen tat ions to encompass l arg e - sc a le h ardw a re (6,7) , har dw ar e /so f t ware co - des ign (8) , AI al gori th ms (9) , compu ta tion a l app roa ches ( 10) , desig n p rinc iples (11) , a nd b ioelec tron ics (12,13) th at c ollec t ively mimic b r a in and ne rvo u s s yste m f un cti on s . A su b set o f neuro morphic eng ineering is sh o w ing prom i s e in m edic a l a pplica t ions (14 – 1 6) , incl uding i n lea rnin g ci rcui ts (17) , seiz ure p redic tio n (18) , seiz ure det ection (19 – 22) , neuropr osthes es (23 ,24) , neur omodulati on (25) , b rai n - m ach ine int erfac e (26,27) wi th on - c hip spike sort ing (28,29) , an d st retchable neuro morphic wea rab le devic es (13,30 , 31) ( F ig ure 1 ). 2 Fig ur e 1 . Vi su al o ver v i ew o f hum a n b od y sy ste ms whe r e fut ure dev ic es ba sed on n eur omorphic prin cip les c ould be a pplied . Th e se a pp licat ions includ e , bu t ar e not l imi ted to , d eep br a i n stimul ati on, vis ua l p rosth esis, c o c hlea r im pl a n t s , v a gus nerv e stim ulat ion, es ophagea l mo nit ori ng , p ace mak e r s , s p inal c ord s timula t ion, res pira t ory pac ing, sen sory res t orat ion , g as tr oin te stin al sti mul ati on , sac r a l ner ve stimul atio n, pr o stheti c limb co ntro l , we ar a b l es , a n d Bra ille rea din g . F igure c redit : adapted with perm is sion from El is a Don ati ’ s Neurom orph ic Princ iples in B iomed icin e and H e a lt hca re w orkshop pres ent atio n in Sessi on 2. In sp ired by recen t ad v a nces in neurom orph ic en g ineering for he a lth, Natio na l Ins titut es of Hea lth (N IH) pr o gram o ﬃci al s — repres ent ing the NI H B rain Resea r ch Th rough Advancing I nnova t iv e Neurotec hnol ogie s ® (BR A IN) I nit i a t ive, the Na tiona l In stit ut e of Biom edica l I mag ing a nd Bioeng ineerin g (NIBI B) , and t he Na tion a l Inst itut e of Neurolo gic a l Dis ord ers an d S troke ( NIND S )— c olla borat ed w ith U.S. Na t ional Sc ienc e Found a tion (NS F ) prog ram direc t ors — rep res en t i ng the D irect or a t e for E ng ineerin g (ENG ) a nd the Direc t ora te for Compu ter a n d I nfo rmat ion Sc ienc es ( C ISE )— to org anize t h e Neurom orph ic Prin c iples in Biom edic ine a nd Healt h (NPB H) hy bri d w orks h op on 21 - 22 October, 2024 , in Ba lt imore , M ary l an d , US A . See T a ble 1 for a lis t of the in vit ed spe ak er s , co - ch air s, p a n elists , and guidin g ques t io ns . Th e N PB H workshop had 171 p a rticipant s (84 in - pers on; 87 v irtual) , ran g ing from ea rly - career t o es tab lish ed scho l ar s and includ ing biom edica l eng ineers , neurom orp hic en g ineers , neu roscien tist s, ma teria l s cien tist s, c linic ia ns, sma ll b us iness e s , federa l fu nder s, a nd priv at e inv estors . Th e p a rt icipa nts c on ven ed wit h t he g oal t o explore t he in ters ec tion of n eurom orph ic en gin eering wit h h e a lth care a p plica tion s. A forw a rd - lookin g communit y ro admap will b e pub lis hed s epara t ely b y w ork s hop att endees. Thi s a rtic le 3 s erves to su m mar iz e wor ks ho p p re se nt ati on s and di scu s sio ns an d hig hlig h t opportunities for neu romorp hic eng ineerin g in neur otec hn olog y a n d biomedic a l eng ineerin g . S e cti o n 1 (To pi cs Cove r ed and Major Di scussions) and Se cti on 2 (Op en Quest ions and Crit ica l Cha lleng es ( B a rriers ) ar e b ase d s olely on w orkshop discuss ion s. S ect ion 3 ( Ou tlook an d V is ion for Next St eps ) c om pris e s pers p ect ives from works hop part ic ipant s an d pro g ra m rep res entat ives tha t were s yn thesiz ed f ollowin g the workshop . Ta bl e 1 : S pea kers , Co - c hai r s, P a nelis ts , and Gui ding Q ue stion s Sp eak er Nam e Title of Talk Co - cha irs/ Pane l ists Guiding Que stio ns Day 1 – 21 Oc tob e r, 2024 Timo th y Deni s on ( Univ ersit y of Oxf ord , Eng l and ) Keynote: Tow a rd s “Neur o/Physio - m orphic ” Dev i c es [G enera t ion 3? ] Se ssio n 1 To pi c 1 - E pile psy & Cort ic al Disorde rs ( Neu romodul ation) Willia m Sta c ey ( Univ ersit y of Mi chi g an ) Data Drive n c lass iﬁca tion wit hout a go ld s tan d a rd: epilept ic HF Os Se ssio n 1 Co - ch ai r s: Gr a ce Hw an g (N IH/ NIN DS) , J e ssi c a Fal co ne (N IH/ NIBIB ) P anel i sts: Syd ne y Cash ( Harv ar d M as s ach use t ts Gen eral H os pita l ) , Natha n Cr one ( JHU) , K endall L ee (M ayo C linic ), Giac omo Va lle (C ha l m ers Univ ers ity of Te chno lo g y, Sweden ) , Chr i s Roz ell (Georg ia In s tit ute of Technolog y , G IT ) 1. If you ha d a low pow er, i m pla nta ble chip that could run any algori thm ( re al - tim e, closed - loop , m ac hine lea rning ( ML) , artiﬁ c i a l intellig enc e ( AI ) , etc. ), how w ould you u s e it ? W h a t are the encodi ng and de c oding c ha lleng e s in ep ileps y, m ovement, and p sy chi atri c di so rd er s? 2. What s trat egies (materia ls , tec h nolog y, a lg orithms, th eory) can we u s e to in terf ace w ith th e brain in a more neuromorphi c m a nner ? Is it c ritic a l for n eurom odulat ion (h a r dwar e o r softw are ) to mimi c the br ai n? W h y? 3. Wha t imp act coul d m a j or neu romod ula t ion c om panies hav e in th is proc ess? W hat is n eeded for c linic a l tra nslat ion? Sr idevi S a rma ( Johns Ho pk in s Un ivers ity (J HU )) A Res tin g Sta te E EG Bi omark er Poi nts to Seiz ure O ns et Zon e V iktor J irsa ( Ai x - Mars eille Un ivers ité , Inse r m , Fr an ce ) V irtu a l Bra in T w in s fo r M edic ine - Enc ep halomorphi c Sys tem s Sessi on 2 To pic 2 - Human- m ac hi ne in te rfa ce s (Pro s the ti c s) Jame s Co t ton ( North wes tern Un ivers i ty ) C linic a l G a ps and Un ders tan ding Rea l Wor ld Outco mes fo r P ro sthe sis U se rs wi th A I T ech nolog ies Se ssio n 2 Co - chairs : Lu ke Osbo r n ( C as e Western Res erv e Univ ers ity ) , F ra n c isc o Valero - Cue v a s (Unive rsity of Southern C alif ornia ) Pa nelis ts : Xing Chen (Un ivers ity of Pit t sb urgh) , N iti sh Th ak or ( JHU) , H elen Huang ( N o rth C aro l in a St ate Un ivers ity ) , Vika sh G ilja (Un ivers ity of S a n Diego 1. H ow can neu romorphic prin cip les imp act t he a bilit y to c onv ey com plex or hi gh - res olu tion s en s ory info rma tion thro ugh neur a l int erfa c es? W h a t’s the c urren t limit a tio n? P rocess ing power, s ensors, neu ral in terf ac es, ph y s iolog ica l underst a nding ? 2. In w h at w ay s co u ld a l ow p ow er sys te m , whic h ad ap t s to an d e vo lv es wit h t he us er , bene ﬁ t h uman - Ranu Jung ( Univ ersit y of Ar kan sa s ) El isa D o n ati (Un ivers ity of Zuri c h & ETH C onv ergence: Dep loyin g Neuroelec t ronic In terf aces Neurom orph ic Ha rdwa re for C los ed - loop in Hea lt hca re 4 Sp eak er Nam e Title of Talk Co - cha irs/ Pane l ists Guiding Que stio ns Zu ri ch, Sw it zerla nd) (UC SD )) , Sh ih - C hii L iu (Un ivers ity of Z uric h & ETH Zu ri ch , Sw itz erla nd ) mac hin e int erfa c e s or reha bilit a tio n tools ? 3. What a re the beneﬁts for us in g neuro morphic co m putin g in neur a l pros t hes i s ? D oes it bring extra ben eﬁ t for rob us t d a ily u se? A re th ere unknown fun ction al bene ﬁ t s ? 4. Is ther e an o ppor tuni ty to u se neu romorp hic prin cip les to rev ea l mec han isms t hat giv e ris e to f unct ion, a nd exp la in dys funct ion , while poin tin g a pat h to c linic a l treat ment? Day 2 – 22 Oc tob e r, 2024 Zhenan B a o ( St anfo r d Un ivers ity ) Keynote: L earnin g f rom s kin : from M at erials , S ens ing F unc tions t o Neu romorp hic Eng ineerin g Se ssio n 3 To pi c 3 - M ate ri al s for Ne u rom orp hic s (Devi ces ) Duygu Kuzu m ( U CSD ) Innov a ting Beyon d E lect rop hysiolog y : M ult imoda l Neura l Interfa c es and Neurom orph ic C o - Des ig n Se ssio n 3 co - Ch a ir s : Duy gu Kuzum (U CS D ) , Shan tanu Ch ak rab ar ty ( Washi ng ton Un ivers ity S t. Lou is ) Pa nelis ts : Gina Adam (Geo rge Washington Un ivers ity ) , Jenn ifer H asler ( GIT ) , Gert C auw enb erghs (U CSD) , Ulkuhan G uler (Wo rces t er Polyt ec hn i c Instit ute ) 1. What a re th e opp ortu nit ies in using new mat eria ls a nd d evic e tec h nolog ies ? How will t hes e in terfa c e wit h t he t iss ue (bio compat ibilit y) a n d compl eme n tary metal - oxide semi cond uc tor ( CM OS ) elect ronics ? 2. Is neuromorphi c t he o nly s ol ution to power, ba ndw id th and thro ughput prob lems for im pla n ts? 3. What a re th e c h a lleng es in ma n ufacturing neur o morphic chi ps for b iomed icin e? C os t? 4. What a re the beneﬁts of on - chi p lea rning or a dap tat ion fo r neuro morphic hard ware? 5. What a re th e c h a lleng es in tran slati on o f neu romo rph ic to the c linic ? W hat role s hould ind ustry pla y in that? T ony Lew is ( Br a inChip ) Flow Ma c hine s a nd Neurom orph ic Devic es Deji Akinw a nde ( Univ ersit y of Te xas Au sti n ) To w ard s He alt h Monitori ng & Com pu ti ng u si ng A dvanced Ma teria ls Se ssio n 4 To pi c 4 - Me dical Imaging, Wea rabl es and A nal ysis Sihong W a ng ( Univ ersit y of Chi c ago ) Bio - elec tron ic inter fa ces: from hi gh ﬁ delit y bios ens ing t o int egra t ed neu romorp hic edg e compu ti ng Se ssio n 4 co - Ch ai r s: R alp h Et ie nne - Cu mmi ng s (JHU) , Jen nifer Bla in Ch ris ten (A rizon a State U niv ersit y ) Pa nelis ts : An dreas A ndreo u (J HU) , R obert Stev ens (JH U) , Pa mela A bsh ire (Univ ersit y o f Mar y lan d , C olleg e Park ) , 1. If you ha d a low pow er, i m pla nta ble chip that could run any algori thm ( re al - tim e, closed - loop, ML , AI, et c. ), how would you u se it? 2. What new stra t e g ies (mat eria ls, tec h nolog y, a lg orithms, th eory) can we u tiliz e? 3. How ca n neu romo rp h ic p rinc iples ma ke t hese idea s a rea lity ? Er ik a S chm it t ( Ca mbr ya ) Novel n eurom orph ic a lgorit hms for a dap tiv e lea rning a t t h e edg e 5 Sp eak er Nam e Title of Talk Co - cha irs/ Pane l ists Guiding Que stio ns J. Brad Ai mon e ( San di a N ati on al L aborat ories ) From new neuro ns to ne w ch ip s: a p ath to a new w a y of thinki ng about the br a in? Amp ar o G üe me s (Univers it y of Cambridge, Eng l and) 4. What im pac t/role d oes m a jor imp la ntab le / w ea rables compan ies ha ve in th is pro c ess ? W hat is the pa t h wa y fo r clin ic al t rans l a t ion? Pu bl icl y a ppr ov ed r ec or din gs w it h tr an scri pt s can be fou n d on th e N PBH w ork sh o p we bsit e at https :/ /2024.ne uro - me d.o rg/ pr o gra m/ se ssi on - vi deo - an d - tr an scr i pt s .P rese ntation sl i des c a n be found at http s :/ /2024.n e uro - me d.o rg/ pr o gra m/ se ssi on - sli de s . A p pendi x 1 con tai n s a syn o psi s of t he me et i ng . A ppendi x 2 pre se n ts t h e p a ne lis ts ’ P osit ion St at eme nt on the Ne eds , Cha lle ng e, Co ntr ib utio n, Imp ac t a nd Inves tme nt fo r the ﬁ eld . T he t opics disc uss ed at t his works hop h a v e reverbera t ed a cro s s rela t ed m eetings spo ns ored b y t he NIH and t he N SF inc lud in g : (1 ) the 23 Oc tobe r 2024 NIH Wed n esda y Aft ernoon L e c tu re S eries ( WA LS ) lec tu re ( https :// videoc a s t .nih.gov/ watch= 55007 ) i n w hich Br a d Aim one i ntroduced hi s Sand ia research gro up’s prelim ina ry implem ent atio n of the ad ul t D ros o p h ila melanogast er con nectome on the L oihi neuromor phic pl atf or m co mprising 14 0K neur ons and 50M syn aps e s (32) ; ( 2) the 12 - 13 November 2024 N IH BR AIN NeuroAI Works hop (33 , 34) ( Day 1 videoc as t , htt ps :// videoc a st.ni h.g ov/w atch=551 60 ; Day 2 vid eocast , htt p s://videocast.nih.g o v/w a tch=55262 ) ; (3) the 27- 29 A u gust 2025 NSF - spo n sor ed Neur oAI +Beyo nd Works hop ( NSF Aw ard No. 2533276 ; ( 35) ) ; (4) the 27 A u g u st 2025 NI H BRA IN M ulti - Coun cil Wor king Group meet ing ( http s :// videoca s t. nih.gov/w a t c h=5687 2 ) a t which the N euroAI Innova tion Domain was publicly intro duc ed b y NI H B R AIN D ire c tor Jo hn Ngai a s par t o f t he B RAIN In itiat ive ’s anti ci pat ed future res e a rc h r oad m ap (36) ; (5) th e 13 N ovember 2025 N IH / BR AIN M inis ymp os iu m on Neu romo r ph ic E ngin eering for C linic a l Care h eld at the B R AIN - sponsored 12 th A nnual Inst itu te of E lec trica l a nd Elec t ronics E n g ineers ( IE EE ) En gin eering in M edic ine an d Biolog y Societ y ’s ( E MB S ) Int ernat iona l C o n ference on Neura l Eng ine eri ng and R e hab ilita t ion ( N ER 25 ) ( http s :/ /neuro .embs .or g /2 025/nih - br ai n - in itia t ive - min isy mposi u m/ ; http s :// www.yout ube.com/ watch?v=ljsd Sc N vrKg ). To pi cs Covered an d Major D iscu ssions T he w orks hop empha siz ed b idirec tio n a l kn owled ge exc han ge b etween en gin eer s and c lin icia ns, with par ti cul ar fo cu s o n whether and h ow n euromorph ic a pproac h es can a ddres s c ritic al c ha lleng es for c lin ical applications. Throug hout t he meeti ng , p a rt ic ip a nts hig hlig hted t he im porta nc e of focu sing o n rea l - wor ld imp act rath er th an m erely m imic kin g b iologica l s tr u c t ure s a nd d y n a m ics. This perspec t ive wa s rein forc ed th roug h cas e s tudies of dev ic es t hat h av e s ucc essf ully trans lat ed t o clin ica l us e, su ch as ad ap tiv e de e p bra in s t imula tion (DB S ) fo r Par ki nso n’ s di sea se (PD) a nd s ens ory feedbac k sy stems f or pros t het ic lim bs . Cur ren t U nd ersta n di ng and Sta te - of - the - art Day 1 Ke ynot e : Tow ards “Ne uro/P hy sio - morp hic” Devices [Ge ne ration 3?] Timo th y Den is on pres ented on the oppo rtun ity a nd challen ges fo r us ing elec tron ic ha rdwa re an d so ftw ar e to intellig ent ly stimul ate the ne rv ou s syste m i n re al ti me t o ame li orate sy mpto ms. He focus ed on a dap tiv e DB S s yst ems like t he M ed tron ic st imu l a t or (e.g ., Perc ept TM (37 ,38) , Sum mi t RC+ S (39) ) tha t ha ve sho wn clin ic al ben eﬁ ts for PD pa tie nts . He em ph asiz ed t he im porta nce of n ot jus t n eurom orph ic but a lso clos ed - loop “p hysiomorph ic” approaches th at u s e ph ysiolog ica l ma rkers to a dj u s t s t imu l a tio n in rea l 6 time to be tter m a tc h ind ividu a l bra in s t ates . Thes e sy stems u tiliz e du al - th res hold hom eosta tic a pproac h e s ins pired by n atu ral p hysio logica l s yst ems, w hic h outp erform s impler th resh old - based c ont rols . A lg orith mic p ro c essin g is perf ormed on ext erna l s yst e ms where p a rameters c a n b e up d at ed th roug h t elemet ry . A lt hough these s yst ems d o not c urren tly em ploy neurom orph ic ha rdwa re or a lgorit hms , Deni s on expla ined that the s e phy siomorphi c approaches a re ins p ired by hu m a n phys iolog y. H e pres en ted a v i si on for “ Genera tio n 3” devic es t hat wo ul d m odel a nd rest ore p hysiolo gy , noti n g th at current neurom odu l a tion dev ices l ack predic t ive ca pabilit ies . De nison fur ther note d tha t a lg orit h m s are of ten r udime ntary , due t o lim ited pow er bu d g et s on curr ent d evic es . Wh ile neur omorp hic tec h nolog ies may c onfer a d v a ntag es in pow er eﬃcien cy , th ey n eed t o de m onst ra te adva nta ges ov er c la ssic dig ital s igna l proc essin g and pati ent - releva nt perfo rma nce m etrics . A gr eat way to eng age the c o mmu nity o n nov el a lgorit hm des ign c ould be t hrough “ b a ttl e o f the a l gori th ms” co mp eti tion s t h at ev al u ate d iﬀerent dev ices running diﬀerent alg orit h m s for diﬀerent dis eas es , allow ing res e a rc her s t o mon etiz e t heir idea s . Topic 1 - Epi leps y & Cort ic al Dis o rde rs ( Ne uromo du lati on) Willia m St ace y exp lain ed that cu rrent surg ical d ecis ion s for int rac t a b le epileps y largely rely on em piric a l ob ser vati o n s and visua l a n a lysis of elect roenc epha log raphy ( E EG ) pat t ern s deriv ed fro m d eca d e s - old met hod olog ies. E xis ting neu ro m odu lation devic es , su ch a s DBS, res pons ive neu r o s timul at ion (RNS), a nd vagus ner ve s timul at ion ( VN S ), w ere no t opt im ized f or epilep sy trea tmen t an d s t imula tion pa r a m eters were in itia lly borrowed from ot her a pplica t ion s . Som e c los ed - loop RN S devic es s ense a nd s tim ulat e wit h a four - ye a r ba ttery life but su ﬀer fr om extremely low proc es sin g p o w er . For exa mple , o nly ﬁve minute s of continuou s data fro m a few c ha nnels are typica lly sa ve d ove r the course of 24 hour s (25) . Iden tif ying biom a rkers to de ter mine epilep tog enic z one s rem ain s an a cti ve are a o f researc h o wi ng i n par t to the lack of gold standard v a li dation . Thr e e promis in g me tho ds for t reatin g ep ileps y w ere pres en ted , including : ( 1) h igh - f req uenc y oscilla t ions ( HFO s ) ( by S ta ce y ) f ro m i ntr a cr ani al E EG ( i EEG ) (40 – 42) ; ( 2) res t ing state iE E G ( b y S ridevi Sar m a ) an al y sis devoi d of s eizu re da t a (43) ; ( 3) v irtu al b ra in m odels ( by Vi k to r J ir sa ) con str u c te d fr om pers onaliz ed neu roima g ing dat a se t s (44 – 47) . Stac e y expla ined th at n euro m or phic t echnol ogies a nd appro a che s are ra rely u sed in ep ileps y in t he U nit ed S tat es . Demo n strate d us e s inc lude real - time seizu re predic t ion (I BM T rueNort h (48) ) and re al - ti me HF O dete cti on from i EE G in Zuric h Swit zerla nd (22, 49) . Topic 2 - Hum an - m ac hi ne int erf ace s ( Prost het ics ) T hree t alks w ere pres en ted o n c linic a l pers p e c tiv es o f neur al in terf aces a n d rehabilit a tio n. Ja mes Co tto n st ar te d th e sess ion w ith em phas i s on patien t p ers pectiv es, rev e a ling tha t pa tient s prioritiz e com fort , redu ced pa in, a nd psyc holog ica l fact ors over advan ced te c h nica l feat ures, with s tu dies s how ing s o c ket ﬁ t is su es an d s kin p roblem s as pat ient - rep orted priorit ies (50) . A s de sc ri bed by Ranu Ju ng (51) , t he f u ndam enta l c h a lleng e of b iohy brid s y stems rema ins la rg ely unsolved : b iolog ica l sy stems a re n on - lin ea r , non - statio n ary, mul ti - scal e and pl ast i c, yet c urr ent tec h nolog ies rem a in rela t ively ﬁxed . Re se ar ch e ﬀ or ts to a ddre ss th i s probl em i n pr ostheti c s d ate back to 2006 an d we r e orig ina lly f und ed a s pa rt of an NI H Bioen g ineering Res earc h Part ners hip Prog ram , DAR P A’ s RE - N ET and HAP TI X pr o gra ms, and the Join t Wa rﬁght er M edica l Res e a rc h Prog ra m . F or exa m ple, a fu lly im pla nt ed syst em was dev eloped th a t pr ovide d sensory feed back fr om se n s or s i n a pro stheti c hand v i a wireles s co m m unicat ion to im plan ted elec t rodes to direct ly s t imula t e nerves (52,53) . In one ca s e, 7 a pr osthe ti c h and p artici pa nt was a ble t o feel s ensa tion in his miss ing h a nd for th e ﬁ rst time in 74 y ears (54) . T he pro stheti c tec hn olog ies p re s ent ed by Ju n g w ere not neur omorphi c , bu t se r ved to represen t an asse ssme nt o f the st ate of the art . El isa Do na ti pres en t ed n eurom orph ic ha rdware ap pr o ach es th at spec iﬁca lly ena ble d eﬃc ient sensory feedb a ck b y d irect ly s timula t in g nerv es wit h patt erns tha t mimic nat ural s ensat ion (24) . These s yste ms hav e c ompara ble ac curac ies to deep learn ing - b a sed ap pr o ach es , ye t us e s ign iﬁcan tly fewer pa ra m eters and less p ow er (55,56) . D on ati al s o describe d a “ Neuromorp hic Tw in ” as an em erg i ng t ool th at p hysica lly emu late s the s o ma tose n s o ry cor tex of hu man p atie nt s to cre ate ne uro sti mulati on patter n s . Thi s emu la tor c ou ld improve t he q ua lity of res t ored feed bac k for n eura l pros t he s es by be tter re produc ing natu r al sen sa ti on s (57) . Ot her p romising bio medica l n eurom orphic in terfac es c overed in t his s essio n inc lud e ca rd i a c a ctiv ity mon itorin g (58) ( f ro m Don ati ) and r e al - tim e pers onaliz ed res p irat ory p a c ing (59) ( fr om Jung ). Day 2 Key not e: Learn ing f rom s kin: from M at eri als, S ens ing F unc ti ons to Ne uromo rp hic E ngi ne eri ng Zhenan B ao pro vided an u p date o n a c ollec tion of s kin - like elect ron ic p olym ers an d mate ri al s wi th enh anc ed elec tric a l ca p a bilit ies co m par ed to tr ad ition a l s e mic on duct or mat eria l s . She p resen ted fro nt - end proc es sing pre s sur e se nso r s o n a stre tc hable sub str ate wit h over 1,000 t r ans i stors in t he int eg rated ci rcui t. The se p re ss ur e se nso rs ex ceed ed hu man ta ctile res olu tion (bet ter t ha n 500 mic rons ) , whi ch wa s de mons trated with im pro vem en t in Bra ille rea d ing ( 60) . Another tech nology, t er m ed “neur omorphic sk in” , h ad n o rigid elect ronic c ompon ents an d e mula te d t he s ensory feedb ack f unct ions of b iolog ica l s kin , inc lud ing mu ltimoda l rec ept ion, n erve - like p uls e - t ra in s ig nal c ond ition ing, an d closed - loop act uation (13) . Ba o no ted t hat prot o t yp ing t hes e nonc onven tio n a l mat eria ls pr e sen t s ma n y c h a lleng es, one of wh ich is tes t ing long - t erm bio c om p a t ibilit y. B ao co nclu de d tha t othe r key c h a llen ge s for the c o mmun ity ar e pr ototyp ing no nconven tional m a teria ls and id entif yin g whic h bio m edica l ap plica tions w ould beneﬁt mo s t fr om ne uromorphi c or hybri d design approa c h es . Topic 3 - Ma ter ia l s f o r N eur o m o rph i cs (D evi ces) Two t alk s fo cu sed on n ovel ma teria ls w hile a t hird t a lk was a bout a commerc ia l neurom orphic p rocess or. F or neu ral in terf ac es, Du yg u Kuz um d escrib ed ho w transp a re nt g r aphene - bas ed mic roelec trod es ena ble d mu ltim odal exp erimen ts by s i m ultan eous ly c ollect ing elec tro phys iolog ica l rec ordings while a llow ing opt ic al ima g in g a nd st imula tion witho ut c rosst alk (61) . As a n ex a mp le, sh e desc ribed r es is tiv e RAM - based neurom orphic p rocessors dev eloped in her la b that perfo rm ed eﬃc ient sp ike - sorti ng wi th 30 - 50 ti me s les s en ergy c onsumpt ion t han trad ition al ap proac hes (62) . Kuz u m sho we d how pri nc iples of neuro morphic co - des ig n c ould be a p p lied t o dec ode in divid ua l c ellula r c a l c ium ac t i vit y in deep er bra in la yers us ing on ly s urfa ce p otentia ls measured f rom graphe ne - based elec trod es an d neuro m or phic c omp u t ation a l mod els wit h rec u rrent neural n etw orks . This t echniq ue su cce ssf u ll y e xp and ed th e spatia l rea ch of d eploy ed elec trod es ( 61) . Deji Akinwa nde p resented two - dim en s iona l (2 D) ma teria ls ( e .g. , graphe ne , less than a nano m ete r thick ) for use i n “or gan - tr on ics ” ( in vivo sen sin g) and “ ski n - tro nics” (direc t s kin in terf ac in g). Ap plic atio ns inc lud ed elec trop hy siolo gica l mo nit oring (63) , im plan t a b le sen s in g ( pac emaker (64) ) , an d a w r ist - wo rn 8 bioim ped anc e sen s or for moni tor ing bloo d pressure w ith c linica l - gr ad e a c cur acy (6 5) . He a ls o pres ent ed “Ato m - r esi sto rs” ( m emris t or s in a t o m ically th in ma teria ls ) th at d e mons trat e both bi na ry a nd analog res ist anc e swit chin g (66) in neu ral n etw orks, wit h c lass iﬁca tion ac c u racy a pproac hing t heoret ica l limits in a Mod i ﬁe d Nati o n al In sti tu te of S tan d ard s a nd Te chn ol o gy d at ab ase ( MN IS T ) digit r ec og nit ion ta s k. On th e c o m merc i a l s ide, Tony Le w is de scribe d Bra inCh ip’s A kid a a rchit ec tu re , wh ich opera t e s as an even t - b a s ed, dig ital, n ear - mem ory co mputin g s ys tem u s in g w eigh ted spikes . Akida a c hiev ed st ate - of - the - a rt res ults in a myr i ad of ta sk s ( gest ure rec ognit ion, eye t rac king , denoisin g, a n d a udio keyw ord sp ott ing ) wh ile o pera ting at extrem ely low po wer (0 .25 milliw a tts ). Le wi s presen ted d a ta sh o w ing tha t Br ai nCh ip ’s T emp ora l Ev ent -B as ed Neura l Net wo rks pro c ess ing unit a chiev ed bet t er perplexit y s c ores tha n st ate - of - the - ar t larg e la n g uag e m odels ( e .g. , M a mba an d Ll a ma 3. 2 ) us ing a fra ctio n of the tr aining d at a an d h al f th e mod el s ize. Topic 4 - Me dica l Im ag ing , We ara bles a nd An alys is On e ta lk f ocus ed on weara bles, while tw o t a lks fo c us ed o n a lg orith ms. Si hon g W an g pres ent ed stre tchabl e , so ft se micon du ctor p oly me rs th at y ielded s tab le sig nal q u a lit y, direc t ly a dh ere d to tis s ue sur fa ce s ( s uch as a b e at in g he ar t ), and suppre ss ed the fo reign body res ponse b y incorpora ting immu no modu la tory pro per tie s int o th e poly mers (64) . H is p re se nt ati on al so su gge ste d the n eed t o shift the proc essing of s en s or data from oﬀ - ch ip , cl ou d co mp uti n g to on - chi p , ed g e co mputin g . W an g then dem onst ra ted a f ully s t retchab le, w ear abl e arr a y for on - chi p cl a ssiﬁ cat ion of elect roc a rdiog r a m ( E CG ) si gn als with elec t ro c hem ica l trans ist or - b ase d ne uro morphic h ar dw ar e (30,67) . Br ad A im o ne pres en ted neurom orph ic a lg orit h m s t h at leverag ed s tructura l rela t ions hip s f rom c onnect om es rat her t han rely ing s olely o n s equen tial t ime - b as ed a na ly s es , and wh i ch e x cel l ed at s amplin g p roblem s (68) a n d ﬁnite elem ent ap proac hes (69) . H is ne uromorphi c ap pr oa ch to s olvi ng partial diﬀeren tia l equ at ions show ed dynam ics rem a rk a bly sim ila r to biolog ica l neura l recordin gs , de s pit e dra stic diﬀeren c es in a rchit ec tu re s . Aimo ne note d th at b oth ne uromorphi c approa c hes a nd t he b iolog ical br ai n utiliz e d het erogen eity , p a ra llel proc es s ing, and s pat i a l t empo ral p roc e s s in g . In a rec ent prep rint , A imon e addit iona lly h igh lights th e se bio logic a l feat ures as crit ica l to dev elopin g the or eti ca l frame wo rk s to char ac te r iz e t he lim its a nd cap abilities of neuromor phic c omp uting (70) . On the c ommer cial s ide, E rika Sc hmit t d escrib ed Ca mbrya’ s w ork dev eloping n eurom orphic a lg orithms to s olve t he au dito ry blin d s ourc e s ep a rat ion p roblem us ing densit y n etworks t h a t do n o t require p rior tra in in g (71) . Reg ulator s and F u nders S ession P rogram staﬀ from the NSF, NIH, A r my, a nd Defens e Adva n c ed Re se ar c h Projec ts Agen c y (DAR P A ) pres ent ed relev a nt fun ding opp ortu nit ies and co mpe titi on s to encour age p arti cip a nt sub m is si ons . A ddit iona l d etails ca n be f ound in App endix 1 ; funders ’ pr esen tat ions ca n b e retriev ed from htt p s://2024.neuro - med.or g /pr ogr a m/ s es sion - slides . T h e Unit ed S tat e s F ood an d Drug A dmin ist r a t ion (F D A) provid ed an ov erview of p rogra ms f or overs eein g nov el med ica l d evic es, ex pla ining a t hree - tier c lass iﬁca tion system a nd hig hlig hting t he need to eng age early thro ug h the Q - sub mi s sio n pr oce s s. Key c onsid era tions rais ed during dis c ussions throug hout t he meet ing include d: ( 1) diﬃcu lty of va lidat ion of a dap tiv e systems ; ( 2) s ec urit y a nd saf ety , including the need for fa ils a fe mec han isms in g enera l , and cyber sec urit y in im pl a nta ble s yst e m s ; ( 3) ma nuf acturing 9 veriﬁ c ation , inc lud in g c h a lleng es in v erifyin g n eurom orphic hard w a re manuf a c t uri ng ; and ( 4) edu c a ting res earc hers abo ut the reg u lat ory pro c ess . It wa s note d th at t he FD A Gu id a nce d oc u m ent on Physiologic Cl ose d - L oop C ont rol D ev ic es (72) an d the I n t erna t ional E lec trot e c hn ical Co mmis s ion ’s ( I EC ) Do cume nt # 60601 -1- 10 ( 73) we re g ood s tarting point s for d evic e dev elopers . T he Open GPU Fou nda tion al so pres ent ed t heir in teres t an d rat ionale for in ves t ing in h a r dwar e th at de mon str ate d ad v ant age s i n tem pora l c o m pu ting, noisy s t oc hastic c o m puting , energy eﬃ cienc y, a nd mult idim e ns ion al c o m putin g . OpenGPU a ddit iona lly disc u s s ed its aims to creat e sta nda rdiz ed int erfac es be t ween ha rdwa re a nd “w et ware” ( e.g., org a noids) . Op en Q ues tion s and C riti cal C halle n ges (Barriers ) T he neuro morp hic biom edi c a l ﬁeld f a c es severa l s igniﬁ cant c halleng es tha t m u s t b e a ddres s ed to rea lize its fu ll pot ent ia l. Thi s se ct io n prov ides a th e mat ic organ izat ion of challen ges dis c u ssed and deb a ted by NP BH w orkshop p an e li st s an d par ti ci p an ts . Clin ic al Cha llen g es Epilepsy , ne urolo gica l , and psychi atr ic conditio ns C lin icia ns exp res sed m aj or challen ges inh erent to th e stud y of epilep s y a nd the identiﬁ ca tio n of th e epilep tog enic z one. T h e state - of - th e - a rt dev ices u sed to st imu l a t e speciﬁc bra in regions to p re dict or a bort s eizur es are not suﬃ cient. I mpla nta ble devic es s uﬀer from s ign i ﬁ can t lim itat i ons in sa mp ling r ate and b and wi dth , bit pre c is ion, and c o m putation a l capa bilities , ham perin g res e a rc h prog ress . C lin icia ns no ted tha t altho ug h t h e ﬁeld is mov in g a w ay from a on e- se ttin g - ﬁt s - a ll mo del a nd to war d s pers onaliz ing th e s ett in g s on d evic e s t o meet the need s o f indiv idu al s , th ere rema ins a n ur gent n eed for devic es wit h more c hanne ls, custo miz able sam plin g rat es, an d g reater s ig n a l - to - nois e ra tios in fully imp la ntab le wireles s form - f acto r s . Over the c ourse o f th e wor ksho p, p a nelis ts ident iﬁed a crit ical need for int ermed ia te bioma rkers that co uld pro vide ea rlier indic a t ions of treat ment eﬃ cacy prior t o th e ch a n g e in c linic a l sym ptoms , p a rt ic ul a rly in ps y chia tric a p plicat ion s. D ata c ollect ion limit at ions T he lim ita t ion s of c urrent data co llect ion were al so d i scu sse d . S om e devic es do not reco rd a nd sav e d a ta continuou sly, and hi gh - res olu tion data is primarily a va ilable on ly d uring inp a tient m onit orin g ra t her t h a n in a mbu l atory se tti ng s. Pa rticip ants prov ided speciﬁc tec hnical req uiremen ts for n eura l record ing s in a mbula to ry a n d inpa tient a c ute a bilit ies , inc luding cust omiza b le elect rode p lacem e n t, stim ulat ion pa ramet ers, sam plin g rat es (0.5 – 4, 000 Hz), an d continu ous d ata c ollec tion . Pa rt icipa nts id ent iﬁed a not her c ritic a l gap i n the need for t im e - s yn c hro nized , s im ultan eous m ult imo d a l rec ordin g wit h hig h s patio tem poral res olut ion f or elec tric a l a nd ch emica l s ignals in clu ding p otass ium, oxyg en, a nd variou s ne uro trans mitte rs. Dev eloping t his multim odal da ta c ollec t ion ca pabilit y w ou ld be a c rit ical im prov emen t for in pa tien t ac ut e s etting s . Eng inee r ing Ch alle nge s W hether neuromor phic a pproa ches o ﬀer re al - wo rl d a dvan tag e s over con ven tiona l met hods rem a ins an open ques t ion . S o me par ti ci p an ts ar gu ed that neuromorp hic systems excel a t s peciﬁc applic ation s that rec over fu nc t ion (e.g ., lim b loss , s igh t or hea ring restora t ion, perip hera l nerv e inj ur y , s pinal c ord in jury ) and requ ir e ultr a - low p o w er or rea l - ti me ad apt ati o n . O th ers s ugges t ed focu sing o n d emon st r a t ed 10 c apa bilit ies ra t her t han p hilos ophica l ad heren ce t o bio logica l prin cip les . Wh ile neur omorphic tec h nolog ies ma y be adv an t ag e ou s in pow er eﬃcien c y or l aten cy , worksh op p ar ti ci p ant s a s serte d th at the re must be d e mons trate d ben e ﬁ ts in va riou s de s ign cons ideratio ns , in c ludin g ma nufact ura bilit y , relia bilit y testi ng , m an ag e me nt of s igna l drift in analo g sys t e m s, syst em veriﬁ ca tio n , ris k ma na g em ent, and op ti miz a tio n ac ross m ult iple pa ramet ers (e.g . , no t only p ower c on s u m ption ). The d ichoto my of o ﬀ - chi p v s. on - chip c omp uting w as ra is ed mult iple t imes . P ar ti ci p ant s d isc u ss ed w hether futur e neuro m or phic c ir c uits shoul d c o m m unic at e wireles sly w ith sens ors an d st imu lators or be imp la nt ed toget her as a n integ rated sys tem. T he la tter ap pr o ach was p referred a cco rdi ng t o mu ltip le workshop p articipant s who m entioned that ra dio t elemet ry remains exp en s ive c ompa red to on - chi p c omp u t ation , with t elemet ry c ons u m ing upwa rds of 80% of t he p o w er bud get (74 , 7 5) . Th e develo pmen t of a ut ono mous (telem etry - f ree) c losed - loop neu romodulation syst e ms remain s a crit ic al c hallen ge t hat neu romorp hic tec hnolog ies ar e p ois ed to ta c kle (25) . F or med ical d evic es , som e part icipa nts exp res sed c yb ersecu rity c on c erns fo r sy st e ms th at require conti nuo us co mmu ni cati on wi t h t he cloud . N euromor phic a pproa c hes were s u ggested to be inher ently les s h ack able due to the pot ent ia l for s elf - su ﬃ cie nt, au tonomou s opera tion and on - bo ar d pr o ce ssi n g . Fo r b rai n - mac hin e int erfac es (B MIs ) , sev eral pa rtic ipan ts as s erted t h at key t ec hnical c ha llen ges inc lud e ad ap tin g to s ig n a l v a riability in n eura l int erfa c es . R e co rdi ng s c a n cha n g e dram atica lly w ithi n w eeks due t o shi ft s in elect rode pos ition or c ellula r res pon s ivity , w hich co uld be addressed t hro ugh on - c hip learn ing and ada pt at io n . Par ti cip ant s n ot e d th at t he c oncept of “ deg eneracy” — t hat biolog ic a l sys tems like th e hum a n bra in ac hiev e res ilienc e by adap tiv ely s elec tin g fro m m ult iple di stinct p at h w ay s and st ru ctu re s to imp lemen t a ny g iven func tion or beh av ior — pres ents a fund a m ental co mpl ica t ion for the encoding an d decoding of neur a l sig nal s . P arti cipa nts ou t lined oth er technic a l c hallen ges t hat re m ai n fo r neural i nt er fa ce s, i ncluding ti ssu e d am age during elect rode imp la ntat ion , lim ited s pat ial r esolu tion of recordin gs , a n d bulky h e ad - st age equ ipm en t for op tic a l imag ing . Re ga r din g h yb rid neu ra l inter faces , one part icipa nt n o ted th at the ﬁeld c urren tly lacks a cr os s - m apping for u nders t a n d ing bidirec t iona l c o n trol b etw een biolog ic a l a nd a rtiﬁ cial s ilico n neur ons . The pa rtic ipan t sug gested that a key measure of th e fun c tiona l embod imen t of a n art i ﬁ c i a l ne uron is w het her b iolog ica l neuro ns integ rate wit h it and trea t i t “ a s one o f their o w n”. Fo r example , a n a rt iﬁcia l em bod imen t of a giv en neu rona l c ell t ype c ou ld be en gin eered t o enhan ce repa ir functi ons a t the ne ural i nterfac e w hi ch imp rove me asu re s of bid irec tion a l c ontrol . Mat e rials Ch al le nges M any pa rtic ipan ts expres s ed t he v iew that i nte grati ng neurom orph i c dev ices at th e biolog ic a l int erfa c e rema in s a c hallen ge. F or bo th we ar abl e and i m pl anted device s , seamles s interfa c e s require bioc om p a t ible, stret cha ble, con formable, ﬂexib le , an d of ten bi odegr a d able mater ials , whi ch e licit a min ima l immu ne res pons e . P ar ti cip ant s v iewed t he organ , locat ion, a nd du ration o f th e int erfa c e to be key f ac t or s t o c on s ider w hen eva luat ing a dev ice’s bio compat ibilit y a nd sta bilit y . Th e re was co n sen su s that p r oto typ in g and a sse ssi ng the long - term biolog ic al su ccess of novel , no nc on ventional ma teria ls wou ld be diﬃcult. 11 Benc hm ar ks v s . D atas et s Work sh op part icip ant s id en t iﬁ ed th e need to develop spec iﬁc, sta n d ar dized d a tas ets rather t han foc usin g on benchm a rk s, w hi ch some pa rti cipants found to be re s tri c tive ( refer t o (76) fo r an e xamp l e ) . Aim on e a rgu ed t h a t benchma rkin g, des pit e its p opu la rity in ML/ AI dev elopm ent, would b e count erprod uc tiv e for the neur omorphic co m m unity as dev elo pers would be forced to u s e ou tdat ed pa radigm s w hile t he ﬁ eld of neuros c ienc e m oves forw a rd. And re o u oﬀered his t oric al cont ext from speec h p r oc es s in g , no ti ng th a t ben ch marks init ia lly h elped s ta ndard ize ev alua tion . H o w ever, inn ovat ion was eventua lly co n s t rain ed s inc e resea rchers opt im ized spec iﬁca lly for ben c hmark d atas ets ra ther tha n so lving r e al - world pr oblems. T he dis cuss ion h igh ligh ted the prec a rious balan ce b etween ﬁndi n g va lue in sta nda r diz ed d a t a s e t s w hile a void ing rest rictiv e ben chma rks . Work sh op part ic ipant s rec o g nized th e need f or d at ase t s to charac ter ize neu roscien ce p rinc iples and ena ble the co - de si gn of mat eria l s , devic e s , a nd a lg orith ms t h a t c o mp ute loc a lly . Thi s crit ical c hallen ge w ill requ ire clos ed - loop integ ra t ion bet w een s ens ing, co mp u t ing , an d a ctua tion . Another crit ic a l cha lleng e that ar ose in di scussions was t he a vaila bilit y o f tim e - syn c hro nized multi - mod al dat ase t s for a s ingle ta sk . F ina lly, the dev elopm ent and us e of stan da rdiz ed m etric s, inc luding en ergy eﬃ c ien cy , com pact ne s s, noise, and r obustnes s of the ha rd wa re and soft ware , w as al so hig h lighted as a c h a lleng e . Inf ras t ru ct u re Chal lenge s Acce ss t o f abricat ion facilit ies , desi gn to ol s, stan dards In fras tru cture lim ita t ions c onst rain prog re s s in the ﬁeld, w ith ma ny wo rks ho p parti ci pan ts ca lling for bet ter a cc ess t o fab ricat ion f ac ilit ies , d e s ign too ls f or asy nchr onous c irc uits , sta ndardized int erfa ces bet ween ha rdwa re a nd “wetwa re” (biolog ic al s ys tems ) , a nd bett er s h a rin g of prot ocols esp ecia lly for n ovel mat eria ls . T here wa s d i s c u ss ion a b ou t limit ed a cc ess t o fab ricat ion f ac ilit ies and a st rong ca ll to es t ablis h prog ra ms s i m ila r to t he h is torica l MO SIS (77) in itia t ive t h a t prov ided f ree ch ip ma n u fa cturin g f or a cad emic res ea rchers . ( The a uthor s note th a t “ M OSIS 2 .0” wa s re - launched in 2023 as a co mpo nen t of a C a li fornia innov ation hub supported by the U.S. CH IP S a nd S ci en ce A ct o f 20 22 ; see http s:/ /ww w.mo sis2 .co m/a bout - us for d etails .) Pa rt icipa nts s ugg e s t ed expa nding beyond s ilicon c hips t o inc orpo rat e new m ateria ls a nd st ruct ural c omplexit y (e.g ., b endab le bioelec tron ic s) wit h a s up port ing infra s t ru c t ure equ iva lent t o MO SIS for th ese novel a pproac hes . Part ic ipant s express ed t he view t h a t d evelop men t w ould a ccelera t e if t ools for des ig nin g as ync hron ou s c irc uits an d t ran s ferrin g s pikin g neura l net wo rks b etween h a rd w a re plat forms were a vaila ble. Acc ess t o st anda rdiz ed tes ting m et hods fo r novel c omp onents like m emris t or a rr a ys wa s note d b y par tici pan ts a s a lim it ation. P rotocol s ha ring t o elimin ate th e need to s tar t from sc ratc h, es pecia lly w hen dev eloping no vel bioh yb rid neu romorphic m ateria ls w a s hig hlig ht ed m ult iple t imes b y p ar ti c i pan ts . Th e gap between th e des ign t ool ec os y s tems ava ilable for tra d ition a l dig it a l s ig n a l p rocess ing v ersus neuromorp hic comp uting was also empha siz ed as a barrier t o broader adoption. Man ufac t uri ng C h alle nge s M anufact uring cha lleng es were dis c uss ed by p ar ti cip ant s , in c lud ing s plit foun dry issu es (e .g. , w here t he low er lay ers a nd u p p er la yers a re m a nu f ac t ured in diﬀ erent l oc a tions ) and tool - chain limit a tio ns . T he high cost of c hip fa brica tion wa s a ls o r a is ed , wi th p a rt ic ipants ’ es tima tes r angi ng fr o m $2 – 30 m illion to produce a ne w ch ip d esign . So me par ti cip an t s ar gue d th at a f ew t ho u s and d ol l ars o n a mu l ti - projec t 12 waf er ( MP W) r un c oul d yield suﬃcie nt pr odu c tion f or hea lt hca re a p p licat ions . F or e x amp le , after the works hop , A ndreou cl ar i ﬁed th at mi ni mum or der s for MP W th ro u gh Eur o pr ac ti ce o r Mu se we re esti mate d at $ 5 , 000 or $6 , 250 depe nding on pr o ce ss node an d tha t Eu ro pr ac ti ce al so sup por ts fab ric ation of cir cui ts u sing org a nic s emic on d u ctors and ﬂ exible s ubst rates . Diﬀeren t pa rt i c ipant s not ed on mult iple oc cas ions t hat neu romorp hic a pplica t ions ca n be d evelop ed w ith st and ar d C MO S t echnology , w hic h is much mo re w idely a v a ila ble . Another c h a llen ge ra is ed by p art ici p an t s w as dev eloping t est ing method s f or novel c ompone nts , like m emris t or a rra ys . Parti cip ant s fur ther note d th at it wa s d i ﬃcu lt to pred ict c hron ic relia bilit y in ha rs h biolo gic a l env ironm ent s, wh ere perform a nc e c a n be impac t ed by temp er atur e ﬂuc tu a ti ons , forei g n body re s pon se , and ma teria l d egrad ation. Commer cial iza tion Regul atory Chal lenges Par ticip ants note d th at t he reg ulat ory p a t h w ay p resen t ed a ddit ional c om plexit y, partic ula rly f or a dapt ive s yst em s t hat learn a nd cha nge over t im e . This r ai s ed ques tions a bout ho w to qu antif y the beneﬁts of a d a pta t ion and esta blis h ap propria t e reg ul a t ory f ramew orks that en abl e res ponsib le i nnova tion. The imp orta nce o f fa il - sa fe m echanis ms in a d a pt ive sy stems to ens u re devic es rever t t o ba s eline fu nc t ion a lit y if pro blems a ris e w as r ai se d mult iple t imes by pa rt i c ipa nts . The sim ila r ity to auto no mo us vehi cl es wi th diﬀeren t levels of a ut onom y w a s al so mad e . O ne par t ici pan t n oted th at sim ilar conc erns a ro s e when NeuroPac e RNS wa s ﬁrst la unched ; d o se lim its w ere emp loyed as g uardrails t o limit the ris k t o pa t ients s hould th e dev ice g o aw ry. M any part ic ipants expres s ed the need for bet ter educ a tion for resea rc hers abo ut reg ul a t ory p rocess . Clin ical Tr ans lat ion C hallen ges Pa rtic ipan ts express ed con ce rn s th at c lin ical t ra nsla t ion is hinde red by long inve st ment horizons that do no t a lig n w ith ty pica l busin ess exp ectatio ns . T here a re diﬃc u lties in dem o ns t r a t in g imp roved rea l - worl d function , be yo nd lab or a tory p erformance . P ar ti ci pan ts d i scu s se d a disc onnect between wha t res ea r c hers priorit ize ( e .g. , adv an ce s in t e c hnica l fea t ures ) versu s what pat ient s va lue ( e.g. , i m pr oved comf or t, redu ced pa in, a n d a vo iding emba rrassm ent from dev ice f ailu res ) . P a ti ent p r i v ac y a nd dat a se cur i ty c ons iderat ions were a l s o m ent ioned as c ha llen ge s for c linic a l tra nslat ion . Investm ent C hal lenges Pa rtic ipan t con se n sus poi n ted to a “ c hic ken a nd egg” p roblem in in ves tment — developers need d a ta t o jus t ify in ves tment in n ovel d evic es, but g ath ering th is da ta c an require s ubsta nt ia l i nit ia l inves t ment in the dev ices . S o m e pa rt icipa nts st ate d th at t his has led t o a c ons ervat ive t echno logy developm ent ap pr o ach in which c om p a n ies inc rementa lly imp rove th eir est a blis hed te c hnolo g ie s r ather th a n pursuing innov at ive de si gns. Mu ltip le indu stry pa rticip ant s no ted th a t l on g in vest ment h orizon s do no t alig n w ith typica l bu sin ess exp ecta t ion s . Tr ans d isc ip lina ry Commu nica tion a nd E duc ation One of the mos t per sisten t th eme s thr ou ghou t NPB H wo rk shop d is cu ssi on s wa s the need for tra nsd iscip lina ry ed uca tion to b ridg e th e c ommun icat ion gap b etween en gin eers, sc ient ists , an d c linic ia ns . F or exa mple, Gina Adam stress ed t he im porta nc e of in terd isc ip lin a ry colla b oration a nd e duca ti on, while Nit ish T ha kor not ed t hat e s tab lis hed n eurosc ienc e c o m munit ies — oft en rooted in 13 experim ent a lly - d riven , m e c han ist i c , a nd reduct ionis t a pp roach es — may ﬁnd ne ur omorphic con cept s un familia r or diﬃ c ult t o co n t extu a lize w ith in their fra mew orks . P a rt icip ants a ls o c a l led for dev elop in g tr an s disc iplina ry team s that include n eurom orph ic engin eers , ma teria l sc ient ists, neuros cien tis ts, c linic ia ns , data scien tis ts, a nd regu l a tory expert s . A mp a r o Güe mes r a ised the que s tion of whether one l ab should hou se m any d o mai n s o f expertis e or s p ec ializ ed la bs s hou ld be enc ourag ed to c ollab orate , a dec is ion which is p art i c ula rly impa ctfu l for hiri ng w hen settin g up a l ab. O utlo ok an d Vi sion fo r N ex t Step s Ne urom o rp hic Pr inc iple s f o r Biome dic al A pp lic atio ns NP BH W orkshop Ch ai r, Ralph Etienne - Cu mming s , pres ent ed a s ummary of t his m e et ing a t the NIH B R AIN NeuroAI Works hop (33 , 34,78) w here he id ent iﬁ ed four key tec hn ical a dvan tages of neu romor phic prin cip les fo r biomed ic al a pp licat ions. These inc lu de pow er eﬃcienc y ; on - chi p , re al - tim e proc e s s ing ; a d a pta bilit y a n d res ilienc e ; a nd a nat ur a lis t ic cod e for int erfa c ing w ith th e nerv ou s s yst em. Pow er eﬃc ien cy — alo n g with reduc ed po w er requirem ent s for b iomed ica l s y stem s — lowers heat dis s ipat ion, wh ic h is part icu l a rly im pac t ful fo r impla ntab le devic es . O n - chip, rea l - t i m e proc essing p rovid es low - latency s igna l proc essin g for c losed - loop a lgo rithms , avo i ds th e complexit y of tr an s ferring data to the c lo ud, and imp roves s e c urit y b y only stor ing d ata lo c a lly . Ada p t abilit y a llow s th e s yst em t o lea r n a n d ta ilor th e trea t ment p a r a digm as t ime p rogress es a nd the biolog ic a l resp onse c hang e s . La stl y , the even t - based s piking c ode of neuro morp hic mod els and d evic es ma y ena ble im prov ed neu r a l int erfa c es by mimi c king th e na tura l s piking pat ter n s of neuronal ci r cuit s in th e t a rg eted reg ions of t he brain . Neuromo rphic a dvan cements may be es pecia lly impo rtant for c lin ica l app licat ions ( s ee Fi gur e 1) where a l arge amo u nt o f da ta is c ollec t ed , pred ict ive a lg orit h m s a re requi red, a nd low he at - dis s ip a t ion a nd e xtended bat t er y - life ar e cr i tical for im pla nta ble s yste ms . Wor ksho p p artici pa nts su gges ted th a t w hen ap ply in g neu romorp hic princip les to solv e biom edic a l prob lems , th e fo c u s sh ould b e on t he c linica l a pp licat ion (whi ch c an of ten be neg le c ted ) . P art ici p an t s di scu sse d t h e ad v an ta ge s in neur a l engin eering of develop ing neuro morp hic s olut ions t hat prec is ely ta rg et spec iﬁc neural c irc uits for a dis eas e rat her t h a n b ro a dly stimu latin g bra in re gion s . On the other hand , parti cipan ts no te d th at en gin eers shoul d be unrestr ained to de te rmi ne th e most appr opr ia te tec h nica l solu tio n for t he p roblem a t hand . Pa nelis ts em phasiz ed ident ify ing sp e c iﬁc a pp licat ions w here neu romorp hic prov ides c lea r a dvan tag e s (e.g ., low - la tency t o a chieve r eal - tim e c lose d- loop s ystems ; extrem ely eﬃc ient and res ilient circ uit s ). Fu tu r e R esear ch A r ea s Bio int erf aces M ult imoda l Bios e nso r s Pa rtic ipan ts d is cussed th e need f or bet ter w ea rable s ens ors t hat mov e bey on d th e lim ita t ions of t he clinic a nd c ollec t d a ta while su bjec ts a re m obile a n d engaged in n atura l b ehaviors . Sev er a l c linic ia n s emp has ized th e need f or tec h nolog ies th at m onit or mult iple biom a rkers . T his includes c he m ic al sig nals , such as pot a s s ium , oxygen, sodi um, and v arious neurotr a n smitters (up to 24 as describe d in (79) ) . Pa rtic ipan ts a ls o disc uss ed the n eed fo r mult im oda l rec ordin g d evi c es tha t can s imult a neously meas ure elec tric a l act ivit y a long s ide c hem ica l sig nals to prov ide a m ore c omplete p ict ure of neura l f unc t ion . 14 Im po rt a n t dev i c e sp eciﬁca tio n s were d isc ussed inclu ding bioc ompat ibilit y, low im mune res po ns e, s tret cha bilit y, ﬂ exibilit y, a dhesiv eness , an d t he a bilit y to eit her d iss olv e or rema in s t a b le in t h e bo dy (e.g., he ar t vs. s to mach) depe nding on the b iomed ical a pp licat ion . T h e potent ia l for neu romo rphic a pproach es to ena ble bet t er sen s or int egrat ion a c ross moda lities ( i.e. , m olecu la r , pres sure, elec trop hy siolo gy ) an d to opera t e at diﬀerent time - s cal e s wi th lo w pow er consumpt ion and min imal la tency r eq uirem ent s was c ons idered a st reng th . Novel M a t eria ls T here were s ev era l pres ent ation s o n nove l ﬂexib le an d s tret cha ble ma teria ls , a nd t h er e w as co nse n su s that th es e charact eris tics are a req uirem ent for nex t - g enera t ion neuro morphi c bio i nte r face s. Bao highli g hted how the ﬁeld is a p p lying mat eria l scien c e to replic a t e rig id s ilic on c ircuits i n soft, e mbedded jelly - like ma t eria ls, of ten w it h c om parable elec t ronic prop erties . Ba o then pr esente d a na noconﬁned poly mer semico n duc tor (D PPT - TT) i n a matr ix ma te ri a l (SE BS) th at c an be ta ilored t o b e s tret c h a ble, s elf - hea ling , or biod egra da ble (80,81) . W a ng a ls o pres ented a s oft , ﬂexible, immu ne - co mpati bl e o rg ani c elec troc hem ica l tra nsist or (O ECT ) usin g a s emicon duct or poly m er (p(g 2 T - T)). Wan g de mons tr a te d ho w t he foreig n b ody resp ons e ca n be m inim i z ed by us ing b ackb one (s eleno phen e ) and si d e - ch ai n ( THP an d TM O gr ou ps) m olec ul a r d esig n st r a t egies (82) . A ki nw a nde sugges ted that the hu man bo dy s houl d be int egra t ed as a n ac t ive c om pon en t of t he d evic e ra th er tha n a pass ive t arge t for tre at ment a nd obs erva tion. His wo r k implem ent ed u ltra - th in g raphen e in elec tron ic tat toos e mb e dded i n the surface of th e s kin (65) . Pan elist s a ls o disc u s s ed the im port anc e of bio compat ibilit y a nd CMO S int egr ation w it h new mat eria ls . Diﬃ c ult ies in c linica l t rans lat ion of no vel ma t eria ls were not ed b y pa rtic ipa nts . NI BIB’s Biomat eria ls Netw ork ( https :// videoc a s t .nih.gov/ watch= 45769 , st arti ng a t ti me 2: 27:1 0) is an examp le of a compl eme n tary fun d ing me ch ani sm for nov el neu romorp hic m ateria ls . The biom at eria ls t ra n s la tion proc es s i s oft en enc umb ered by th e s o c a lled “va lley of d ea th ” , beyond w hich f ew bioma t eria l tec h nolog ies mak e the t r a ns it ion from benc h t o bedsid e . Th e B iomat er i als Netw ork ai ms t o ac celera t e biom a teria ls - ba sed t ech nology d evelop men t vi a a c oordinat ion c enter and tr an sl at i o nal re so ur ce s that s upp ort t e c hn olog y d evelopment projec ts . The go al of this infr ast ruc ture is t o fac ilita t e c olla borat ion, imp rove d iss em inat ion, a nd prope l clin ical t ra nsla t ion a n d c o m mercia lizat ion of b iom ateria ls . Neuromo rphic Neur om odula tio n Duygu Ku z um c on cluded th at neur omorp hics can h elp neu ros cien c e a nd neurom e d icin e in t wo m a jor way s: (1) neurom orph ic com putat iona l m odels ca n int eg r a te in form ation from mu l tip le mod a lities t o c reat e acc urate, low - d i men s ion a l n eura l repres ent atio ns to im pro ve neu r a l pros t h et ic s or B MI perf orma nce; (2) n eurom orph ic ha rd w a re can tran sf or m neura l rec ord ing methodolog ies b y en ab ling on - chi p, r eal - time s ig nal p rocess ing w it h lea rnin g ca pabilit ies . Th ese ca pabilit ies ca n shift t he p roc ess ing l ocat io n o f th e d at a from oﬀ - chip to on - chi p a n d increas e the lon gev ity and st a bilit y of neura l in terf ac es . Pa rtic ipan ts express ed t hat ext erna l relia nc e on in tern et res ou r c es a lso p ose s security ris ks , whi ch can be o bvi at ed by o n - chip pro cess ing . P ar ti cipan ts a s ser ted th at n euromorph ic n eurom o d ula tio n dev ices c a n be d eveloped us ing leg acy pr oce ss no de s (e .g., 180 n m, 6 5 n m) th at wo ul d no t dep e nd on acce s s to co st - proh ibit ive new infr a stru c ture. O ne exam ple wa s proposed b y par ti ci pan t Tobi Del br uc k wh o des cribed an approa c h to s hort - t er m lea rn ing via the us e of h idden pers is tent s ta te s th a t sto r e s m emories fo r a rbit rar ily lon g t imes u si ng co nte xt - dep enden t con trol o f a fo rge tting ti me - c ons tant , whi ch a llow s r apid 15 memo ry up d a te s ( 83) . Del bruc k conten d ed tha t i t wa s fe a s ible to build on - chi p mixed - sign al CM OS c ont roller s using aﬀ o rd abl e 65 – n m tech no logy that r un on 1 V supply, utilizin g 1 mm 2 of s ilic on and a 2 - la yer rec urrent neur a l n et wor ks com pris ing 50 k we igh t s . This syst em w ould consu me fe w er tha n 5 µ W ( at 100 Hz upd ate ra te wi th a 250 k Hz clo ck including a non - optim ized low - po wer sta ti c r a nd o m - ac ce s s memo ry ) an d coul d p owe r m an y re al - t im e det e c tion and c ontrol t a s k s in s tan d a rd iEEG elect rode a rra ys , es pec ia lly fo r use i n ta ndem wi t h opto gene ti c sti mul ati on . Real -T ime , On -Chip Lea rning Pa rtic ipan ts h igh ligh ted tha t B MI si gn al s ar e kn own to d rif t ove r ti me (84) , su gge sti ng th at on - ch ip lea rning c ould b e a fast a nd e ﬃc ien t w ay t o correc t s ignal d rift reg a rdles s of the source of t he drif t . Par ticip ants a lso n o t ed t h a t int erpers onal a n d int r a pers ona l va ria bilit y in b iolog ica l rec ord ing s c ou ld a ls o be addres s ed by on - c hip learn ing . This w ould beneﬁ t lon g - te rm ex per ime nt s th a t tr ack th e same type o f biolog ic a l dyn am ic s t o study behavior and cogni tion in both an i mal s an d h u m an s. An advant a ge of the r eal - time co mpone nt is a reduct ion in pr o cessi n g tim e, a s there is n o long er a wait t o t ran sf e r th e dat a to the cloud. Thi s c ould m inim i z e dela ys in clos ed - loop a l gorit hms. As decodin g alg orithms utiliz ing mac hin e lea rning a nd s pikin g n eura l netwo rks b ec o m e more prev a lent (85 – 87) , p a rticip ants c onclude d that s hift ing from c loud st reamin g to rea l - time , o n - chi p h ardwa re pro cess ing (25) can be com plem enta ry and expand the te c hn ical ca pab ilities of neuromo dula tion, B M I, a nd clo sed - loop biom e dic a l s yst e m s . Nex t -G eneration P ros theses Par ticip ants en t husias tica lly not ed t h a t n euromo rphic ha rdware a nd prin cip les a ppear rip e for pro st het ic a pplica t ions that req uire neu ra l rec ording , dec odin g, en cod ing, a nd closed - loo p c o ntrol t o e nha nce c omp lex bidirec t iona l s ensory inform at ion pro cess ing . Pan elist s highlig ht ed t he potent ia l for h igh - den sity neu ral p robes t o s imulta neous ly rec ord and s timula t e , em pha s iz ing tha t n euro m orph ic ap proaches exc el a t ext ract ing s pars e inform ation for rea l - t ime, low - la t en c y feedbac k sy stems. Preli min a ry pay oﬀ s are not a ble in v is u a l p ros thet i c s where t he ﬁ eld c an levera g e extens iv e knowled ge of t h e func t ion a l a nat omy of t he v isu a l cor tex and apply dim ensiona lit y red uct ion m eth od s t o pred ict the loc ation s at w hich phosph e ne s will a ppea r in stea d of relian c e on pat ient report s (88) . Parti ci pan ts arg ue d tha t t he perip hera l nervou s sys tem c ould s tan d t o beneﬁ t from neu ro m orph ic st ra te g ies but is a n “ ignored c ous in” de s p ite it s c ritic a l imp ortance to b rain res earch . There w as gen eral c onc urr en ce t hat neu romorp hic s is a pprop riate f or th e repla c emen t of lost func t ion . T his has been o bs erv ed in t he h ist ory of c oc hlea r im plants , in a ddition to emerg ing evid enc e in s ight rest or a t ion an d s o ma tosens ory neuro prosthese s (89) . Time - S ynchro nized M ultimoda l Da t aset s The NIH BR AI N B BQS (Br ain B ehav io r Qu a n tiﬁcat i on a nd S ynchroni zation) program ( htt ps://br a in - bbqs.org ) — wh ich a ims t o ca pture s ync hroniz ed n eura l and behav iora l da ta in natura lis tic env ironm ents ac ro ss both cli nica l and non - clin ical p opu lat ions — was m ent ioned by mu ltiple par t ici p ant s . T he imp orta nce of sim ulta neous, multim odal rec ordin gs with d iﬀerent sens or t yp e s (e.g . , E EG , elec trooc u logram ( EO G ) , elec t romyog raphy ( E MG ) , elect roc ortic ogra phy ( ECo G ) ) c aptur in g the s ame even ts and be haviors w as e mp h a s ized . Thi s neur al - behav iora l s ynch roniza tion would allow f or c orrela tion ac ross m oda lities a nd enable unc on str a ine d algor ithm re se ar ch and de velopment. An d r eas A ndreo u not ed t hat res e a rc her s s uch as Pa ri s a Rashid i a re co llect ing data from Intensiv e C a re Units ( ICU s ) pot ent ia lly w ith even t - b as ed ca m er a s that c ould be of grea t v a lu e to t he n eu rom orp hic 16 com muni ty (90) . Andre ou f urther noted the i mpor tance of look ing b eyond the neur omorphi c worl d to the br oader com muni ty for d at ase t s. Sy d ney C ash noted t hat subst a nti al intracranial r ecor ding dataset s a lread y exis t thro u g h BRA IN Init ia tiv e d a ta rep os itories like D a ta A rc hiv e for t he BRA IN I nit i a t ive ( DAB I ; https:// da bi.lo ni.usc.edu ) an d the Dis trib ut ed A rc hiv e s for Neu roph ys iolog y Data I n tegr a ti on ( D AND I ; https:// da ndi arc hive .org ) . T hese d atas ets hav e been p reviou s ly lev erag ed in a lgorithm dev elopm ent c omp etit ions t o s olve b iomed ical p roblem s su ch as seiz ure p redic tion . Mult iple p a rtic ipa nts s ugg est ed th a t futu re a lgorit hmic or da ta c ha lleng e s us ing t hes e m ult imo d a l da taset s c ould feature ne uromorphi c a lgorit hms , h ar dw ar e , or cod e s ign , an d ev a luat e sp e ci ﬁc adv ant a ge s over con vention a l ap proac hes. Standa rds Pa rtic ipan ts express ed t hat the imp ort anc e of open s tanda rd s ca nno t be overlooked . It w as no ted th at s uccessful op en st and ar d mo de l s lik e USB a nd O penGL tr a n sfo rmed co mpu ting thr ou gh co mmon int erfa c es. T he O penGPU F ou n d ation hig hlig ht ed t h e n euro m orph ic inter mediat e repres ent a tio n (N IR (10) ), a co mmo n t ool - ch ai n lay er th at tr an sla te s betw een frontend sim ul a t ed app lica t ions an d neuro morphic mod eling so f tw ar e and b ack e nd digit a l h a rd ware pla t forms . NI R cleverly dec ou ples th e s oft ware from t he ha rdware la y er , ena blin g the p oten tial for sea m less in teg r a t ion with mixed - s ig nal chi p s, an al og h ar dw ar e, an d hy br id di gi t al - an al og sy ste ms. Th u s far , NI R ha s su cce s sfu ll y e val ua ted four har dw ar e pl atf or m s and seven si mul at or s. N I R repres ent s a n exc itin g direc t ion for t he ﬁ eld t o promo te wid er a dopt ion and fas t er develo pm ent o f neu romorp hic s yst ems. Benc hmarks The val u e of b enchm arking for adv an c in g tec hnolog y dev elopm en t w as de b ate d b y par ti ci p ant s : while com mon me tri cs f or h ead - to - head comparis on of d iﬀerent ap proac he s ena bles ob j e ct ive a ss essmen t, s ome pa rtic ipan ts express ed c oncern s t h at s uch metric s ca n lead to in c rementa l progres s a t t he ex p ens e of p oten tia lly tra nsfo rm a t ive in nov ation . T hrou ghout the c y cle of ne uromorphi c deve lopment, funder s need to m a ke evid enc e - b ased dec is ion s and c o m pa re nov el a p p roa ches to est ablished b enc hma rks . T hes e cyc les m ay be d iﬀerent depend ing on th e neurom orp hic pl a tf orm (e. g., c hip - l e ss we ar abl e s v s. on - chip CMOS device s). F or exa mp le, Neuro Ben ch (76) is a c ollabo rat ively des igned framew ork for neuro morphic benc hmark s th at re sulted f rom open com munity i nput a c ros s acade mia and indus try , supported by c umula t ive eﬀort s of t he NSF - fun ded T ellu ride Neuro morp hic C ognition E ng ineering Work sh op, t he Neu ral I nsp ired C om putation E lemen t (NIC E ) c onf erenc e, an d rec ently p romo ted at the Internati ona l Confe rence on N euromorphi c S y st em s (ICON S) in 2025 . Ne uroBench provide s ﬂexib le benchmark ing th at i s int end ed t o c o - ev olve w ith the ﬁ eld as it g ro w s t h roug h c om p et i tions , workshops, a nd c entra liz ed lea derb oards . I n its c u rrent form, Neu roBen ch h a s t wo trac k s : a lg o rit hm ic and s yst e m s . T he a lgorit hm trac k is hardwa re - in depen dent, ut ilizin g d atas et, a lg orithm, a n d metric s , while t he s y stem tr ack pro vides en d - to - end ev a lu a t ion of an inte grate d syste m. Ne uro Be nch h as sup por ted a co mmu ni ty - driv en Gra n d C h a llen ge on neu ral d ec oding for m ot or c ont rol of n on - huma n primates in IEEE BioCA S conferences in 2024 (91) and 2025. Be nchm a rk ing tools like NeuroB enc h may pr ove to be ke y ena b lers of com muni ty - driven ad v an ce s for the neurom orp hi c biom edica l ﬁ eld . O ther impo rta n t c onsid era tion s f or dat ase ts a n d bench marks d e s crib ed by t he embodied neuro morp hic int elligen c e comm unity ( see Box 6 in (92) ) will a ls o need to b e c ons idered giv en t he inext ric a ble link betw een em bod iment a n d biomedic a l eng ineerin g. 17 In fras tr u ctu re to Adva nce Neu romorphic Engine ering T he im portan c e of t emp or a l proc ess ing a nd th e need t o inv est in resea rch a n d desig n m eth odolog ie s — like c o - des ig n a lg orit h m s wit h sy nchron ous c irc uit s or even t - b ase d (a syn ch ro no u s) ci rcu its — w ere str es se d by p artic ipa nts , noti n g th at curre nt d esign auto ma tio n too ls c a nn o t sup por t a uto mate d wo rkﬂo ws from alg orit h m s to c hip des i g n. Neura l c hip lets (93) w ere pos ed a s a way forw a rd g iven t heir mo dula rit y, sc alab ility , an d reu sab ility . The Na tional Institu te of Stand a rds and Te chno logy (NIST) N ati o n al Adv an ce d P ack agi n g M an uf a ctu ri n g Pr o gr am (94) , ini tia ted as par t of the U.S . C HIP S and S cien ce A ct of 2022 , w as hi ghlighted as a new ly releas ed funding opp ortu nity relev an t t o th e ne ur omorphic co mmu ni ty ; it wa s note d th at b iom edi c a l pac k a g ing is wit hin th e s co pe of thi s Progra m . T he N eur omor phi c C o mmo n s - T HOR p rojec t , a warded by NS F in 202 4 (A w ard No. 2346527 , http s:/ /ww w.ne uro morp hic com mon s.co m ), aims to d evelop a nd d eploy la rge - sca le neu rom orphic c omp u t ing researc h infras truct ure oﬀerin g c ommun ity a ccess to h eterog eneous ne uro morp hic com pu ti ng h ar dw ar e sy st em s th ro u gh cl o se - knit p a rtner s hip s wi th i nd u str y. TH OR o ﬀe r s (1) re mote ac ce ss t o l ar ge - sca le n eurom orph ic sys tems ; (2) open - s ou rce hard ware/s oft ware co - d es ign fr a m eworks and t ool s ; (3) co mmon be nch mark s an d com pe titio ns; and (4) ra pid al gorithm deve lo pment by provi ding a ccess to a collec t ion of learn ing mo dules, netwo rk mod els , an d examp le fra mew orks. In a ddit ion, t he NS F h as sponsored the A CCE SS ( A dvanced Cyb erinfra s truc ture Co ordina t ion Ec osys t e m : Ser vic es & Supp ort ) P ro gra m, w hi ch pro vides acc es s to gr adu ate s tud en t s, researc hers , an d e du c ator s i n the Unite d St ate s t o ut ilize a dva nced c ompu ting s ys tems and s ervic es at no c ost . F o r det a ils ab out th e pr o gram and elig ibilit y , s ee h ttp s:/ / acce ss - ci. or g . Computa tiona l Infrastruc ture to Imp rove Cli nica l Care V iktor J irs a’s pr esen tat ion s howed t hat virt ua l bra in mo dels derived fr om pati ent - s p eci ﬁc d at a se ts c an de mons trate su rp ri si ng ly acc urat e predic t ions for individ ua l resp onses to diﬀ erent st imu l a tion para mete rs an d for det erm ining t he locations of epilep tog enic z ones ou t sid e a reas a ccess ible b y imp la nt ed elec trod es (44 ,45) . S imula tio n - based inferen ce f or pers ona liz ation requ ires t hou sands of mo del s imu lation s , whi ch pres en t s a su bstan tia l c o m putat iona l bot t lenec k (e.g ., Ji rs a c o mme nted th at 20 min utes of f M R I rest ing - s ta te rec ordin gs co ul d require m ore th a n t wo da y s of proc es sing t im e even wit h Gr ap hi cs Pr o ce ssi n g Uni t ( GP U ) acc elera tio n). Thes e v irtual b ra in m odels hav e not been imp lemen ted on c urre nt neuromorphi c a r chitectur es su ch as Sp iNNa ker or Br ai nSc al e S . I t r e m ain s an open que s tion whe ther othe r neur omorphic ar c hite c t ures — such a s In tel ’ s Loi hi p l atf or m or othe r spi k ing neural network c hips ; for rev iew s o f c urr ent and emer g i ng pl atform s , s ee (9,95, 96) — co ul d address th e compu tat ion a l b ott lenec k of br ai n - sc a le pred ict ive m odelin g f or epileps y a nd other neu rologic a l dis eas es . New App r oa ch Methodo logi es (N A Ms) New A p proac h Met hod ologies (NAMs ) a re int end ed t o a ccu r a tely m odel hu man bio log y , e nco mp as s ing in vi tro or gan - on - a - c hi p sy ste m s and in silico appro ache s per for med b y co mpu ting p la tfor ms or cu sto m ha rdwa re, inc lud in g mat hematica l m odeling , si mu l atio n , an d M L /AI ap p ro a che s (97 ) . Comp ut ati o n al models (e.g., virt u a l b rain t wi ns ( 47) ), neuromorphi c emulator s (e. g ., neuro mor phic twin (57) ), an d an al og or mixed - s ig n a l neu rom orphic d evic e s m ay p ro vi de a b asi s f or future e nergy - eﬃc ient , perso n a liz ed, and safe N AM s wit h the p oten tial t o a ccelera t e neu roscien ce an d b rai n - health tech nolo g ies . NA M s w ere not 18 explic it ly dis c uss ed at th e NP BH work s hop, b ut s ev er a l relat ed technique s w ere pres ent ed, inclu ding Jirs a ’s v irt u a l bra in twi n s and Don ati’s c ortica l em ulato r. Th e NIH rec ently (29 A pril , 2025) announced th at it is p rioritiz ing N AMs - b ase d ap pr o ach e s : htt ps :// www.nih. gov/new s - even t s /n ews - releas es /nih - priorit ize - hum a n - b ase d - researc h - techn olog ies . Tr ansd iscip li na r y Educ atio n Bas ed o n com munity inp ut , t arg eted edu ca t iona l init iativ es to b ols ter n eurom orphic en gin eering in neu rosc ien ce an d hea lthca re m ay require bu ilding t ran sdis cip lina ry team s w ith a sha red voca bu l a ry, c omp lementa ry expert ise, a nd a ccess ible t ra inin g t hat s p a n s the res ea rch – t r a ns lat i on p ipeline. In the few po stgr ad u ate neuro morphi c eng ineerin g p rogram s that exis t , tr aining is heav ily w e ig ht ed t oward phys ics, mat eria l s cien ce, a nd elec tric a l and c ompu ter en gin eering (98) . Man y n euromorph ic d evic e eng ineers d o not have form al training in neur o a natomy a nd neur oph ys iol ogy, in vi t ro or in vivo n euros c ienc e method s, c linic a l neurosc ienc e, pat ient - c ent ered d esign, the reg ula t ory p roc e s s, and /or neu roeth ic al c onsid era tions for m edic a l dev ice d evelop men t (98,99) . F or exa mple, as of 202 0, there w ere no s p ec iﬁ c po s tgra dua te eng ineerin g p rogra ms tha t t r a in ed studen ts in t he ﬁ eld of imp la ntab le and w earab l e elect ronic s for s ens ing a nd t reat ing diag nosed neurolo gica l dis eas es (98, 9 9) . Theref ore , n eurom orph ic eng ineering tea ms may not a lign techn ical ca pab ilities , s u c h as des ignin g in terfac es a nd a lg orithms, w ith rea l - wo rl d c linic a l impac t . Devel oping hum a n - and b r ai n - m ach ine int erfa c e s requ ires ex pertis e a nd tr aining in m ic ro - elec t rode des ig n, biocom p a t ible ma teria ls, low - power c ircu it d es ign, an d rea l - t i m e neura l s igna l proc ess ing . Fo r exa mp le, an emerg ing them e from the w orksho p , di scu sse d b y Duy gu Ku z um and others, w as th e call to sh ift process ing from oﬀ - ch ip , cl ou d - b ased co mputin g to on - c hip, ed ge com puting i n human - an d br ai n - mac hin e int erfa c e develo pment (s ee Enginee ring Chall enges , Topic 4 - Medical Imag ing, Wear ables an d Anal ysis, D ay 2: Ses sion 3 Ma t e rial s f or Neu romo r phic s (Device s) ). Traini ng in t he r ecording and an al ysi s of hi gh - d imensio na l , mu ltim odal data from BMI rec ordin gs ( s ee Clin ical C hallen ges , Data collect ion l imita tion s ) was al s o identiﬁ ed as an i m portant challen ge . On the c linica l s ide , so me atten dees express ed t he view t h a t c linic ia ns w orking w ith B MI s may h ave limit ed f am ilia rity w ith neurom orphic pl atfor ms or h a rdwa re/sof twa re co - de s ig n princ iples . Thi s kn owl edge ga p m ay potentia lly hi nder clin ica l input into the d evelopm ent of n eurob iologic a lly - inspired arch it e ctur e s and dev ice d e s ign , wh ic h i s c ritica l t o ensuri ng respo n s iven ess t o pat i en t n eeds (see C linical Chall enges ). A s im ilar neuromorphi c kno w ledg e ga p exist s in trad ition a l compu tat io n al n euros c ience training , whi ch ty pically foc uses on t heoret ica l mod eling and an alys is , wi th min i mal emphasis on the eng ineerin g d ema nds of deploy ing a l g orit hms in r e al - wor ld c linica l s ett ings in medica l dev ic e s or robotic pl atf or m s. Br ad Aimone di scu s sed ho w neuromor phic compu ting can help u s unde rst and ne uros c ie nc e bet ter b y m ovin g b eyond t radit iona l s equentia l p rocess ing m odels t o m ore bra in - like co mp utat ional fr ame wor ks , inc luding t hos e in co rpor a t ing heterog eneit y, para llel s amplin g and s pat ia l tempora l pr oce ssi n g (s ee Day 2 S ess ion 4: M edi c a l Imag ing, Wea r a bles and Ana lysis ). F ut ure direc t ions t o a dd re s s thes e kn owled ge g ap s an d challen ges c o u ld inc lud e s e vera l tra ini ng and edu cat iona l act iv ities to fo ster coll a bora tion and unde rstandin g of neuro morphic approache s . T hese a ctivit ies c ould foc us on trainin g clin icia ns an d neu rosc ientis ts in neur omorphic appr oa che s, while tra in ing mat eria l scien tis ts an d neuro m or phi c engin eers in neu ros cienc e. S uch c ross - disc iplin ary 19 a pproac h e s ha ve b een em ploy ed in b iophys ics, whic h b e g an by a p p lying qua nt itat i ve m et hods to biolog ic a l prob lems (100,101) . Th e t rai ni n g an d e du c ati o n al a ctiv ities illus t rat ed b el ow ha ve b een s upp orted by U . S. fun ding agen c ies an d c ou ld inf orm fu tu re tra nsd iscip lina ry trainin g o pport un ities for c linic ia ns, computa t ional n eurosc ient ists , mat eria l s cien tist s, a nd eng ineers . (1 ) Inte r disc iplina ry S umm e r Scho ols, Ha ckathons , and Hands - on Wo r ks hops The N SF - fun ded Tellu ride Neurom orp hic Co gn ition Eng ineering Wo rksh op, Neur o - Ins pired Compu tat iona l E lemen ts (NI CE) prog ram (s ee Ben c h ma rks sect ion), a n d t he I n t ernat iona l C onf eren c e on Neurom orphic Sy stems (I CONS ) tuto rials are exa m ples of join t c ommu nity m eetin gs led b y t ra n s disc iplin ary expert s in the neur omorphic spa c e. The se works hops c a n ra ng e f rom a 1 d ay ov erview ( e .g. N IC E ( https :// nice workshop.org ) a nd ICON S ( https:// icons neur o morphic. c c ) ) t o a 3 w eek int ens ive d eep div e ( e.g. Tellurid e ( ht tp s ://s ites .g oogle.c om /view /tellu ride - 2026/home ) ). To pi cs o ve r t he y ear s acr o ss th e se thr e e wo r ksho ps h av e rang ed f rom “ neu ro s c ienc e bas i c s” fo r eng ineers (e.g ., s yna p s es , dendrites , tra v elling wav es, osc illa tio ns , spa tia l nav igat ion, a ud ito ry pat hw a y s, dec od in g s peec h , m usic perc eption , mic rosa ccad es, n eurot ra n s m itters , sy napt i c plas t icit y, n eurom odulat ion , b iologica l neura l net wo rks , se nsor y - motor s yst ems, reinfo rcem ent lea rning , and dec i s ion making , etc. ) to m ore f oc us ed neuro morphic co mput atio n al han d s - on w orks hop top ics (e. g. , Neu roBen ch, Neng o, Bra inS caleS , Sp iNNa k e r 2, sp iking neu r a l net works on FPG A for em bed ded sys tems , neurorob otic s, dy n a m i c vis ion se nsor s , on - c hip lea rn ing for s igna l - d rift c orrec t ion, m ult imoda l bios ens or in teg r a t i on, a nd re a l - time c los ed - loop c on t rol s trat eg ies). (2 ) Algori thm and Da tase t Chal lenges During the di scussion on Be nc h m a rks and Da ta, s eve r a l particip ants voiced t he advant a ge s of c omp etit ive cha llen ges for ad v a n cing neuromorphi c alg orit hm s and d ata se t s — a so - ca lled “ ba ttle of t he a lgorit hms ”. T he dis cuss ion inc luded ut ilizin g m ult i m oda l data sets to develop new n eurom orphic a lgorit hms and h a rd wa re, wh ich c ould t hen be c om pared t o c onvent iona l app roaches to h igh ligh t the neu romorp hic a d v anta ge. As referen c ed a bov e (s ee Tim e Synch roniz ed Multimo dal Datase ts ) , a w ide variet y of ne ura l and b eh avi o ral mul timo d al d a ta sets , in c lud ing int racran ial recordi ng s in h umans, are curre ntly a v a ila ble thro ugh NI H BR AIN Ini ti ativ e d at a arch iv e s su ch as DABI a nd D A ND I, whic h c oul d be in corpo rate d into co mmo n be nc h m ark s for c ha llen ges . C h a lleng es us ing dat a s cienc e met hods c ou ld bu ild a bett er und ersta nd in g of h igh - d imen sio na l mul timod al neu ra l an d b eh a vio ral d at aset s by co mp ar i ng d iﬀerent alg orit h m s a nd their perfo rma nc e. Du ring th e works h o p, it was n oted tha t t he A m erican E pilep sy S o c iety host ed a n algorit hm develo pmen t c ha lleng e for s eiz ure pred ict ion th rough Kag g le ( http s:/ / www. ka ggle .co m/ c / sei zure - predic tio n ). Th ese typ es o f challen ges enco urage entries f rom t ra n s d isciplin a ry t e a ms an d inc orp orate benc hma rks t hat c ons ider b oth c omp utation a l and c linica lly - releva nt c onst ra ints . (3 ) Clin ical Imm ers ion Pro gram s fo r Enginee rs and Ma teria l Scientist s C linic a l imm er s ion p rog rams for b iomed ical en gin eers c u rrent ly exis t at t he un dergra duat e a nd gra duate level a nd a re fun ded b y t he NIH (102) . The pur po se of t he se prog ram s is t o expos e b iomed ica l en gin eers to clin ica l op era tions and m edica l wor k cu ltu re, wit h t he end goal o f iden tify ing c linic a l cha lleng es. T his mo del c ould be ta ilored for n eurom orphic en gin eers a n d m ateria l s scient is ts t o identify un met biom edic a l need s. C linic a l s ettings for pot ent ia l s hort - t erm rota tio n s c ould inc lude epileps y mon itorin g 20 un its , reh a bilit a tion clin ics, m edica l ima g in g facilit ies , and pros t het i c ﬁttin g c en ters . Pot ent ia l areas of int eres t c ould inclu de (bu t a re not limit ed t o) pa tien t workﬂ ows, immun o - compat ible bioel ect ronic dev ices , c hallen ges in m edica l im age rec ons t ruct ion , a nd h u man f a c tors in devic e use. (4 ) Mat erials – Devi ce – A lgorit hm C o - D esign Worksho ps fo r Bi omedi cal E ngine er i ng Pa rtic ipan ts n oted tha t t here is a l a c k of data s ets t o fac ilita t e co - des i g n in neu romorphic mat eria ls , dev ices , a nd a lg orith ms. W orks hops h a ve been h eld t o a ddress this co - design g a p for neur omorphic ha rdwa re a nd s oftwa re (exa mples inc lud e, b ut a re not lim ited to: "N SF Workshop o n M a chine Lea rni ng Ha rdwa re Brea kt hroughs T owards Green A I a nd U biqu itous On - Dev i c e In tellig enc e " ( https :// s ite s.goo gle.com/rice.e du/ml - hw - green a i - on - devic eai/ho me ), t he " Worksho p on Ne uromorphi c C omp u t ing : fro m M a t eria l to A lg orit h m (NeuMA ) (h tt ps://s it es.go ogle. co m/v iew/nuema /h ome), and t he NSF Works hop on AI and HW 2035: Sh a ping t he Nex t Deca de ( https ://p ublis h. illinois .ed u/a i - hw - works hop/ ). How ever, there is a la ck of focus on biom edica l applic at ions . D evelopment of w orkshops th at brin g t og eth er mat eria l s scien tis ts , chip des ig ner s , a nd a l g orith m d evelopers fr om a ll s ec tors a nd en gagi ng te am s to wo rk th rough in teg rated des ign prob lems — s uch as bioc om p a tible s t retc hab le int erfa c es wit h on - chip s pi ke so rti ng, syn apti c tran sisto r s, or me mri sto r - bas ed n eur o morp hic h a rd ware vs . c onvent ion al d eep neu ra l network sys tems — i s one approa ch t o f a cilit a te t r a nsdisc ip lina ry t e a mwork . It wa s als o n oted d u ring th e works h o p t hat mat eria l scien tis ts are n ot w ell rep resen ted in t he neuro morphic co m mu ni ty, w hich is a mi ssed o ppor t unity to inc orpor a te th eir expert is e in t he c o - de si gn pr oce ss. Continued support of th e se type s of tr aining and educ at io n act ivit ies w ould fost er an un derstan ding of neuro morphic metho ds, jump - s t ar t cro ss - d iscip lina ry colla b ora tions, and enc o u rage the d evelop ment of a sha red v ocab ul a ry for t he next g enera t ion of n eurom orph ic res earc hers. Co nclu sion: The NPBH W orks hop su c ce ss full y broug ht toge ther engin eers, sc ient ists, clin ic ians , an d fu nders t o explore h ow n eurom orph ic eng ineering c an a dva n c e m edica l ap plica tion s. The wo r k shop inc lu ded se ssi on s on neuromod ulation, pro stheti cs, d evic es, an d we ar abl e s . P a rtic ipan ts emphas iz ed t h a t when dev eloping neu rom orphic t ec h n ologies , the foc us sho uld b e on s olv ing real wo rl d, c linic a l pro blems . T he key is ident ify ing sp eciﬁc a pp licat ions w here neu romor phic approac hes pro vide c le a r, demo n s t r a ble adv an t age s , p art ic ul a rly for rea l - time , clo sed - lo op s yste ms th at requir e low la ten c y a nd high ly e ﬃ c ient, res ilient circ u its . P ar tic ipants ident iﬁ ed s evera l s ign i ﬁ can t challen ges tha t the neuro morp hic biom edica l ﬁ eld fa c e s , whic h w ere in t he ar eas of c linica l ap plica tion , en g ineerin g, infr a str uctur e , an d c ommerc ia lizat ion . Throu ghout the wor kshop , parti c ipan ts al so emphasiz ed the ne ed for tra nsd iscip lina ry comm unicat ion a nd educat ion bet w een engineers , neu roscien tis ts , ma teria l sc ient ists, and c li nici a ns. Thi s wo r ksho p h as h i g hlig ht ed res e a rc h a rea s of int eres t an d ident i ﬁed exis tin g fed era lly funded p rog rams that ca n b e levera ged to furt her adv anc e the ﬁeld. Appe nd i x 1: Me etin g Synop sis The two - d ay NP BH wo rk sho p w as a si n gle - tr ack mee tin g stru ctur ed to pro gre ss fr o m cor tic al neu romod ula t ion a p proac he s t o perip hera l tec h nolog ies a nd p rosth etics , follow ed b y a mo der ated 21 dis c u s s ion t o genera t e a r o ad m ap , a n d a funder s pane l to discuss regul at or y c o nsi derations and fundin g opp ortu nit ies . Day t w o foc u s e d o n a dva n c es in ma teria ls re s ea rch a nd en a blin g te ch nolog ies , culmina ting in a mode rated di scu s sio n wi th fo ur the me s: (1) b en ch mar k s an d d at a , (2) clin ic al need: senso r technology , devic e spec iﬁcat ion, data collec t ion, (3) reg ulato ry c onsid erat ions a nd b a rriers , an d (4 ) infra s t ru c t ure need s . D ay 1 - Op e ni ng Rema rk s Ralph E tie nne - Cum m ing s int roduc ed th e co n cep t of neuro m or phic princ iples by dr aw ing in sp ira tion from nat ure w it h exa mp les f rom A n toni Gaudi ( Naut ilus shell for a rc hit ec tu r a l des ig n s ), L eo nar do da Vi nci ( sy cam ore s eed - ins pired helic opt er ae r od yn ami cs ) , John Hopﬁ eld (a ssoc iat ive memories ) , Pa ul Mu eller and Carv er Me ad ( tr an si sto r - b a s ed neuron mod els ( 4, 5) ) , an d Mi sh a Ma ho w ald and Er i c Vit to z ( nature - ins pired elect ron ics ). He pres ent ed G ert C auw enb erghs ’ neuro morphic anal y sis b y synt he si s fram ework (103) hi ghlighting t he diﬀeren t mo deling gran ula rities acro ss m an y sp ati al sc ale s , fr om s ing le - c h ann el tran sisto r mod el s ( t he wa y transis tors operate in the subthre shold) to whole - br ain comp u tati onal proc es s e s , em p h a s iz ing th e imp ortan c e of c onsiderin g t he en tire c on tin uum. R al ph as ser ted th at neu romorp hic prin cip les can impact hea lthca re in m ult iple dom a ins inc lu ding diagnos is , neural int erfa c es, b iologic a l a n alys is, a nd t he dev elopm ent of spec ia lized neu romorp hi c t o ols for m edic a l a pplica t ions . H e co n clu ded b y remind ing t he audience t hat the deﬁ n ition of “neu ro m orph ic” h a s evolv ed from st ric tly h a rd w a re - sp eciﬁc imp lementat ions to enco mpass low - level se nsor y p r oce ssi n g, AI a lgorit hms , d a ta scienc e ap proac hes, and c o m pu tat ional ac c elerat ion f or hea lt h c a r e a pp lic ations . A ndrea B e c kel - M itch ener , Deputy Direc t or of the N IH B RAIN In itiat ive, expla ined that BRA IN is a g lobal eﬀort focu s ed on disc overy a n d tec hno logy develo pmen t fo r bett er und ersta nding th e huma n bra in . Sinc e 2014, B RA IN h as in vest ed ov er 3.5 b illion d ol la rs in hu ndred s o f projec t s, y ielding rem a rkable dis c overies a bout b ra in fu n c t ion. B uildi ng o n the s e s igni ﬁcant ac hiev emen ts, B R AIN is int erest ed in levera g ing know ledg e to exp lore new a reas su ch as neu romorphic eng ineerin g a nd s y stems b ased on neu romorp hic princ ip les as pot en t ia lly tr ans format ive app roach es . T his emerg ing ﬁ eld prom is es ne w a rch itect ures, energ y - e ﬃ cien t c om putin g dev ices, a nd novel a l g orit hms poised to e nhan c e our un ders tan ding of t he b r a in, while a ls o co nt ributing to know ledg e a bou t h e a lth a nd dis ea s e a nd inno v a t ive biom edic a l dev ice d es igns. S h e joked “ I kn ow we’ re foc used on b iomed ical a pplic at i on s here, how ever, w e mig ht just sav e th e planet a long t he wa y ... ”. B R AIN ma intain s a n as s ertiv e d a ta sha r ing polic y (104) , ma kin g wiring dia g r a ms o f mu ltip le org an i s ms’ brain s, fu nct iona l elec trop hysio logica l m easu remen ts, a nd c ellula r “p a rts lis ts” a v a ila ble t hrou gh NI H - supp orted arch ives for res ea rchers t o access a nd utiliz e and enc ourag ed t heir u s e of BR AIN kn owled ge b ases . A d istin ct ive f eature o f BRA I N lies in it s int erdis c iplina ry a pproac h, rec ru iting no t ju st t raditio na l NIH g ra ntees like biolog is ts , bu t als o res e a rc hers from ph y s ica l s cien c e s and en gin eering , whic h h a s prov en ext remely fru itfu l for d evelopin g n ew t ec hn ologies and approache s. P aul Lan e , A cting Dep uty D ivis ion D irec tor, NSF E NG Div is ion of Elec t rical, Commu n ic a tio n s , a nd Cyber S ys te ms (EC CS) Div isio n st ar te d his welc om ing rem ark s wit h a fam ous line f rom T he Gra du ate w ith a sli ght upd ate : “Ou r prot ago nist, Ben jamin , is a t a pa rty w it h a bu nch of older in divid ua ls . O ne ma n c o m es up to him an d talk s abou t the futur e. He says one wor d: pl asti cs. We 're her e to update th at. I migh t s ay one w ord: energ y ” . He a sser te d t ha t the t rans forma tiv e a dvan c e s in s oc iety, in m edicine , f rom a r tiﬁ c ia l int elligen c e, g enera t ive AI , to larg e la nguag e mod els all req uir e e normous a moun ts of e nergy. A nd so 22 a pply ing neu ro m orphic p rinc iples t o ne w c irc uit d e s ig ns, devic e s a nd mat eria ls fro m t he bot t o m up w ill be es s ent ia l. He h as been int erest ed in neuro morp hics sin c e 2018, pa rtic ipat ed in the Bra in I ns pired Dynamic s for E nginee ring E ner gy - eﬃcien t ( BRA ID ) ( 105) E merg ing F ront iers in R e s ea rch a nd Innova tion (E FRI) t opic , a nd fore s aw neuromor phic in the F uture of S e mico ndu cto rs ( F uSe 2 ) (106) prog ra m s i n ter ms of s pin tron ic s, m e m rist ors, phas e cha nge ma teria ls, nov el gat ing c o nc epts, ion g a ting, even strin g ga tin g. He h ope d t o s ee new p artnership s an d i de as e merg e fr om thi s w ork s hop . Da vid Rampulla , Dep uty D irect or of the E xtramu ra l Scien ce Pro gram , A ctin g D irect or of th e Divis ion of Dis c overy Scien ce an d T e c hn olog y (Bio eng ineering ) at N IBIB , view ed n eurom orph ic eng ineering and neuro morphic co mputi n g as a w indow t o next g enera t ion biom edica l t echnolog ies , ba c kreferenc ing Ra lph’ s openin g rema rks t ha t highlight ed be ne ﬁt s of ne uromorphi c techn olog ies to the w hole body. He ec hoed th a t t he s e beneﬁ ts inc lude powe r c ons ump ti o n, h e a l t h, an d p ri vac y highlig hting the v al ue of on - chi p co mp uting th at neg ate s th e need t o sen d d at a to po we r - hungry da ta f ar m s . H e con clud e d with the ob s e rvatio n t hat n e ur omor phi c comp u ting oﬀers t he poss ibilit y of c ommu ni c at ing direc tly wit h hu ma n biology whi ch h as the pote nti al to open up ne w w a y s of e ngineer ing biolo g y for human he alth . Day 1 Key no te - T o wa rd s “ Ne u ro /P hys i o - mo rp hic ” D evic e s [ G eneration 3 ?] T imot hy Den ison , Profess or of E ng ineerin g Sc ienc e, Un ivers ity of O xford , pre s en te d his vis ion for nex t gen era tion ada ptive DBS d evic e s s tart in g w ith a retros pect ive on Med tron ics’s Percept TM for whic h a c linic a l tria l wa s rec ently c omplet ed . T he jour ney f rom op en - loop to c losed - loop ne uro modulat io n is signiﬁcant; h e emp ha s iz ed the impor tance of not ju st neuromorphi c but also clo se d - loop “ phy siomorphi c ” approache s us in g physiolo gic a l markers t o adju st s t imu l a tion in rea l - time to bet te r ma tch br ai n st ate s. A s emin al pa per by Pet er Brow n a nd Sim on Lit tle (107) lo ng esta blis hed a correla t ion b etw een b eta - b and po wer and sy m pto m s in Pa rkins on’s pat ients , ena bling a t herm o s tat - like a p pro a c h where s t imu l a tio n would be ac tiva te d w hen power exc eed ed a pre - se t t hreshold. H o w ever, mo st ad apti ve sti mul atio n re s earch w as con ducted be t ween “no on and te at ime ,” rai s ing que stions a bout ho w t hese sy s te ms perfo rm a c ross diﬀeren t t imes in t he cir cad i a n cyc le, especia lly durin g s leep. Fu ture devic es s hou ld th us inc orpo rat e bot h reac t ive and predic t ive ca pabilit ies , sim ila r to ho w bees p redic t se a s onal c hanges rat her than j ust r ea cting to im medi ate co ndi ti on s, cr e ati ng a tr uly “ ph ysiomorphi c ” syst em. D evelo ping a dapt ive d evic es requ ires ba lanc ing c linic a l nec essit y w ith p ract ica l elemen ts s uch as ec onom ic vi a bilit y, wo rkﬂow c onsidera t ions, a nd n a viga ting c omplex regu la tory pathway s . A m aj or c ha lleng e in d evelopin g n ovel b ioelec tron ic devic es is th e “chic ken a nd egg ” p roblem - needin g da ta t o jus t ify in ves tment but req uirin g in ves tm ent t o gat her t he neces sa ry da t a . De nison advo ca te d f o r an inc rem ental a p proac h usin g “s cienc e pla tf orms” that b uild u pon exis ting t echnolog i es w ith gua r a nt eed th era pies while ena blin g exp lora tion of nov el c onc epts wit h minim a l a dd ition al in vest ment, hig hlig hting a c olla bora tio n that en ta iled j ust o ne ﬁ rmwa re eng ineering (< $1M dolla rs) o n s pinal cord s tim ulatio n using neuro morphic proces s ing ( 108) . Deniso n is a proponen t of ﬁg uring out what one c an do to incremen t a lly imp rove exis t ing tec hn olog ies giv en t he “ hid den ca pab ilities ” in them. Deni s on reminis c ed a bo ut Neu ro P a ce ’s Resp ons ive Neu ros timula t ion ( RN S ) w hile it was s t ill in dev elopm ent , rem indin g that a u d ienc e tha t ea rly sti mul ator s w ere c onc eptua liz ed as “rev ers ible les ions ”, 23 wh ic h m ay ha ve anch ored th e ﬁeld to stat ic a pproac hes rather t han dy namic physiologic a l int ervent ion s. A key des ign c onst ra int i s po we r c onsumption . Sti mul a tor s typi call y run a t aro und 100 micro watt s ( wh ile car di a pa ce m ak er s r un ar o un d 1 mi cr ow at t) requir ing a dapt ive s ensin g an d c ontrol s yst ems t o opera t e a t about one - t ent h that pow er level. Tra ined as a circ uit des igner and inﬂue nc ed by m any hour s of ultr a - lo w pow er des ig n c onversa tion s wit h R ah ul S arp esh k ar (109) , De nison dev eloped a “ b ra in ra dio” (110) to ex tract micr ovo lt - lev el sig nals from the bas a l ganglia w hile c on suming o nly 5 mic rowatt s, wit h a n other 5 – 10 mi cro watt s for the al gori th m to adj u s t stimul atio n. O the r e arly suc cess e s inc lu d ed us ing c ortica l s ens ing for b rain - c o m puter in terfa c es in ALS pat ients (111 , 112) and es sentia l t remor (113) , bu t c hallen ges rema in in s ens ing from deeper bra in s tr u c t ures like th e bas a l ganglia a nd tha lamus w here s ig na l s w ere 10 – 20 t im es sma ller than s eiz ures . T he B R AIN - fu nded Su mmi t R C + S de vice (39) repres ent ed a tech nolog ical a d v a ncement , mov ing from a n a log preproc essing to on - chip f a st Four ie r transfo r ms t hat r edu ced po wer c o nsu mpti on to n anow atts per c ha n nel w hile ena b ling meas urem ent a c ro s s m ult iple freq uenc y b a n ds . Thi s sy stem a llow ed one t o set lower and up per bo unds a nd an in - ra nge “ no a cti on req uired ” bound , s imila r t o t he t heory of Wa lt er Cannon to a dhere to how hu ma n p hy s io logy w or ks. At is su e is ris k man agem en t a n d way s for reg ula to r s to ensure that neuro morphic or phy siomorphi c technologie s w oul d be safe . ( I nt eres ted rea d ers are referred t o Guidance document 60601 -1- 10 (72) an d Appe ndix 1 - Da y 1 M ee t t he Funde rs .) Denison c lariﬁ ed that Sum mit R C+S d ev i ce was embedded with actua tion limi t s, uppe r a nd l ower bounds, and fa llba c k m ode t o ens u re safet y . “With th e s e saf ety systems in p lac e, it crea t ed a sa n d box w here w e c ould explore n ew a lg orit hm s. ” Deni son su gge sted th at a g reat wa y t o engag e the c om muni t y on no vel a lgorit hm de si gn co uld be t hro ug h “the bat tle o f th e a lgorit hms ” w hich w o uld a llow diﬀerent devic es to run diﬀerent alg orit hm s for d iﬀerent dis eases a nd allo w res e a rc hers to mon eti z e their idea s . T he o ngoi ng AD AP T - PD s tudy aim ed at demons trat ing s af ety and eﬀec tiv eness of ada ptiv e DBS a lgorit hms in Pa rkin son’ s di se a se p at ients ( M ed t ronic Perc ept TM dev ice ) (37,38) com pa red t wo c ont rol a lgorit hms : a s ing le - thre s hol d “ b ang - b ang ” appr o ac h ve rsu s a du al - t hreshold ad ap t i ve DBS home os tati c a pproac h inspired b y p hy siolo gica l s yst e ms that a llow pa ramet ers t o vary w ith in a cc eptab le limit s . S tu dy r esu lts s ho wed th at the d ual - thres h old homeos tat i c a pproach d emonst rated sta tistic ally s igniﬁca nt c linic a l imp rovem en t s over t he s ingle - t hre s hol d approa c h , s u gg e sting tha t m imicki ng na t ural ph ys iolog ical res pon s e s m a y be more eﬀec tiv e. Denison n o ted th at 30 - 60% of t he va ria n c e in t he b eta band for PD p atien ts is rela ted to tim e of da y, sugg e s t ing future d evic es s hou ld incorp ora te c irc adia n rhy th micit y in to t heir d esig n. I n co n cl usi on, e nergy eﬃ c ien c y rem a ins a c ritica l co nsidera t ion, wit h neu romorp hic a p p roac hes needin g t o d e mon strat e a dvan ta g es over c las s i c dig ital s ig n a l pro cess ing , pa rt icu l a rly f or imp la ntab le med ical d evic es . Fut ure c h a llen ges inc lude des ig ning for ma nu f a ctu rabilit y, tes t reliab ility , m a nag ing s igna l drift in an a log syst ems, s yst em v eriﬁca t ion, risk m ana g e m ent , an d opt imiz ing a c ross multip le dim ens ions rath er th an foc usin g o n a s ing le para m e ter like pow er consu mpti on. Da y 1 K eyno te – Q&A Highl ight s Syd ne y Ca sh from Ma ssac hus etts Genera l H o s pita l inqu ired a bo ut the o n - chip ve rsus oﬀ - c hip p roc essin g dic hot omy, not ing t he t en s ion bet ween min i a tu rizin g com puting po wer on to ch ips vers u s st reamin g da ta to perform proc es s in g ext erna lly in th e c loud. De nison a c knowledg ed sy stem distr ibut ion as a n import a n t des ig n c onsidera t ion, referen c ing his wo rk wit h Cort ec’s b r a in int erc hang e s y stem w hic h s t reams data out 24 for ext erna l pro ces s ing bef ore sen din g comman ds b ac k, with la t ency being a c ritical fa c t or. Deni s on no ted a m u lt i - tiered a pproac h d e m onst rat ed by G reg W orrell a t Ma y o C linic us in g S u mm it RC + S , emp loyin g s imple em bedd ed a lg orithms for quic k res pons es while s im ulta neous ly s tream ing data to pa t ient - ca rried dev ic es a nd c loud proc e s s in g for m ore s oph is ti c at ed a lgo r i th ms . He emp hasi zed th at pa t ient experienc e mus t b e cons idered w hen des igning dis t ributed s ys tems, n otin g pra ct i c a l cha lleng es suc h a s r adi o l in ks b r eakin g when pat ient s ro ll over d uring s leep a nd th e burd en of requirin g p at ient s to ma nage m ult iple ext erna l c omp onents . Duygu Kuzu m from UCSD quest ioned the relia nc e on loca l ﬁeld poten tia ls (LF Ps ), h i gh ligh tin g their c omp lex natu re as vaguely deﬁ ned s pat iot emp or a l av erages w hos e inf ormat ion c o nt ent depen ds hea vily on m eas urem ent met hods, part ic ul a rly wit h la rg e DBS elec t rodes tha t in troduc e signiﬁ ca nt s p a t ia l aver a ging. Denison c lariﬁ ed t h at wh ile LF Ps hav e limita t ions, they hav e proven reliab ility over long - ter m rec ording s c o m p a red to a lt ernat ives, ackn owled ging t hat th ey rema in s uscept ible t o a rtif act s s uch as EC G si g nals t hat c a n b e pic ked up wh en d evic e s a re plac ed in the p ec toral reg ion. In r e spon se t o T obi D elbru c k’s onlin e ques t ion abo ut ep ilept ic seiz ure d etec t ion a nd qu enc hin g, Den i so n s ugges t ed t he ﬁ eld needs t o regrou p with fres h p ersp e c tiv es on epileps y tre a t ment, pa rt icula rly qu est ionin g whet her c u rrent a ppro ach e s like RNS m igh t be t rig gerin g in terv ent ions t oo la t e in s eizu re pr og re ssio n. Denison crit iqu ed c urren t app roac hes in c lu ding th e S A N T E tria l’ s a pp li c ation of v aga l nerv e s tim ulat ion pa ram e t ers t o n ew n etworks a nd dis e a s e s tat es, n ot ing u n int ended cons equenc es su ch a s s leep s tat e c orrupt ion when s timula t in g t he ant erior nu c leus of t he t h a lam us during nightti m e. Sihong W a ng from th e Univ ers ity of Chica g o a s ked whet her fu tu re neuro morp hic ana ly sis c irc uits should wireles s ly comm unica t e w ith senso rs an d stim ulat ors or be im pla nted t og eth er a s a n in teg r a t ed syst em. Deni s on empha siz ed that ra dio t elemet ry rema ins rela tiv ely exp ens ive c ompared to on - chi p co mp u ta tio n, with much of hi s current DyNeu Mo (114) s t udy foc used on prac t ical is sues like t ea ching families a bo ut rec ha rg e m a nag ement rath er th an a lgorit hms or neu roscien c e princ iples . He pref er red incorpo rat ing mo re proc es sing w it hin im plan ted devic e s g iven current limit a tio n s , not ing tha t t el em etry ty pica lly con sumes a n order of mag nitud e more pow er th a n on - c hip a lg orit hms . This w as c o nﬁ rmed by Ral ph Et ie nne - Cu mmi ng s w h o cla riﬁ ed tha t pu blis hed litera t ure show ed t elemet ry us ed approx ima t ely 80% o f on - chip po we r con sump ti on ( 74, 75) . D ay 1 S ess io n 1 : Ep il epsy & Co rti cal Disord e r s (Ne uromo dul atio n) Se ssio n 1 , mode rated by Gr ac e H wang of NI NDS an d Je ssi c a Fal co ne of NIB IB , w as a d eep div e int o epilep sy d iagn o si s and car e us ing DBS and rec ording techn olog ies and th e virt ual brain . The ﬁr st spe ake r wa s Willia m S ta cey , M.D. , Ph.D. , of t he U nivers it y of M ic higan . Hi s t alk , Da ta Dri ven cl assiﬁcatio n wi t hout a gold standa rd: epileptic HFOs , intr oduced th a t epileps y a ﬀect s one in a h und red peop le wit h on e - thir d no t respo ndi n g to tre atme nt whi ch present s s ubsta ntial c hallen ges in un ders tan ding s eizu re origin s a nd develop ing eﬀective t rea tment s. The c urren t s u r gic a l dec ision s f or int ra cta ble ep ileps y rely hea vily on em piric a l observ a tions us in g vis ual a na ly s is of EE G pa tt erns deriv ed from decade s - old methodolog ies. E xis t ing neuromod ula tion dev i c es ( VN S , DB S , RN S ) were n ever p roperly opt imiz ed f or epileps y trea t ment, w it h stim ulat ion p aram et ers of ten borrow ed f ro m oth er ap plica tio n s . So me c los ed - loop d evic es s ens e a nd stim ulate w it h fou r - yea r b a tt ery life b ut suﬀ er from ext reme ly low pr oce ssi n g po wer : “ 24 hours of recording res ults in j u st ﬁve m inutes of s a ved data and o nl y f r om a f ew 25 channel s . The prob lem is t hat t he st ate of the art for c linic ia ns is ju st not g ood en o ug h. We keep inven tin g a nd add ing ne w thin gs t o the clinic a l pip eline. A nd it rea lly has n't made m uch of a diﬀeren ce. ” He inv est iga tes high f requen cy o s c illa tions (HFO s ) w hic h rep resen t a p romis ing b iomarker fo r ident ify ing epilep tog enic z ones , b u t d isting uish ing p a t holog ica l from normal H FO s rema ins c hallen gin g d ue t o lack of go ld st and ar d val id at io n. Ad va nce d d at a - d riven approa c hes , includi ng non - ne gati v e matri x f a c tor i za ti on , graph theo ry, one - cl a ss su ppo rt ve cto r mac h ine s a nd Uniform Ma nifold A p pr oximat ion a nd Projec tion ( U MAP ) clus t ering of m illions of HF Os s h o w pot ent i a l for iden tif ying ep ilepto gen ic t is s ue w ith great er s pec iﬁcit y [ (41,42) , and unpubl ished d a t a ]. A r ecen t mul ti - c ent er H FO stu d y al so s howed p romis in g re sul t s for d etec t ing s eiz ures a nd other i nt racr an ial EEG p a ttern s to im prov e post surgica l seiz ure out c ome (40) . N e uromorphi c i s nas c en t in epilep sy , uses inc lu de r eal - time s eiz ure pred ict ion ( IBM TrueNo rth (48) ) an d r eal - time H FO de te ctio n from int racran ial EE G ( 22, 49) . In co n cl usi on , St ace y asse rt e d th at c urrent imp la ntab le devic es for epileps y su ﬀer from s igniﬁc ant lim ita t ions in sa mp ling freq u enc y, bit prec isio n, a nd c omputat iona l ca p a bilit ies , hamp ering res e a rc h prog ress . The ﬁ eld needs new a naly tica l t ools to pr oce ss hi gh - res olu t ion, h igh - throu ghput E EG d at a , and low - pow er dev ices t o bett e r int erfa c e wit h t he bra in, regar dles s of whether t hey speciﬁc a lly m imi c bra in st ructu re. The se co nd sp e ak er wa s Dr. Sridev i S arma of Joh ns H opkins U niv ersit y. H er tal k , A R e stin g S ta te E EG Bi omarke r Po ints t o Sei zure Onse t Zone , f ocused on dru g - res istan t epilep s y pat ients wh ose surg ical out c omes vary sig niﬁca nt ly, wit h succ e s s r a tes a verag ing a roun d 50% for s eizu re freedom follow ing res ect ion. Her la b d evelop ed t he “ EZt rac k - RS ” soft w a re tool that u ses res t ing s tat e int racra n i a l E EG to iden tify pot ent ia l epilept og enic zo nes w ith out relyin g on s eizu re da ta, func t ioning as a de c is ion s u p p ort tool. Her approach refra m e s th e problem by foc usin g on t he inhibit ion hyp oth es i s : epilep tog enic z ones may be s t rong “s in ks” bein g in hib ited b y “s ourc es” in th e bra in’s netwo rk durin g n on - seiz ure st ates . Netw ork n odes ca n be d eterm ined b a sed on jus t 5 minute s o f i nt ra cr ani al EE G da ta . She ha s t ested EZt ra ck - RS on 65+ p a t ients across s ix c ent er s dem on s t r a t ing tha t overlap b etw een predic ted epilep tog enic z ones and s u r g ically trea t ed a re a s c orrela t ed w ith s ucc essfu l s ur g ica l out co me s (43) . Sh e a ssert ed t hat neuro morp hi c c o m pu ting is releva n t fo r r e al - ti me closed - loop c on t rol such a s in 1) using a neu romorp hic feedb ack con troller for d rug releas e t o preven t s eiz ures vi a a no tio n al no n - invas iv e ultrasou nd neuromodul ation s ys te m , or 2) for m edic a l ap plic ation s wh ere hig h s pa t ia l a nd t e m pora l prec is ion is need ed t o ra pidly a lter b ra in or perip hera l nerv e temp eratur e u sin g a t hin - ﬁ lm therm oelec tric (115) devic e . Sri exp lain ed t ha t her Ph.D. trainin g w as i n pur e con t ro l theory ; s he b ecam e int eres ted in neu rosc ien ce in grad u a t e sch ool w hen sh e st udied w ith th e lat e S uza nn e Co rkin on mot or s yst ems and PD . It is her v iew t hat advan ced c ont rol t heory a pplica tio n s in n eurom odu lation have n ot kep t p ace w it h dev elopm ent s in other ﬁ elds, with m ost biolog ic a l control s yst ems still u sin g s im ple t hres hold or propo rtional a ppro ache s . Modern cont rol meth od s (opt im a l con trol, rob u s t c o nt rol, n etwork c on t rol) requ ire more s ophis t ic ated c omputa tio n a l a pproac hes t h a t could ben eﬁt from neuromorph i c imp lemen tat ion. Sh e co n cl ude d th at n euromorp hic c o mp ut ing c ould p rovid e th e n ec es sa ry c omp u t ation a l capa city f or imp lemen tin g s ophis ticat ed feedb ack con tro llers in a pplica t ions r a ng ing from foc us ed u ltras ound t o t herm oelect ric neu ro m odulat ion. The ﬁn al spe ak er wa s Dr. Vik to r J ir sa o f In se r m . He wa s heav ily inv olved wit h t he Hum an Bra in Projec t that used neuromo rphi c a r chitecture s i nc ludi ng S pi NN ak e r an d B r ai nS cal e s waf er . While the Human Bra in Pro ject jus t en ded in 202 3 , the ne uromorphi c infr a stru ctur e rema in s ac tiv e in the d igit a l neu rosc ien ce in fras tr u ctur e, EB R AIN S (https://ww w.ebrain s .eu/ ) . Ji r sa’ s t alk , Virt ual Br ain T wins for Med icine 26 - Encephal omo rphic Sys t ems , f ea t u red dig ita l t w in s usin g an in div idua l’s neu roimag in g da ta (inc lud in g diﬀu sion ten sor weig h t ed imag ing ) to model bra in d ynam ics at va rious sca les, enabling p redic tio n of epilep tog enic z ones and r esp onses to n eurost imu l a tio n. H e expla ined that c urren t v irtu a l bra in m odels opera t e wit h 2 00 - 300 n odes repres en ting b rain reg ions, but a re sc a ling to 2 00,000 no des to ac hiev e millim et er - s cale res olu tion needed for n euros tim ula t ion ap plic ation s. A lread y v irtua l bra in m odels h ave pred i ct ed epilept og enic zo nes that lie ou ts ide a rea s ac c e s s ible by imp lanted elec trodes (44 – 46) , pot ent ia lly exp lain ing cases of u nsucc ess ful s urg i c a l outc omes (2 9% of reg ions lin k ed t o fa iled c as es ). Dig ita l t wins built from pat ient - s pec iﬁc d a ta sho w surp risin g pred ic tiv e power fo r indiv idua l res pons es to diﬀeren t s tim ula t ion param eters in t erm s of s pat ia l - tempo r al dyn ami cs. H o w ever, m aj or co m put at io n al cha ll enge s ex ist in r unning high - res olut ion v irtual b ra in s imu l a tio ns : 2 0 minute s of f MRI res tin g st ate rec ording s require mor e than t wo days of pro cess ing t ime ev en w ith GPU acc elera tion. Current neu romorp hic a r c hit ec tures like S piNN aker a n d Br ai nS c ale S ha ve n o t been s ucc e s sfu lly a pp lied t o th ese full - bra in m odels du e to lim ita tion s in hand ling tim e delay s, n ois e, an d high d imen siona lity . H e conc luded that s imul atio n - based inferen c e for pers on aliz a tio n requ ires t ho u sa nds o f mod el s imu lat ions, c rea t in g a s ub st an ti al co mp u tati onal b ottle ne c k tha t ne uro morp hic arch ite ctu re s c ould pot en t ia lly a dd ress. D ay 1 S e ssio n 1 : P a nel D is c uss ion Hi ghl ig hts Thr e e clinic ia ns ( Sydney C a sh of Mass Gener a l, Na than Crone of Jo hns Hopki ns, Ke nda ll L ee of Mayo C linic ) an d t wo neur a l engin eers ( Giac om o V a lle of C halm ers Univ ers it y of T ec h n ology in S w eden, a nd C hris Roz ell of G eorgia T ech ) joined th e Ses s ion 1 Panel dis c uss ion. W hen a s ke d w h a t th ey would do if giv en a low pow er im planta ble c hip , mo st c linic ia n s em phasiz ed t he need for dev ic es wit h m ore c h a nnels , bet ter s am pling rates , a n d g rea ter s igna l - to - n oi s e ra t ios in fully im pla n t a ble w ireless f orm f act ors . Kenda ll Le e im pres s ed u p o n the pane l tha t t he br a in i s elect rochem ica l ; n euroch emica l s ensing pres en ts prom is ing opp ortunit ies a lon gsid e elect roph ysiolog ic a l approac hes but g enerat es eno rmou s am ounts of da ta r equirin g a dva nced p roc essin g. Lee further m ent ioned mu ltip le cyc lic s qua re w a ve v oltammet ry ( MCS W V) an d N - sha ped mult iple c yclic s qua re wave volt amm etry (N - MC SW V ) th at n ow al lo ws moni tor ing of a bs olu te va lue neu roc hem i c al ch ang es in th e br a in that ca n be us ed for elec t roc hem ica l clos ed - loop neuro modulat io n (116 , 117) . Vik to r J ir sa po in te d out th at m ult iple c hips w o u ld be n eeded b ec aus e th e bra in is a s p atia l tem poral s ys tem . N athon Cro ne agreed , expla inin g t hat there is a gr owin g ap prec iation th a t ep ilepsy a ris es fro m broa d er bra in n etworks tha n prev ious ly thoug hts, a nd th us it i s i mpor tant to s ample as much of th e netw ork a s po s s ible t o develop eﬀect ive a blat ive a nd n euro m odu lation thera p ies. The p anel w as in concur rence th a t t im e s cales pres ent a signiﬁ ca nt cha lleng e in optim iz ing trea tments , wit h m onths oft en need ed to det er m ine if int erven tion s are eﬀec tiv e, pa rticula rly i n ps ych iatric a pplica t ions . C ollec tiv ely, t he panelis ts ident iﬁ ed a crit ica l need for int ermedia t e bioma rkers tha t could prov ide ea rlier in dic a tio ns of treat ment eﬃ cacy b efore c linic al s ymp tom s cha ng e. Sri Sa rm a hi g hli ght ed th e impo rta nce of ma inta in in g s p a c e for scien tiﬁc dis covery alon gsid e AI a nd n euro m orph ic a p proac hes, a s f undamen tal un ders tan ding of bra in h ealth is s t ill lac kin g. Th e limit ation of c urren t d ata c ollec t ion in epilep sy wa s al so dis cuss ed, w ith hig h - reso lutio n dat a primarily a vaila ble on ly d uring in patien t mon itor ing r ather th a n i n ambul ator y s e tting s ; d at ase t s from am bu l atory device s are ext remely l imit ed . 27 D ay 1 S e ssio n 2 : Hu man - m ac hine I nterf aces ( Pr osthe ti cs ) Se ssio n 2 , modera ted by Lu ke Os b or n of C as e Wester n Res erv e Univ ers ity and Fr an cis co V al er o - Cue v as o f th e Univ ers ity of S outhern Calif ornia , focused on how neuro morphi c principle s and t ec hnolo gies m ay enhan c e H MIs , we a ra ble s ensors, a nd reha bilit a tion techn olog ies . The ﬁr st spe ake r wa s Dr . Ja me s Cotto n , a physic ia n sc ient i s t at S hirley Ryan Abilit yLa b . His talk , Clin ical Ga ps and U nderstandi ng Real Wo rld O ut co mes for P rosthesi s Users w ith AI Techno logi es , fo cu sed on the g aps in under s tandi ng r e al - w orld fu nct ion a nd how tec hnolog y rea d iness levels help context ua liz e how prep ared inn ova tio n s a re for a ct u a l im plem entat ion in pa t ient c a re ( 118) . He em pha siz ed the c ru cial n eed for p rost het ic dev ices to dem onst rate im prov ed rea l - w or ld fun ct io n t hro ugh home trials rather th a n labo rator y asses smen ts , poi nting o u t th a t l ab - b a s ed eva luation s oft en don’ t p redic t how t ec hn ologies will perfo rm in da ily life sit ua t ions (119) . H e expla in ed tha t pa tients priorit iz e com fort , r edu c ed pa in, a nd ps ych olog ical fa ct ors like a voidin g em barras smen t f rom falls ov er a dvan ced t ech nica l fea tures , wit h s tu dies s how in g s o c ket ﬁ t iss ues a n d skin problem s a re what p a t ients m ost freq uent ly rep o rt (50) . H is r ese arch fo cu se d on dev eloping a cc essib le mov emen t ana lys i s s yst ems for c lin ica l set tin gs t h a t c an c aptu re ric h biom ec hanica l dat a, wh i c h c ould provid e bet ter in div idua liz ed un derstan ding of p a t ients’ mo vem ent pa tterns for p redic t ing f a ll ris k a nd im prov ing int erven tions (120) . He proposed a fra m ewor k us ing c om putat iona l metho ds to lin k c linic a l a ssess m en t s w ith real - world a ctiv ities , c rea tin g “dig ita l twin s” th a t c ou ld simula t e ho w diﬀeren t int erven tion s m ight a ﬀe c t p a t ient outcom e s wit hout requ iring exten siv e randomized cont rolled t ria ls . He c onclu ded b y empha siz ing t he im porta nc e of ca us a l m odelin g approache s to unders ta n d ing th e c o m plex int era ction s b etween b ra in s igna ls, p rosth etic dev ices, a n d reha bilit a tio n outc omes for i mpr ovi ng lon g it udin a l p a t ient d a ta . The se co nd sp e ak er wa s Dr. R a nu Jun g , a n Ass oc iat e Vi ce C han cello r an d D isting uish ed Profes s or a t the Un ivers ity of A rkansa s . Her ta l k, Con verge nce : Deploying N eu r oelec t ronic Inter faces , em phas iz ed that tec h nolog y d evelopmen t m u s t be g in w it h t he desired impac t in m ind, add ressin g not o nly t he t ra dit ional “v a lley of d ea th” b etw een la b an d c ommerc ia lizat ion, b u t a lso th e gap bet ween c o m merc i a liz a tion a nd wid es prea d d eployment whic h in volv es ec onomics , polic ies, polit ics, a nd beha v ior. Jung explaine d t h at t he hea lthca re s pendin g t rend is p rojec ted to increa s e by 74 8% for w ellness tec hn o log ies by 2040 , while s pend ing on d i s ease t rea tment is dec re a s ing, hig hlig h t ing th e imp or t a n c e of d evelop ing tec hn olog ies that ma ke ec on o m ic s ense a nd c ontribu te t o w ell - bein g ra ther t ha n just t reating dis ease. She a ss erted th at th e fun dam ental c hallen ge o f bioh ybr id s yst ems, a s de s c ribed in her book (51) rem a ins la r g ely un s olved: biolog ic a l sys tems a re non - lin ear, n on - s tati onary , mul ti - s ca le and pla st ic , while exis t ing tec hnologies , des pit e so me a dap tiv e alg orit h m s, rema in rela tiv ely ﬁ xed. J ung pres ented h er wo rk on res pira t ory p acin g us ing neu romo rphic c ontrollers t ha t in c orporat e ada p t ive p a t t ern g enera t or s t o res pond to ch ang ing met abolic need s , su ch as CO 2 levels , aut omat i c a lly adju s ting t ida l vo lum e a nd res p i ra tory ra te (59) . Th i s r esp ir ator y p aci ng sy ste m u se d a c omputa tion a l mo del w ith a n euromorph ic c los ed - loop c ontrol s ystem th a t c a n automa t i c ally pers onaliz e s tim ulat ion pa tt ern s a nd make rea l - tim e predic t i on s. Fo r pr os theti c limb s , J ung desc ribed a fully imp la n t ed s yst e m develo ped s inc e 2 006 (a s p ar t o f an NI H B ioeng ineerin g Re se ar ch Par tne r shi p Pr og ram an d D ARP A ’s RE - NET and H AP TI X pro gra ms, a nd the Joi nt Wa rﬁ ght er M edica l Res ea r c h Prog ram ) that pr ovi de d s ensory feedb ack b y d irect ly stimu l a t ing nerv es, w ith s ens ors in th e pros t het ic ha n d co m m un i c atin g w ireles sly w ith imp lant ed elec trod es . The pr ost h eti c sy ste m use d a dis t ribut ed in terfas c icula r stim ulat ion a pp ro a c h with long it udinal w ires plac ed a long n erves , a llow ing for 28 s elect ive s t imula tion by v a ry in g elec trod e loc a tion (52,53) , co n ﬁg ur ati o n , an d w ave f o rm sh ape . Jung emp has ized that w hile resea rc h ers focu s on sp ec i ﬁ c sens ory mo da lities like pres s u re or pro pr ioc ept ion, pa t ients ma y ca re m ore a bout ot her as pect s o f s en sa tio n t h a t a re pra ctic a l a n d usefu l to t hem , s uc h as s imp le tin glin g s ensat ions w ith go od c la rit y and lo c ati on . Sh e c onc lud ed her p res en t a tion with t wo ca se s tu dies de m onst ra ted profou nd impac t : on e p a rt icipa nt s eamles s ly in tegra ted t he s ens ory feedbac k aft er ye ar s of non - sens orized p ro sthe ti c use as p ar t of th e NIH fu nded NEP H Stud y (121) , while a n oth er experien ced sen sat ion in his m issin g han d for t he ﬁ r s t t ime in 74 yea r s , de scr i b ed by his w ife as a “mi r acl e .” T he t hird s peaker w as Dr. E lisa Do n ati of th e Univ ers ity of Zu ric h a nd E TH Z uric h. Her t a lk, N eurom o rphic Ha rdw are f or Cl ose d - loo p in Heal thcare , describe d neuromorphi c h ar dw ar e as an id e al app ro ach t o com muni cate nat ively with t he human n er vo u s sy ste m. It c a n a d dress key c hallen ges in n eura l int erfa c es by work ing in a low - p ower reg im e th roug h ev en t - based proc essin g, w hich o nly p roc esses releva nt info rma tion r a t her than enti re signals continuo usly. She arg ue d th at n eurom orph ic sy stems en a ble loc a l proc es s in g of n eura l dat a, red uc ing bandw idth requ iremen ts for w ireles s tra nsmis sion, wh ile al so a llowin g f or rea l - t ime ada ptat ion a nd lea rning w ith s uperio r sp atia l - te mpor al re sol u tion c o mpar ed to co nve nt io n al app ro ache s. She pres ented h er wo rk on m ixed - s ign a l ne uromorphi c arc hite ctures t hat us e a n a log s ubthres h old c ircuit s t o emula t e neural b eha vior, with tra nsist ors func tion ing like “L eg o bloc ks ” tha t can be combi ned to crea te neurons and s ynapse s with customiz a ble fea t ures i ncluding a continuo us pr ostheti c h and c o ntro l th at m a p s elec trom yogra phy sig nal s to ﬁn ge r mo veme n ts . Thi s sy ste m ac hiev ed c ompa rab le accu racy to d eep lea rnin g ap proac hes usi n g sig ni ﬁ ca nt ly fewer pa rameters an d low er pow er requ iremen ts ( (56) , Manne s chi 2024 i n prep ). She describe d her n eurom orph ic a pp roa c h f or tac t ile feedb a ck in pros thet ics, show ing how a s im ple s pikin g n eur a l n etw ork w ith under 1, 100 neuro ns c ould mat c h the 90% c lass iﬁca tion accurac y of a C NN for iden tify ing 17 diﬀeren t g ra sp ed ob ject s (55) . She al so dev eloped a “ N e uromorphi c T win” w ith just 200 ne urons to e mul ate c ort ica l lay er 4 of the s omatos en s ory c ortex t hat ac c urat ely rep lica t ed a dap tat ion p henomena seen durin g in vas ive st imulat ion in hu man pa t ients, where s ub jects stop feeling s ensa tio n s a ft er co n t inu ou s s t i mu la tion (57) . Dona ti co ncl ud ed by applying neuro mor phic t echn ologies t o cr e ate adaptiv e pac emaker b a sed on cent r a l pa tter n g enera tors for c a rdiac ac tiv ity mo nit oring (58) , long - term m on itorin g in p a t ients w it h dement ia foc us in g on m ulti - day s an d m ul ti - su bject ada p t a bilit y, a nd p o s t s troke reh a bilit at ion . D a y 1 Se s si on 2 : Panel D is c uss ion Hi ghl ig hts Xing Chen of t he U nivers ity of Pitt s burgh , Nit ish T h a kor of Joh ns Hop kins Un ivers it y , Hele n H ua ng of Un ivers ity of Nort h Ca rolina , Vika sh Gilja of th e Univ ers ity of C ali fo r ni a, S an Dieg o ( and Par ad ro mi cs) , and Shi h - Chii L iu of t he Un ivers it y of Z uric h join ed t he pan el alo ng wit h the s peakers an d se s sio n co - ch ai r s . When aske d abo u t what im p ac t n eurom orph i c prin cip les c ou ld hav e in c onv eying high ly c om plex in for mation b ack to hu mans t hrough u se r in ter face s, Xing Chen fra m ed th e challenge in term s of eﬃ cacy , fea s ibilit y, and safet y for vis ual p ros theses . O n e need s a d evic e that ca n be ca librated q uic kly during a c linic a l tria l wh ere it w ill no t t a ke mon ths o f man u a l rep ortin g t o det erm ine phosph e ne loca t ions . In ter ms of s afety , o ne ne eds to ens ure tha t s ti mu la t ion d oe s not lea d t o propaga t ing a ctiv ities an d g enera t e epilep tic s eizu res. Fran cisc o Va lero - Cueva s em phasiz ed foc usin g on enablin g p erc e pt s rather t han replic a tin g s i g na l s , not ing that biolog ica l syst ems nat urally ﬁlt er s ens ory in format io n an d a dapt to c re a t e mea n in g fu l perc ept ions regard les s o f inpu t modalit y. Nit ish Th ako r id entiﬁ ed three key neura l int erfa c e 29 com pon en t s (de co d in g, en co di ng , an d co ntr ol ) ackn owl edgi n g un resolv ed q uestions ab out mod eling rec eptors in t he p eriphery , and unders ta nding n erve a nd f a sc icle en c oding precis ion . Sh ih - C hii L iu hig hlig ht ed t he exc it ing po t ent ia l of hig h - dens ity n eura l pro bes (10,00 0 elec tro des ) for s imu lta neo u s rec ording a nd s tim ulat ion, e m phas iz ing t h a t neu romorp hic a pp ro a c hes exc el a t extra c ting s pars e informa t ion for real - t im e, low - lat enc y clos ed - loop sy ste ms. Sh e s pec ulat ed t hat w it h 10, 00 0 elec trod es, one coul d i magi ne s tim ulat ing diﬀerent lay ers of the cort ex a nd ho w th at c ou ld dram atic a lly ass is t a blin d ind ividu a l one da y . When aske d if low pow er s yst em t h a t ada pts t o evolv e wit h users wou ld beneﬁt th e us er or reh ab ilita t ion, Helen Hu a n g disc uss ed t he im portanc e of pers on aliza t ion in m edica l dev ic es, n oting t hat int ersu bject va ria t ion requ ires a daptiv e ap pro a c hes at diﬀeren t t ime s ca les, s ugg esting tha t reinf orc ement lea rning c ould c on tin uously update d evic e pa ram eters . H elen a ls o exp ress ed en thus i a sm fo r neuromor phic an al og appr o ac h es o win g to its latency , esp e c ia lly if t hes e c irc uits c ould direc t ly in terfac e w ith th e h u man ne r vou s sy ste m . Mu ltip le pa nelis ts add ressed th e c ha lleng e of s par se an d varia ble d a ta in neura l int erfa c es, w ith E li s a Donati p oint ing out tha t sig n a ls c han ge ev en w ith in the s ame su bjec t ov er tim e du e to elect rode s h ifts or c ellula r t urn over. J ame s Cotto n s uggest ed t h a t c orrec tly implemen ted AI a pproach es c ould h elp b y a lig ning l a t ent ma n ifolds b e tw een in div idua l s t o ena ble m ore ra pid ca libra t ion (referenc in g L ee M iller (122) ) , wh ile V ika sh Gilja desc ribed h ow s p eech res torat ion s yst ems lev era ge f ounda tiona l mo dels from hea lthy in divid ua l s . F ranc isc o V alero - C ueva s argued that neuro mor phi c /physiomor phi c c omp u t ing exc els a t em bed ded log i c (c ont r a st ing with b ig d ata /opt i m izat ion) b ecaus e it c an lev erag e ins ig hts f rom t he p hysica l s ubst r a t e wh ere lea rning occ ur s . When aske d a bou t ha rdware a nd s of t ware lim ita t ions, Ni ti sh Th ak or no ted s ig niﬁcant challen ges in conveying neuro mor phic c on cept s to esta blish ed neu ros cienc e comm unit ies that are ba s ed o n experim ent a lly dr iven, m echa nistic , reduct ion ist neu ros cienc e. E ven the esta blis hed neuro morp hic DAVI S ca me ra ( htt ps ://iniv a tion . c om / s olut ions /camera s ) is a sm a ll p a rt of t he la rg er cam e ra ec os yst e m . Xi ng Che n high lighted prelimin a ry pay oﬀs in vis u a l pro st hetic s, w hereb y c a rryin g out rec ordin gs from t he bra in s im ulta neously du ring stim ula ti on , t he ﬁ eld is ab le to lev era ge ext ens iv e knowledg e of t he fun c tional ana tomy o f th e vis ual c ort ex a n d ap ply di m ens ionalit y reduc t ion m et hods to predic t t he loc a tions a t w hic h p ho sph e ne s ar e goi n g t o ap p ear i nste ad of relia nc e on patien t rep orts . Th e p a nel c onc lud ed w ith c a lls for dem onst ratin g c lea r clin ic al b eneﬁ ts of neu romo rphic a pp r oa c he s , bridg ing gap s wit h t ra dit ional n eurosc ienc e, a nd d evelop ing mu ltid i s cip lina ry te a ms that inclu de neuro morphic eng ineers , neur oscie nt ist s , a nd mat erials sc ient is ts t o cr e a te b ett er neu ral in terf aces to co mmun i cate with the nervo us s y st em . D ay 1 : M o derate d Di sc ussio n T his disc u s s ion was mo derated by Sunny Bains of th e U n ivers ity C ollege L ondon w ho be gan wi th explorin g t hree p oten tia l a rea s w here neu rom orphic c o m pu ting c ou ld c o nt ribu te to bio medic a l a pplica t ions : a dva n c in g sc ientiﬁc un derstan ding of neu ral s i g na l s , develo ping imp l ant a ble devic es w ith low pow er requirem ent s , a nd ena bling pers ona liza t io n thro ugh adaptive circuit s that can ch ang e as pa t ients’ n eed s evo lve ov er tim e. Gra c e Hw a n g ra i s ed a crit ic al q uest ion a bout h ow n euromorph ic approac hes migh t ad dress th e problem of va ria bilit y with in ind ividu a l s , wh ere pros th etic train ing ca n be c ome in eﬀect ive a fter jus t t wo weeks due 30 to neura l c han ges or elec t rode m ovem ent . Eli sa D o na ti expla ined t ha t despit e c h a ngin g b ra in rec ordin gs, s table “ la tent dynam ics” can b e ident iﬁed in lower - di m ens ional s pac e, a nd neu ro morphic s yst ems a re pa rt icu l a rly g ood at ext ra cting t he s e un derly ing dy n am ics an d w orking w ith them. Vikt or Jirs a i nt ro duced th e c onc ept of “ deg enerac y” a s a fun d am ent a l c ha lleng e in n eura l syst ems, where t he s ame in divid ua l ca n gen era te iden tica l beha v iors t h roug h diﬀeren t neu ra l p a t h w ays, creat ing diﬃcu lties for bot h en co ding and decodi ng neural si g nal s. Jame s Cotton n o t ed t h a t biolog ical sys tems n a t ur a lly ha ve va ria bilit y in dim ens ions t hat don’t aﬀect t a s k perfor m a n ce, s u g g esting tha t eﬀectiv e ass istiv e sys tems sh o uld mirror th is ﬂexibilit y rat her t h a n t rying t o elimina t e a ll va ri a bilit y. El isa D o n ati ask e d w h y Spi NN ak e r an d Br ai n Sc al eS w ere una ble to support V ikt or Jirs a’s virt ual b rain twi ns t o wh ic h V ikto r cla riﬁ ed t ha t thes e syst ems w ere op tim ized to us e sp iking neurons wh ile th e virt ua l bra in o pera tes at t he mea n ﬁeld level g ivin g ris e t o a fundame ntal m athe ma tical i n comp at i b ility issu e. Ralph E tie nne - Cu mmings emp h as ized tha t n eurom orph i c ap proaches s ee m pa rt i cula rly w ell - s u ited for pros t het ic ap plica tion s that requ ire th e full n eura l rec ordin g, dec od ing , enc oding, and c on trol lo op, ask ing th e pa nel w hether t his is w here neu romo rphic t ec hno logy is most rea d y for d eploy ment . Ni ti sh Thak or hig hlig ht ed t he need for b ett er tac t ile modelin g ac ros s t he ent ire neu ra l ch a in from rec ept ors , to f as cicl es , to c ortex for s ens ory pros t h etics highli g ht ing Luk e Osb or n ’s neuro morphi c prosthe s i s wor k ( 23) . Th ak o r further noted tha t s o me mod els a re ig nore d (123,124) w hile t he perip hera l nerv es rema in a n “ignored c ous in” c om p a red t o bra in res e a rch despit e t heir c ritic a l im portanc e. Xing Chen enthus ias tica lly expres s ed h er int erest in c om bin ing neuroreg enerat ion t hera p ies an d pha rma c eu ti c a ls w ith c losed - loop stimul atio n in t he perip hera l or c ent ra l nerv ous s yst ems, referen cin g Greg oire Co u rt ine’ s wo rk i n w hich c hem ical c o c kt a ils we re injec t ed in to t he s pin a l c ord for c lose d- l oop sti mula tio n (125) . Syd ne y Ca sh prop o s ed t h a t while th e int erfa c es w ith t he nerv ous syst em need t o be neurom orph ic o r ph ys iolog ically c ompatib le, t he c omputa t ion a l c om ponents in between cou ld p otentia lly us e a n y tec h nolog y t hat works eﬀec tiv ely. Mu ltip le pa rt icipa nts d eb a t ed w he t her t he v a lue of neurom orp hic a pproac h e s sh ould be evalu a ted based on p ract ica l perform a n c e me tric s ( siz e, weig h t, power, la tency, throughpu t, eﬃ cienc y ) r athe r th an p h ilos ophica l adh erenc e to biolog ic al p rinc iples . R og er M iller emp has ized th e imp ortance of phy s iolog ically a p p ropria t e sig nals based o n experi en ce wit h c o c hlea r imp la nts . W hile c ochlea r imp la n t s w ork well w hen elec trod es a re p roperly pos it ioned t o rep lica t e na tura l frequ enc y ma pp ing, m i s plac ed elec t rodes (w here freq uen cies do n't m atc h their n orma l loc atio ns ) requ ire pa t ient ad ap t ati on th at d oe s n' t al w ay s su c cee d . S o ther e really is a v a lu e in ﬁn ding w hat t he ph ysiolog ical s igna l is a nd u si ng th at to des ig n a ne uroprost hesis . C hris Roz ell obs erved tha t t hose w orking on r eplacem ent of lo st fu n c t ion ( e.g . , pro t h es is, B M I , co chl e ar imp la nt ) see m ed m ore ent hus iast ic a bou t n euro m orph ic solu tio n s t h a n t hose f ocus ed on mod ul a t ing a berra nt neu r a l act ivit y for c ond itions like epileps y or DBS for psychiatri c neur omo dulat io n . Ja mes Cotto n countered th at as a cli ni ci an h e’s ag no sti c if the solution i s neur omorphic , he would pres c ribe a wh eelch a ir in t he fores eea ble f utu re beca us e exos keleton s , thou g h a re more ne uromorphi c t h an wh eelc ha irs , a re fa r fro m usable f or mos t p a t ients . Br ad Aimo n e encou raged c linic ians to “drea m big” a b o ut wha t alg orit hms t hey wo u ld wa nt in im plan ts if po wer a nd la tency cons tra ints weren’ t an iss u e , sugg e s t ing neur o mo rphic hard ware opens a wid er a pert ure of c om putat iona l poss ibilit ies . Ken da ll L ee describ ed his n euros urg i c a l perspec t ive t hat the idea l 31 neu romorp hic so lut ion would prec is ely t a r g et t he spec iﬁc neu rons inv olved in a diseas e rather t ha n broa dly s timulat ing bra in region s, c o m parin g it to un dersta nding th e princ iples of ﬂigh t ra t her than s imp ly m i m ickin g b irds . Ken d a ll further a ss erted th at we can’t c h a nge t he brain but we ca n c h a nge t he compu te r to bette r ma tch how the br ain work s , a r guing th at i t is better t o int era c t w it h t he bra in t h roug h s pikes . D av id Wy att c autio ned ab o u t d icta ting t he en gin eering s olut ion and s ugg ested t ha t it is best to focus on the proble m an d p a tien t needs , and cha lleng e eng ineers to p ropose s olution s tha t bett er so lve th e prob lem, whet her th at i s neur omorphic , qu ant u m , an al og, MIMD , v on N euman n , or othe r . Tim Deni s on empha siz ed the im porta nc e of keep ing a s yst ems m indset a nd s h a re d the s tory of ho w M edt ronic revolu tion ized ada p t ive c a rdiac pacemaker u s ing motion c ry sta l w ith piez o elect ric to a void hea rt fa ilures . T he dis cuss ion c onclu ded wit h Grac e Hwan g empha siz ing t he im porta nc e of id entif yin g b oth s hort - term and long - term tech nolog ica l solu tio n s for s pec iﬁc d ise ase s and su gg es ting that neu romor phic a ppro aches s hould n’t be c onst ra ined t o human biolog y b u t c ould draw in s pir a t ion fro m ma ny species . Day 1: Mee t the Fund er s Dr. Z a ch McKinney of th e Uni ted Sta tes Food a n d Drug A dminis tration ( FDA ) prov id ed a h igh - lev el overv iew of FD A’s reg ulat ory f ramew ork a nd pro g rams fo r nov el medica l dev ic es. He expla ined t ha t th e F D A reg ula t e s a p p roxima t ely 2 0 % of t he U S ec o n omy, wit h the C en ter for Devic es and Ra dio logica l H ealth (C DR H) s pec iﬁca lly over seein g m edica l devic es thro u g h a ris k - based a pproa ch consi dering both intende d use and technolo g i cal char ac teris t ics. F D A uses a t hree - tier c lass iﬁca tion system : C l a ss 1 (low ris k) devic e s a re gen era lly exem pt from pre - ma rket r eview ; C lass 2 (mod era te ris k) t yp i c ally follow t he 51 0(k) p at hw a y to m a rket c leara nce v ia d emons trat ion of subs tantia l eq uiva len ce t o on e or mor e lega lly m a rket ed “p redic a te” devices ; and C l a ss 3 (hig hest risk) require p re - mar ke t ap pr ov al . T he Q - s ubmis sio n pro gra m prov ides free op port unit ies for d evic e dev elopers to eng ag e w ith F D A ea rly and of ten, a llow ing for feedb a ck on scient iﬁc approac hes before form a l subm ission s . I nvest igat iona l D evice E xemp tion s (IDEs) a re required for sig niﬁ c a n t risk clinic a l stu dies, wit h a ppro v a l bas ed prim a rily on patien t s a fety , ben eﬁt /ris k, and inform ed c onsen t c ons id erations rat her th an scien tiﬁc merit . F D A rec og niz es th ree c linic a l stud y sta ges : early fea s ibility stu dies (sma ll - sca le pro tot yp e test ing ) , tra d itiona l fea sib ility stu dies (pilot s tud ies wit h ﬁ nal dev ice), a nd pivot a l invest igat ions ( ﬁ na l dev ice fo r m a rk e tin g a pp licat ion ). The Brea kt hroug h Devices Prog ram exp edit es d evelop men t and review for h igh - impac t devic es trea t in g life - th rea ten ing or irrevers ibly deb ilita t ing co ndit ions, prov idin g fast er review c yc les a nd acc ess to t he Tot a l Produ c t L ifecy cle A dvis ory Pr ogram ( T AP) . He encour ag e d d evelop ers o f neu romorp hic t echnolog ies to eng age ea rly a nd often through t he Q - su b m issio n proc ess , a s regu l a tory a pp roac h es w ill be det erm ined ca se - by - cas e bas ed on compreh ens ive ris k - be ne ﬁt as se ssm en ts in view of eac h d ev ic e’ s techn olog ical c ha r a cteris t ics a nd int end ed us e . F DA h as a l s o b een w orking to develop Gu ida n c e on “p redeterm ined cha nge co n tro l pl ans (P C CP s) ” t hat (when ﬁ n a liz ed) (126) may provide a sy ste m ati c me ch a ni sm to ac cou nt for t he a da ptiv e na ture of n eurom orph ic t ec hno logies in reg u l a tory subm i s s ions. Dav id W ya tt , f ou n de r of the O penGP U F o und a tion , de scribe d his m o tivati on for fundi ng neur omorphic stand a rds a nd te chnologie s by p oin tin g to t he uns u s ta ina ble powe r co n s u mpti on o f curr ent AI d ata c ent ers, with indiv idual serv er ra ck s con suming 1 28 kilow a t t s eac h (eq uivalen t t o s i x h omes ) plu s 60+ liters of wa t er per m inut e ; expon ent ial g rowth in A I m odel c omplexit y a nd computat iona l d emand s, dou bling a pp roximat ely ev ery 3 .5 m onths ( Mo o re ’s La w s quar e d) , c rea ting un sustai na ble energ y 32 requ iremen ts t h a t can no t be s olv ed s imply by expan ding the po w er infra s t ru c tu re ; a nd b iolog ical s yst ems th at perf orm rem a rka ble c o m pu tat ion s wit h m inima l energ y ( ty pica lly abo u t 20 watt s), lev era ging enorm ous int erc onn e c tiv ity a nd t e m pora l proc e s s in g tha t fu ndam en t a lly diﬀers from tra d itional v on Ne um ann c om puting a rchite ctures. The Ope nGPU Founda tion is focu s e d on fundin g t est c hips , es pe c ia lly neuro morphic exemp la ry a pp licat ions t hat demo n strat e ad van ta ge s i n te m po r al , st och ast i c, or mul ti - dim ens iona l compu tin g , an d energy eﬃc iency . Open GPU ai m s to c rea t e s tandardiz ed har dw ar e - sof t w are int erfa c es t o ac celera t e broader adoption by application a nd s y s te m s develop ers , a s th e Open GL ( Ope n Graphics Libr ary ) a pp licat ion p rogramm ing int erface ( API ) did for ea rly GPU s and N vid ia ’s prop rietary CUD A ( C o m pu te Un i ﬁ ed Dev ice A r c hit ec ture) fra mew ork did fo r g en eral - p ur po se co mputi ng on GPU s ( GPGP U) . Wy att env ision ed in terfac es that c ou ld pot en t ia lly ext end to “w etwa re” (b i olog ic al a nd hybrid ) sys te m s , an d sug ge st ed this co uld all be a c hiev ed by follow ing t he s ucc ess ful m ode l of nim ble open st and ar ds as so ci ati o ns — like US B , RI SC - V, and OpenG L — throu gh indu s try - wide gro u ps which pr opose , dev elop, a nd ma inta in open protocols t o ena b le vibra nt an d in terop era ble tec hnol og y ec osystems . Rec ent a d v a nc es in n eurom orph ic computin g in c lude t he d evelopm ent of n euromorph ic interm edia t e repres ent a tion ( NIR , (10) ) f or intero pera ble s oft wa re dev elopm ent ac ross va s t ly diﬀeren t neu rom orphic ha rdwa re proc ess or arch itec tu re s a n d eﬃc ient sta t e space mo dels b a s ed on t ern a ry repres ent at ions , th a t hav e demonst ra ted reducing a LLM (la rge la ngu age m odels a s in Ch atGPT, Grok, D eepS eek) from a 700 - w at t NVI D I A - GPU wor kl oad, do wn i nto 1 3 w a t t co mmodi ty FPG A chip (127) . Op enGP U ( https :// www. O penGP U.com ) s eeks to in ves t in s olut ions f or cus tomer - deﬁned pro ble m s and pr ojects th a t ﬁ ll fun ding gaps betw een NSF , D OE and other ag enc ies, w ith focus on c ommerc ia lizat i on and eng ineerin g rig or. Fed eral Fund in g O ppo rt uni tie s The BR AIN Init i ativ e is d ise a se - ag no sti c an d foc uses on develop ing new t ools , t heo r ies a nd m ethods, and c ollect ing la rg e - sc a le d at as et s , whi ch it is int erested in levera g ing in co nv er ge n t ap p ro ache s an d tec h nolog ies a t t he in ters e c t ion of AI and neuro science ; see t he NIH H i g hlig ht ed T o pi c that e nc our age s releva nt pr op os al sub mi ssio n s to p ar e nt a n nounce m ent s at http s :// grants.ni h.gov/ funding/ﬁnd -a- ﬁt - for - your - res ea r c h /hig hlig hted - top ic s/ 18 . Add ition a lly, Neuro A I - relat ed a pplica t ions may be s u bmitt ed t o BR AI N I nit ia tiv e: Theories , M o d els a nd M ethods fo r Ana lysis of C o m plex D a ta from t he B rain (R0 1 C linica l T rial N ot A llow ed); see ht tp s :// grants.ni h.gov/gr a nt s / g uide /rfa - ﬁl e s/R FA - DA - 27 - 004.html . NI NDS su pports re s e arch using bi g data approache s a n d developi ng d evi c es to pro vide real - t ime, rea l - lif e da ta, with fun ding op port unit ies inc lud in g T r ansl ati o n al N eura l D evic es (R61/ R 33), Bioen gin eering Re se ar ch Gr an ts ( BR G ) , and C olla bor a t ive O pp or t unit ies for M ult idisc iplin ary, Bold, and Innov ativ e Neuros c ienc e (CO MBINE ) . F or m ore, vis it h ttp s:/ /ww w.gr an ts. gov/ and se arch for NI H opportunitie s via the K e ywo rd (s ) or Oppo rtunity Numbe r ﬁ eld s. NIB IB sits bet ween fundam ental sc ienc e/en gin eering at NSF and c linica l t rans lat ion at ot her NI H ins t itu te s , s up port ing broad platf orm t echno logy a nd biom edica l a p p lica t ions ac ross im ag ing, b ionic s, pros t het ics, w earab les, a nd A I/ML . F undin g c oul d be se c ure d t hro ugh m a ny mecha nism s includi ng p arent R01, R21 , R03 , a nd P41. F or m ore, vis it h ttps://ww w.nibib.nih.gov/fundi ng . 33 T he Bra in Beh avior Qu ant iﬁcat ion a nd Sy n c h ronizat ion (BBQS) p rogram integ rat es dens e record ing s of neu ral s ig nals , behav ior, a nd environ ment, w ith pa rt i c ula r int eres t in energ y - eﬃcien t s en s or t ech nolog ies for m inim a lly inv as ive m onit oring . F or m ore, vis it htt ps :// brain - bbq s.org/ . NIDA’s dat a scie nc e a nd infor m a t ic s program su pports dat a sharing , infra s truc ture, standard s dev elopm ent , an d int eg r a tion too ls f or BRAI N I nit i a tiv e res ea rch, w ith m ult iple w eb - acc e s s ible d ata a rch ives a lrea dy e s tab lis hed. Fo r more, v is it https://ni da . nih.gov/a bout - nid a/org a niza tion /div isions /divis ion - neuro science - b ehavior - dnb/da t a - sci en ce - res earc h The NIH Com mon Fund’ s Precis ion M edic ine w ith A I : I n t egrat ing Imag ing w ith Multim od al Da ta (PR IMED - AI ) p rog ram w ill co mbin e c linic a l imag ing w it h ot her t ypes of hea lth data (“m ult i m o da l dat a ”) t o dev elop inn ova t ive a r tiﬁ c i a l int elligen c e (AI ) - p o w ered c linic a l dec is ion su pport ( C D S) t ools t o ena b le new pers ona liz ed m edic ine st rateg ies . F or more, visit https://co mmonfund.ni h.g o v/pr imed - ai . N SF E NG Direct orat e has inves t ed ~ $58 5 m il lion in 875 brain - in sp ired res ea rc h aw ards a nd ~ $37 m illion in 65 neu romorp hic projec t s, w it h the Div isio n of E lec trica l, C o mmun icat ion s an d Cy ber S yst ems man agi ng ~ $107 mil lion in 225 b rai n - ins pired resea rc h aw ar ds an d ~ $19 m illion in 50 neuromor phic pr oj e cts (5 0 a war d s) . Th e mome nt um in th is spac e a s reﬂect ed by recent inv es t m e n ts mad e by N S F EF RI prog ra ms in clu ding BR AID w hi ch invest ed ~ 28 m illion in 16 a wa rds ( ( 128) , Ta ble 1 ) and Bi oco mputi n g th roug h En gin eering Orga noid Intellig enc e (Beg inO I ) whi ch foc use d on under standi ng fundamen ta l biolog ic a l princ iples a nd the u s e of biolog ic a l c o nst ruc ts w ith h ardwa re in terfac es for com pu ting ra t her th a n ex a c t biolog ical m im i c ry t o a d d res s uns usta inable en ergy dem ands of cu rrent com putin g approache s. F or m ore, vis it htt ps://w ww. n s f. gov /fun ding/opp ortu nit ies /emerg ing - front iers - res earch - innovation - efri - bio computing/13708/ nsf24 - 508 . NS F CI SE D irec torat e’s F oundat ion of E merg ing T e c hno logies (F E T) program fo c uses on going beyon d von Ne um ann com puting t hro ug h bio c o mpu t ing, q uantu m com puting, and ne uromorp hic computi ng approache s. F or m ore, vis it http s :// www.nsf.gov/fundi ng/o pportunities/fe t - ccf - fou n da t ion s - em erging - tec h nolog ies . NS F’s SBIR prog r a m p rovides up to $2 million t o seed initia t ives tha t t ra nsla te sc ien t iﬁ c brea k th roug hs into pro duc t s or s er vices, funding hig h - ri sk, h i gh - i mp act te chno lo gy deve lo p ment. F or m ore, vis it https:// s eedf und.nsf.gov/ . A rmy Resea r c h La b or a tory s eeks b as ic foun dat ional s c ien c e opp ort unit ies to en ha n ce s old ier/hum an s yst em s p erform anc e, wit h prog rams in Neurop hy siolo gy of Cognit ion a n d Bion ic Elect ronic s s up port ing hi gh - ris k p rojects wit ho u t prelim ina ry data req uirem ents. The f undi ng i nc lude s e ﬀorts to under stand bra in c ompu tat ions for g rou n db rea king dev elopm en t of t heories , a lgor ithms, and ha rdware, as w ell a s eﬀorts t o develop novel hybrid com put in g s yste m s t o h a rness the a dvantag e s of b oth living and non - livin g n etw ork s . F or more, vi s it http s:/ /ww w. arl. a r my. mil /col labo rate - with - u s/o ppor tuni ty/arl - b a a/ . D ARP A purs ues neuro - inspired com putin g throu gh m u ltip le oﬃc es in c lud ing I2O (AI a nd cy bers ecurit y) , M TO (m icrosys tems inc ludin g s ca lable a n a log neura l net w ork s ), and DSO (Defens e S c ienc e s Oﬃc e with broa d solic ita tion s ). F or mo re, v is it ht tp s:// www. darp a. mi l/ wor k - wi th - us/oppo rtunit ies . A ne w N S F - N IH bip has i c awa rd mec han ism ca lled “ B RING Syn B io” allow s projec ts t o s t a r t w it h NSF funding for tw o ye ars then tr a n s iti on to NIH for two ye a rs , c urr ently focused o n syn th eti c bio logy but 34 pot ent ia lly exp an ding t o ot her a reas . F or mo re, v is it https://w ww.nsf. gov/funding/ oppo rtunities/br ing - syn bio - biomed ical - resea rch - init i a tiv e - next - gen - b iotec hn olog ie s /ns f24 - 603/solicit a ti on . E xis ting NSF - NI H colla bo rations in clu ding t he C ollabora t ive R es e a rch in Com p u tat iona l Neu rosc ien ce (C RC N S ) prog ram, a nd th e Sma rt H e a lth and Biom edica l Res ea r c h in t he E r a of Ar ti ﬁ c i a l In tellig enc e a nd A dvanced Dat a Sc ien ce ( SCH ) prog r a m ar e als o in scop e for n eurom orphic app roaches . F or more on CR CN S: ht tps :// www. ns f. gov/fundi ng /o pportu nities/crcn s - c olla borat ive - res earc h - co mp ut ati o n al - neuro s cience . F or more on S CH: https://w ww.nsf.gov/fun ding/oppor tun iti es/ s ch - s mar t - h ea lth - biomedi cal - res earc h - era - arti ﬁ ci al - int elligen c e . Day 2: Key no te – Lea r ning f r o m S kin: from Mater ial s, S e ns ing Func tio n s to N eu romo rphic En gi nee rin g Zhenan B a o, K.K . Lee Prof ess or of Ch emica l E n gin eering of S tanf ord U nivers it y , pres ented her w ork on sk in - ins pired ma teria ls and elec tron ic sys tems that process s igna ls at the se nsor interfa ce. Th is mi mic s how skin me chanore c eptor s re s po nd to ex t er na l s timuli thro ug h fre quency - en c od e d s pike t ra in s ig n a ls . Bao ’ s te a m cr e ate d b re ak t hr ou gh m ate r ial s w ith n ano - con ﬁne d s truc ture s th at for c e po lyme rs to adapt planar c onﬁ gurations , s igni ﬁcantly enh an c ing cha r ge tr ansp ort mo b il ity comp are d to traditi onal s emic on du cto r ma teri als. The res earc h has p rodu ced a c ollect ion of s kin - lik e elect ronic m ateria ls inc lud ing stret cha ble s emicon duct ors, biodeg ra dable c on d u ct ors, s elf - h ea ling polymers , a nd tran spa rent elec trod es w it h ord ers of mag nitud e lower im ped anc e than meta l elec trod es wh en in con tac t w ith ti ssue (60) . Ba o’s grou p has d e velop ed in teg ra ted circu its w ith over 1 ,0 00 t r a ns ist ors on s t retc h a ble s ubst r a t es opera t ing a t m ega hertz freq uen cies , enablin g sig nal p roces s ing direc t ly at th e s en s i n g f ront . Thi s l ed to th e c reat ion of press ure s ens ors w ith high s pat i a l res olut ion (les s t h a n 500 m ic rons ) t ha t exc eed human tac t ile reso lut ion, dem onst rating t he po tentia l t o develo p elec tron ic skin s w ith ca pabilities bey ond biolog ic a l limit a tions inc ludin g acc urac y impro vem ents in Bra ille rea ding (60) . He r r e s e a r ch de mon str ate d neu romorp hic s igna l proc e s s ing by us ing ring osc illa tors t o gen er a t e sp ike tra in s igna ls w it h pres sure info rma t ion enc od ed a s freq uency a nd sy napt ic transis t or s t h a t comb ine m ultip le sens or inp ut s (129) . Fo r imp la ntab le a pplica t ion s , the team dev eloped “Neuro St ring ” - one - dimens iona l ﬂ exible prob es w ith 150 - 300 sens or s in a n arrow (150 - 250 μ m) dia mete r str u c tur e th at c a n b e imp l anted i n the b rain a n d gut f or s imu ltaneo us neuro transm itter m easu remen ts in r ode nts (130) . She then intro duced the la t e s t v ers ion of her elec tron ic sk in, which she c oined “ neuromorp hic s kin ” wh ere th e frequ enc y en c oded sig nal, th e s ens or, a n d c irc uit s are a ll m onolit hica lly int egrat ed int o a s ens ing s heet where pres s ure or t em pera ture info rma t ion c ould be enc oded. T h is sy stem is d evoid of any rigid elec tron ic com pon ents , can em ulat e th e s en s ory feedb ack f unct ions of b iolog ica l s kin, in clu ding mu ltim odal rec ept ion, n erve - lik e puls e - t r a in sig nal c ond ition ing, and c los ed - loop actua tion (13) . S he remin ded th e a udien c e t h at m ost of th e mat eria ls that she develop s are r elat ively new with un certa in long - ter m bio compat ibilit y. Bao conc luded by identify ing key c hallen ges inc lu ding the need for lon g - te rm bioc om p a t ibilit y t es ting, c ha llen ges in prot oty pin g no nc onve ntio na l m a ter ials, and det ermin ing wh ich a pplica t ions w ould ben eﬁ t m ost from th e hyb rid or neu romorp hic de s ig n appro ach. A s a m ult idisc iplina ry com m uni ty , we nee d to unders t a nd t he stren gt hs a n d th e limit ations for diﬀ erent d esig n approa ch es . 35 Day 2 Key no te – Q &A Hi ghl ig hts When que stione d a bout wh a t co ns tit utes “ne uromorphi c,” B a o clari ﬁ ed that her ap p ro ach mim ic s sp ike tra in commu nicatio n ra th er than precisely replic at ing neu r a l act ion pot ent ia l proﬁ l es . F ocusing on a s imp list ic un ders tan ding of n euron s and s ynapse s resu lts in s im ple r circ uit des ign s . Whe n a ddre ssing pow er eﬃc iency , B a o not ed tha t the des ig ned tr ansi sto rs h av e a lower c ha r g e carrier mobi li ty than c ryst allin e s ilicon . How ever, the t rans ist ors a ls o c ond u c t less c urr ent wit h m inimal hea tin g dur ing opera tion. P oten tia l app lic ations w ere disc uss ed, inclu ding enh anc ed tac tile s ens ing for hum a ns as well a s robots , pro s t het ics w ith c yb ernetic p orts, s kin - to - s kin c ommun icat ion mo da lities, and ﬂ exible s ens ors for t he gut . Da y 2 : S es sion 3 M a teri als f or Neurom orphics (Devices ) Se ssio n 3 was moder ated by Dr. S ha n tanu C ha kr abartty of W ashington U niver s ity St. Louis a nd Dr. Duygu Ku zum of the U niv ersit y of Ca lifornia S an Dieg o (U CSD) and c overed mult im od a l n e ura l int erfa c es and neuro morphic co - des ig n , Brainc hip ’s n eurom orp hic de v ic e, and n ovel ma teria ls for b iomed ic al app li c ati o ns and neuromor p hic co mputin g. The ﬁr st spe ake r wa s Dr. Duy gu Ku z um . Her talk, In no vat i ng Beyond Electroph ysio logy: M ultimo dal Neural Interfac es and Ne ur om orp hic Co - De sig n , add ressed limit a tions in c urren t n eura l rec o rdin g t ec hn ologies and expla ined th e va lue o f neu romorp hic c o - d es ign . While elec tr ophy s iolog ical rec ordin g t ec hn ologies ha ve ma d e s teady prog ress ion, re cor di ng ca pa ci ty is st ill a limit a tion , wit h even advan ced Neurop ixel pr ob es only capturi ng about 1 , 000 ne urons simult aneou sly. O pt i c al ima g in g tec h niqu e s , i n comp ar i son , h ave dou bled in rec ording c a p ac it y a pproxima t ely ev ery 1 .5 y e a rs a nd can no w ima ge te ns of thou sa n ds of neuro ns acro ss 3D brain volu mes (131) . Cur rent n eural in terfa c e cha lleng es includ e t iss u e d a mag e during imp la ntat ion , lim ited s patia l res olu tion of record ing s , and bul ky he ad - sta ge eq uipment for op tic a l im a gi n g. Kuzum ha s deve loped tran sparent gra phene - based m ic roelec trod es that a llow for mu ltim odal experim ent s by in teg r a t ing opt i c a l imag ing, opt o g enet ic st imu l a t ion, a nd elect rop h y sio logic a l rec ordin gs with out cro sstalk (61) . T hese t ra nsparen t elec t rodes s i mu ltan eous ly monit or neu ra l a c tiv ity a t mult iple s patia l sc a le s , inc luding rec ordin g loca l ﬁ eld pot ent ia ls fro m the c ort ical s urfac e while mon itorin g c a lc iu m s pikes from deeper lay ers (132) . F or rea l - t ime n eur a l s ig n a l p rocess ing, Kuzu m h as d evelop ed neuro morphic brain m achine interf ac e s b ased on r es istive ran do m acce ss me mo ry (R RAM ), -- a ls o ca lled memr isto r cr o ssb ar arr ays -- to p erform eﬃc ient , i n - me mor y s pike s ort in g (28) . She has de mon str ate d th at RRAM cro ssb ars c ould proc e s s rec ordin gs from 10 0 c han nels within 4.8 mic ros e c onds while c ons uming 30 - 50 tim e s les s energ y than F PGA approac hes (62) . Kuzum ha s al so led resea rc h on neur omorphi c c o - des ig n t o extrap ola te add itio n a l s pat ial inform ation from neura l rec ord ing s wi th ne u ro- inspi red approaches. Neural rec ordin gs fr om t he surf ac e of th e br a in include s p at ial ly in teg r a t ed d at a of the ne ura l act iv ity furt her b elow t he elec t rode. Using neuro morphi c c omp u t ation a l mod els wit h rec u rrent neural n etw orks , Kuzu m success fully dec oded ind ividu a l cellu la r c alc ium ac tiv ity in d eeper bra in la yer s u sin g on ly su rface p ot entials , expan ding s p a tia l rea ch (61) . Kuzum c on cluded th a t neuro morphi cs ca n help ne uros c ie nc e and neu romed icin e i n two m ajo r wa ys: [1] neuro morphic co m put a tion al mod els c an int e g ra te in forma tion from mu ltiple m odalit ies to create 36 a ccu rate, low dim en s iona l neural repres enta tion s to impro ve neu ra l pros thet ics or BM I performa nc e ; [2] neuro morphic hard ware can tr ansf orm n eural rec ordin g m et h odologies b y ena bling on - chip , r e al - time s igna l proc essin g with learning capab ilities tha t can sh ift the p roc essin g loca t ion of th e da ta an d inc reas e long evit y a n d s ta bilit y of n eura l int erfa c e s . The se co nd sp e ak er wa s Dr. To n y Le wi s , CT O o f B rai n C hip . H is ta lk, Flo w Mac hines a nd Neu romo rphic Devices , pres ented Bra in Chip’s c ommerc i a l neu romorp hic tech nolog y . The Ak i da 2. x chi p w as featured , which is a dig ital, ev ent - based, ne a r - memo ry co mpu t e sy ste m. Th e A kid a a r c h itec tu re has a di st ri but ed array of proce ss or s conne c ted in a mesh networ k . E ach pr o ces so r cont a ins four no des and e a c h node co nt ai n s a Tem p or al Eve n t - b as ed Neu ral Net works ( T EN N s ) pro c e ss in g unit. Eve nt - ba sed con volut ion dra mat ica lly red uc es c omputa tion al req uirem ents by o nly p roc e s s ing non - z ero values . F or e xample , even t - b a s ed conv olut ion ca n reduce t he m ult iply - acc umu late opera t ions (MACs ) from 225 to 2 7, w hile compu ti ng the sa me r esult. L e w is introd u c ed t he c oncept of “ﬂow m ach ines ” t o dis tingu is h fut ure ed ge AI (which include s neuromor phic s and TENNS ) from conven tiona l AI s y stems that levera g e s tat e m emory . Bra inC hip’ s TE NNs m odel is a s tat e s p ac e m odel t hat lev era ges ev en t - ba sed compu tat ion and ca n rep lace tran sfo rme r ta s k s , su ch as languag e m odels , spa ti ot em po r al dat a, an d t i m e series d ata , w hile dra mat ica lly red uc ing power c onsum ption. Th e a ppro ach u s es C heb yshev poly nomia ls to c omput e c onv olut ions in kern el mod e or rec urs ive m ode, a llow ing for t ra ining of ext rem ely d eep net wo rks ( 100+ la yers ) . Bra inChip ’s TEN N s h ave many uses inclu ding aud io denois in g, au d io ke ywor d spotti ng , g enerativ e AI l arg e - la nguag e mod els, and indust ria l art iﬁcia l int ellig enc e of th ings. T h e a udio keyw ord s po tting o pe rat es a t 0.2 5 milliw atts w ith 97 % ac curac y , and t he a u dio den ois in g s yst e m c an r emo v e back g ro und nois e from s peec h sam ples while ma in tainin g intellig ibilit y , p oten tia lly lea ding to “ su per int ellig ibi lity ” a pplica t ions . Br ai nChi p ’s TEN N s ach ieved b ett er perp lexity sco re s ( a bilit y o f a larg e - lan gua ge model t o pred ict th e next wo rd) th an Ma mb a an d Lla m a 3. 2 wi th a fra ction of the da ta at h alf the m odel s iz e. Le wi s emp has ized tha t succ ess ful neu rom orphic implem enta tion requires b oth clev er hard wa re a nd so ftwa re th at can t ake adv ant a ge o f th at h ar d war e to cre ate sp arse ne t wor ks . The thi r d spe ak er wa s Dr. Deji A kinw a nde of th e Univ ers ity of T exas at Aust in. H is ta lk, Tow ards Heal t h Moni toring & Com put in g using Advanced M a t erial s , focus ed on fu n d amen t a l m ateria ls , p a rt ic ul a rly tw o - dim ens iona l (2D ) mat eria ls like g rap hen e , which a re atom ica lly t hin, ﬂ exible, s tret cha ble, a nd p oten tially bioc om p a t ible fo r biom edic a l ap plica tions and n eurom orph ic com putin g. Thes e 2D ma t eria l s ar e le ss than a n a no mete r thi ck and oﬀer mu ltif unct ional prop erties a s elect rodes, senso rs, a m pliﬁ ers, and tr an si sto rs. A kinwa nd e ’ s lab explore d t hes e mat eria l s for b oth “orga n - tr o nics” ( in vi vo sen sin g) and “ ski n - tro nic s” (direc t s kin in terf ac ing ) by build ing on prior wo rk demon s trat ing gra p hene’s eﬀec tiv eness for imp la ntab le sensin g and elec tr ophys iolog ica l m o n itorin g (63) . Gra phen e’s t r a nsp arenc y enab les c omb ined h igh s pat ial resolu tio n opt ica l imag ing w ith hig h t e mp ora l res olut ion elec t rical rec ordin g , an d c an a ls o be used for t rea tm ent a p p lica t ions s uch as a t rans p a ren t g r a ph ene pac e ma ker (64) . Aki nw an de has d evelop ed ult r a - thin gra phene elec t ronic t att oos th at a re barely det ecta ble on ski n an d can b e u sed as a wr is t - worn b ioim peda nce s ensor f or blood pres s ure m onit oring . In addition, h e a c hiev ed clin ical - gr ade a c cur a cy of blood pres s ure m o nit oring through an AdaBoost regre ssion mod el (65) . Wi th re spect to neu romorp hic mat eria ls, A kinwa nde ha s cr ea ted “a tom - res is tors” (memris to rs in atomic a lly t hin ma teri als) th at de mon str ate d both bin a ry and a nalog re s i stance sw it ching (66) . Thes e a tom - re si sto rs ar e s uit able for n eura l net wo rks , with c lassiﬁ ca tion a ccurac y a pproac hing t heoret ical li m its in a n M NI ST dig it rec ogn ition tas k . He al so develo ped ﬂ exible s yn aptic tran s i s tors using g ra phen e wit h naﬁ on (a 37 ﬂ uorop olym er ion c onduc tor) as th e dielect ric lay er, en a blin g po tentia t ion a nd dep res s ion fu nct iona lity wit h ext rem ely low energ y d is sipa tion compared to oth er mat eria ls (133) . Th e mo st in tim a te in terfac e bet ween s y n th etic a nd biolog ical m a t eria ls t hat Akinw a nde ’s te am ha s creat ed is a “skin - g a te d tr an sistor for neurom orphic t a t t oos ” . Here, th e huma n skin direc tly s erves a s t he ga te d ielect ric la yer of th e tra nsis t or, a llow ing synap tic w eig ht ch ang es d irectly th roug h s kin int erfa ce ( unp ublished work ). Th eir fut ure g oa l is t o move from oﬄ ine pro cess ing to loca l sens ing and c o m put ing to en abl e d irect m onit oring of p hys iolog ica l indic a t or s , like s odiu m or ca lc ium ions , w ith m inimal d evic e requ ire men ts . D a y 2 Se s si on 3: P a nel D iscuss ion Hi ghl ig hts Gi na Ad a m of George W ash ingto n Un ivers ity , Jenn ifer H asler o f Georg ia T ech , Gert C auw enb erghs o f UCS D , and Ulku h a n G uler of W orc es ter Poly t echnic I n s t itute join ed t he pa nel a lon g wit h t he s peakers and se ssi on co - ch ai rs. Shant a nu Chakr abar t y high lig hted t he im porta nc e of embo di me nt an d c on t rol in hyb rid n eura l int erfa c es . C ross m a pping between biolog ic a l a nd synt het i c neu rons is miss ing . He ponde red if biolo gic a l neu rons wou ld t rea t silic on neuron s as t heir ow n. Jenn ifer Has ler ar gue d t ha t neur omorphic a pplic at ions or wea rab le, im pla nta ble sy stem cou ld very w ell be d one ent irely wit hin st a ndard co m ple men ta r y metal - oxide s em ico ndu ctor ( CMO S ) (7) . Gert C a uwenb erghs em phasiz ed retur ning t o the physical fo undations th a t Ca rver M ead es tab lish ed when c oinin g " n eurom orphic engineerin g, " noting tha t despit e diﬀerent s ubst rates (bio logy vs . silic on ), s imila r phy s i c a l princ iples g o v ern b oth sys tems ( 4 ,5) . Both biolog ic a l ion mo vem ent thro ug h c han nels a nd elect ron /hole m ovem ent in t r a ns i s t ors a re go verned by th e sam e foundation al physics - Boltzma nn s tat i s t ics and th ermodyn amics . T hese prin c iples ext end t o hig her levels , including networ k - level beh a viors a nd ener gy - based o ptim izat ion dynamic s . Th e ﬁe l d b en eﬁts fr om compl eme n tary to p - do wn and botto m - up approa ches t h a t c rea te sy nerg ies b etwe en comp u tati onal sys te m s n eurosc ienc e a nalys is an d ne ur omo rp hi c sy ste m s engin eering synt hesis . Thi s appro a ch en a ble s s ca ling from car riers u p t o wh ole bra in - level in terf ac e s w ith isomorp hism at every level and h a s le d t o mo re eﬃc ient neura l int erfac e circ uits ins pired by neur omorp hi c c omput ing princ ip les (134) . Ul kuha n Gu ler des cr ibed a bio morp hic transc utaneo us s en s i ng sy ste m tha t her lab p roto typ ed to det ec t oxy gen and carbon diox id e molec u les in a p at ch or w atc h f or mf ac tor . Gina A dam express e d th e need f or: 1) dat ase ts to co - d e s ign mat eria l s , devic es, a nd a lgo rithms tha t ca n com pute loc a lly. This requ ires a c los ed loop b etw een s ens ing, computin g, a nd act uat ion; 2 ) met rics, inc lud in g energ y eﬃcien c y, c ompact ness, nois e, a n d robus tness of t he ha rdwa re and s oftware; 3 ) int erdis c iplina ry c olla bor a t i on an d educ ation . When asked abou t new neuromorphi c opp ort unit ies in using new m ateria ls , panelis t s dis cuss ed t he imp orta nce o f bioc om pat ibility a nd CMOS int egrat ion fo r new mat eria ls . Deji A kin wan de su gge st ed t h at the hum a n body should b e int egrat ed a s an active co mponent of th e devic e rat her t ha n just v iewin g t he hum a n body a s a pass ive tar ge t to pr obe , tre at, and/o r stud y . Wh en dis cuss ing w het her n eurom orph ic is the on ly s olut ion to p ower an d ban dw id t h c hallen ges , Shant a nu Chakr abar t y emphasiz ed iden tif yin g s pe c iﬁc applic a tions where neu romorphic prov ides c lear adv an t age s (e.g., laten cy t o ach ieve rea l - t ime close d- loop s yste ms , ext remely eﬃc ie n t a nd res ilient ci rcui ts ). When a s ke d a bou t c halleng es in ma nufact urin g n eurom orph i c ch ips for b iomed ic ine, To ny L ew is expla ined tha t he ta rgets CMOS becaus e it is widely ava ilable . Ot her manuf a ctur ing challe ng e s in c lud e 38 s plit foun dry issu es, t ool c h a in limit a tion s, an d th e hig h cost s ass oc i a ted with c h ip fabrica t ion, w it h es tima tes r a n ging fr om $ 2 - 30 m illion jus t to sta rt a chip (w hile ot hers a rgue tha t a few t housa nd dolla r s on a mul ti - p roject waf er run is suﬃc ient a t la rg e volume p roduc t ion lev els ). When a s ke d a bou t be ne ﬁt s o f on - chi p learn ing , Duy gu Kuzum expla ined t hat neural sig na ls drift ov er time for a va riet y of biolog ic a l a nd n on - biolog ic al reas on s. On - chip lea rning c ould be a fas t an d eﬃci e nt wa y t o co rrect for s igna l drift ; th is wou ld ben eﬁt long term experim ent s th a t tr a ck t he s a me type of neur al dy namics ov er tim e to study beh avior and c ognitio n in both a nim al a nd h um a n s . Ul kuhan Guler added th at interpers ona l an d int ra per s on a l v a ria t ion s c ould als o b e addre ss ed by on - chip lea rning . Ger t Cau wenb er gh s oﬀered th at from a c irc uit pers pect ive, one a d d ition al a dvanta ge of a dap t a t ion is to c ircu mvent imp erfec tions . Wh en as ked a bout c hallen ges in clin ical t ra nsla t ion , pan elist s expla ined tha t a key cha lleng e is that l ong inv est m en t horiz ons d o no t a lig n w ith t ypica l bu s ine s s expect ation s, as well as reg ulat ory p athwa ys, priv ac y c onc erns, a nd s ec urit y c on sid eration s. D a y 2 Se s si on 4 : Medic al Im a gin g , W earables and A n alys i s Se ssio n 4 , modera ted by Jennifer Bla in Chr ist en of A riz ona S tat e Un ivers ity a nd Ralph E tien ne - Cu m min g s of Johns Hopki ns , cov er e d we ar ab le s, biomed ical in terf aces , an al ysi s , a nd c o mpu ta tio n using neu romorp hic dev ices . The ﬁr st spe ake r wa s Sihon g Wang of t he U nivers ity of C hicago. His ta lk , Bio - el ectr o ni c interfaces: from high ﬁdel ity biosensi ng to in tegra t ed ne urom orphic e dge com puting, presented new bio - elect ronic in terfac es that ena bled prec is ion m edic ine t hrough la rge - sca le ph ysiolog ica l da ta m onit orin g and ne uromorphi c compu ti ng . He ass erted that c u rrent comm ercia l wea rable a nd imp l a ntab le dev ices fa c e s everal c ha lleng e s th at limit dat a collec t ion ca pabilit ies . The se includ e limited func tion and l a ck of a cce ss to imp orta nt p hysiolo gic a l bioma rkers in w ea r a ble dev ic es and re duc ed lifesp an of im pla nta ble dev ices (typi ca ll y < 2 ye ars in th e case of b ra in elec tro des ). Ac a demic researc h is in creas ingly ut ilizin g A I to ana lyz e wearable a nd impl antable data , bu t the r aw se nsor da ta i s t yp ically proc ess ed thro ugh c lo ud compu ting, which in cre ase s lat ency , cons um e s more p ower, an d ris ks privacy brea ches . He po s ed two g ra nd ch alleng e s for the ﬁ eld a s follows : 1) how to dev elop c onfo rmable, elas tic , low proﬁl e s ensors for c ollect ing hi gh - ﬁd elity, mult i - mo dal, and lo ng - t erm st a ble data a t s kin /tis sue s urf aces? And 2 ) how to inc orpo rat e neu romo rphic edge c omput ing func tions int o weara ble/im pla ntable s ys tem s? To ad dr e ss gr and c h a llen ge #1 , Wan g has d evelop ed a so ft , str et ch ab le , im muno - compat ible, bioa dh es ive , b ioelect ronic int erfac e to imp ro ve the s ig nal - to - nois e ra tio of th e sensor . Th e prop erties of th e int erfa c e preven t s igna l deca y by con for m ing to the ti ssu e and preve nt ing ﬁ br o si s for long - ter m imp la nts . The h ydr oge l interf ac e is c om pris ed of a n i mmu ne - c ompa tible, se m icond ucting po ly m er , whi ch res ult ed in a sup pr e ssed fo re ig n bo dy r e spo n se ( FB R) when impl anted subcut a neo usly for 4 w ee ks (135) ; when selen oph ene is inc orpor ated s uppres s ion of th e F BR reac h ed 68% (82) . An or g a ni c elec troc h emica l tra nsis t or (OE CT) wa s de s ig ned us ing th e h y drog el s emic onduct or to recor d bot h elec t rica l an d chem ical si gn als. Thi s bioelec tron ic interf ac e can d irect ly at tac h t o tis sue s urfac es, lik e a b ea ting ra t hea r t , a nd ma in tain st able s ig na l qualit y t hro u g h st rong mechan ical in terf a cial a dh es ion (136) . 39 Reg a rding g ra n d c h a lleng e #2 , Wang d emo n s t rated a fu lly st retc hab le , neuro morp hic c omp uting 3x3 ar ray with inte gra te d me mo ry and comp u ting fu n c ti on s direc t ly on the ha rdwa re u sin g O ECT s . Th e neu romorp hic dev ice h as hi g h cla ssi ﬁc ati o n ac cur a cy f or ab n or m al EC G si gn al s and mai n t ain s that ac cur acy dur ing inc reas ing me ch an i cal str ai n ( 30) . He th en fea tu red Gin a Adam’ s w ork o n neuromorphi c de tecti on of propagation w avefront in car di ac ven t ric ula r ﬁbrilla tion to tri gger prec is e a blation (137) as an ins pira t ion t o his rec ent res ea rch on la rg e - sc al e , s t ret cha ble neu romor phic c irc uits for on - body edge pr oce ssi n g of sen sor y da ta for p rec ise t reatm ent d elivery ( 67) . He c onc lu ded his talk b y pos in g fou r addit iona l quest ions for t he ﬁeld to res olve, a s follow s : (1) Which part of m ac hine lea rn ing ( ML ) / neu romo rphic data proc ess ing funct ions sh ould b e phys ically int egra t ed with s en s ory dev ices (i.e. edg e vs c loud c o m put ing) ? (2) How ca n w e bet ter p hysica lly integ rat e an d func t ion a lly merg e s ensing fu n c t ion s w ith neur omorphic compu ti ng? (3) Wh ich ty pe of ML alg orit hms an d neu romo rphic dev i c e t ypes could be t he m o s t s uit able for int egra t ing w ith sen s ory d evic e s to provi de t he low es t com plexit y and pow er c onsumpt ion? (4) Wh at s h ould b e th e des ign of p eriphera l c irc uits that a ls o hav e tiss ue / sk i n - like p ropert ies ? The se co nd sp e ak er wa s Er ik a S ch mitt of Cambry a . Her talk, Novel neuromor phic algorithms for ada ptive lea r ni ng at the ed ge , foc used on t he au di t or y blind so urce separa tion probl em using synchro ny loop prop aga t ion t hrough bidirec t iona l s i g na l exc hange ac ros s gra ded s pikes in den s ity net w orks (71) . Ca mbr y a’ s network a rc hit ectu re wa s in s p ired by th e a udit ory s yst em with ton o t opic org a nizat ion a nd inc rea s in g ly c omplex c og nit ive fu nct ions . As a re sul t, t hi s net work archite cture can se par at e mu ltiple a udio sig nals from c row ded , no i sy env ironm en t s wit h no prior t ra ining dat a , a lea r n ing tim e of 0. 5 se cond s, and a m odel s ize o f only 12,000 par am eters. Wh ile over 1. 5 m illion peo p le glo ba lly a re aﬀ ected by so me fo rm of hea ring los s, f ew p eople w ith mild to m oderat e hea ring loss use as s istive d evic es . Ca mbr y a’ s tec hnology c an p rovide an a lternat ive to addres s he a rin g loss by allo wing u sers t o s elec t an d a mp lify spec iﬁc voic es in cro wd ed enviro nm ents insta ntaneously . L ea rning is h appening direc tly on t he dev ice with no cl ou d pr oce ssi n g required . He r tea m is pu r s uin g ma rket en try through tra dit ional ha rdwa re wit h a ﬂ exible S DK t h a t c an in t egrat e wit h exis tin g h e a ring devic e s . Lon g ter m, Ca mbry a wo ul d like to imp lement t heir a rc hit ect ure on neuromor phic hard wa re . F inally , S ch mitt no te d tha t the neuro morphic nature of their algor ithm is n o long er div ulg ed t o pot entia l inves t or s bec a use the te rm “neuro m or phic” mak e s the te ch nol ogy appear more ris ky. Th erefore, th ere is a n eed in this c urr ent funding e co syste m to gen erat e exc item ent a roun d neuro mor phics and t heir nea r - t erm , com me r ci al c apa bilit ies . T he t hird s peaker w as J. B ra d A im one of Sand ia N a tiona l L abora to r ies . Hi s t al k, Fro m ne w neurons to new chips: a path to a new w ay of thinki ng about t he b rain? , di scussed ho w neuromor phic compu ting can help us underst and n eurosc ienc e bet ter b y m o v ing bey ond t r a d itiona l seq uent ia l proc e s s ing m odels t o m ore br ai n - like c om putation a l fra meworks . He ob s erved th at c urrent theoret ica l n euroscienc e fra m eworks as overly sim plis t ic , often a pp lying t he sam e fundam ental prin c iples to diﬀeren t bra in region s des p ite t heir div erse a rc hit ectu res a nd f unct ions . Ai mone ’s te am ha s ex plo re d spati al - t em p ora l pro ces s ing a p p roac hes th a t trace cau sal c hain s thro ug h s pikin g n etw orks, levera g ing s truct ural rela t ions hips from c onnec tomes ra th er tha n relyin g s olely on s eq uent i a l t ime - b ased an al yse s . The y h av e f ound that ne uromorphi c s yst em s exc el led a t samplin g p roblem s ( i.e ., Mo nte C a r lo me thod s (68) ) a nd p hysica lly mapp ing spatia l 40 int era ct ions ( i. e. , ﬁ nit e elemen t a pproach es (69) ) , whic h ma y reﬂ e c t fu n d amenta l compu tat ional prin c iples of the br a in. Th eir neurom orph ic app roac h to s olv in g pa rtial diﬀeren tia l equ a tion s show ed d yna mi c s rema rka b ly s im ila r to b iolog ica l n eur a l rec ordin gs , de s pit e dras tic diﬀeren ces in arc hi tectur e s. A imo ne con clu de d t hat “ w hat m a ke s neu romorp hic hard ware s pecia l a re th e th in g s that ma ke bra ins spec i a l .” Bot h u tiliz e hetero gen eity , p a r a llel proc es s ing, a nd s pat i a l tem pora l p roc e s sing . T hus , thes e att ribu tes des erve m ore fo cu s when creatin g t heoret ica l fra meworks for u nders ta nding h ow the bra in w orks . D a y 2 Se s si on 4 : Panel D is c uss ion Hi ghl ig hts Andreas Andr eou and Robert Steven s of Johns Hopk ins U nivers ity , P a m ela A b s h ire of t he Un ivers ity of Mar y lan d , Co ll e ge Par k, an d Amp ar o G üe me s of t he U nivers it y o f Cam bridge i n the United Kingdom joined th e pa nel a lon g w ith the s pe a kers a nd s ession c o - ch ai rs. An dre as Andreo u h igh ligh ted the imp ortan ce o f con str ucti ng algor ithm s th at ex plo it tem poral proc e s s in g or time , an d th e need to in ves t in res ear ch a nd des i g n m eth o dolog ies , like co - desi gn a lgor ith m s w ith s ynchrono us cir cui ts or even t - ba se d ci rcui ts. However, de s ign aut omat ion tool s a re not able to sup por t auto ma ted wo r kﬂo ws f rom a l gori th m - to - chip de sign. He ass ert ed tha t n eur a l c hip lets a re the way t o go , a s t hey a re mod ul ar , s cal ab le , a nd reus a ble . Neura l c hip lets can be in teg rated int o hetero gen eous dev ices of va rious c om plexit ies wi th diﬀeren t m at eria ls a nd s t ructu res . He conclude d with a newly relea s ed f und in g opp ort unit y ca lled N ati on al Adv an ce d P ac k agi n g M anu f a ctur i ng P r ogr am (94) an d su gge ste d th at t he N IH pay s atten ti on to this funding m ode l. T he pa n el discus sed c ha lleng es in b ring ing neu romorp hic t ec hnolog ies t o hea lt h c are, inc lud ing regu la tory ba rriers , integ ratio n wit h exis tin g clinic a l wo rk ﬂ ows, b a rrier w ith mu ltiple dis cip lines need ed t o be s ucc essfu l (neu ros cienc e, ma teria ls, data sc ienc e, eng ineer ing ), a nd clea r dem onst r a t ions of the b en e ﬁt s of ne uromorphics . Pow er requirem ent s w ere deba t ed , w ith m an y par ti c ip an ts argui ng t hat ne uromorphi c approache s (l ow po wer , low heat diss ipa tion ) a re es sen tial fo r impla n tab le devic es , w hile a fe w su gge ste d th e pow er a dvan tages might be ov ers old for many med ical a pplic ation s . In di s c u ss ing ﬂexib le bioelec t ronic s , p anelis t s hig hlig ht ed the p oten tia l for neurom orph ic a ppro ach e s to ena ble bet ter sen so r fus ion acro ss m odalit ies (molec ula r, press ure ) a nd diﬀerent ti me - s c ale s . Pan elis ts stro ng ly iden tiﬁ ed a need f or ne w des ign too ls, fab ricat ion p lat forms, syst e m s int e g rat ion beyo n d ch ip d esig n, an d st an d ard s for neurom orph ic d evic es to acc elera te d evelop men t a n d redu ce tim e s pent on c usto m im plem entat ions . The term “neuro morphi c” was noted a s a pot ent ial h ind er a nce to war d s technology ad option, and sug ge sti on s were mad e to f ocu s inste ad o n de mo nstr ating the “ma g i c al” c a pabilit ies tha t ca nn o t be a ch ieved with c onv entiona l app roac hes . Day 2: Modera ted Di s cussion Openi ng D isc ussio n on Be nchmarks and Data Work sh op part icip ant s id en t iﬁ ed th e need to develop spec iﬁc, sta n d ar dized d a tas ets rather t han foc usin g on b enc hma rks, w hich some pa rt i c ipa nts fou nd to be res t ric tiv e. Bra d Aim one arg ued th at benc h m a rking, des pit e its p opularit y in A I d evelop ment, w ould be c oun terp roduct ive fo r th e neurom orp hic comm unity by forc ing develop ers t o us e outdated pa r a digm s w hile t he ﬁ eld of neu rosc ien ce moves forw a rd. And reas A ndreo u oﬀered his t oric al c ontext from c o m pu ter vis ion , notin g t h a t benchma rk s init ia lly help ed st an d a rdiz e eva luat ion. How ever, in no v a tion w as ev ent u a lly co n s t r a ined s in ce res earc hers optim ized s pec iﬁcally for benc hma rk da taset s r a ther t han s olv in g real - wo rld prob lems . Th e di s c us s ion hig hlig ht ed 41 th e prec a rious balan ce b etween a c knowledg ing the v alu e of s tan dardiz ed da t a s ets while a v oidin g res tric t ive ben c h m a rk s . Sev era l pa rt icipa nts emp h asi ze d t he imp ortanc e of simultaneo us, mult imodal r ecord ing s wi th diﬀerent se nsor ty p es (e .g . EEG , EOG, EMG, ECoG) c ap tur ing th e sam e even ts. This w ould a llow for c orrela tion acr o ss m ult iple mod a lities and facilit a te a lgorit hm develo pment w it hout c onst ra ini n g in nov ation . An dr e as A ndreo u not ed t hat res e a rc her s s uch as Pa ri s a Rashid i a re co llect ing data from ICUs p oten tially wit h even t - b a s ed cameras t h a t c o uld be of g rea t v a lu e to t he n eurom orph ic commu nit y. It is crit ica l to look ou t s ide t he n eurom orph ic world to the br oade r co mmu ni ty for dat asets. Syd ney Cas h note d th at s ubsta nt ial in trac r a n ial rec ord ing d a ta sets a lready ex ist th rough BR AIN d at a repos it ories lik e DA BI a nd DANDI . A ls o , th ese d a tas ets h a v e been p reviou sly levera g ed in algorit hm dev elopment co mpet it ion s like Ka ggle t o s olve b iomed ica l p roblems s uch a s s eizu re predic t ion. Mul tipl e par ti cipan ts s ugge sted th at fut ure a lg orit hm i c or da ta c hallen ges usin g t h es e mu lti m oda l data s ets c o u ld featur e neuro morphic a lgorit hms or ha rd w a re and t he sp e c iﬁc a dva n t age ov er con ven tiona l ap pro a ches . Clin ical N eed: Sensor Te chnolo gy, Device Speciﬁc ations , and Data C ollect ion Pa rtic ipan ts d is cussed th e need f or bet ter w ea rable s ens ors t hat mov e bey on d th e lim ita t ions of t he clinic and c ollec t data while s ubjec ts are m obile a n d engag ed in n a tural b ehavio rs. Co rne lia F erm uller hig hlig ht ed ev ent - b ased vis ion s ensors in th e neuromorp hic comm unity (e.g ., https://iniv at ion .c om/) as a suc ce ssf ul tec h nolog y with pract ical a p plicat ions in low - pow er eye t rac king a nd act i on reco gn ition . Dan a Sc h loesser mentioned t he BRAIN BB QS (Br ain Be ha vior Quanti ﬁcat io n a nd S ynchroni za tion) program, wh ich a ims to ca pt ure syn chroniz ed n eura l a n d beha v ior a l da ta in nat ura list ic en viron men ts ac ross b oth c linic a l and n on - clin ica l p opulat ions . It was n oted tha t com p a nies like T u cker D avis T ec hno logies pro vide mu ltim odal rec ordin g s yst e m s in clu ding elec trop hysio logy , ﬁb er phot om etry, au ditory, a nd b eh a v ior on prec linic a l a nimals . Th is d i s cus sio n highlig ht ed imp ortan t dev i c e sp ec i ﬁ cat ions inc l u d ing bioc ompat ibilit y, s tret cha bilit y, a dh e s iven ess, power c on sumption , lat ency requirem ents , a nd t he ability to eit her d is solv e or rem ain s table in th e bod y (e. g. , h e a rt v ers u s s t o mac h) depen ding on th e applic ation. Willia m Sta c ey prov ided spec iﬁc techn ical requ iremen ts for neu ra l record ing s in ambula tory and inp a tient acu te setti n g t h a t would beneﬁ t neurolo gy, p syc hi a t ry, a n d neurom odu lation ap plica tion s . These c apa bilit ies includ e cust o m izab le elect rode p la cem ent, s timul ati on p a r ame ters , an d s amp ling rates (0. 5 - 4000 Hz), a s w ell a s da ta c ollec tion on c on tinuou s recordi ng s . Sev era l c linic ia n s em phasiz ed t he need for t ech nolog ies th at moni tor m ult iple b ioma rkers . T his i nc lude s c hem ical s ig n a ls, s u c h as potas s iu m , oxygen, s od ium, a n d v a rious neu rotransm itters (up to 27). Ro ber t St ev ens s pec i ﬁ cally ident iﬁ ed th e immun e sys tem as a c ritic a l but u nder - su rveilled biolog ica l s yst em th at lacks continuou s , re al - time mo ni tori ng . Ti m De ni so n not ed t h a t c ompa nies s uch as SetP oint M edical and Ga lvan i ha ve on go in g c linica l tria ls in t he immunomod ul a tion s pa ce , cau ti oni ng ag ains t reinven ti ng the wh eel. Pa rtic ip ants d i s cussed th e need f or mu ltim odal rec ord ing devic es t h a t ca n s i m ult aneous ly meas ure elec t ric al ac t ivit y alon gside c h emica l s ignals t o pro vide a m ore c o m plet e pic tu re of neu ra l function. Regul atory C on side r a tions a nd Barrier s Pa rtic ipan ts ca lled f or bet ter ed uc a tion on reg ulat ory p roc esses for a da p t ive s yst e m s . T im D enis on no ted th e exis ten ce of regu lat ory gu ida nce doc umen ts for p hysio logic c ont rol s yst e m s (7 2) , docu men t # 60601 - 42 1- 10 (73) , and AI - based med ical d evic es (126) th a t o u tlin e i mpor ta nt r egul ator y c o n si der ations f or dev elopers . Pa rtic ip ants iden tiﬁed ma nufact urin g c halleng es s pec i ﬁ c to neu romorphic dev ices, inc ludin g tes t ing m eth od s fo r novel c om pon ents like mem ris tor a rray s an d ensu ring relia bility in t he h a rsh biolog ic a l environ men t w here tempera t ure ﬂuc tuation s a nd t he foreig n bod y res ponse ca n a ﬀe c t perform a nc e. P am Abs hi r e ra is ed c onc erns abou t reg ulat ing ad apt ive s y s t e m s, not ing the diﬃc u l ty in qu antif yin g t he ben eﬁt s of a d a p t a t ion and esta blis hing a pp ropriate reg ula t ory fram ewo rk s for sy stems that lear n a nd c hang e over t im e. Sh an t anu Ch akr ab ar ty d i s cussed th e imp ortan ce o f fa il - sa fe m echa nisms in ada pt ive s yst em s , s imila r t o auton omous vehic les w ith diﬀerent levels of au ton omy, to ensure devic es can revert to bas eline fu nc tion ing if prob lems a ris e . Other parti cipan ts sugge ste d th at neu ro mo rphi c syste ms are r esilien t t o ha c king, wh en com p a red to t oday ’s elec t ronics , owing t o t he s t ochas tic n a ture o f th e bra in. T his led t o a disc u s s ion o n the s tabilit y plas tic it y dilemm a an d how w ould th e F D A reg ula t e an un stable syste m. Tim Den is on no ted th a t when N euroP ace RNS was ﬁr s t la unche d, si milar c oncerns a ro s e, and dos e lim its w ere emp loyed as g uard rails to lim it ris k t o the user shou ld t he d evi c e go a wry . Inf rastructure Needs M ult iple pa rt ic ip a nts emphas ized th e need f or bet ter in fra s truct ure t o sup port neuromorp hic eng ineerin g, inc lu ding fabric at ion ac c e s s, d esig n tools , prot o c ol s harin g, a n d s t andardiz atio n. T here wa s dis c u s s ion ab out limit ed acc ess t o fa b ricat ion fac ilit ies a n d a ca ll to esta blis h p rograms s imila r t o t he his t orica l MOS IS init ia tiv e tha t p rovided fre e chip manufa c t uring for academi c researchers (77) . Andr e as A ndreo u s u g g e s ted expa nding bey ond silic on c h ip s t o inc orpo rate new mat eria ls and st ruct ural c omp lexity (e.g ., ben d a ble bio - elec t ronics ), with a s u pport ing inf r a s truct ure eq uivalen t t o MO SIS for t hes e novel a pproa c hes. Sunny B ain s reitera t ed Jen nifer B lain Ch ri ste n ’s sug g est ion for protocol s ha rin g to elimin a te t he n eed to sta rt from sc rat ch, es p e c ia lly w hen d evelop ing nov el biohy brid neuro morp hic mat eria ls . Ulk uhan Gule r not ed t he im port anc e of c hip de s ig n t o ensure U S c ompetit ivenes s in t he g lob al ma rket . C ornelia F erm uller ident iﬁ ed t he la ck of d esig n tools for th e neurom orph ic commu nit y a s a c ritic a l gap . Dev elopm ent wo uld acc elera te if t ools for d es ign ing asy nchro nous c ircuit s and t ra n s ferrin g s piking neu ra l net works b etw een har dware p la tfor m s w ere a vaila ble. Grac e Hwa n g poin ted out th e s ign iﬁcant gap between t he des ig n t ool eco sys tems a va ilable for tr ad ition a l digit a l signa l proc ess ing vers us n eurom orphic c ompu ting, em p h asiz ing that s ubs tantia l w ork rema ins t o make neuro morp hic approache s m o re acc essib le a nd sca la ble. Appe nd i x 2: P ane lists ’ P os iti on Sta temen ts To pi c 1 - Epil epsy & Cortic al Disorders (N e uromo dulatio n) Vi kt or Ji r sa: In the past twent y y e a rs, w e h a ve ma de s igniﬁca nt prog res s in c reat ing dig ita l m odels of an ind ividu a l’s brain , s o called virt ua l bra in tw ins. B y com binin g b r a in im agin g da ta w ith mat hema tica l mo dels , we c an pr edic t ou tcom e s mo re acc urat ely t han u s ing eac h me th od sep ar at e l y. Our app ro a ch h as help ed us un ders tan d n or m al b ra in s ta tes, th eir opera t ion and c ondit ions like h ea lthy a ging , d e m entia a nd ep ileps y. U s in g a c ombinat ion of com putat iona l m odelin g an d dy n am ical s ys tems a n a ly s is we provi de a mec han istic desc ript ion of th e format ion o f res tin g s tat e m anifo ld via t he n etw ork c onn ectiv ity . We demo nstr ate t ha t the sy mme try b re a k ing b y the co nne ctiv ity cre ate s a cha r ac te r istic ﬂo w o n the 43 ma n ifold, w hic h produc es the ma jor da ta fea tu re s acro ss sc a les an d imag ing m oda lit ies . T hes e inc lu de spontaneou s high ampli tude c o - a c tiva tions , n eurona l casc ades , s pectra l c ort ical g r a d ient s, m ult i s tab ility , a nd ch aract eris tic func tiona l con nect ivit y dy nam ic s. Wh en agg rega ted a c ross c ortica l hiera rc hies , t hes e mat c h the proﬁ les from em pirica l da ta a nd expla in fea tures of the b rain ’s m i c ros tate orga n izat ion. E xamp les f or clin ic al t r a n slat ion a re ta ken from dru g r esist ant epileps y and m en t a l dis ord ers . T he d ig ital bra in t w in a u gments th e v a lue o f emp iric al d a ta by c om plet in g m issing data , a llowin g clin ical h yp oth e s is tes t ing a nd optimiz ing tr eatm en t s trat egies for t he in div idual p at ient . Virt u a l B rain T wins a re p art of the E urop ean infras t ru c t ure ca lled E BRA INS, w hic h s u pports res earch ers w orld wid e in d igit a l neuro scien ce. Kendal l L ee: M od ern neu r a l eng ineerin g t ec hno logies now a llow for c on trol of neurot ra nsm itt er levels by DBS . A rt i ﬁ cial I nt elligen ce w ill allo w for a ut omat ed n eur a l c ont rol alg orit h m s in the fut ure. F urt her c omp lex elect ro - ana lytic al met hodo logies suc h as MC SWV, will a ls o be av a ilab le for n europs y chia tric th era pies . C hris Roz ell: C urr ently , c ommerc ia l neu rotechn olog ies a re d esign ed w ith v ery minim a l proc ess ing c apa bilit y, g eared almos t entirely tow a rd very speciﬁc not ions of w h a t onboard proc essin g is nec e s sa ry for clos ed - lo op s timula t ion to tre a t speciﬁc dis orders suc h as epileps y an d PD . There are p res um ably ma ny c onst r a int s driv ing th is n a rrow prod uct des ign, in c luding p ower cons umptio n , form fac to r limit a tio ns, a nd t he t ime n ecessa ry f or new eng ineerin g d esigns that meet reg ulatory s tan d a rd s. F ut ure innovation s w ould ben eﬁt from a n exp a n ded s et of cap abilities in a pa c k a ge that improv es the user experien ce in wa y s that w ill inc rea s e acc ess ibilit y (e. g. , sma ller form fa c tors, long er b at t er y life, et c .). Sr idevi S a rma: Durin g ro ut ine ph y s ica l exa mina tions, va rious ph y s iolog ica l pa rameters s uch as age, weig ht , height , a nd b lood p ressu re a re meas ured to ga uge ov erall h ea lth statu s . Am ong th e s e, Body Ma ss In dex s erves as a w idely ut ilized tool t o s trat ify indiv idu als in to d iﬀerent weig h t catego ries, a iding in t he iden tiﬁ ca tion of pot entia l healt h ris ks . H owev er, a pertin ent inquiry a rises regard ing th e ident iﬁca t ion of ind ica t or s for b ra in hea lt h and wheth er a quan tita tiv e a ssess m en t of b rain hea lt h ca n b e derived from ima g ing dat a . I n t his pro po s ed st ud y , w e dev elop t he “B ra in E nt rop y I nd ex (BE I ) , " a no vel qu a nt it a t ive meas ure o f bra in h ea lth, ra ngin g b etween 0 to 1 , w here a h igh BEI (e.g ., > .7 5) w ould indic a te a h ea lthy bra in a n d a low BEI wou ld in dica te a ris k fo r a n eurolog ic a l dis order. Th e B E I is not deriv ed b y t ra ining mac hin e lea rning m odels on larg e dat a s ets, whic h is th e typica l a pp ro a ch us ed today t o diag no s e neu rologic a l dis ord ers from im a g ing data . In c ont r a st, the BE I is deriv ed from a un ifyin g the ory th at c omb ines princip les from t herm odynam ics, dynam ica l sys tems a nd mat rix th eory. T his un ifying t heory (i) iden tiﬁ es propert ies that a hea lthy hu man br a in a t rest must h a ve, a nd t h en (ii) der iv es m etric s t hat c ha r a cteriz e w het her t hes e prop erties exis t in a p a t ient u s ing ima ging d ata . The id ea is t o ca pture res t ing - st ate E EG f or 10 - 1 5 m inu tes from pat ient s vis iting t heir p rima ry care p hysic ians f or a nn u a l exa ms an d c omp u t e th e BEI in m inut e s a ft er th e E E G s ca n is ca pt ured . Furt herm ore, we illu st r a t e how dev i a tion s in BE I ca n eﬀectively clas s ify va rious neurolo gica l dis orders , in c lud ing epileps y , A lz heimer’ s dis ease, and fron tot emp ora l dementia . T h e BE I a ls o f a c ilit ates th e ass ess ment of t rea t m ent eﬃc a cy , w ith succ ess ful tre atmen ts re sto ri ng the B EI t o a hea lt hy r a ng e. Fina lly , we s howca s e ho w linea r t ime - vary ing mo dels can be lev erag ed t o s im ulate v irtu a l surg eries f or epileps y ba s ed on neurosu rgeon 's pla n s, wit h s ur g ica l s ucc ess p redic ted by meas urin g the BE I pre - and po s t - virt u a l s urg ery. Willia m Sta c ey: Hig h F requ en c y O scilla t ions (HFOs ) hav e been in ves t igat ed for o ver 20 yea rs a s a pot ent i a l biom a rker of ep ileptic t is su e. H owev er, H FO s a re a ls o present in n ormal b ra in. T here ha s been g reat 44 int eres t in dis tin gu i s hing n ormal from pat holog ica l HFOs . Ea rly s tudies on sma ll data sets sugges t ed that HFO s c ould be c las siﬁed by pea k freq uency , a s en t imen t that persis ts to th i s day. H o wev er, ma ny lat er rigo rous s tu die s ha v e refut ed t his, a nd th ere is no tru e go ld stan da rd for comparison . We c ollec t ed a la rge da tas et w ith millions of H FOs and us ed sev era l stra tegies to ﬁ nd patho logica l fea tu res . We ﬁ nd c ons i s t ent feat ures in man y pat ients , but the la r g e het erogen eity betw een pat ients ma kes c lass i ﬁ ca tio n very diﬃ cu lt. La rger da tas ets a nd bet t er controls (w hich a re extrem ely diﬃ cult to ac quire) a re need ed for bet ter c las siﬁca tio n. Syd ne y Ca sh: Neura l int erfac es, n eurom odu l a tory syst ems a nd br a in - c omput er interfa c es, h ave mad e s ign iﬁcant strid es ov er the p ast deca de , eviden c ed by on g oing clin ical t ria ls and increa sin g clinic a l a pplica t ions . How ever, s ubsta nt ia l cha lleng e s rema in in ha rn e s s in g th eir full p oten tia l fo r clin ic al u s e. O n th e eng ineering s ide, challen ges s uch as energ y eﬃc ienc y, pac kagin g, dat a manag e m ent , an d s ec urit y a re iden tiﬁ ed a nd t h e focus of resea rc h in mult iple g roups . Convers ely, our u nderst anding of h uman br ai n ph ys iolog y and its in terfa c e w ith th ese t e c hn olog ies often lags b ehind our en gin eering cap ab ilities , w hic h tra nsla t e s t o a poor g r a s p of the f undamen tal c onst ra in t s underlyin g n eurom orph ic a p pro a ches . For ins t anc e, in w ell - s tu died con dit ions like epilep s y, we s t ill gr apple w ith crit ica l ques t i on s: Wh ere do s eizu res o rigin at e? Ho w d o they s tart ? Wha t is t he opt i m a l scale fo r int erac t ing w ith a nd ma pping s eiz ure a ctivit y ? What a reas shou ld w e targ et f or mo dulat ion? F ur t herm ore, we rema in u ncerta in ab out t he s hor t - and long - term c ellula r res ponses t o variou s forms of neuro mod ul a t ion. To f ully levera ge t he p oten tial of br ai n - c om puter int erfa c es a nd neuromod ulatory t echnolog ies , contin ued invest ment a n d int erdis c iplina ry dia log ue are es s en t ia l. T his ca nn ot b e a ch ieved withou t in teg r a t ing eng ineering a dvan cements wit h b iolog ical u nders ta nding in a co llabora tiv e feedb ack lo op. N ath an Cr o ne : Re c ent c linica l tria ls of im plan table b ra in - compu ter inte rfac e s ha ve shown tha t both single un it a ctiv ity and elect roc ortic ogra phic s igna ls ca n be d eco ded in to t ext a nd sp eech t o res tore c ommun icat ion in peop le wit h v ary ing deg rees of L ocked I n Sy ndro me. Th ese ac hievemen ts hav e relied hea vily on ad v a nces in ma c hin e lea rning t h a t were origin a lly ins p ired by neu romorphic prin cip les . Ho wev er, cu rrent dec odin g models requ ire va st ly g rea ter c o m pu ting p ower t han is pra ct i c a l for f ully imp la ntab le, wireles s B MI s yst ems. M oreover, t he lim ita t ions of cu rrent ly a vaila b le neu ral in terf ac es h a ve hig hlig ht ed th e need for ﬁ ner a nd mor e com prehens ive sa mplin g of neura l elements , further inﬂ at ing dem a n d s o n data transm issio n a nd c omputat ion . Neurom orph ic princ iples cou ld potent ia lly b e us ed in th e des ig n of n europ rost hetic dev ices t hat meet thes e grow ing need s . Furt herm or e, th es e princ ip les ma y be p art ic ul a rly help ful w hen mov ing bey on d dec odin g the n europh ysiolog ic a l sign atures of bra in hea lt h and di se ase , to r est or ing he althy br a in function t hro ug h e ﬀective ne ur a l stimul atio n. Gia c o m o Va lle : Th e needs for t he n eurom orph ic te c h nolog ies in neu ropros t he t ic s r ela te t o t he exploit a tion of b iologic a lly ins pired p roc ess t o int erfac e, c ommun i c ate a nd int er a c t wit h t he nerv ous s yst em at multiple levels . Neuro morp hic tec h cou ld p oten tially allo w fo r a b e tter i nf or matio n tr ansfer bet ween the a rt i ﬁ cial a nd biolog ical s yst ems. Mo reover, th e hig h lev el of po rta bilit y a n d em beddab ility c ould b oos t the t ra ns a b ility of t hes e tec hnolog ies into clin ic al ap plica tion s (24) . The ch a lleng e s a re rela ted to the exploit at ion of th e s ophist ica ted n eurom orph ic fr a mework g iven th e limit s o f t he c u rrent ava ila ble int erfa c ing techn olog y. Mo reover, th e neu romorp hic t echnolog y is ba sed on the em ula t ion of t he biolog ic a l co unt erparts . Not ab ly, im porta nt n a tura l proc ess es rela ted to t he t a r g et n eura l s truc tu res ha ve not been deﬁ n ed or not exha u s t ively inves t i g ated y et, m ak in g th e va lid a t ion qu ite cha lleng ing . The 45 c ont ribut ion of neurom orph ic t e c h has s hown int eres tin g app licat ions in art iﬁcia l sens ing (vis ion and to uc h) a nd compu ting. Th e ﬂexib ility of th e tec hn ology , it s emb eddab ility , c ould fa vor its a pplica t ion in every da y life s et tin gs (e.g . pros thet ic s). T he beneﬁt to exp loit a c om p u tat ion bas ed on neu ra l emu lat ion wo uld a llow for a bet t er human - mac hin e int erfa cing. Human neura l p roc e s sin g has been s ha p ed by evolut ion over thou sa nds of y e a rs , an d it is ﬂ exible a nd eﬃc ient . I ts e m ulat ion th roug h neu rom orph ic te ch ope n s im p ortan t app lic ation s in eng ineering a nd m edic ine (n eurom odulat ion, neuropro st het ics, etc .). T he d esign o f a fut ure g enerat ion of bio - in s pired neu ropros t het i c devic es a llow ing the a rt iﬁ cial c onv ey a n ce of more c omplex in form at ion and t herapies w ith neurost im ul a t ion c ould dra s tica lly impr ove th e ﬁeld . T he neuro morp hic tec h cou ld pot ent i a lly increa s e , not only the eﬃ c acy b ut a lso the accepta n c e of more bio - ins pired neurop ros th etic dev ic e s . T he use of neuro st imu l a t ion is a l s o critic a l t o bra i n– c omp u t er int erfa c ing, as w ell as bioelec tro nic med icin e, in whic h t he elec trica l s tim u la tion targ ets the c ent ral, or t he au ton o m ic, n ervous sy stems, res pect ively . As t he h ea lthcare s ecto r inc reasin gly adop ts A I a nd ad v a n ced c omputa tion a l method s, n eurom orphic t ec hno logies a re likely to p lay a piv otal role, ma kin g them a c o m pellin g a rea for in ves tment . I nves ting in neuro morp hic tech nolog ies in m edic ine pres ent s s ever a l prom is ing opportu nit ies: neur oprost hetics , diag nost ics, m enta l hea lth applic at ions , we ar ab le devic es , rehabilit a tion tech nolog ies , an d A I in m edic ine. Topi c 2 - Human- m a chi ne in ter fa ce s (P ro sth e ti cs) Xing Chen: Needs : Clin ica l s olution s t hat a re sa fe, long - liv ed and eﬀec tiv e, provid ing a ssis tanc e in da ily tas ks , object recog nit ion, n avig ation , an d ind epend enc e. Cha llen ges: Techn olog ical c ha llen ges in int erfa c ing wit h t he bra in w it h suﬃcien t res olut ion a nd s pec i ﬁ c it y , a llowin g neu romodula t ion w ith out ind uc ing s eizu res and a berran t ac tiv ity, build ing dev ices t hat a re easy to impla n t a cros s c linica l s ites a nd ind ividu a l pat ients . Co n t ribu tion s: Develop ment of b ioco mpat ible mat eria ls, s c ient iﬁ c understa ndin g of neu ral en cod ing a nd d ecoding to im prov e s p eciﬁc it y a nd eﬃcac y o f neuro modul a ti on techniques. Im pact : Prov iding hop e a nd ins pira tio n for b lind a n d visu ally im p a ired p eople, a nd l a yin g the fo und ation for b reakt hro ug h dev ic es. Inv estm en t : Ha rness ing so lut ion s fro m t he la b an d brin ging t hem to p a t ients. Rec ruit m ent of p a tien ts and other s tak eholders t o shap e a nd guid e the ﬁ eld from an ea rly s tag e. Rais ing a waren ess am ongs t the g enera l pu blic of b oth th e prom i s e a nd p itfalls of th is t ec h n olog y. Helen Hu a n g : 1. T he d e s ign princ ip les f or neura l - mac hin e int erfac es (NMI s ) s hou ld be an c hored in es tablis h ed hu man mo tor c ontrol a nd lea rning t heories . 2 . The des ign of an eﬀerent neura l dec oder should be guide d by hum an neur ome chanics . 3. Th e design of an aﬀ erent n eura l enc oder s hou ld a im t o enh anc e sen s orimotor in teg ra tion 4. Neuro morp hic c omput ing a nd a rt iﬁcia l int ellig enc e are po werfu l to ols f or a ddress ing NMI ada p t ation s to in div idual us ers a n d for ev olvin g ov er t ime, w ith low p ower consumption and dis t ri buted proce s sing. Ranu Jung: Neurot ec h n olog y hold s the prom is e of enhan cin g our understa nd ing a n d oﬀering innov at ive trea t ment o f bot h p hysica l a nd menta l hea lt h c on dit ions . How ever, transit ion ing these a d v a ncem ents from th e la bora t ory to c linic a l a nd hom e s ett ings to t r a nsf orm h ea lt hcare rema in s a s ig ni ﬁ ca nt cha lleng e. T his talk d isc usses t he use of a c onverg enc e a pp roac h for p ioneerin g a dva nc e m ents in neu rotec h nolog y to revolut ion ize hea lt hca re. T he jo urney from c onc ept to develop men t an d dep loyment of a Neura l - En ab le d P ro stheti c Han d Syste m in an FDA appr ove d ﬁrst - in - h u ma n t ria l will b e us ed t o illus t rat e t he exten s ive c olla bora t ion ac ro s s g overn men t, ind ust ry, a ca de m ia , an d the v olunt eer pa rt ic ip a nts nec e s sa ry to mak e m aj o r ad v an ce s in hu m an - mac hine in terf a ces. 46 Luke O s born: Neuromorp hic enc od in g is nec e s sa ry for r educ ing c om plex, hig h d i mens iona l s ensor s igna ls (e. g. , t ac t ile, v is ion, a ud itory ) int o m e a nin gfu l feat ures for driv ing us efu l s en s ory feedb ack stimul atio n par adig ms i n se nso ry - ena bled prost hes e s , br a in - c ompu ter i nt er fa ces , and hum an - m a chine int erfa c es . A key c halleng e is und ersta nding w h at t hes e meanin gful fea tu res a re and how to eﬀec tiv ely communic at e them bac k t o a h u man o r int elligen t s yst em . N iti sh Th ak or : Prost hes e s , wheth er up per or low er lim b for a m put ees , h a ve con c en t ra ted on a c hiev ing mec hat ron ic mot ion a nd co nt rol. C ontrol is a chieved th roug h musc le, nerve or dire c t c ort i c a l int erface, a nd us ing eith er nonin vas ive o r inva s ive ap proac hes . How ever, prost hes i s in g enera l a n d neu ropros t hes i s in p art ic ular a re only now a t ten ding t o inc orpo rating s ensory enc o d ing a nd feed bac k. I will d isc uss th e need t o develop diﬀerent sen s ory m od a lit ies and mim i c king th e senso ry mo d a lit ies in biom imet ic man ner. Th is w ould inv olve s ensors t h a t m imic rec eptors in t he s kin and enc od e th em a s neu ral s p iking act ivit y. T his a pp ro a c h has ma ny a dvan tag e s in t erm s o f enc o d ing , but a lso po s es c ha lleng e s in t erms of nov el alg orit hms a nd n ovel ap proac hes to neural in ter fa ce and fee dba ck. T herefore, m y g o a l is t o pres ent s om e ideas an d p rovoke d iscuss ion on s ensors, s e ns ory enc oding , n eur a l int erfa c e and ac hiev ing s en s ory prost hes is solu tio ns for bu ilding fu tu re c lo s ed - loop brain - m achine int erfa c e tec hnolog ies. To pi c 3 - M ateri al s fo r Ne urom orp h ic s (De vi ces) Gert Cauwenber ghs: Our wo rk oﬀers a n eﬃcien t neuro morp hic h a rdwa re - sof tw ar e co - desi gn fra me work rec tify ing t he energ y - eﬃc iency imbala nc e betw een the tra in ing an d th e inferen ce pha s e s in m a c hin e lea rning (ML ) art iﬁc ial in tellig enc e (A I) sy stems. It levera ges t he great t olera nce of tas k perf orma nce a ch ievable d urin g in ferenc e t o mem ory leakag e du ring train ing, t o im plem ent neuromorp hic pa r a digm s for energ y - eﬃc ien t m emo ry th roughou t the lea rnin g pha se. Hence it s ubs tantia lly red uc es t he overall energy footprint of ML - b as ed A I, as it is dom in a t ed by t he energ y costs of t rain ing w hic h req uires s ign iﬁcant ly more para meter up dates and me m ory w rites c om p a red to inferen c e. Op erat ing a t m ult iple time - sca les to im plem ent s yna ptic m etap last i c it y w here fas t epis odic - me mor y work in conjunct io n with one - sh ot con tin u a l on line lea rnin g a rc hit e c t ures, th e s e lear ning - in - m emory s yst ems s t rike a b a la nce bet ween th e a d a ptatio n ra te o f th e s y n aptic elem ents and t heir p aram eter reten tion ca pab ility to realiz e an o pt im al tr ade d - oﬀ w ith respec t to t h e energ y d issip at ed p er memory writ e. Shant a nu Chakr abar t ty: Hybrid neur al in terf ac es, c om pris in g biolo gic a l a nd synt hetic neu rons , prov ide a pa t h wa y fo r des ign ing t he nex t g enera t ion of n euroen gin eering s y s t e m s. E xis ting wet wa re - h ard war e imp lemen tat ions hav e non - over lap ping a rc hit e c t ur a l a nd operat iona l bou ndaries , wi th e ach su bsy ste m opt imiz ed in depen den tly u s ing plat form - spec i ﬁ c s igna l repres ent at ions. In cont r a s t, a tru e hyb rid neu r a l int erfa c e sh ould f a c ilitat e c ross - map ping between biolog ic a l and sy nt het i c neuron s, en a bling the s yst em to s eam less ly optimiz e a nd deleg at e c o m putat iona l ca p a bilit ies bet ween th e two doma ins. T h i s ty pe of int erfa c e wou ld no t on ly b etter c ont rol th e capa bilit ies of biolog ic a l sys tems at t he neura l c ircu it lev el but also aug m ent a nd repair ne ural functio ns a t th e s y stem level. C u rrent ly, s evera l c hallen ges exis t in a ch ievin g s uch a hybrid interf ac e. Besid e s the d iﬃcult ies in des ign ing bioma teria ls, elect rodes , and opt rodes t h a t can s eamles s ly st imula te an d rec ord fro m b iologic a l neu rons ov er long period s , an d t he c ha lleng e of lon g - term pow ering of su ch s yst ems, a c ritica l c o m pu tat ion a l c hallen ge lies in t he m is ma tc h a t t he in terfac e bet w een the neu robiolog ic a l a nd n euromorph ic d o m a in s . F irst , t here is a ba n d w idt h an d int erfa c e challen ge d ue to t he lack of sp eciﬁc ity in s t imulat ion a nd record ing , wit h eﬃ cacy that drift s ov er 47 tim e. S ec ond, t here is a repres en tat ion cha llen ge a ri s in g fro m t he m i s matc h betw een t he c omp lex s patio tem poral repres en ta t ion s in cou pled neuron a l systems a n d th e simpler s p at i ot em pora l dy n am ic s in c urrent neu romo rphic s yst ems. T he s e c o m pu t a t ional c hallen g es prev ent eﬀec t ive cros s - m a pping bet ween biolog ic a l and sy nth etic n euron s. We believ e t hat em ploy ing a n en ergy - b ased or Ham ilto nia n - like j oint m odeling a nd synt hesis of b iologic a l and neuro morp hic s yst ems m i g ht a ddress t he s e c ha lleng e s . S uch a hyb rid neu ra l int erface c ou ld t hen c ontin uously evolv e a nd d elega t e func tions ac ross both neuro biolo g y and neuro morphi c s y s tems. Ulkuhan Gule r: Biom edic a l s ens ors g enera t e c ont inu ous strea ms of dat a, often requ iring real - time proc es s in g t o ens ure t imely, accu rat e a n alys is an d pro mpt ac tion bas ed on t he int erpret a tion of t his info rma t ion. B y im plem en t ing neu romo rphic t echniq ues direc tly on - chi p , da ta can be pro c essed eﬃc ien tly wit h low po w er cons umptio n, en a blin g w ireles s t r a nsm ission of the p rocess ed s i g na ls. T his redu ces th e need for oﬀ - c hip data han dling , resultin g in c o m pac t, energ y - e ﬃ c ient, a n d hig hly respons ive biom edic a l syst ems, wh i c h a re c rucia l for w ea r a ble or im pla n table h ea lth monitorin g d evic es . Duygu Kuzu m: T he n ext lea p in imp l a nta ble neu ra l int erfa c e s requires t ec hnolo gical a dva nc es in mat eria ls , d evic e s , a nd c omp u t ing para digms. Mult imoda l ap proac hes int eg ratin g optic a l a nd elec trica l s ens ing m odalit ies c an o verc ome sp atiot emp or a l res olut ion lim it s of neura l s ens ing as w ell as op en u p new avenue s for non - inva siv e neu r a l rec ording . I nt eg r a tion of s ensin g, c omputa tion a nd m e m ory o n a s ing le a rray ca n en ab le rea l - t i me p roc essin g of neura l s igna l s for com pact , low - power a nd high - throughpu t neuromor phic br ain - mach ine in terfac es . H ere, I will pres ent t his vis ion, it s cha lleng es, and dis c u s s rec ent a d v a n c es in t he a reas of t ra nsparen t n eura l int erfac es f or mu ltim odal rec ord ing s, neurom orphic a pproaches for on - chip ne ural proce ss ing a nd comput at io na l co - desig n a t the s y s t em lev el for m inim a lly inv as ive neu ra l int erfa ces. Ton y Le wis: Int elligen t d evic es, from w e a ra ble h ea lth m onit ors to im plants , in c rea s i ng ly su rroun d us. As th es e tec h n olog ies ev olve, more eﬃ c ient AI s y s t ems a re cruc ial. Tradit ion al d eep learn ing , wh ile pow erful, s tru ggles wit h edge d eploy men t d ue t o hi g h c ompu t a t ional a nd energy d eman ds. W e intr oduc e F low M ach ine s , a no vel pa ra digm for bio medica l w ea r a bles levera g ing s tat e spac e mo de ls a nd neurom orphic c omp u t ing . U nlike c onv ent ional fra m e - b ased deep lea rn ing, F low Ma c hin e s creat e int erna l repres ent a tions of th e extern a l world inc rem entally , ut ilizin g p rocess ing h is tory . T his a p pro a c h con tras ts wit h t yp i c a l edg e - d eploy ed deep lea rning net w orks th at rec om pute fro m sc ratc h, ignorin g p ast c omp u t ation s . F low Mach ines a cc umula te in terna l repres ent a tions o ver t ime, enabling more c om pact a nd eﬃc ient AI mod els suita ble for res ourc e - cons trained wea rab les. By t r a nsc en di ng t ra dit ion a l pa ra digms, w e a im t o enab le s ophist icat ed an alys is a nd d ec i s ion - ma king in a n ew cla ss of int ellig ent biom edic a l dev ices. Ou r resea rc h co - opt imiz e s F low Ma ch ine s u sing s ta te spac e models a s their co re a rch itect ure. Th i s met hod a dd resses edge c omput ing limita tio ns by levera g ing int e rna l s tat e and inc rem ental lea rn ing , ac hiev ing s tat e - of - the - a rt p erforman ce ac ross variou s edg e cha lleng es, fro m heal th mo nito ring to a daptiv e th era peut i c s . F low M ac hines p ro m ise t o rev olut ioniz e wea rable hea lthca re tec h nolog y, pav ing th e w ay for p ers ona liz ed, c ont ext - awa re s olut ions an d im proved pa t ient outcomes . To pi c 4 - Medi cal Imaging, Wearabl es and A nal ysis Br ad Aimo n e : I nsp ired by experien ces in mod eling a dult hip po c am p a l neu rog enes i s , in t his t alk I w ill des c ribe ho w t he cha llen ges of m odelin g the d ivers e t ime an d s p a t i a l s cales of t he bra in’s plas tic it y and dy namics led m e to look to wa rds t he emerg ing ﬁ eld of neu romo r phic c o m pu ting as a pot entia l so lutio n. 48 T oday’s neu rom orphic syst ems a re b egin ning t o rea ch bra in - like s cales in t erms o f nu mbers of neuro n s , a nd in terac t in g w ith these s yst e m s pres ents s imila r c hallen ges . I w ill des c ribe a ra nge of u nexp ect ed res ult s t hat w e ha ve lea rn ed from brain - like c omp uting syst ems operat ing a t this sca le. F ina lly , I will rev isit my orig ina l g o a l of h ow u nders ta nding n eura l c o m putat ions at b iolog ica lly rea list i c s ca les on neuro morphic syste ms c a n, in turn , help us unders tand how a nd why t he br ain op e ra tes the w a y that it d o es. Sihong W a ng : Prec isio n m edic ine, the fut ure of hea lt hca re, p romises pers ona lized dia gn o s is and trea t ment b y c onsid ering each ind ividu a l' s uniq ue g enet ic ma keu p , ag e, medic a l hist ory, and env ironm ent . Ac hievin g t his v isio n hin ges on tw o c rit ica l ca pab ilities : (1 ) c ont inuo us , high - res olu t ion, mul ti - m o da l b iosens ing a t t he in terf ace of skin a nd tis sue, and ( 2) rea l - time , hi gh - th ro ug hput a n a lysis of la rge, c o m plex da tas ets to enable t imely int ervent ions . Ho w ever, t he s e a dvances com e wit h s igniﬁca nt c ha lleng e s . F irs t, bio - elec tron ic in terfa ces m u s t esta blish an d ma in t a in s ta ble, long - te rm co nt act with biolog ic a l tis sues to ensu re eﬀect ive s ens ing. S eco n d , ac hiev ing e ﬃc ient d a ta proc e s s ing an d power ma nagem ent requires t he s e a m le s s int egrat ion of neurom orp hic c omp u ting di re ctl y onto the b io - elec tron ic s en sing p la tfo r m, ena bling edg e c o m put ing nea r t he sens or. In this t a lk, I will p res ent our w ork on d evelop ing orga n ic sem ic on d ucto r - a nd tran sis tor - bas ed bios ensors t hat inc orporat e biom im etic pr op er ti es such as s kin - like s t retcha b ility , bioa dhesiv enes s, an d im mune c ompatib ility . I will a ls o int rodu ce o ur prog res s in c re a t ing int rin s ic a lly st retcha b le neurom orp hic devices an d a rra ys ca p a ble of perfo rmin g n ear - s ens or edg e computin g, a dva n c ing th e ﬁ eld of bio - elect r onic in terfa c e s t oward rea l - ti me , pers ona liz ed h ealthca re. T h e success fu l deploy men t o f th e s e tec h nolog ies c ould revolut ion ize pa t ient c are, enab ling ea rlier det ect ion of dis eas es a nd m ore pr ecis e int erven tio n s . R ea lizing t his futu re will requ ire sig niﬁ c a nt i nve stme nt in cro ss - disc iplina ry res e a rc h , inf ra struc ture, a n d long - ter m sup por t for s cala ble b io - elec t ronic ma nufac t uring . Dis c laimers T he opin ions expres s ed in th i s ar tic le are t he a ut ho r s own an d do no t reﬂ e c t t he view of th e Nat iona l Instit ute s of He al th , the De par tmen t of He al th a nd Hum an Se rvi ce s, o r the Unite d S tate s Gove rnme n t. Dr. M cKinn ey’s rema rks in the N PBH w orks hop repres ent his p erso n a l pers p ectives based on his reg ulat ory experien ce a nd do not cons tit ut e a ny sta tem en t of oﬃ cial pos it ion or polic y of the U S F ood and Drug Admin i stra tion or th e US gov er nmen t. Hi s co ntri but ion t o the pres en t ma nuscript w as lim ited to en s uring a n acc ura te su m ma ry o f his rem arks in th e work sh op rega rding a p plica ble reg ula tory pro ces s e s . An y opin ions , ﬁndings , conclus ions , or recommen dat ion s expres s ed herein are those of the author (s) and do not nec essarily reﬂec t the v iews of the U.S. Nation a l Scienc e F oundat ion. Ackno w ledgement The NPBH Work s ho p was supported by N SF Confe rence Aw a rd ( 2433739 ) through c o - fundin g fro m the follow ing N SF prog rams: E lectron ic s, Phot on ics & M ag netic D evi c es (E P M D), D isab ility a n d Reh abilita t ion Eng ine eri ng (DA RE ) , C o mm unica tion s, Circ uits & S ens ing - S ys te ms ( CC SS) , a nd F ou nd a t io ns o f E m e rg ing T ech nolog ies ( F ET ) . N IH NIN DS and N IB IB co - spo n sor e d th is wo rk sho p vi a trav el su pp or t. The au tho rs ac k no wle dge th ou ghtful co mmen ts fr om Hol ly Moo re , V i ck y Whitte mo r e, Ralph Etie nne - Cum mi ng s, D avi d W yatt, Andre as Andreou, Luke O s born, Tob i Delbruck, A ndrea B e c kel - M itch ener, John N gai, Stev en Zeh nd er, Chou Hu ng, P aul Lan e , So o - Sian g Li m, Jeﬀrey Kop s ick, M i c hele F erra nt e, Y vonne 49 Be nnett, Da n a Sc hloes s er and referen ce h elp fro m Meg an Fr an ko w ski . The a uthors further thank the mem bers of t he NIH B R A I N NeuroA I Wo rking Group and t he F edera l NeuroAI F und ers Workin g G roup . G enerative A I D is closure T he A nthrop ic C laud e S on n et 3 .7 la rg e - la ngu age m odel wa s us ed t o gen er a te the in itia l d ra ft of Appendix 1 based on the correct ed a ud io tran scrip ts of p res ent ations an d disc uss i on s at the worksho p . Th e draf t was subs equent ly rev iewed , ed ited , an d veriﬁ ed fo r a ccur ac y by th e auth or s by cr os s - referenc in g c la ims , pers pect ives, a nd oth er st ate me nt s a gai n st th e tr a nsc rip ts, s lid es , and ot her sour ce s. In a dditio n, the a ut hor s inv ited the w o rk sho p spe ake r s to review th e ir respec t ive sect io ns i n A ppendix 1 and to provide fe edbac k to ve ri fy ac cur acy , r e me dy any omi ssi on s, an d improv e the overa ll qu a lity of t he pres ent atio n summa ries . Th e au thor s th ank the work sh op par ti cipan ts fo r contrib uting th eir feedb a ck and helping ens ure that their v iews a re a cc urately reﬂ ec ted . Ref er ences 1. M ea d CA , M ahowa ld MA . A s ilic on m odel o f ea rly v isua l proc ess ing . Neura l Net w. 1 9 88 Ja n ;1(1 ):91 – 7. 2. Mahowald M, Dougl a s R. A sili c on neuro n. Na ture. 1991 Dec 1;3 54(6354):515 – 8. doi:10.1038/354515a0 3. M ea d C . Neu romo rphic E n gin eering : I n M em ory o f M is h a Mahowa ld. Neura l C omp ut . 2 023 Feb 17;35(3):343 – 83. doi:10.1162/ nec o_ a_01553 4. Mea d CA. Neuromorphic el ec tronic s ystem s. 1990. 78(10):1629 – 36. 5. Mea d C. Analog V L S I and neur al s y s te ms. Bo ston , MA: A ddison - W e s ley L on gm a n Pu b lish ing C o. , I nc.; 1989. 6. Boa hen K. Dend rocentric lea rn ing for s ynth etic int ellig enc e. N a t ure. 2 022 Dec ;612(79 38): 43 – 50. 7. Has ler J, Ma rr B. Fin din g a roa dma p to ac hieve la rg e neu romo rphic hard ware s yst e m s . F ron t Neur os ci. 2013 Sep;7:118. 8. Sc hu m a n CD, Kulkarni SR, P a rs a M , Mitchel l JP, Date P , Kay B. Oppor tunitie s for neuromor phic computing algorithms and applic at io ns. Nat Com put Sc i. 2022 Jan 1;2(1):10 – 9. doi :10.1038/ s 43588 - 021 - 00184 -y 9. Kudithipudi D, Sc hu man C, Viney ard CM, Pan dit T, M er kel C, Kuben dran R, et al. Ne uro morphic comp u ting at scale. Nature. 2025 Jan;637(8047):801 – 12. 10. Peders en J E , Ab reu S , Jobst M , L enz G, F r a V, Bauer F C, et a l. Neu romo rphic int erm e dia t e repres ent a tion : A un iﬁed instru ction set for inter opera ble bra in - inspire d c o mputing. Nat C om mun. 2024 Sep;15(1):8122. 11. C ass idy AS , G eorgio u J, A n dreou AG . Des ign of silic on bra ins in t he na no - CM OS er a : Spiking neur ons, lea rning s yna p s es a nd neural a rc hit e c t ure opt im izat ion. N eura l Netw . 2 013;45: 4 – 26 . doi:https ://do i.org/10.1016/j.ne unet. 2013.05. 011 12. Kim Y, C hort o s A, Xu W, L iu Y, Oh JY, So n D, et a l. A bioins p ired ﬂex ible orga n ic art iﬁcia l a ﬀerent nerve. Science. 2018 Jun;360(6392) :998 – 1003. 13. Wang W, Ji a ng Y, Z hong D, Zh a ng Z , Chou dhury S, Lai JC , et al. Ne uromorphi c sensor imotor l oop emb odied by m on olith ically int egrat ed, low - vol tage , sof t e - s kin. S c ie nc e. 2023 May;380(6646) :735 – 42. 14. Don a ti E , I n d iveri G. Neurom orph ic bioelec tro nic med icin e for n ervous s yst em int er fa c e s : from neu ral c om putat iona l prim it ives to medica l a pp licat ions. Prog B iomed E n g . 20 23 F eb 17;5 (1): 0130 02. doi:10.1088/2516 - 1091/ ac b51 c 15. A boumerh i K, Güem es A, L iu H, T en ore F, E t ienne - C u mmin gs R. Neu romorp hic applic a t ions in medicine. J Neur a l Eng . 20 23 A u g;20(4):041004. doi:10.1088/1741 - 2552/ a ce c a3 50 16. Indiveri G. Ne uromorphic is d ead. L ong live neuro m or ph ic. Neur on. 2025 Oct 15;11 3(20):3311 – 4. doi:10.1016/j .neuron.2025.09.020 17. Ru bino A, C a rt iglia M, Payva nd M, Indiv eri G. Neurom orph ic a nalo g c irc uits for robu s t on - chi p al w ays - on learning in spiki ng neur a l ne t work s. In: 2023 I EEE 5 th I n ternation al Confe rence on A rtiﬁc ia l Int ellig enc e C ircu its a nd S ystems (A IC A S ). IEEE ; 2023 . p . 1– 5. 18. Tian F, Yang J, Zhao S , Sawan M. Ne uroCAR E : A g ene ric neuro m or phic edge c ompu ting frame w or k for healthcare applic a tions. Front Neuro sci. 2023 Ja n;17: 1093865. 19. Ba rt els J , Ga llou O , I to H, Co ok M , Sarn th ein J, I nd iveri G, et a l. Even t d riven n eura l netw ork on a mixed s ign al n eurom orph i c proc ess or for E EG b ased epilept ic s eizu re detec t ion. S ci Rep. 2025 May 7;15(1):15965. doi:10.1038/ s 41598 - 025 - 99272 - 6 20. Ga llou O, Bart els J , G h os h S , Schin dler K , Sa rnthein J, In div eri G. O nlin e epilept ic s eizu re detec t ion in long - ter m iEE G rec ordin gs us ing mixed - s igna l neurom orp hic c irc uits . I n: 2 024 IEE E Biom edic a l C ircu its and S yst ems C onf erenc e (BioCA S ). I EEE ; 202 4. p . 1– 5. 21. L i R, Zhao G, M uir D R, L ing Y, Burelo K, Kh oe M , et a l. R ea l - time sub - milliw a t t epilep s y det ectio n imp lemen ted on a s pikin g neura l n et w ork ed ge in ferenc e pro ces s or. C ompu t Bio l Med . 2024;183:109225. doi:ht t ps://doi.o rg/10.1016/j.compbio m ed.2024.10 9225 22. Sh a rifshaz ileh M , Burelo K , Sarn thein J, I nd iveri G. A n elec tron ic neuro morp hi c s ystem f or rea l - time det ec tion of hig h f requen cy o s c illa tions (HFO ) in in trac ra nial EE G. Nat C ommu n. 2 0 2 1 May;12(1):3095. 23. Os b orn LE, Drag omir A, Bet tha u s er J L, Hunt CL , Nguyen HH, K a liki R R, et a l. Pros th e s is wit h neu romorp hic mu ltilay ered e - der m is perceives tou c h a nd p a in. S c i Robot. 2018 Jun;3( 19):ea at3818. 24. Don a ti E , V a lle G. Neurom orph ic ha rdwa re for s omat os ensory neuro pros th e s es . N a t C o mmu n. 2 024 Jan ;15(1):556. 25. Herbo z o Contrer as L F, Truon g ND, Eshr a ghi a n JK , Xu Z, Huang Z, Ber sa ni - V eron i TV, et al. Neurom orph ic neurom odu lation : T owa rd s t he next g eneratio n of c los ed - lo op neur ostimul at io n. PNAS Nexus. 2024 Nov;3(11): ga e488. doi:h ttps://doi .org/10.1093/pnasnexus/pg ae 4 8 8 26. Det hier J , Nu yu jukia n P, E lia smit h C, Stewa rt T, E lassaa d SA, Sh enoy KV , et a l. A Br a in - M achine In terf ace O perat ing wit h a R e a l - T i m e S piking Neura l Net w ork C ont rol A l g orith m. A d v Neura l I nf Process Syst. 2011;2011:22 13 – 21. PubMed PMID: 253091 06; Pub M ed Central PM C I D: PM C4190036. 27. In div eri G, L ina re s - Barr a n c o B, Hamilton TJ , va n S c ha ik A, Etienne - C umming s R, Del bruck T, et al. Neur om or ph ic silicon neuron circuit s . Front Neuros c i. 2011;5:73. doi:10.3389/f nins. 2011.00073 PubMed P M ID: 21747754; Pub Med C entr al PMCID: PM C 3130465. 28. Shi Y, Ananthakr ishnan A , Oh S , Liu X , Hot a G , Cauw en be rghs G , et al. A neuromor phi c brain inter fa ce based o n R R AM cro s sb a r arr ays for high thro ug hp ut real - t ime s pik e sort ing. I EEE Tran s Elect ron De v ice s. 2022 Apr;69(4):2137 – 44. 29. Yoo J, S hoa ran M . Neur a l in terfa c e syst ems w ith on - devic e c omput ing: ma chine lea rn ing a nd neuro morphic ar c hitect ures. Curr Opin Biotechnol . 2021 Dec;7 2:95 – 101. 30. Da i S , D a i Y, Zhao Z , X ia F , L i Y, Liu Y, et al. I nt rins ica lly st retc hab le neuro morp hic devices fo r on - body proc es s in g of hea lt h d a ta wit h a rt iﬁcia l int elligen ce. Matter. 2022 Oct; 5(10 ):33 75 – 90. 31. C hoi Y, Jin P, L ee S , S ong Y, Ta y R Y, K i m G, et al. All - print ed c h ip - less w ea r a ble n eurom orph ic sy s tem for m ult imod a l phys icoch emica l hea lth m onit oring . Nat C o m mun. 2 025 J ul 1; 16(1 ):5689. doi:10.1038/s41467 - 025 - 60 854 -7 32. Wa ng F, T heilman BH, R ot h ga ng er F, S ev era W, V iney a rd C M, A imo ne JB . Neurom orp hic Sim ulat ion of D roso phila M elanoga ster B rain C onn e c tom e on Loih i 2 [I nt ernet ]. a rXiv ; 20 25 [c ited 202 5 S ep 7]. Available from: http://arxiv.org/ab s /2508.1679 2 doi:10.48550/ar Xiv.2508. 16792 33. BR AI N Neur oAI W orks hop Meet ing S ummary [I nt ernet]. Nation a l Inst it u t es of H ealth; 202 4 [c ited 2025 Dec 7]. Available from: ht tps ://bra inin itia t ive.n ih. gov / s ites /def ault/ﬁ les /doc umen t s /BRA IN%2 0NeuroA I% 20Works hop%2 0S u mmary_2024_508 c.pd f 51 34. NIH BRA IN Neu roAI W orks hop 2 024 — P rogr am Boo k [In te rne t]. Natio nal In stit ute s o f Heal th; 202 4 [c ited 2025 Dec 7]. A va ilab le from : ht tps ://bra inin itia t ive.n ih. gov / s ites /def ault/ﬁ les /doc umen t s /20 24%20 NIH% 20BRAI N % 20Neu roAI% 20 Workshop % 20Fin a l%20Pro gram%20Boo k_508c%201.pdf 35. F ellous JM, Ca u w enb erghs G, F ermu ller C , Sa ndam irs k a ya Y, S ejnows ki T . Neuro AI an d Bey ond . 2026. doi:https://doi .org/10.48550/arXiv.2601.19955 36. NIH BRA IN R es e a rc h R oa d ma p [I nt ernet ]. Natio na l Ins tit ut e s of H ea lth; 20 25 [c ited 20 25 D ec 7 ]. A v a ila ble fro m: h tt p s ://b ra ininit ia tiv e.nih .g o v /sit es /defau lt/ﬁ les /2025 - 11/BRAI N %20Re s ear c h % 20Ro ad m a p % 20I nnovation%20Dom a ins_50 8c.p df 37. Br onte - St e w a rt H, Beud el M , Os t rem JL, F a sa no A, A lm eida L , Hass ell T , et al. A dap tive DBS A lgorit hm for Pers on a lized T hera py in Parkins on ’s D is e a se: A DAPT - PD c linica l tria l m eth odolog y an d e arl y d at a (P 1 - 11.002). Ne urology. 2023 Apr;100(17_s upplement_2). 38. Br onte - St e w a rt H M , Beud el M, Os trem JL, L it tle S , A lm eid a L, Ram irez - Z am or a A , e t al. Lon g - Ter m Pers ona liz ed A dapt ive D eep Bra in S tim ul a t ion in Pa rkins on D is e a se: A Nonra ndom i z ed C linic a l T ri a l. JAMA Ne urol. 2025 Sep 22. doi:10.1001/ja maneurol.2025.2781 39. Oeh rn C R, C ern era S, Hamm er LH, S hcherb a kova M, Ya o J, H ahn A, et al. Chron ic a dap tiv e deep bra in s t imula tion versus conv ent ional s t imulat ion in Pa rkins on’s d i s ease: a b lind ed ra nd omiz ed feas ibi lity t ri a l. Nat Med. 2024 Nov 1;30(11):3345 – 56. doi:10.1038/ s 4159 1 - 024 - 03196 -z 40. Dim a kopou los V , G o t man J , Klim es P, von E llenrieder N, T a n S B, S mith G , et a l. M u lticent re a naly sis of s eiz ure out c ome predic t ed by remo va l of h igh frequ en c y oscilla t ions . B r a in [I nt ernet ]. 202 4 Nov . Available f rom: htt p s://pubmed.n cbi.nlm.nih. gov/3953026 2/ 41. Glis ke S V , Irw in ZT, C he s t ek C, H eg e m an GL, Brin kman n B, Sagh er O, et a l. Va ria bilit y in th e loca t ion of hi g h fre quency oscill a tions during prol onged intracr a nial EEG recor dings . N at Co mmun. 2018 Jun 1;9(1):2155. doi:10.1038/ s 41467 - 018 - 04549 -2 42. L in J, Sm ith GC , G lis ke SV, Zoch owski M, S hedd en K, S tacey W C . H ig h freq uency oscilla t ion n etw ork dy namics pred ic t ou tcom e in n on - pa llia tiv e epileps y s urg ery. Bra in Commu n. 2 024 F eb 1;6(1):fca e032. doi :10.1093/br a inco mms/fcae032 43. Gunnar s dot tir KM, Li A, Smith RJ, K a n g JY, Korzenie wska A, Cro ne NE, et al. Source - s in k co nnec t ivit y: a nov el int erict a l EE G m a rker for s eizu re loc a liza t ion. B r a in. 2022 Nov; 145( 11): 3901 – 15. 44. Jirs a V, W a n g H, T riebkorn P, H as hem i M, J h a J , G onza lez - Mart inez J, et al. Pers ona lised v irtu al b ra in models in epil eps y. Lanc e t Neurol . 2023 M ay;22(5):443 – 54. 45. M a k ha lova J, M edin a Villa lon S, Wang H , G ius iano B, W oodman M, Béna r C, et al. Vir tua l ep ilept ic pa t ient br a in m odelin g: Rela t ionship s wit h s eiz ure ons et a nd s urg i c al o utcom e. Ep ileps ia . 202 2 Aug ;6 3(8):1942 – 55. 46. Wa ng HE , W oodm a n M , T riebkorn P, L em a re c ha l JD, Jha J , D olloma ja B, et a l. D elinea tin g epilep tog enic net w orks u s ing bra in imaging data an d person a lized modeling in dru g - re sis tan t epil eps y. Sc i Tr a nsl M ed. 2023 J a n;15(680):eab p8982. 47. Wa ng HE , Trieb korn P, Brey to n M , Dollom a ja B , L e m a recha l JD, Pet kos ki S , et a l. V irtua l b rain twins : from ba s i c neu roscien ce t o c linica l us e. N at l Sci R ev. 202 4 M ay 1 ;11 (5):n wae0 79. doi:10.1093/nsr/nw a e079 48. Ki r al - Korn ek I, R oy S, Nurs e E , Ma s hford B , Ka roly P, C a rroll T , et a l. E p ilept ic s eiz ure predic tion us ing big da ta an d deep lea rning : T ow a rd a m o b ile sy s tem. E BioMed ic ine. 2 018 J an;2 7:103 – 11. 49. Burelo K, Sh a rifs h a z ileh M, In div eri G, Sarn thein J. A ut o m atic det ection of hig h - frequenc y osc illa tio ns with neuro m or phic spiki ng neural net works. Front Neur osci. 2022 Jun;16: 8614 80. 50. M anz S, Va lett e R, D amont e F , A van c i G a ud io L, Gonza lez - Va rgas J, S a r t ori M , et a l. A review of us er need s t o driv e th e dev elopm ent of low er limb pros t hes e s . J Neuroeng ineerin g R ehab il. 202 2 Nov 5;19(1):119. doi:10.1186/ s 12984 - 022 - 01097 - 1 Pu bM ed P MID: 363 35345; Pub Med Central PMCID: PMC 9 636812. 51. Ju ng R , edit or. Biohy brid S y st e m s . W einh eim, Germa ny : W iley - V C H Verl a g; 2011. 52 52. Tho t a AK , K un tae go wd an ah al li S, St ar o s ciak A K, Abb as JJ , Or b ay J , Hor ch K W, e t al . A sys te m and met hod to inter face w it h m ult iple g roup s o f a xons in s ev era l fascic les of p eriphera l n erves . Bra in Comp u t I n terfac e s T ribute Gre g Gerh ardt. 2015 Apr 15;244:78 – 84. doi:10.1016/j .jneumeth.2014.07.020 53. Pena AE, Kunt aegowdan a h alli SS, Abb a s JJ , Pa t ri ck J, Horc h KW, Jung R. M ec hanical f a tigue r esistance of a n im pla ntable b ra nc hed lead s y s t e m for a dis t ribut ed set of long it udina l int raf asc icula r ele c trodes. J Neural Eng . 20 17 Nov 10;14(6):066014. doi:10.10 88/ 1741 - 2552/a a 814 d 54. 75 Yea rs a ft er A ccid ent , U AM S S urg eon s , I 3 R R e s ea rchers Rest ore Greenw ood Pa st or’ s Sen se o f T ouc h, A b ility to G rip [I nt ernet ]. A vaila ble fr om: htt ps: //news . u am s. edu /2024 /11/08/75 - year s - af te r - a ccid ent - uams - surgeo ns -i 3 r- res earchers - restore - gre enwo od - p as to rs - sen se - of - to uch - ability - to - gri p/ 55. Wa ng X , G eiger F, Nic ulesc u V, Magn o M, B enini L . L evera g ing tac t ile sen s or s for low la tenc y embedde d sma rt h a nd s for prostheti c a n d robotic appl ications. IEEE Tran s In s tru m Meas. 2022;71:1 – 14. 56. Za nghieri M , O rla n d i M, D o nat i E , G ruppioni E , Ben ini L, Bena tti S . O nlin e unsu perv ised a rm p o s tu re ad ap t atio n f o r sE MG - bas ed ges ture rec ogn ition on a para llel u ltra - low - po wer m ic ro c ont roller. I n: 202 3 I EEE Bio med ical C irc uits an d Syst ems C onf eren c e (B ioC AS). IEEE; 2023. p. 1 – 5. 57. Ram irez H , Donat i E, Behren s W von der, I nd iveri G, Va lle G, Va n a tto u - Sa ifoud ine NVS . A p hysica l emu la tion of s o m atos enso ry c ort ex as a Neu romorphic Twin for neu ra l pros t hes e s . 2024. 58. De L uca C , T inc a ni M, I ndiv eri G, Do n a t i E. A neu romo rphic m ult i - sc al e appr o ach f or r e al - ti me he a r t rate an d state dete ct io n. Npj Unconv Com put. 2025 Apr;2(1) :6. 59. Siu R, Abba s J J , H illen BK , Gom es J, Coxe S, Ca stelli J, et a l. Res t oring ven tilat ory control us ing a n adaptive bioel ec tronic syste m . J Neur ot rauma. 2019 De c; 36(24):3363 – 7 7. 60. Jia ng Y, Zhang Z , Wa ng YX, L i D, C oen CT, Hwa un E, et a l. T opolo gica l supram olec ular net w ork ena bled hig h - con duct ivit y, st retc hab le organ ic b ioelec tron ics . S cien ce. 202 2 Ma r;3 7 5( 6587 ):14 11 – 7. 61. Rameza ni M , Ki m JH, L iu X, Ren C, Aloth man A, De - E knam kul C, et al. Hig h - de n sity tr ans pare nt gra p hene arra ys for pred ict ing c ellula r ca lciu m act ivit y at dept h f rom su rface p otentia l rec ord ing s. Nat Na no tech nol. 2024 Apr;19(4):504 – 13. 62. Shi Y, Ananthakr ishnan A , Oh S , Liu X , Hot a G , Cauw en be rghs G , et al. Hi gh throug hput neuro morphic brain inter face with C uO x resist ive c rossba rs f or real - tim e s pike s ort ing . I n: 202 1 I EEE In tern ation a l Elect ron D evic es M eetin g (IEDM) . IEEE ; 2 021 . p. 16.5 . 1 - 16.5.4. 63. Huang H, Su S , W u N, W an H, Wan S, Bi H, et a l. Gra phene - ba s ed sen sors for hum an health monitor ing. F ront Chem. 20 19 Jun;7:399. 64. L in Z, K ireev D, L iu N, Gupta S, La Pia no J, Obaid SN, et a l. G ra phen e bioint erfac e for c a rdiac arrhyt h mia d ia gn o sis and t re atment . A dv Mat er . 2023 Jun ;35(22):e 2212190. 65. Kireev D, S el K , Ib rahim B, Kuma r N , A kb a ri A, J af ari R, et al. C on t inuous c u ﬄess mo nit oring of a rteria l b lood pres s ure v ia grap hene b ioimp edanc e ta ttoos . Nat Na no t ec hno l. 202 2 A ug ;1 7(8) :864 – 70. 66. Che n S , M ah m oo di MR , Sh i Y , M aha ta C, Yu an B , Lia n g X , e t al. W afe r - sca le int egra t i on of t wo - dim ens iona l materia ls in h igh - dens ity mem ris tiv e crossba r a rra ys fo r art iﬁcia l neu ra l netw orks . N a t Elect ron. 2020 Oc t; 3(10):638 – 45. 67. L i S, Zha o Z, W eires M , H u S, L i Y, Tang L, et al. Larg e - scale s t retc hab le neurom orp hi c c irc uits f or on - bod y ed ge p rocess ing of s enso ry data . C hemRxiv . 20 24 D ec. 68. Sm it h JD, Hill A J, R eeder LE , F ranke BC , L ehou cq RB , P a rekh O , et al. Neu rom orphic sca ling adv an t age s fo r e ne r gy - eﬃ c ient r a n dom w alk comput a tion s. N a t E lectron. 2022 Feb ;5(2):102 – 12. 69. T heilm an B H , Aim one JB. Solv ing s p a rs e ﬁnite elem ent problem s on neur omorp hic ha rdwa re. Na t Mac h I nt el l. 2025 Nov 1;7(11):1845 – 57. doi:10.1038/ s 42 256 - 025 - 01143 -2 70. A imon e JB. Neurom orph ic Com putin g: A T heoret ical F ram ework f or T ime, S pac e, and En ergy Scaling. A rXiv Pr epr A rXi v 2507178 86. 2025. 53 71. Sc hmit t E, Gupt a S , Ab b s P. C on tinuou s lea rn in g for rea l - t im e a u dit ory blind s ourc e s epa rat ion a pplica t ions . In: 2024 Intern a tiona l Con ferenc e on Neurom orph ic Sys tems (ICO NS) . I EE E; 202 4. p. 90 – 3. 72. C ent er for Dev ic es, Rad iolog ical H ea lth . T echn ical c on sid er a tio ns for m edica l devic es w ith ph ys iolog ic c los ed - loo p c on trol technolo g y. U .S. F oo d and Drug A dm ini stration [In ter net]. FDA ; 2023 [cited 2023 Au g 7] . A v a ilable from: htt p s://www.fd a.gov/regul ato ry - infor mation/se a rch - fd a - gu id an ce - do cu me nt s/ tech ni cal - consid era tion s - me di cal - devic es - ph ysiolog ic - c los ed - loop - c on t rol - technology 73. IEC 60601 -1- 10 E d. 1 .2b, 2 020 . M edi c a l elec trica l equ ipm ent - P ar t 1 - 10: G enera l requ iremen ts for ba s i c sa fety a nd essen tial p erforma n ce - Co llat er a l S tan d a rd : Req uirem ents f or th e dev elopm ent of ph ys iolog ic c los ed - loop c ontrollers . [In tern et]. I nt ernat iona l E lectro tec hn i c a l Commis sion (I EC); 202 0 [c ited 2025 Sep 4]. A vaila b le from : ht tps ://web store. iec. c h/en/p ublica tio n/673 83 74. F. Ch e n , A. P . Cha n dr ak as an , V. M. S toj an ovi c. De si gn and An al ysi s of a H ard w are - Eﬃ c ient C omp ressed Sens ing A rch itect ure fo r Da ta Com press ion in Wireles s Sens ors . IEEE J S olid - St ate Circuits . 20 12 Mar;47(3):744 – 56. doi:10.110 9/JSSC.2011.2179451 75. J. Zha ng, Y. Su o , S . M itra , S . Chin, S. Hs iao, R. F . Yaz icio g lu, et al. A n eﬃ c ient a nd c o m pact c omp ressed sens ing m icros ys tem for im pla nta ble neu ra l rec ordings . I EEE T rans B iomed Circ u its Syst. 2014 Aug;8(4):485 – 96. doi :10.1109/TBCAS.2013.2 284254 76. Yik J, V a n den Berg he K, d en Bla n ken D, Bo uhad jar Y, Fa bre M, Hu eber P, et a l. T he neurob enc h fr amewor k fo r be nch marki ng ne uro mo rph ic co mputi n g a l gori th ms and sy stem s. N at Co m mun. 2025 Feb 11;16(1):1545. doi:10.1038/ s41467 - 025 - 567 39 - 4 77. Tom ovi c h C. MO SI S - A gat eway t o s ilic on. I EEE Circ u its Dev ic e s Mag. 1 988 Mar; 4(2):22 – 3. doi:10.1109/101.936 78. A rbib MA . A r t iﬁ cial int elligen c e meet s bra in t h eory (a ga in). Bio l Cy bern. 2025 Jun 28;11 9(4 – 6):16. doi:10.1007/s00422 - 025 - 01 013 - 5 PubMed P MI D: 405 79583; PubMed Cen t ral PM C ID : PMC 122 04934. 79. Lun a - Mungu ia H , Z e s tos AG, G liske S V, Ken nedy RT, S tac ey WC . Chem i c a l bioma rkers of epilep tog enes is a n d ict og enesis in exp erimen ta l epilep sy . Neu robiol D is. 201 9 Ja n ;121:17 7 – 86. doi:10.1016/j .nbd. 2018.10.005 Pub Med P M ID: 30304705; Pub Med C entr al PMCID : P M C6242767. 80. Xu J, W a ng S, W a ng GJN, Zhu C , Lu o S, Jin L, et al. H ig hly st ret c h a ble polymer sem ico n duc tor ﬁ lms through th e nano c onﬁ nement e ﬀect . Sc ie nce. 2017 Jan 6;3 55(6320):59 – 64. doi:10.1126/science. aah 4496 Pub Med P M ID: 28059762. 81. Xu J, Wu HC, Zhu C , Ehr lich A, Shaw L , Niko lka M, et a l. Multi - s ca le orderin g in hig hly s t retc hab le polymer semicon ducting ﬁ l ms. N a t M ater . 2019 Jun;18(6):594 – 601. doi:10.103 8/s41563 - 019 - 0340 - 5 PubMed P M ID: 30988452. 82. Li N, Ka n g S, L iu Z , W a i S, Cheng Z , Da i Y , et al. Immune - c om p a t ible des ig n s of semic ond u c t ing poly mers fo r bioelec tro nic s w it h s u ppress ed fo reign - bod y respon s e. N a t Mat er . 2025 A pr;1 – 9. 83. Zhou S, Li Z, Del bruck T, Kim K , Liu SC. An 8.62 μ W 75dB - DRSo C E nd - to - End Spoken - Langu age - Under s t a nding SoC With C h a n nel - L evel A GC a n d T e m pora l - Sp ar si ty - Aware Stre ami n g - Mode RNN. In : 20 25 IE EE Intern at ional S olid - Stat e Circ u its C onf erenc e (ISSCC) . 202 5. p . 238 – 40. doi:10.1109/ISSC C 49661.2025. 10904788 84. Perg e JA , Homer ML , Ma lik W Q, Cas h S, Eska nda r E , F rieh s G, et a l. I nt ra - day signa l i n sta bilit ies aﬀec t dec odin g p erform anc e in a n in trac ort i c a l neural in terfa c e sys t e m . J Neu ra l E n g . 20 1 3 Jun;10(3):036004. doi:10.1088/1741 - 2560/10/3/ 036004 PubMed PM I D: 23574741; P ubM ed Centra l PMCID: P M C3693851. 85. Ka rpo wicz B M , A li YH, W imalas ena L N, S edler A R , K e s ht k a ra n MR, B odkin K, et a l. St a biliz ing bra in - comp u ter interfaces through alignmen t of l a tent dyn a mics. Na t Co m mu n. 2025 May 19;16(1):4662. doi:10.1038/s41467 - 025 - 59 652 -y 86. Deg enh art A D, B ish op WE, O by E R, Ty ler - Kab ar a E C, C ha se S M, B ati st a AP , e t al . St a bi l iz ati on of a br ai n – compu ter in terf ace v ia th e a lign ment of low - dim en s ion al s paces of n eura l activ ity . Na t Biomed Eng. 20 20 Jul 1;4(7):672 – 85. doi:10.1038/ s 41 551 - 020 - 0542 -9 54 87. Erbslöh A , Buron L, Ur - Reh man Z , M us a ll S , Hryc ak C, Löh ler P, et a l. T echn ical s urv e y of end - to - end signal processing in BCIs using inv a sive M EA s . J Neural Eng. 2024 Oct 15;21(5):0510 03. doi:10.1088/1741 - 2552/ ad8031 88. Lozano A, Chen X, La Gro uw M , Li B , Wang F, v a n der Grinten M , et al. L arge - sca le RF map pin g without visual i nput for neuropr ostheses in macaque a n d human visual corte x. m edRx iv. 2024 Jan 1;2024.12.22.2431904 7. doi:10.1101/2024.12.22.243 19047 89. Liu SC, Delbr uc k T. Ne uromorphic s en sory systems. Sens Sy s t. 2010 Jun 1;20(3) :288 – 95. doi:10.1016/j .c onb.2010.03.007 90. Da vo udi A, M a lhotra KR, Sh i c kel B, Sieg el S , Willia ms S , R upp ert M, et al. In tellig ent ICU for A ut onomous Pa t ient M onit oring Usin g Perv asiv e S ens in g a nd Deep L earnin g. Sci R ep. 2 019 May 29;9(1):8020. doi:10.1038/ s 41598 - 019 - 44004 -w 91. B. Zhou, P. - S . V . Sun , J. Yik, C . Fren kel, V . J. R eddi, A. B asu . Gran d C ha llen ge on Neural D ecod ing for M otor C on t rol of non - H uman Prim a tes. I n: 202 4 I EEE B iomed ical C irc uit s a nd S y s t e m s Con ferenc e (Bio C AS ). 2024. p. 1 – 5. doi :10.1109/BioCAS61083.2024.107 9837 3 92. Ba rt olozz i C, In diveri G , D onat i E . E mbod ied neu romorp hic int ellig enc e. Na t C omm u n. 2 022 F eb 23;13(1):1024. doi:10.1038/ s 41467 - 022 - 28487 - 2 93. A. G. Andreo u, T. Figlio lia, K. S a nni, T . S. M urra y, G. Tog net ti, D. R. Mend a t, et al. N euromorphi c C hiplet A rchit ect ure f or Wid e Area Mot ion Im agery Proc essin g. In : 2024 A rg en t ine Con ferenc e on Elect roni cs (CAE ). 2024. p . 160 – 71. doi:10.1109/ CAE 59785.2 02 4.10487119 94. The Vis io n for the N ational Adva nced P a ck a ging Manuf a ctur ing Progra m [ I nterne t]. Nat io na l In s titut e of S tan d a rds an d T ec hno logy; 20 23 [c it ed 20 25 Sep 4 ]. A v a ila ble f rom: ht tps ://w ww.n is t.gov /sys t e m/ﬁ les /docu ments /2 023 /11/19 /NA P M P - Vis ion - P ape r - 20 231120.pd f 95. M uir DR , S heik S . T he roa d t o co m m erc ia l s ucc ess f or neuro morp hic tech nolog ies . Na t Commun . 2025 Apr 15;16(1):3586. doi:10.1038/ s41467 - 025 - 57 352 - 1 96. T ha k u r CS, M olin J L, Ca uwenb erghs G, I ndiv eri G, Ku mar K , Qia o N, et al. Larg e - Sca l e Neurom orph ic Sp iking A rray Proc ess ors : A Q uest to M imic th e Bra in. Fron t Neu rosc i [In tern et]. 2 018 ;V olum e 12 - 2018. Available from: https:// www.frontie rsin.org/ journals/ne uro science/artic le s/10.3389/fni ns .2018.0 0891 97. Su nd eric K , Wrig ht A M, Kleins t reuer N, L edb e t ter V , M ilora KA, Hap pel C, et a l. C omplemen t - ARI E: C ata lyzing t he dev elopm ent a nd ad op t ion o f new a pproach met hod ologies . N AM J. 2 025; 1:1 0002 6. doi:https ://do i.org/10.1016/j.na mjnl.2025.100026 98. Gh an n am R, Curia G, B rante G , Khosra v i S, Fa n H . I m pl a nta ble a nd W earab le Neuroen gin eering Education: A Review of Postgra du a t e Pr ogra m m es. I EE E Acce s s. 2020;8:212396 – 408. doi:10.1109/ACC E S S .2020.30400 64 99. Rom m elfang er KS, Porc ello D, S alles A, Na lb ac h T ou rn as L . Neu roeth ics h ackat hons bridg e th eory to prac ti ce. Neuron. 2024 Dec 18;112(24):3994 – 8. doi:10.1016/j .n eur on.2024.11.013 100. St a c y RW. T he S t a tus and Devel opment of Biophysic s . Science. 1951 Feb 16;11 3(2929):169 – 75. doi:10.1126/science.113 .2929.169 101. Dü zg üneş N. H istory and Philo sop hy of a Bio phys ics Depar tm en t . T he Biop hysic ist. 2022 Dec;3(2):115 – 20. doi:10.35459/t bp.2022.000218 102. Gu ilford W H, Ko tche M, Sc hmed len RH. A Su rvey of C linica l Imm ersion E xperienc es in Biom edic a l Eng inee ring. Biomed E ng Edu c . 2023 Jul 1;3(2):113 – 22. doi:10.1007/ s43683 - 023 - 001 07 -5 103. C auwen berg hs G. Rev erse eng ineering t he c ogn itiv e bra in. Proc N a t l Acad Sc i. 20 13 S ep 24;110(39):15512 – 3. doi:10.1073/ pnas.1313114110 104. Da ta Scienc e and In forma tics . htt p s ://b ra ininit ia tiv e.nih .g ov/res e a rc h/da t a - sci en ce - and - inf or mati cs [In tern et]. [cit ed 2 025 Ma y 26]. Ava ila ble from : h ttps ://b rain init iativ e.nih .g ov/res e a rc h/d ata - sci e nce - and - inf orma tics 105. Em er gin g Frontie rs in Rese a rch a nd I nnovation ( EFRI - 2022/23) 1. Enginee red Living System s (ELiS) 2. Br ai n - I n s pired Dy n a m i c s for E n g ineerin g E n ergy - Eﬃc ient C ircuit s a nd A rt iﬁ cial I nt elligen ce (BR AID ) 55 [go vernm ent ] [I ntern et]. [cit ed 2 023 Nov 23 ]. Emerg in g F rontiers in R es earch and I nnova t ion (EF RI - 2022/23). Ava il able fro m : https://ww w .nsf.gov/pub s/2021/nsf21615/ ns f2161 5.htm 106. Fut ure of semi conductors ( FuS e2) https://w ww.nsf. gov/funding/o pportu nities/fuse 2 - future - semic ond uctors/506141/n sf24 - 521. N SF - Na tional S c ience F oun d a t ion [I nternet ]. Av aila b le from : https:// www.nsf.gov/ funding/oppor tuni t ie s /f us e2 - future - sem i con du cto r s/5 0 6141/n sf24 - 521 107. L itt le S , Po gosyan A , Nea l S, Zav a l a B, Zrinz o L, Hariz M, et a l. Ad aptiv e deep bra in s timula t ion in advanc ed P arkinson dise a se. Ann Neurol . 2013 S ep;74(3) :449 – 57. 108. Wagn er F B, Mig n a rdot JB, L e Goﬀ - M igna rdot C G, Dem esmaeker R , Kom i S, C a po gross o M , et a l. T arg e t ed n eurot ec hnolo gy res tores w a lking in h u m a ns w ith sp in a l c ord injury . Na t u re. 2018 Nov;563(7729):65 – 71. 109. Sa rp eshka r R. A nalog V ersus Dig ital: Ext r a po lating from E le c t ronics to Neu robiolog y . Neura l Comp u t. 1998 Oct 1;10(7):160 1 – 38. doi:10.1162/ 089976698300017052 110. A vest ruz AT, San ta W , Ca rlson D, Jens en R, Stans laski S , Helf ens tin e A , et al. A 5 μ W/ c h annel s pect ra l a n a ly sis IC for c hron ic bidirec t ional bra in – mac hin e int erfac es. IEEE J S olid - S t a te Circui t s. 2008 Dec;43(12):3006 – 24. 111. V ans teensel MJ, Pels E GM, B leich ner M G , B r a nc o M P, D enis on T, F reuden burg Z V, et a l. F ully imp la nt ed brain – compu ter in terf ace in a lo cked - in pa tien t w ith AL S. N E ng l J Med. 201 6 Nov;375(21):2060 – 6. 112. V ans teensel MJ, Lein ders S, B ra n c o MP, C rone NE, Den ison T , Freu denb urg Z V, et al. L ong evit y o f a br ai n - c om puter int erfa c e for a myot roph ic latera l sc lerosis . N Eng l J M ed. 2 024 A ug ; 3 91(7 ):61 9 – 26. 113. Op ri E , Cern era S, M olin a R , E i s ing er R S, Ca gle J N, Alm eida L, et al. Ch ronic embedded c ort ico - th a l a mic c los ed - lo op deep bra in st imu l a tio n for t he t rea tment of ess ent ia l trem or. Sc i T r a ns l M ed. 2020 Dec;12(572):eaay7 680. 114. T oth R, Zam ora M , O ttaway J, Gillbe T , M a rt in S , Benja b er M , et al. D yNeu Mo M k - 2: A n in vestig at iona l ci rcad i an - locked neu romod ulat or wit h res po n s ive s t i m ula tio n for a pplied c hro nobio logy . C onf Proc IEEE I nt C onf Syst M a n Cyber n. 2020 O c t;2020:3 433 – 40. 115. Os b orn LE, V enk a tasu braman ia n R, H immta nn M, M or a n CW, Pierc e JM, Ga jendiran P, et a l. E v oking na t ur a l t herma l perc ept ions usin g a t hin - ﬁ lm thermo elect ric devic e wit h h i gh coo lin g pow er dens it y and speed. N a t Biomed Eng . 20 24 Aug;8(8) :1004 – 17. 116. Oh Y, Heien ML, Park C , Ka ng YM, Kim J, Boschen SL, et a l. T rac king to nic dopamine lev els in viv o us ing mult iple c yc lic s quare w a ve v oltamm etry . Bios ens B ioelec tron . 20 18 Dec 1 5;121:1 74 – 82. doi:10.1016/j .bios.2018.08. 034 Pub Med PMID: 3021892 5; PubM ed C entr al PMCID: PMC 6 775780. 117. Shin H, Goyal A, B arnett JH, Ru s hee n AE, Yue n J, Jha R, et a l. Toni c Serotoni n Mea s ur eme nt s In Vivo Usi ng N - Sh a ped Multiple Cyc lic S qu a re Wa ve V oltamme try. A n a l Che m. 2021 D ec 2 8;93(51):16987 – 94. doi:10.1021/acs. a nalchem.1 c 02131 118. Hua ng H H, Ha rgr ove LJ, Ortiz - Ca t a la n M , S ens ing er JW . I nt egrat ing Upp er - Lim b Prost hes es w ith the Human Body: Technol ogy Adv a nce s , Re a dine ss, and Roles in H um a n - Pro sthe s is In ter a ction. Annu Rev Biomed Eng. 2024 Jul;26(1):503 – 28. doi :10.1146/ a nnurev - b ioeng - 110222 - 09 581 6 PubM ed P MI D: 38594922. 119. L ang C E, H ollera n CL , S tru be M J , Ellis T D, Newman CA, Fa hey M , et a l. Im prov e m ent in t he C a pa c it y for A c tiv ity V ers us Imp rovem ent in Perf orma nce of Act ivit y in Daily Lif e Durin g O utpa t ient Rehabili ta tion. J Ne urol P hys T her JNPT. 2023 Jan 1;4 7(1):16 – 25. d oi:10.1097/ NPT.0000000000000413 PubMed P M ID: 35930404; Pub Med C entr al PMCID: PM C 9750113. 120. C imorelli A , Pa tel A, K arakos tas T, C otton RJ . V a lidat ion of port ab le in - c lin ic v ideo - ba se d g ai t an al ysi s for pro s thesis u s er s . Sci Rep. 2024 Feb;14(1):384 0. 121. Kunt a egowd a n ahalli SS . The L ong - t erm Fu n c t iona lit y of a Ha pt ic S ens ory F eedbac k S yst em w ith a n In tra fasc icula r Nerv e In terfa c e [I n t ernet ]. F I U E lec tro nic T hes e s a nd Diss erta t ions; 2022 [c it ed 20 25 Ju n 18]. Availa b le from: htt ps://d igit a lc o m mons.ﬁ u.ed u/et d /5180 122. M a X, R iz z og lio F , B odkin K L , M iller L E . U n s up ervis ed, piec ewis e linea r dec odin g enables a n acc urate pred ict ion o f mu scle a ct ivit y in a mult i - task br a in com puter i nt er fa ce. J Ne ural E ng. 2025 Ja n 56 29;22(1). doi:10.1088/ 1741 - 2552/ adab93 Pub M ed PM I D: 39823647; Pub M ed Central PMCI D: PMC 1 1775726. 123. Saa l H P, Delhay e BP, Ray haun B C , Bensma i a S J. S i m ula t in g tac t ile s ignals fro m the w hole h a nd wit h milli s econd precision. Proc Natl Acad Sci. 2017 Jul 11;114(28):E 5693 – 702. doi:10.1073/pna s .170485611 4 124. Ron gala U B , Ma zzoni A , C hiu razz i M, Cam bon i D, M ila zzo M, Massa ri L , et al. Ta c tile D ec oding of E dg e Orien ta t ion With A rt iﬁcia l C un eate Neuron s in Dyna mic C ond itions . Fro nt Neu rorob otic s [Internet]. 2019;Volume 13 - 2019. A v a il a ble fr om: http s:/ /ww w.fr onti ersin.org/jo urna ls/neuror obotics/artic le s /10.338 9/fnbot.2019.00 044 125. C ourt ine G, G erasim enko Y, van den Brand R , Yew A, M usien ko P, Zho ng H, et a l. T ra ns formatio n of non fun ction al s p in a l c irc uits in to f unct iona l st ates a ft er th e los s o f bra in in put. Na t Neu ros ci. 2009 Oc t 1;1 2(10):1333 – 42. doi:10.1038/nn.2 401 126. M a rketing S ubmiss ion Rec ommen dat ion s for a Predet ermin ed C hang e Contro l Plan for Art iﬁ c i a l In tellig enc e - Ena bled D evic e S oftwa re F unction s. FDA - 2022 - D- 2 628 [Internet]. U.S. Foo d & D rug Adm ini strati on; [cited 2025 Sep 2]. Av a ilable from: https://w ww.fda.go v /regu lat ory - i nf or mati o n/ se ar ch - fda - g ui dan ce - do cu ments/ mar ke ting - su bmi ssi on - re commenda ti ons - pred eterm ined - c ha ng e - co n tro l - p lan - ar tiﬁ ci al - in tellig enc e 127. Res ea r c hers Ma ke L o w Energ y LLM Brea kthrou gh [In tern et]. [cit ed 2 025 Jun 18]. Ava ila ble from : https:// www.santa cruzw or k s .or g /ne ws/re volutioni zing - energ y - in - lar ge - lan gu age - models 128. Hwang GM , S imoni an AL. Spe c ial I ssue — Bios en s ors a n d Neuros c ience: Is Bios ensors E ngin eering Rea dy t o E mbrace D es ign Princ iples from Neuro scien ce? Bios ensors . 2024 ;14 (2). doi:10.3390/bi os 14020068 129. Kh at ib M, Zha o E T, W ei S , Ab ram s on A, B ish op ES, Ch en C H, et a l. S piral Neu roSt ring : H ig h - dens ity s oft bioelec tron ic ﬁ bers for mult imodal s ens ing a nd st imula tio n. bioR xivorg . 2 023 O c t. 130. Li J, L iu Y, Yu an L, Zh a ng B , Bishop ES, Wang K , et al. A tis s ue - like neur otra nsm itt er sens or fo r th e brain a nd gu t. N at ure. 2022 Jun;60 6(7912):94 – 101. 131. Rameza ni M , Ren Y, C ub ukcu E , Kuzu m D. Innovatin g beyond el ect ro physiol ogy t hr ough multi modal neu ral in terf ac es. Na t Rev E lec tr En g. 2024 Dec ;2 (1): 42 – 57. 132. L iu X , R en C, L u Y, L iu Y, K im J H, L eutg eb S, et a l. Mu ltim od a l n eura l rec ordin gs w ith Neu ro - FI TM un cov er div ers e patt erns of c ort i c a l - hi ppocampal interactions. N a t Neuros c i. 2021 J un;24(6):886 – 96. 133. Kireev D, L iu S , Jin H, Patric k Xia o T, Ben net t CH, A kin wan de D, et a l. M eta plas tic a nd energ y - eﬃcient bioc om p a t ible g rap hen e art iﬁc ial s yna p t ic tr a n sist ors for enhan ced a ccu r a cy neuromorp hic comp u ting. Nat C o mmun. 2022 Jul;13(1):4386. 134. Broc ca rd F D, Jos h i S, Wan g J , Ca u w enb erghs G . Neu romorp hic neu r a l int erfa c es: fro m neu rophy s iolog ical in sp ira tion to b iohybrid cou pling w ith nerv ou s syst ems. J Neura l E ng . 2 017 Jun 2;14(4):041002. doi:10.1088/ 1741 - 25 52/aa67a9 135. Da i Y, Wai S , L i P, Sha n N, C a o Z , L i Y, et a l. Soft h ydro gel s e m ic ond u c t or s wit h au gm en ted biointer a ct ive fun ctions . Sc ience. 2024 Oct;386(6720):431 – 9. 136. L i N, Li Y, Chen g Z , Liu Y, D a i Y, Ka ng S, et a l. B ioa dhesiv e poly mer s em i c onduc tors an d t rans ist ors for i nt imat e bio interfac e s. Science. 2023 Aug;381(6658):68 6 – 93. 137. Yan g Z, Z ha n g L , Ara s K, Eﬁmov IR, Adam GC. H a rdware - map pable c ellu lar n eura l netw orks for dis t ribut ed wa vefro nt d etect ion in next - g ener a t ion ca rdiac imp la n t s . A d v Int ell Sy st. 2022 Aug ;4(8):220 0032.

The role of neuromorphic principles in the future of biomedicine and healthcare

Original Paper

Comments & Academic Discussion

Leave a Comment

Original Paper

Related Papers

Comments & Academic Discussion

Leave a Comment