In vivo dynamic optical coherence tomography of human skin with hardware- and software-based motion correction

In vivo application of dynamic optical coherence tomography (DOCT) is hindered by bulk motion of the sample. We demonstrate DOCT imaging of \invivo human skin by adopting a sample-fixation attachment to suppress bulk motion and a subsequent software motion correction to further reduce the effect of sample motion. The performance of the motion-correction method was assessed by DOCT image observation, statistical analysis of the mean DOCT values, and subjective image grading. Both the mean DOCT value analysis and subjective grading showed statistically significant improvement of the DOCT image quality. In addition, a previously unobserved high DOCT layer was identified though image observation, which may represent the stratum basale with high keratinocyte proliferation.

💡 Research Summary

This paper addresses a critical obstacle in applying dynamic optical coherence tomography (DOCT) to in‑vivo human skin: bulk motion of the sample during the long acquisition sequences required for dynamic imaging. The authors propose a combined hardware‑and‑software motion‑suppression strategy that is low‑cost, easy to implement, and open‑source.

The hardware component is a 3‑D‑printed fixation attachment that contacts the skin via a flexible TPU spacer, an optical cage plate, and an adjustable objective spacer. The device applies gentle pressure to the forearm, stabilizing the tissue without the need for expensive motion‑tracking hardware. The design files and assembly instructions are made publicly available.

The software component is a 2‑D intensity‑based image registration and shift correction pipeline implemented in Python. Using scikit‑image’s phase_cross_correlation function with an up‑sampling factor of 10, sub‑pixel (0.1 pixel) lateral shifts are estimated by cross‑correlating each frame with a central reference frame (the 16th of 32 frames). The estimated shifts are then compensated using SciPy’s ndimage.shift with third‑order spline interpolation, applied directly to dB‑scaled intensity images. This step further reduces inter‑frame misalignment after the physical fixation.

The experimental setup employs a swept‑source Jones‑matrix OCT system (center wavelength 1310 nm, A‑line rate 50 kHz, lateral/axial resolution 18/14 µm). DOCT data are acquired using a 32‑frame repeated raster‑scan protocol (6.55 s per block, total 52.4 s for a 6 mm × 6 mm field) yielding 4,096 frames. For comparison, a standard 4‑frame OCT‑angiography (cmOCT A) protocol is also performed. Ten healthy volunteers (3 M/7 F, age 24‑29) are scanned on both outer and inner forearms. Four motion‑suppression configurations are tested: hardware + software (HS), hardware only (H), software only (S), and no correction (NC).

Quantitative analysis uses the logarithmic intensity variance (LIV) as the DOCT metric. Surface detection is performed with the Segment Anything Model (SAM) to flatten volumes, and four depth slabs (20‑220 µm, 220‑420 µm, 420‑620 µm, 620‑820 µm) are defined to correspond to epidermis and dermal layers. Mean LIV values are extracted from three central regions of interest (ROIs) within each slab. Paired t‑tests reveal that both hardware fixation and software registration independently reduce mean LIV variability (p < 0.05), and the combined HS condition yields the greatest reduction (p < 0.01).

Subjective evaluation involves three expert graders rating image quality on a 1‑5 Likert scale. The HS condition receives the highest average score (4.7 ± 0.2), significantly outperforming H (3.9), S (3.6), and NC (2.3) (p < 0.01).

A notable finding is the consistent appearance of a high‑LIV layer in the superficial 20‑220 µm slab, which was not discernible in conventional OCT intensity images. The authors hypothesize that this layer corresponds to the stratum basale, where basal keratinocytes exhibit high mitotic activity, thus generating stronger intracellular motion signals detectable by DOCT. This observation suggests that DOCT can provide functional contrast of cellular proliferation in skin, a capability absent from polarization‑sensitive OCT or OCT‑angiography.

Limitations include potential tissue deformation caused by the pressure of the fixation attachment, possible discomfort during prolonged scans, and the fact that the software correction addresses only lateral (2‑D) misalignments, leaving depth‑wise non‑linear deformations uncorrected. Future work is proposed to integrate pressure sensors for feedback‑controlled fixation, develop 3‑D deformation correction algorithms, and extend the method to pathological skin conditions such as psoriasis or atopic dermatitis.

In conclusion, the study demonstrates that a simple, inexpensive hardware fixture combined with an open‑source 2‑D registration algorithm can substantially improve the quality of in‑vivo skin DOCT. The approach yields statistically significant enhancements in both quantitative LIV metrics and expert visual assessments, and it reveals previously unseen high‑LIV structures likely associated with basal cell proliferation. This combined motion‑suppression strategy paves the way for practical clinical deployment of DOCT as a non‑invasive tool for monitoring skin metabolism and disease progression.