We apply three optical coherence tomography (OCT) image analysis techniques to extract morphometric information from OCT images obtained on peripheral nerves of rat. The accuracy of each technique is evaluated against histological measurements accurate to +/-1 um. The three OCT techniques are: 1) average depth resolved profile (ADRP); 2) autoregressive spectral estimation (AR-SE); and, 3) correlation of the derivative spectral estimation (CoD-SE). We introduce a scanning window to the ADRP technique which provides transverse resolution, and improves epineurium thickness estimates - with the number of analysed images showing agreement with histology increasing from 2/10 to 5/10 (Kruskal-Wallis test, {\alpha} = 0.05). A new method of estimating epineurium thickness, using the AR-SE technique, showed agreement with histology in 6/10 analysed images (Kruskal-Wallis test, {\alpha} = 0.05). Using a tissue sample in which histology identified two fascicles with an estimated difference in mean fibre diameter of 4 um, the AR-SE and CoD-SE techniques both correctly identified the fascicle with larger fibre diameter distribution but incorrectly estimated the magnitude of this difference as 0.5um. The ability of OCT signal analysis techniques to extract accurate morphometric details from peripheral nerve is promising but restricted in depth by scattering in adipose and neural tissues.
Deep Dive into Extracting morphometric information from rat sciatic nerve using optical coherence tomography.
We apply three optical coherence tomography (OCT) image analysis techniques to extract morphometric information from OCT images obtained on peripheral nerves of rat. The accuracy of each technique is evaluated against histological measurements accurate to +/-1 um. The three OCT techniques are: 1) average depth resolved profile (ADRP); 2) autoregressive spectral estimation (AR-SE); and, 3) correlation of the derivative spectral estimation (CoD-SE). We introduce a scanning window to the ADRP technique which provides transverse resolution, and improves epineurium thickness estimates - with the number of analysed images showing agreement with histology increasing from 2/10 to 5/10 (Kruskal-Wallis test, {\alpha} = 0.05). A new method of estimating epineurium thickness, using the AR-SE technique, showed agreement with histology in 6/10 analysed images (Kruskal-Wallis test, {\alpha} = 0.05). Using a tissue sample in which histology identified two fascicles with an estimated difference in mean
In the field of neural prosthetics, the performance of nerve cuffs for recording and stimulating bio-electric signals can be improved using physiologically accurate volume conductor models of nerves [1,2]. The morphometric details required for such models are the size, number and location of fascicles, the thickness of ultra-structural tissue layers, and the spatial variations in fibre diameter distribution. It is common to use simplified morphometric details, such as in [2][3][4], however, the results are not transferrable to patients due to patient specific tissue morphology. Another approach is to use destructive imaging methods, such as light microscopy, to acquire micrometre resolution histological images of the nerve cross section at one location and then extrude this along the length dimension, such as in [5,6], however, this does not account for the length variation in tissue morphology caused by fascicle bifurcation [7]. Magnetic resonance imaging (MRI) enhanced with gadolinium-DTPA (diethylenetriamine penta-acetic acid) contrast agent, a non-destructive imaging method, has been used to image the size, number and location of fascicles in an extracted nerve tissue with a voxel size of 30 x 30 x 250 µm 3 [1], which is promising, particularly if replicated with in-vivo measurements. For a patient-specific and physiologically accurate model, a non-destructive volumetric imaging method is required with a resolution of several µm.
The structure of peripheral nerves comprises one or more fascicles bound together by epineurium tissue 10’s of μm thick [8]. In humans, the median nerve is several mm across and can contain 10 or more fascicles at the elbow, with each fascicle ranging in size from 0.12 to 2 mm 2 [9], whereas, in comparison, the rat sciatic nerve is approximately 1 mm across and can contain 3 to 4 fascicles ranging in size from 0.05 to 1mm 2 . Each fascicle contains several thousand nerve fibres bound together by endoneurium tissue, and encompassed by a layer of perineurium tissue several μm thick [10]. Nerve fibres are long cylinders ranging in size from 1 μm to 22 μm diameter and are heterogeneously distributed within fascicles [9]; they are also highly aligned, densely packed, and usually sheathed in lipid rich myelin from encasing Schwann cells. The orders of magnitude of the dimensions of the nerves under study place Optical Coherence Tomography (OCT), a non-destructive imaging method, well as a potential means to acquire morphometric details without damaging the nerve.
Qualitative OCT techniques of distinguishing neural tissue from surrounding tissue [11][12][13][14], identifying different neural tissue layers [11-13, 15, 16], and analysing levels of myelination [17] do not provide quantified values nor confidence levels. On the other hand, quantitative OCT techniques, such as the depth-resolved analysis of optical properties [18] and statistical analysis of spectra [19,20], provide quantified values for nerve tissue morphometry but have not been validated. Other quantitative OCT techniques, such as analysis of Mie scatter spectra [21] and optical scattering properties [22], have been used to classify tissue but have not yet been applied to peripheral nerves. There is therefore a need to evaluate and validate the performance of OCT techniques in imaging peripheral nerves, which builds on preliminary work in ref [23].
In this paper, we present results from three quantitative OCT signal analysis techniques that we identified in the literature and replicated with some improvements on images of rat sciatic nerve acquired with a swept source OCT (SS-OCT) system. Of the three OCT techniques -average depth resolved profile (ADRP), autoregressive spectral estimation (AR-SE), and correlation of the derivative spectral estimation (CoD-SE) -one, the former, was developed specifically for estimation of epineurium thickness while the latter two were developed for scatter size estimation applications. With each technique, we attempt two tasks: 1) extract the epineurium layer thickness, and 2) distinguish adipose tissue. In addition to this, for the two techniques based on scatter size estimation we attempt a third task: 3) estimation of fibre diameter distribution. We compare our results to histological analysis performed on light-microscopy images, a step which is absent in the original reporting of the OCT techniques [18][19][20][21]24]. When replicating each technique, parameters were selected using unbiased methods, or otherwise noted as biased, to ensure fair comparison and practical application. We demonstrate new applications of the two scatterer size estimation techniques by using them to evaluate the combined thickness of epineurium and perineurium tissue, and to differentiate adipose tissue from neural tissues. The scatterer size estimation techniques were also used to evaluate the fibre diameter distribution of nerve fibres within fascicles. The results help establish the abilities and li
…(Full text truncated)…
This content is AI-processed based on ArXiv data.
