2.1.

Imaging and Stimulation System

A custom-developed polarization-sensitive optical coherence tomography (PSOCT) imaging and stimulation platform, shown in detail in Fig. 1, was developed and optimized to observe and characterize PNS tissue during and after recovery from electrical stimulation. The system used a near-infrared (1310 nm) swept-source laser focused into tissue to generate depth-resolved optical scattering profiles. A 110-nm bandwidth swept-source laser (HSL-200-50LC, Santec Corp., Japan) provided an axial and lateral resolution of $\sim 7$ and $20\mu \mathrm{m}$ (in air). Three-dimensional imaging was achieved by scanning the imaging light across tissue using a pair of galvanometer scanners. A green tracking laser was added to help operators visualize the scan location incident on tissue. To measure birefringence independent of optical axis, this Jones matrix-based system modulates the input polarization state of light on alternating A-lines³⁶ using a custom-fabricated semiconductor-based polarization modulator (Boston Applied Technologies, United States). OCT-A scans are simultaneously collected with PSOCT data by taking five sequentially repeated scans at each physical location. All OCT imaging was performed without the use of exogenous stains or dyes, and each dataset was collected in $\sim 2\min$.

Fig. 1

Optical coherence tomography (OCT) imaging: full system layout. This custom-built swept-source polarization-sensitive OCT system was designed to capture volumetric scans of the sciatic nerve during electrical stimulation. This system noninvasively extracts tissue features (structure, birefringence, and perfusion) without stains or dyes. (a) Nerve stabilizer 3D-model and (b) stabilizer with sciatic nerve in situ. Deinsulated portions of the electrodes are noted with yellow bars, nearest to the rightmost fascicle.

A custom-designed 3D-printed nerve stabilizer was modified from the previous work³⁴ to include PSOCT calibration phantoms^37–39 and embedded stimulation electrodes [Figs. 1(a) and 1(b)]. The stabilizer supports the nerve during imaging to reduce motion artifacts from breathing and electrical stimulation. The pair of phantoms, further described in Fig. S2 in the Supplementary Material, serves as external markers to calibrate the optic axis of the imaging system at 0° and 45° with respect to each other. These fixed external references allow measurements to be directly compared to one another, even when collected on different days or even different imaging systems. For stimulation, two Teflon-coated platinum/iridium hemi-cuff stimulation electrodes were installed (Microprobes for Life Science, United States). Both electrodes had a diameter of $\sim 0.05\mathrm{mm}$, with deinsulated regions of 0.2 mm (positive) and 0.4 mm (negative) within the nerve stabilizer holder. A MATLAB-controllable precision AC/DC current source (6221, Keithley Instruments, United States) delivers customizable stimulation waveforms. To reduce the appearance of bands of Fontana, a rodent leg extender (RLE) was created. The RLE, discussed in further detail in Fig. S3 in the Supplementary Material, gently and safely applies force around the ankle using a rubber band to pull the leg and sciatic nerve into its naturally extended position. A nearby syringe pump administered sterile saline ($0.05\mathrm{mL}/\min$) over the course of experiments to keep tissue hydrated. Finally, a sterile, thin piece of plastic film, approximately the dimensions of the nerve channel, was placed over the nerve to reduce back reflections and prevent dehydration effects over the imaging period. Custom-develop GPU-based processing and polarimetry-reconstruction code⁴⁰ was developed to process raw data, account for system-specific noise, and accurately reconstruct tissue morphology in PSOCT images. A more complete description of the imaging and stimulation platform design, specifications, and testing is available in the Supplementary Material.

From the OCT signal, six distinct imaging outputs (channels) were calculated from each volumetric acquisition. Together, these six channels quantify a range of tissue optical properties. Imaging metrics that characterize the optical and birefringent properties of tissue and vessel perfusion were compared across treatment groups during stimulation and after recovery. The six channels are illustrated in Fig. 2. Birefringence-weighted optic axis (BwOA) is derived from a combination of phase retardation and optic axis³⁹ and provides a color-coded visualization of the vectorial birefringence, effectively the orientation of the microstructure of tissue. Angiography (OCT-A) is a tissue vascular map and is extracted from temporal signal fluctuations due to blood flow. Reflectance data, the typical structural representation of OCT data, derives contrast from the scattering magnitude of various tissues. Phase retardation and optic axis are both related to the birefringent or optical polarization properties of tissue. Finally, the degree of polarization uniformity (DOPU) measures the consistency of tissue properties as the imaging beam propagates through tissue. In aggregate, these channels are sensitive to both the structural, physiological, and birefringent optical properties of tissue.

Fig. 2

Platform output: processed PSOCT data, animal #2: SC, D7. Representative data from a rat sciatic nerve as observed in this study of the six data channels captured. En face representations and cross-sectional scans for each channel are shown. BwOA, birefringence-weighted optic axis. Scale bar: $500\mu \mathrm{m}$.

2.2.

Experimental Design and Protocol

Animal work performed in this study was approved and conducted with oversight by the U.S. FDA (White Oak campus) Institutional Animal Care and Use Committee. 15 female Lewis rats ($279\pm 12\mathrm{g}$, Charles River Laboratories, Inc., United States) were acquired for this study. Upon arrival, all subjects were allowed a 5-day acclimation period with 12 h day/night cycles before experimental use.

Figure 3 shows an overview of the experimental protocol used for this study. Surgery and imaging were performed by the same team to limit experimental variance. The animals ($n=15$) were equally and randomly assigned into three groups: sham control (SC, $n=5$, $0.0\mathrm{mA}/0\mathrm{Hz}$), stimulation level 1 (SL1, $n=5$, $3.4\mathrm{mA}/50\mathrm{Hz}$, Shannon $k=2.57$), and stimulation level 2 (SL2, $n=5$, $6.8\mathrm{mA}/100\mathrm{Hz}$, Shannon $k=3.17$). For stimulation, two Teflon-coated platinum/iridium hemi-cuff stimulation electrodes were installed into the nerve channel (Microprobes for Life Science, United States). Both electrodes had a diameter of $\sim 0.05\mathrm{mm}$, with deinsulated regions of 0.2 mm (positive) and 0.4 mm (negative) within the nerve stabilizer holder. The stimulation waveform used here is a cathodic-first charge-balanced square wave ($100\mu \mathrm{s}$ pulse width, $400\mu \mathrm{s}$ interphase delay), with damage thresholds identified using the Shannon model.⁴¹ Shannon $k$ values above $\sim 1.85$ have been shown to cause damage,¹² with higher values producing more severe damage. For the SC group, the output of the stimulation system was connected via electrodes to the animal, but the output was set to $0.0\mathrm{mA}/0\mathrm{Hz}$.

Fig. 3

Surgical and imaging experimental protocol. Day 1 began with a baseline assessment of nerve function, followed by sciatic nerve exposure surgery. Afterward, the animal was placed into the imaging setup for a 30-min acclimation period (PREP). A baseline (B) OCT scan of the nerve was captured before stimulation. Stimulation was active for 1 h (ON) followed by a 1-h recovery (OFF). Simultaneously, OCT scans of the stimulation region were collected every 30 min (during stimulation: S1, S2; and recovery: R1, R2). On day 7, nerve assessment tests were repeated, and the left sciatic nerve re-exposed for imaging. OCT scans were collected using a tiled collection strategy to ensure capture of the stimulation region from day 1. Nerve exposure and tiled imaging was repeated on the contralateral (right) leg to serve as an internal control. Finally, the sciatic nerves from both legs were harvested for histological analysis.

On day 1 (D1), a functional baseline nerve assessment for each animal was collected immediately prior to experimentation using both walking track⁴² and Von Frey analyses.⁴³ To collect a measure of gait, specifically through sciatic function index (SFI) and tibial function index (TFI), each animal was placed onto a clear plexiglass platform (FreeWalk, CleverSys, United States) and restricted within an $8\mathrm{cm}\times 40\mathrm{cm}$ rectangular area.⁴⁴ At least five continuous steps for each foot were recorded using a digital video camera. Afterward, animals were transferred to a plexiglass enclosure with a perforated metal floor ($25.4\mathrm{cm}\times 25.4\mathrm{cm}$) for a Von Frey assay. Animals were acclimated within the setup for 5 min, and then a rigid filament esthesiometer was used to measure 5 trials on both hind limbs by applying centered point-stimulation on the plantar surface of the paw. Forces resulting in a clear paw withdrawal were recorded for each foot.

Animals were then anesthetized and prepared for surgery. The complete surgical method is described in the Supplementary Material. During surgery, animals were thermally stabilized, fully anesthetized, and hydrated during and after the procedure. After surgery, animals were transferred to the imaging and stimulation platform and loaded into the RLE and nerve stabilizer (see Fig. S3 in the Supplementary Material for more details) for a 30-min acclimation period. Animals were stimulated for 1 h, following the parameters specified for each group, then allowed to recover for 1 h. Simultaneously, PSOCT volumes of the stimulation region were collected, starting with a baseline scan collected just prior to stimulation, and then every 30 min thereafter. The second scan was taken at 55 min to avoid scan acquisition during the stimulation system shutoff at 60 min, which can otherwise lead to movement artifacts. A digital photograph of the nerve within the holder was taken immediately prior to the baseline scan and after recovery to assist with nerve identification. Animals were returned to the surgical microscope field to close the incision and recover from anesthesia. Postsurgical antibiotics and analgesics were given to each animal.

On day 7 (D7) after the recovery period, functional assessments were repeated, and the left sciatic nerve was surgically re-exposed. Overlapping PSOCT volumes of the nerve were collected to capture the original stimulation region and surrounding nerve tissue. Capturing 4 to 5 tiled volumes ensured acquisition of the stimulation region and aided in co-registration between D1/D7. Digital photographs of the nerve at each imaging location were also taken. After imaging the left leg, sciatic nerve exposure surgery followed by the D7 imaging protocol was repeated on the contralateral (right) leg. Finally, the sciatic nerves from both legs were harvested for histological processing and analysis.

2.3.

Histological Processing

Each nerve was separated into three segments, relative to where stimulation was applied and placed into separate vials with a fixative agent for histology. Distal and proximal segments ($\sim 2$ to 3 mm length portions) from the stimulated nerve region (e.g., closer to hip or limb extremity) were outside of the OCT scan area. The median segment was $\sim 7$ to 10 mm in length and known to contain the stimulated nerve region with confirmed representation in PSOCT image data. Using surgical photographs and OCT en face reconstructions, the stimulation region was identified in the median segment and further sectioned prior to paraffin embedding. Median segments, after this sectioning, were $\sim 4$ to 5 mm. Median segments were fixed in 4% paraformaldehyde, rinsed in phosphate-buffered saline (PBS), embedded in paraffin blocks (ASPS300S, Leica Biosystems), and manually sectioned into $5\mu \mathrm{m}$ sections. Two slides, each containing three sections, were taken from the median segment for immunohistochemistry (IHC). One set of cross sections was stained with hematoxylin and eosin (H+E) for coarse analysis.

For IHC, sections were put into a 60°C oven for 30 min, rehydrated in 100% xylene and several ethanol solutions (100%, 90%, and 70%), and incubated in citrate buffer antigen retrieval solution. Afterward, slides were incubated in a 4% blocking solution of goat serum (Thermo Fisher Scientific) and washing solution [0.5% Triton X-100 (Sigma) in PBS] for 1 h at room temperature and then incubated overnight with primary antibodies at 4°C. Myelin protein zero (P0) and $\beta$-tubulin-III were used as primary antibody targets to observe the myelination and axon changes in the nerve. The primary antibodies used were rabbit anti-P0 (1:150; Millipore) and mouse anti-β-tubulin-III (1:150; Sigma) immunoglobulin G1 antibodies (IgG1). Slides were then rinsed and incubated in a secondary antibody for 1 h at room temperature. These steps were carried out in the dark to prevent photobleaching of the secondary antibodies. Fluorescent secondary antibodies used were goat anti-mouse-IgG1 (1:100; Jackson ImmunoResearch) and goat anti-rabbit-IgG (1:100; Jackson ImmunoResearch). These secondary antibodies were tagged with AlexaFluor 488 and AlexaFluor 594, respectively. After a final incubation, excess washing solution was removed, and slides were mounted with gold antifade mount containing DAPI (Thermo Fisher). DAPI was used as a nuclear counterstain. Slides were then imaged on an FV 1000 confocal microscope (Olympus) at $10\times$ magnification.

2.4.

Data Extraction and Analysis Methods

2.4.2.

PSOCT data processing, co-registration, and analysis

The data organization, processing, and ROI selection are visualized in Fig. 4. With 6 channels, 15 animals, and 6 timepoints ($n=6\times 15\times 6$), the data library was first organized then co-registered across imaging timepoints. The optimal D7 tile was carefully chosen that best overlaps with scans from D1 using structural cues from the angiography map and overall fascicle structure, ensuring the stimulation region from D1 is fully visible in the D7 tile. Four of the six channels were chosen for in-depth analysis: attenuation maps (calculated from reflectance data⁴⁵), phase retardation, DOPU, and optic axis [top to bottom, Fig. 4(a)]. These contrast mechanisms have been successfully used in other studies to distinguish between healthy and diseased tissues.^46–49 For nervous tissue, electrostimulation-induced damage to myelination is of primary interest.

Fig. 4

Data extraction methods. (a) Data channels were organized and co-registered across each imaging timepoint for analysis (from top to bottom: intensity, phase retardation, DOPU, and optic axis). (b) A 3D representation of nerve volume is plotted using BwOA channel. ROI selection: five analysis regions within the stimulation region were selected to extract representative frames. At each location, 25 frames were averaged to generate a resultant frame. An arithmetic or coherent mean (CM) was used to ensure data integrity during processing. For each of the five resultant averaged frames, ROIs were manually segmented using histology and PSOCT image morphology to select only the fascicle (teal) and not the epi- or perineurium (red). From each ROI, a mean value for each fascicle can be extracted and group statistics computed. (c) To analyze angiography data, the full en face dataset from each timepoint was processed to extract both vessel diameter and density metrics. Changes in vessel diameter (top) were calculated by manually selecting at least 30 segments per timepoint, extracting the mean FWHM. To compute vessel fraction (bottom), a coverage mask was computed. A ratio (%) of vessel to total (vessel + background) pixels is calculated for each frame.

To capture ROIs representative of the data and to limit selection bias, five analysis regions in four data channels ($n=5\times 4\times 15\times 6$) were selected for each dataset in Fig. 4(b) (top, dotted boxes) that span the stimulation region within the electrodes of the D1 volume and overlap through all other timepoints. In each of the 5 regions, 25 cross-sectional frames were averaged ($\sim L\times W\times H:100\mu \mathrm{m}\times 2.5\mathrm{mm}\times 2\mathrm{mm}$) into 1 resultant frame [Fig. 4(b), lower left). Depending on the data type/channel, a coherent mean (optic axis^39,50) or standard arithmetic mean was used to maintain data integrity during processing. Within these five resultant frames, fascicle ROIs were then manually segmented based on structural cues cross-referenced from both PSOCT image morphology and histological sections [Fig. 4(b), lower right). Although a more complete extent of the nerve profile can be visualized in the reflectance channel, multiple scattering causes depolarization and degradation in polarization-sensitive information in deeper regions. ROI depth was thus conservatively chosen using PS data to avoid including invalid data or regions beyond the imaging depth of the system, which would detrimentally add noise during analysis. A unique ROI selection for each fascicle and for each timepoint was saved to ensure that any temporal changes in tissue have been captured. Fascicles were labeled based on proximity to the stimulation leads: left, center, and right [L/C/R, shown as green/blue/yellow in Fig. 4(b), lower right], where the rightmost region is closest to the stimulation leads ($n=3\times 5\times 4\times 15\times 6$). Finally, average values were computed using the entire fascicle 2D ROI. The statistics for D7 to D1 values were calculated for each animal and averaged within each experimental group for comparison. This calculation aims to remove animal–animal variability using D1 values as a baseline correction for changes at D7. Generally: $\mathrm{\Delta }\text{metric}={\mathrm{avg}}_{\text{stim}\mathrm{grp}}(\mathrm{D}{7}_{2\mathrm{D}\text{fasc}}-\mathrm{D}{1}_{2\mathrm{D}\text{fasc}}$).

To analyze angiography maps [Fig. 4(c)], volumetric scans were converted into en face vessel map average intensity projections ($\sim 400$ to $600\mu \mathrm{m}$ in-depth to include signal from the entire nerve). Customized MATLAB scripts aligned the en face images using a cross-correlation and peak finding method for optimal co-registration across all timepoints, using major vessels as landmarks. Across D1 timepoints (B, S1, S2, R1, and R2), co-registration was straightforward, as movement was minimized using the nerve stabilizer and RLE. The co-registration of D7 data was challenging but feasible due to some slight rotation and compression/stretching of tissue between timepoints. Regions outside of the co-registered D1 to D7 region were cropped from the stack. Finally, horizontal lines caused by any motion artifacts were reduced using a local vertical profile smoothing method.⁵¹

Vessel diameter values from the dataset were manually extracted using a custom-developed LabVIEW program (National Instruments Inc.).⁵² The software operates on en face OCT-A images and using vessel ROIs set by the operator, extracts the full-width at half-maximum (FWHM) vessel diameter. Thirty vessel measurements were made for each timepoint. Vessel density coverage maps were extracted and quantified using a separate custom-developed LabVIEW program.⁵² This software also operates on en face OCT-A images and applies several filtering steps (scaling, thresholding, and morphological noise reduction operators) within a selected en face vessel density ROI to separately mask vessel and background pixels. The filtering settings and ROI were empirically determined to segment all vessels at all timepoints while also reducing the influence of spurious motion artifacts and noise. Within the density ROI, the ratio of the number of pixels within the vessel mask is compared to the total number of pixels in the density ROI (vessel + background) to provide a measure of vessel coverage (i.e., vessel density). With values extracted from every timepoint for these metrics, experimental changes were calculated by taking the D7 value referenced to D1 as a baseline. The results were averaged within each stimulation group for comparison and plotting.

3.1.

Surgical Results

Animals were closely monitored for 5 days after surgery, with additional analgesia provided for 2 to 3 days. One animal died unexpectedly prior to recovery from surgery. All data from this animal were excluded from the analyses. All other animals recovered without complication.

3.2.

Imaging Results

Representative images from stimulation group level 1 (animal #4) are shown in Fig. 5. The complete dataset collected for this animal, encompassing BwOA D1 and co-registered D7 tiles, is shown in Fig. 5(a). It is possible to observe the calibration phantoms (4A/upper right *, cropped in all other panels), which are captured at the start and end of every scan. The stimulation lead location is marked in yellow on D1, and on D7 they were not active. The BwOA colormap links tissue orientation with color. The use of 0 deg and 45 deg polarization phantoms allowed a precise standardization of the presentation of the optic axis orientation across imaging timepoints and animal subjects.³⁷ Figure 5(b) shows all data channels for the baseline (B) timepoint, whereas Fig. 5(c) shows the co-registered tile from D7 (purple boundary). As visually demonstrated in Fig. 3, overlapping tiles were collected of the nerve on D7 to capture the original stimulation region and surrounding tissue. D7 scans were then co-registered to one another and to D1 scans. Angiography perfusion maps helped to identify the stimulation region from D1, even though some rotation and compression of the nerve occurred between D1 and D7. ROI regions were then carefully drawn to compare the same fascicle and region across timepoints. In this manner, data across these timepoints can be compared and quantified, capturing both temporal dynamics during stimulation on D1, and electrostimulation-induced changes after a 1-week recovery.

Fig. 5

Representative imaging results, animal #4: SL1. (a) (left) OCT scans capture temporal dynamics of the left sciatic nerve during stimulation and recovery. Polarization phantoms (*, red/purple) shown in (a, left series), cropped in all other figures. (a, center): Overlapping scans are co-registered using cues from the structural organization of fascicles and perfusion maps. Tile #3 (purple, solid outline) corresponds best to day 1 scans (teal, dotted outline). BwOA colormap [(a), right] relates color to the physical orientation of tissue subunits. (b) Day 1 baseline, including en face projections of the structure (left), birefringent-weighted optic axis (BwOA, center), perfusion map (right), and cross-sectional scans (below) from all data channels. (c) Day 7 tile #3, co-registered data showing changes after a 7-day recovery. Note: yellow lines mark electrodes and stimulation region (day 1). Orange circles demonstrate one perfusion map visual cue for co-registration. White-dotted lines denote matching locations in cross-sectional data. Cross-sectional scans, from corresponding en face projections: (row 1) structural, BwOA, and angiography/perfusion and (row 2) phase retardation, optic axis, and degree of polarization uniformity. Distal and proximal labels indicate sciatic nerve orientation. Scale bars (white, solid): $500\mu \mathrm{m}$.

This dataset can be examined qualitatively for visual changes across D1 and D7. Across D1 timepoints from baseline to recovery, no changes in the structure or birefringence of the tissue were observed due to SLs used at above-threshold values. No severe or immediate changes to tissue were expected, which is confirmed in the observed data in D1 and compared to more severe stimulus, such as nerve crush models as previously demonstrated.³⁵ At D7, a visible increase in nerve diameter was observed, likely caused by postsurgical and poststimulation swelling in the epineurium. While swelling and increased vascularization were observed overall, the relative extent or lack of swelling and vascularization was a function of the inherent biological variability between the animals. At D7 and within the stimulation region from D1 [Fig. 5(c)], an increase in vessel density was observed, as well as a loss of “BwOA” signal, visible both in the en face and cross-sectional images. This signal loss was seen only in the optic axis and phase retardation channels. At equivalent points along the nerve in other channels, such as reflectance, there was no loss in signal, indicating this is a result of a change in birefringence (tissue) and not because of low SNR, absorption, or other artifacts. Representative 2D cross-sectional images and 1D depth profiles from all three experimental groups are comprehensively presented in Fig. S6 and S7 in the Supplementary Material.

Figure 6 shows final results of OCT metrics separated by stimulation group (and fascicle for phase retardation and attenuation). For any data measuring a change, $\mathrm{\Delta }\text{metric}={\mathrm{avg}}_{\text{stim}\mathrm{grp}}(\mathrm{D}{7}_{2\mathrm{D}\text{fasc}}-\mathrm{D}{1}_{2\mathrm{D}\text{fasc}})$. D7 to D1 differences were averaged for each fascicle and group and two-way ANOVA and Tukey’s tests applied to determine significance. For angiography values, mean values for each timepoint are shown to demonstrate the time-dependent nature of these effects, with repeated measures ANOVA and Tukey’s test applied to determine significance.

Fig. 6

OCT metrics: changes after 1 week. (a) Phase retardation, D7 to D1. The rightmost fascicle nearest the stimulation electrode has the greatest effect, with the SL2 group with a significant reduction in phase retardation average value (SL2, ***$p<0.001$) after recovery. (b) Optical attenuation, D7 to D1. Results show a decrease in mean attenuation in the center and right fascicles in both SL1 and SL2. Left/center/right (L/C/R) denotes fascicle position, with R closest to the deinsulated portion of the electrode. (c) Blood vessel diameter (whole nerve), all timepoints across D1 and D7. Mean vessel diameter results demonstrate the stimulation group values increase in diameter under stimulation as expected and slowly decrease during the recovery period. Notably, a small increase over baseline values persists at 1 week for SL1 and SL2. (d) Blood vessel fraction (coverage, whole nerve), all timepoints across D1 and D7. Vessel fraction results show no appreciable change in the number of vessels over the course of day 1 as expected, though an increase in the total amount of vessels is seen at day 7 (SL1 **$p<0.005$ and SL2 *$p<0.05$). Error bars: ±SD. $\mathrm{\Delta }\text{Metric}={\mathrm{avg}}_{\text{stim}\mathrm{grp}}(\mathrm{D}{7}_{2\mathrm{D}\text{fasc}}-\mathrm{D}{1}_{2\mathrm{D}\text{fasc}}$).

A significant decrease was observed in the phase retardation [Fig. 6(a)] in the SL2 stimulation group compared to SC in the rightmost fascicle. The SL1 group largely follows this trend, although one animal had larger phase retardation values than others, which diminishes the decrease. For optical attenuation [Fig. 6(b)], the right-most fascicle closest to the stimulation lead had the greatest change, although it did not reach statistical significance. Angiography data followed expected trends. During stimulation on D1, vessels dilated across the entire nerve, leading to a measured increase in diameter, but, not density, as seen previously.³⁴ At D7, a small increase from baseline vessel diameter is retained in the stimulation groups, which was unexpected, as well as a large increase in the vessel density at D7 after recovery in all groups, with a larger increase in density in stimulated groups. Additional results and analyses from OCT data (optic axis and DOPU) and the functional assessment tests (walking track, Von Frey, and nerve volume) are shown in the Supplementary Material.

3.3.

Histological Findings and Correlation to PSOCT Results

Representative IHC and H+E images from each group are shown in Fig. 7. In the IHC SC group, the fascicles appeared intact and healthy, with a regular distribution of fibers and blood vessels in each fascicle. In the stimulation groups, there is an increasing amount of disorder, manifested in irregularly shaped myelin sheaths, a reduction in myelination counts (red channel) and an uptick in axon pixel counts (green channel), especially closest to the stimulation lead in the rightmost fascicle, which is magnified and inset in the dotted boxed areas. IHC findings were quantified and plotted in Fig. 8, taking the same averaged D7 to D1 comparison as in Fig. 6, using two-way ANOVA and Tukey’s test. H+E images follow a similar trend to IHC data. The SC group showed intact and regular epineurium and perineurium boundaries. The stimulation groups seemed to have thicker regions in those boundaries due to the growth of additional and increased diameter blood vessels. In the stimulation groups, the increase in blood vessel counts and size was evident over the SC. Finally, the fascicle nearest the stimulation lead, especially in SL2, showed an increasing number of vacuoles with the stimulation group, which is a known marker of nerve injury.⁵³

Fig. 7

Histology co-registration and comparison. Representative IHC and H&E-stained sections from each stimulation group. Insets are expanded views of the fascicle nearest to the stimulation leads (right side). Sections were co-registered with PSOCT data and evaluated for signs of injury. IHC and H&E scale bars represent $125\mu \mathrm{m}$ ($35\mu \mathrm{m}$ inset), acquired at $10\times$.

Fig. 8

IHC image analysis results. Relative structural components are quantified in fluorescent IHC images by summing the fluorescent pixels within a fascicle ROI (D7 to D1) and comparing between stimulation groups. (a) Red pixels (myelin, SL2 **$p<0.005$), (b) green pixels (axons), (c) blue pixels (cell nuclei/DAPI), and (d) the ratio of myelin/axon pixel counts (SL2 *$p<0.05$). Left/center/right (L/C/R) denotes the fascicle position, with R closest to the deinsulated portion of the electrode. Error bars: ±SD. $\mathrm{\Delta }\text{Metric}={\mathrm{avg}}_{\text{stim}\mathrm{grp}}(\mathrm{D}{7}_{2\mathrm{D}\text{fasc}}-\mathrm{D}{1}_{2\mathrm{D}\text{fasc}}$).

Overall, changes observed in OCT and IHC data followed the stimulation intensity. When comparing values from the stimulation group against the SC, there was (SL1/SL2): $+4\%/-309\%$ ($p<0.001$, SL2) average change in phase retardation, and $-79\%/-148\%$ reduction in optical attenuation correlated to a $+1\%/-36\%$ ($p<0.005$, SL2) change in myelin pixels, $-13\%/+29\%$ change in axon pixels, $+20\%/+35\%$ increase in cell nuclei pixels, and $+7\%/-41\%$ ($p<0.05$, SL2) change in the myelin/axon pixel ratio in the stimulated region of the rightmost fascicle.