2.1.

Theory

The interferometric expression of the detected spectral signal of the A-scan, $I(k)$, for a single moving scatterer acquired from a spectral-domain OCT scanner, can be formulized as

The inverse Fourier transform of the backscattered light interferogram is a function of the sample depth.³⁴ Performing an inverse Fourier transformation along $k$ to the detected interferogram in Eq. (1), the depth-dependent signal $I(z)$ of the sample can then be reconstructed as

To implement the enhanced STdOCT, a time-series signal expressed in Eq. (2) is acquired. Performing the Fourier transformation to the time-series signal in Eq. (2) along the time $t$, the resulting frequency function of the pixel at depth $z_0$ can be expressed as

Equation (3) can be simplified as

Equation (6) provides the first explanation how the center frequency and the axial velocity are related to each other.

2.2.

Experimental Setup

OCT scans were acquired using an upgraded TELESTO II scanner (Thorlabs Inc., New Jersey, United States) to demonstrate the OCTA methods on healthy human skin and human burn scars, in vivo. This system has a center wavelength of 1300 nm with an axial and lateral resolution of 5.5 (in air) and $13\mu \mathrm{m}$ (as defined by the vendor), respectively. In addition, the system offers several A-scan rate options, ranging from 10 to 146 kHz, with a sensitivity of 109 dB at 10 kHz and 91 dB at 146 kHz (according to the vendor specifications).

A custom-made silicone flow phantom was fabricated by mixing Elastosil P7676A and P7676B fluid (Wacker Chemie AG, Germany) with titanium dioxide in a 3D-printed resin container. The container had 2 holes in the sidewalls to hold a small glass cylindrical capillary (outer diameter: $330\mu \mathrm{m}$; inner diameter: $200\mu \mathrm{m}$) to mimic a blood vessel. After curing, the capillary was embedded in the solid silicone matrix that mimicked the static tissue. The scattering properties of the silicone matrix were adjusted by tuning the ratio of titanium dioxide to Elastosil P7676A and P7676B, which was $\sim 1:250:250$ in weight. The phantom had an attenuation coefficient that approximately matched the attenuation of healthy human skin.^11,35 To mimic blood flow, the capillary was connected to a syringe filled with a polystyrene microsphere (diameter: $0.5\mu \mathrm{m}$) aqueous suspension (Polysciences, Inc., Pennsylvania, United States). The syringe was connected to a pump (Fusion 200, Chemyx Inc., Texas, United States) to control the flow speed. The phantom was fixed on an angle-adjustable stage to allow easy adjustment of the flow angle relative to the OCT beam.

In vivo scans were collected from the arms or legs of 10 human subjects (including 1 healthy subject and 9 patients) with approval from the Human Research Ethics Committee of The University of Western Australia (RA/4/1/4131) and South Metropolitan Health Service Ethics Committee WA (RGS0000004980). Written informed and voluntary consent was obtained from all subjects prior to OCT imaging.

Two types of in vivo skin OCT scans were acquired from each patient. The first was a 3D scan with two repeated B-scans at each location across a field of view (FOV) of $6.0\times 3.6\mathrm{mm}$ or $6.0\times 4.5\mathrm{mm}$. This 3D scan was processed with the conventional speckle decorrelation method,^16,28,36 to generate a map of the microvasculature. The resulting 3D image facilitated the selection of vessel locations for subsequent 2D data acquisition. The second type of scan was a 2D scan, in which 1000 A-scan locations were scanned over 6 mm with 1000 repeated A-scans acquired at each A-scan location. This 2D scan was then processed using both DOCT and enhanced STdOCT methods. To mitigate the effects of bulk tissue motion during in vivo data acquisition, an imaging spacer was attached to the skin surface to couple the OCT probe and the skin.³⁷ Ultrasound gel was applied to the skin surface for refractive index matching, which reduced the surface reflection and imaging artifacts.³⁸ The dataset comprised 27 groups of scans, including 3D scans and 2D scans located at the scar, the boundary between the scar and healthy skin, and healthy skin located $\sim 10\mathrm{cm}$ away from the scar area. The acquisition of scans at the scar and healthy skin boundary was motivated by the aim of validating the potential to detect scar edges. This validation was performed by investigating whether the features of blood vessels, as observed in burn scars and healthy skin, could also be detected at the burn scar boundary. The burn types included electric shock, contact, laser (for wart treatment), and open flame burns, and were located on the arm or leg of the patients. The age of the patients ranged from 19 to 60 years.

Blood flow in microvessels with a diameter ranging from $\sim 10\mu \mathrm{m}$ to more than $500\mu \mathrm{m}$ has been reported to have flow speeds ranging from $\sim 100\mu \mathrm{m}/\mathrm{s}$ to more than $10\mathrm{mm}/\mathrm{s}$.³⁹ Considering this flow speed range, the OCT scanner operated at 10 kHz for all in vivo 2D scans, and the acquisition of one in vivo 2D scan took $\sim 109\mathrm{s}$. The velocity measurement range by DOCT and enhanced STdOCT is $-2.2$ to $+2.2\mathrm{mm}/\mathrm{s}$ (assuming a refractive index of 1.5 for the blood),⁴⁰ which generally matched the expected range of blood flow velocity in the axial direction.

2.3.

Data Processing

The TELESTO OCT system acquired a spectrum for each A-scan, which was linearly resampled in $k$-space before performing a Fourier transformation to generate complex OCT depth signals for all scans in this study. Complex OCT signals from repeated A-scans were utilized to form a time-series signal at each depth. Zero-padding was applied to the time-series signal before the Fourier transformation to achieve higher resolution in the resulting frequency domain signal. A Gaussian window was then multiplied with the signal to minimize sidelobe impact during the Fourier transform.

In accordance with Eq. (5), a narrow Gaussian band-rejection filter (center frequency: 0 Hz; full width at half maximum: 3% of the full frequency range) was applied to eliminate the Dirac delta function at 0 Hz from the frequency spectrum. The non-flow velocity signal displays symmetry about the $y$-axis in both the positive and negative frequency spectra. By computing the difference between the positive and negative spectra, we can effectively attenuate the non-flow velocity signal, thereby increasing SNR of the flow signal.

Two techniques were evaluated to determine the center frequency of the Gaussian-shaped function that represents the flow in the processed frequency spectrum. The first technique was curve fitting, which involved fitting a Gaussian function to the filtered spectrum. Through this technique, the center frequency of the Gaussian-shaped function was identified. The processing time for 1 depth pixel was $\sim 0.11\mathrm{s}$, totaling 110 s for one B-scan. The second method was peak finding, which directly located the maximum amplitude in the processed spectrum and considered its corresponding frequency as the center frequency. This method was faster, taking only ~0.10 s to process one B-scan. When applied to the same scans, both methods produced nearly identical results, with a difference in the measured frequency of $<4\%$. The center frequency was then used to calculate the axial flow velocity according to Eq. (6) while the positive or negative frequency range indicated the flow direction.

When a comparison was made between enhanced STdOCT and DOCT methods, the axial flow velocity measured by DOCT was calculated using the equation $v(x,y,z)={\lambda }_0\mathrm{\Delta }\phi (x,y,z)/4\pi nT$,⁴¹ where ${\lambda }_0$ is the center wavelength, $\mathrm{\Delta }\phi (x,y,z)$ is the phase difference located at coordinate $(x,y,z)$, and $T$ is the time interval between A-scans. An example flow velocity image created using the enhanced STdOCT is shown in Fig. 1(b). Subsequently, an ellipse-fitting algorithm was applied to identify the boundaries of individual blood vessels in the diagram, as shown in Fig. 1(d). Each ellipse represents a cross-sectional view of a vessel, and the number of ellipses corresponds to the number of vessels within the imaged area. It can be assumed that blood vessels are shaped like round tubes. When an OCT image is taken, it captures a slice of the round tube, which appears as an ellipse when the slice is not perpendicular to the blood vessel. The minor axis of the ellipse (the shorter diameter) is equal in length to the diameter of the circle. By measuring the minor axis of the ellipse, the diameter of the blood vessel can be determined. In addition, by analyzing the OCT image in Fig. 1(a), the position of the skin surface can be determined based on the region of the signal with the highest change in intensity of OCT signal along the depth. The depth of each blood vessel can be determined as the distance between the fitted ellipse and skin surface. It is important to note that due to signal attenuation, only vessels located from the skin surface to a depth of 1 mm (in optical path length) will be counted and analyzed.

Fig. 1

Flow chart illustrating the process of extracting blood vessel information. (a) OCT cross-sectional image; (b) OCTA image processed by enhanced STdOCT using the data as in panel (a); (c) OCT cross-sectional image with detected surface; (d) OCTA image with the blood vessels marked with ellipses; and (e) final result: skin surface, vessel locations, dimensions, and flow velocities. Scale bars: $200\mu \mathrm{m}$.

The speckle decorrelation for the in vivo 3D skin scans was calculated with a previously reported method¹¹ with a window size of $3\times 3\text{pixels}$, corresponding to a depth and width of 11 and $18\mu \mathrm{m}$, respectively. The decorrelation was further weighted by the corresponding original logarithmic OCT signal amplitude to remove the artificially high decorrelation signal created by the noise in the OCT signal. Prior to weighting, the logarithmic OCT signal was thresholded by an empirically chosen threshold of 10 dB. The vessels from the skin surface to a depth of $\text{600\hspace{0.17em}\hspace{0.17em}}\mu \mathrm{m}$ into the skin were then projected for visualization of the microvasculature using the weighted decorrelation.

3.1.

Phantom Results

Figure 2 shows the flow signal generated by applying enhanced STdOCT to OCT scans from the phantom. As explained in Sec. 2.1 [Eq. (5)], the spectrum of the time-series signal consists of a Dirac delta function and a Gaussian-shaped function for a flow region, in contrast to a static region, where only a single Dirac delta function exists. An example of such a comparison is shown in Fig. 2(a), taken from a static region outside the capillary (blue) and a flow region (135 deg and $20\mathrm{mm}/\mathrm{s}$) at the center of the capillary (green). The magnitude of the spectra across the detectable frequency range ($-73$ to 73 kHz) was normalized by dividing the original magnitude by the total power of the original time-series signal. The normalized magnitude in Fig. 2(a) shows the total signal power percentage [the vertical axis in Fig. 2(a) is cropped to display a range from 0 to 0.16]. Both spectra in Fig. 2(a) present a Dirac delta function with a width $<0.5\mathrm{kHz}$ centered at 0 Hz, referred to as the static peak hereafter. The maximum magnitude of the static peak in the static region is more than 0.98 as shown in the inset of Fig. 2(a). In the flow region, a Gaussian-shaped function is observed corresponding to the flow, which is referred to as the flow peak. The blue and green spectra significantly overlap, except in the frequency range surrounding the static and flow peaks. The center of the flow peak from the flow region is located at $\sim 42.6\mathrm{kHz}$, identified through curve fitting (dashed line). An axial flow velocity of $19.8\mathrm{mm}/\mathrm{s}$ at the center of the capillary is obtained by Eq. (5), which agrees well with the set flow velocity. In the experiment, the average flow velocity of $20\mathrm{mm}/\mathrm{s}$ was controlled by the pump while the refractive index of the aqueous suspension is assumed to be 1.4, and the flow is assumed to be laminar. There are two possible reasons why the measured velocity is slightly lower than the theoretical velocity. First, the measured velocity may represent the average velocity of a small area at the center of the capillary, as the resolution is not infinitely small. Alternatively, the actual flow angle could be slightly smaller than the set angle.

Fig. 2

(a) Comparison of spectra from a static region (in blue) and flow region (in green) with a Gaussian fit to retrieve the peak induced by the flow centered at 42.6 kHz for the flow region (dashed red); (b) spectra from flow regions with different axial flow velocities; (c) velocity distribution along the depth across the capillary with enhanced STdOCT (blue), DOCT (red), and pump setting (black); and (d) velocity comparison between enhanced STdOCT (blue) and the DOCT (red) from all phantom scans. The dashed line indicates equal flow velocities for the OCT flow measurement (with enhanced STdOCT or DOCT) and pump setting. The error bar represents the standard deviation over 60 experiments.

The ability to measure both the flow velocity and direction in the phantom scans is further demonstrated by the spectra in Fig. 2(b). These spectra were obtained from four different flow regions in the phantom with the same average flow speed but different flow angles. The two spectra (yellow and red), with the flow peaks located in the positive frequency range, are respectively from the center and side of the capillary. The pump controlled the average flow speed at $20.0\mathrm{mm}/\mathrm{s}$, and the flow angle was set at 135 deg. The laminar flow model was used to estimate the flow velocity of 21.5 and $8.0\mathrm{mm}/\mathrm{s}$ for these 2 flow regions, based on the average flow speed and flow angle. Enhanced STdOCT demonstrated accurate measurements, with recorded flow velocities of 21.7 and $7.9\mathrm{mm}/\mathrm{s}$, as shown in Fig. 2(b).

Similarly, the two spectra (green and brown) with negative frequency peaks are from the center and side of the capillary, with an average flow speed of $20.0\mathrm{mm}/\mathrm{s}$ and a flow angle of 45 deg. First, the flow peaks are in the negative or positive frequency range, respectively, when the flow angle is smaller than 90 deg (i.e., axial flow has the same direction as the OCT beam direction) or larger than 90 deg (i.e., axial flow direction is opposite to the OCT beam direction). This feature helps to resolve the flow direction, similar to DOCT. Second, a faster axial flow results in a larger shift of the flow peak relative to the static peak in Fig. 2(b). The center frequency of the flow peak is proportional to the set flow speed and helps to quantify the axial flow velocity based on Eq. (6).

Figure 2(c) shows a comparison of the flow velocities measured by enhanced STdOCT and DOCT. The graph in Fig. 2(c) shows the flow profile as a function of depth across the capillary. The dashed black curve is the theoretical laminar flow profile, estimated from an average speed of $20.0\mathrm{mm}/\mathrm{s}$ with an angle of 135 deg in a straight tube with a circular cross-section.⁴² Both flow profiles determined by enhanced STdOCT and DOCT correspond well with a theoretical parabolic laminar flow profile. The enhanced STdOCT better matches the theoretical values, especially at the locations close to the center of the capillary. Measuring the velocities within the capillary at each depth from phantom scans with 4 average speeds (0, 5, 10, and $20\mathrm{mm}/\mathrm{s}$) and 5 flow angles (45 deg, 60 deg, 90 deg, 120 deg, and 135 deg), Fig. 2(d) provides a summary of the flow measurement accuracy compared to the velocities calculated from the pump setting. The velocities measured with enhanced STdOCT are more accurate and stable, as demonstrated by the smaller standard deviation in Fig. 2(d), compared to those measured with DOCT.

Figure 3 shows the results from a 2D phantom scan. Figure 3(a) shows the cross-sectional OCT B-scan, where the capillary wall shows an extremely low signal (black) due to its transparency. The flow region is located inside the capillary wall while the static silicone matrix is located outside of it. Figure 3(b) shows the flow velocity image generated with enhanced STdOCT from the same area shown in Fig. 3(a). Figure 3(c) provides a 3D perspective of Fig. 3(b), with the flow velocity encoded in both color and height. The laminar flow structure within the capillary is distinctly visible in Fig. 3(c), demonstrating the high resolution and accuracy of this method.

Fig. 3

(a) and (b) OCT B-scan and the resulting flow velocity image of the flow phantom, respectively; (c) 3D view of the flow profile of panel (b). Scale bars: $100\mu \mathrm{m}$.

3.2.

Normal Skin Results

Figure 4 shows flow imaging of the forearm skin of a healthy 29-year-old subject. The projection of the speckle decorrelation image in Fig. 4(e) shows the distribution of the blood vessels from the skin surface to a depth of $600\mu \mathrm{m}$. Within this OCT FOV, a 2D scans for flow measurement was obtained from the location indicated by the green dashed line. Figures 4(a)–4(d) are the OCT, unweighted speckle decorrelation, enhanced STdOCT, and DOCT images from the location at the dashed line in Fig. 4(e), respectively.

Fig. 4

Images of the forearm skin of a healthy subject, in vivo. (a)–(d) Cross-sectional images of the OCT signal, speckle decorrelation, enhanced STdOCT, and DOCT, obtained from the location of the green dashed line in panel (e); the blue arrows mark three vessels with shadow artifacts in panel (b); the red arrows mark two vessels located under other vessels, mainly visible in panel (c); the green arrows indicate two vessels that are only visible in panel (b); (e) projection of the blood vessels from skin surface to a depth of $600\mu \mathrm{m}$ with a dashed line marking the locations for 2D scanning; (f)–(h) magnified images of the regions outlined in panels (b)–(d), respectively; the green arrows in panel (g) mark two vessels that are not visible in panels (f) and (h). Scale bars in (a)–(e): $500\mu \mathrm{m}$. Scale bars in (f)–(h): $200\mu \mathrm{m}$.

Blood vessels are marked with colored arrows to highlight the different signal features. First, the blue arrows mark three blood vessels with strong shadow artifacts in Fig. 4(b), which are almost eliminated by the enhanced STdOCT method in Fig. 4(c). This observation suggests improved immunity to shadow artifacts for enhanced STdOCT.

The red arrows in the right section of Figs. 4(a)–4(d) mark blood vessels, which are confounded by the shadow artifacts induced by an overlying vessel in the speckle decorrelation [Fig. 4(b)]. This vessel is barely visible but can be identified in the projection image of the vessels in Fig. 4(e) at an intersection of three vessels. In contrast, the flow image in Fig. 4(c) clearly shows this vessel, which has a flow direction opposite to the top vessel. Another two blood vessels marked with red arrows on the left in Figs. 4(a)–4(d) demonstrate a similar case of overlapping vessels with opposite flow directions. These cases show that enhanced STdOCT, due to its high immunity to shadow artifacts, enables flow measurement in deep vessels that are overlapped by superficial vessels, which cannot be visualized with speckle decorrelation.

In the speckle decorrelation image [Fig. 4(b)], several vessels are visible, which are marked by green arrows, but they are hardly visible in the enhanced STdOCT image [Fig. 4(c)]. We speculate that these vessels may be perpendicular to the incident beam, leading to a low axial velocity, which results in a low signal in the enhanced STdOCT image.

Figure 4(d) shows the DOCT image as an alternative for measuring blood flow velocity. The axial velocity and direction were also acquired using the DOCT. However, there is significantly more noise than in the enhanced STdOCT image, particularly in the low OCT SNR region. This further demonstrates that the DOCT requires a higher phase stability.

The region outlined by the dashed yellow squares in Figs. 4(b)–4(d) is magnified in Figs. 4(f)–4(h) with the color maps adjusted to enhance the visibility of these vessels for each method. Green arrows mark 2 vessels with a diameter smaller than $50\mu \mathrm{m}$, and they are poorly visible in speckle decorrelation and DOCT images, but clearly visualized with the enhanced STdOCT method, demonstrating the superior performance of the enhanced STdOCT method for imaging small vessels.

The 3D structure of the microvasculature can be clearly seen in the healthy skin image when speckle decorrelation is used. The speckle decorrelation method cannot measure blood flow speed, but it can help in understanding the blood vessel orientation. Both the DOCT and the enhanced STdOCT methods measure the axial velocity within blood vessels; however, the enhanced STdOCT approach clearly outperformed the DOCT in terms of image quality and the capacity to identify small blood vessels, and its ability to quantify flow in the presence of a shadow artifact.

3.3.

Burn Scar Results

Figure 5 shows images of second-degree burn scars and healthy tissue on the left forearm of a 58-year-old male. A photograph of the burn scar is shown in Fig. 5(a), where in vivo scans were acquired from both the scar, as indicated by the black box in Fig. 5(a), and adjacent healthy tissue. Projection images of the scar and nearby healthy tissue, generated by speckle decorrelation, are shown in Figs. 5(d) and 5(e), respectively. Figure 5(d) shows that the burn scar has a higher blood vessel density, characterized by a greater number of small blood vessels. In addition, large blood vessels within the scar appear more disorganized and tortuous, exhibiting increased curvature. In contrast, large blood vessels in healthy tissue shown in Fig. 5(e) are more extensive.

Fig. 5

In vivo images of a second-degree burn scar on the left forearm. (a) Photograph of the scar with the black dotted box indicating the scan location; (b) and (c) cross-sectional flow images generated by enhanced STdOCT from the scar and adjacent healthy tissue, respectively, as indicated by the green lines in panels (d) and (e); (d) and (e) projection images generated by speckle decorrelation from the burn scar indicated by the black box in panel (a) and adjacent healthy tissue, respectively. Scale bars in (b)–(e): $500\mu \mathrm{m}$.

Cross-sectional flow images from the green lines in Figs. 5(d) and 5(e) are presented in Figs. 5(b) and 5(c), respectively, from the skin surface to a depth of 1 mm. Comparing these images emphasizes that, within the scar area, a number of blood vessels are present in the 0 to $200\mu \mathrm{m}$ range, with some exhibiting larger diameters. The majority of vessels in the scar are concentrated in the 0 to $500\mu \mathrm{m}$ range. In contrast, healthy tissue demonstrates fewer vessels in the 0 to $200\mu \mathrm{m}$ range, with blood vessels evenly distributed from the surface to a depth of 1 mm.

Figure 6 shows a series of images from a third-degree burn scar and healthy tissue on the upper right arm of a 48-year-old female. Figure 6(a) shows a photograph of the scar, with data collected from both the scar, as indicated by the black box in Fig. 6(a), and adjacent healthy tissue. The speckle decorrelation method was used to generate projection images of the vessels in the scar and neighboring healthy tissue, as shown in Figs. 6(d) and 6(e). These images show a higher blood vessel density in the burn scar, characterized by a greater number of large blood vessels aligned vertically. In contrast, large blood vessels in healthy tissue appear more evenly oriented.

Fig. 6

In vivo images of a right upper arm with a third-degree burn scar. (a) Photograph of the scar with the black dotted box indicating the scan location; (b) and (c) cross-sectional flow images generated by enhanced STdOCT from the scar and adjacent healthy tissue, respectively, as indicated by the green lines in panels (d) and (e); (d) and (e) projection images generated by speckle decorrelation from the burn scar indicated by the black box in panel (a) and adjacent healthy tissue, respectively. Scale bars in (b)–(e): $500\mu \mathrm{m}$.

Figures 6(b) and 6(c) exhibit cross-sectional flow images from the green lines indicated in Figs. 6(d) and 6(e), respectively, from the skin surface to a depth of 1 mm. The images indicate that blood vessels in burn scars are concentrated more closely to the skin surface, and the blood vessel density in the burn scar is higher than that in healthy tissue.

A total of 602 blood vessels were extracted from 2D scans from burn scars (including vessels at scar and scar side of the scar margin) and 361 from healthy tissue (including vessels at healthy tissue and healthy side of the scar margin) using the method described in Sec. 2.3. Since blood vessels were obtained in a cross-sectional image, a density unit of blood vessels was introduced as vessels (v) per square millimeter ($\mathrm{v}/{\mathrm{mm}}^2$). However, due to the limited imaging depth of the technique, $\sim 95\%$ of the blood vessels are located within 1 mm of subcutaneous tissue. Therefore, when calculating the density of blood vessels, we excluded vessels that were located beyond a depth of 1 mm. The results show that the density of blood vessels in the burn scar on average was $7.8\mathrm{v}/{\mathrm{mm}}^2$ while it was $6.7\mathrm{v}/{\mathrm{mm}}^2$ in the scar area near the scar edge. The blood vessel density in healthy tissue was $4.4\mathrm{v}/{\mathrm{mm}}^2$ and $4.3\mathrm{v}/{\mathrm{mm}}^2$ in the healthy area near the scar edge, demonstrating a 67% increase in vessel density in the burn scar.

Figure 7 shows the distribution of blood vessels as a function of their depth. In Fig. 7(a), the depth information encompasses all blood vessels, including those within the scar, healthy tissue, and the scar margin. The analysis indicates that, in the case of burn scars, most of the blood vessels are concentrated in the superficial layer. As the depth increases, their number significantly reduces. This decrease is attributed to a combination of factors, including the existence of larger blood vessels at deeper depths. As a result, with a consistent cross-sectional area, the presence of these larger vessels contributes to an overall reduction in the number of vessels observed. Moreover, it should be emphasized that this initial decrease is further compounded by the effect of signal attenuation at greater depths. In contrast, healthy tissue shown in Fig. 7(b) displays a relatively uniform distribution of blood vessels throughout the depth, with only a reduction in their number at depths below $700\mu \mathrm{m}$. This decrease is mainly caused by the difficulties created by image signal attenuation at these deeper levels. However, the distribution of blood vessels in the burn scar area was different, with a significantly greater number of vessels observed at a superficial depth of 100 to $600\mu \mathrm{m}$ compared to healthy tissue. This observation suggests that scar tissue may exhibit increased superficial vascular growth. Figure 7(c) shows that at the boundary between healthy tissue and burn scar, the healthy tissue exhibits a higher concentration of blood vessels at a depth of 200 to $300\mu \mathrm{m}$. This is possibly due to the presence of blood vessels in the scar that is more densely distributed in the depth range of 100 to $500\mu \mathrm{m}$.

Fig. 7

Histograms of blood vessel depth. (a) Distribution of blood vessel depths in the tissue, including vessels within the scar, healthy tissue, and the scar margin; (b) comparison of blood vessel depth distributions in scar and healthy tissue; and (c) blood vessel depth distribution at the scar margin.

The distribution of vessel sizes is shown in Fig. 8. The histograms in Figs. 8(a) and 8(b) show a dense distribution of blood vessels in healthy tissue, centered at 20 to $25\mu \mathrm{m}$ for the blood vessels in healthy tissue. In contrast, the blood vessels in the scar tissue are more concentrated below $30\mu \mathrm{m}$, with a higher density of vessels in the 15 to $20\mu \mathrm{m}$ range. Specifically, the number of blood vessels in the 15 to $20\mu \mathrm{m}$ range is 130% higher in scars compared to healthy tissue. The scar tissue has more vessels larger than $80\mu \mathrm{m}$ while in healthy tissue, vessels larger than $80\mu \mathrm{m}$ were rare. In addition, the number of blood vessels larger than $80\mu \mathrm{m}$ increased significantly in the healthy tissue at the scar boundary, as seen in Fig. 8(c). It should be noted that the minimum threshold for vessel size was set at 3 pixels in diameter, equivalent to $\sim 11\mu \mathrm{m}$, to mitigate the impacts of noise. This setting was made based on observations during the blood vessel analysis in the B-scan. Certain signals appearing to be vessels in the B-scan did not always correspond accurately to the actual vessels in projection view. These variations were attributed to noise, often appearing as one or two pixels. Applying this threshold effectively removed the noise artifacts and enabled the clear identification of blood vessels. Figure 8(d) shows the relationship between blood vessel size and flow velocity. As shown in the Fig. 8(d), the scar area exhibits the highest flow velocity across different vessel sizes. This suggests that blood flow velocity in the scar area is faster than in healthy tissue.

Fig. 8

(a) Size distribution of blood vessels in the tissue, including vessels within the scar, healthy tissue, and the scar margin; (b) comparison of blood vessel size distributions in scar and healthy tissue; (c) blood vessel size distribution at the scar margin; and (d) relationship between blood vessel size and flow velocity.

Figure 9 shows the distribution of blood vessel flow velocity. Figure 9(a) shows the axial flow velocity of all blood vessels, including those within the scar, healthy tissue, and the scar margin. The analysis reveals that the majority of blood vessels exhibit flow velocities ranging from 0.06 to $0.20\mathrm{mm}/\mathrm{s}$. The average flow velocity in the scar area was $0.15\mathrm{mm}/\mathrm{s}$, representing a 7.1% increase compared to the average flow velocity in healthy tissue ($0.14\mathrm{mm}/\mathrm{s}$).

Fig. 9

Histograms of blood flow speed. (a) Distribution of blood flow speeds in the tissue, including vessels within the scar, healthy tissue, and the scar margin; (b) comparison of blood flow speed distributions in scar and healthy tissue; and (c) blood flow speed distribution at the scar margin.

Figure 9(b) shows that the flow velocity in the scar region is 25% higher compared to the healthy tissue area. Notably, the scar region exhibited a number of blood vessels with a flow velocity $>0.3\mathrm{mm}/\mathrm{s}$. In contrast, healthy tissue did not show any blood vessels with a velocity exceeding $0.5\mathrm{mm}/\mathrm{s}$. These findings suggest that the scar tissue exhibits an increased average flow velocity and increased number of blood vessels with high blood flow velocity.

In Fig. 9(c), we observed that at the boundary, the average velocity in healthy tissue was 6.7% greater than that in scars. This difference can be attributed to the uneven skin surface observed at the scar border, which leads to a more parallel distribution of blood vessels in the measurement direction. As a result, a relatively larger axial flow velocity is observed. These findings indicate that scars have a distinct vascular network that affects the blood flow velocity and distribution in comparison to healthy tissue.