2.1.

Imaging System

In this study, we used a fiber-based swept source OCT (SS-OCT) system consisting of a Mach-Zehnder interferometer and a balanced detector. The overall specification of the system is described in our previous work.²³ Briefly, a swept laser source (VCSEL laser, Thorlabs, Newton, New Jersey) having a central wavelength of 1300 nm with a 100 nm bandwidth at 10 dB cut-off point was employed in the system, producing an axial resolution of $\sim 12\mu \mathrm{m}$ in water. The sample arm incorporated a stand-alone probe head containing X-Y galvo scanners, a collimator and a microscope objective (LSM03, $5\times$ magnification, Thorlabs) mounted on the probe head [Fig. 1(a)]. A plastic plate with a 25-mm diameter hole was affixed to the probe head using movable space bars to enable adjustment in distance between the objective and the plate. A 1-mm thick flat glass slide (optical transmission: $>90\%$ at 1300 nm) was attached to the plastic plate using double-sided tape and was in contact with the skin surface, which allowed for applying pressure normal to the skin surface during OCT imaging. In this work, the nailfold region was chosen as the tissue site for imaging under pressure because that the region contains well-defined capillary structures parallel to the skin surface, which, however, appears perpendicular to other skin sites.^24,25 The parallel arrangement of microvascular appearance may permit observation of significant flow change in the skin capillary when being pressed. Particularly, the nailfold of the fourth finger was tested to achieve the best visibility because of greater transparency than those of other fingers.²⁵ The fingertip was placed under the probe head and the top surface of the nailfold was carefully positioned in parallel to the glass slide for uniform pressing across the surface area. The fingertip was then stabilized with supporting clay to reduce artifacts caused by involuntary body movement. Ultrasound gel was applied between the nailfold surface and the glass slide as a coupling material to mitigate the refractive index difference.²⁶ The probing beam from the OCT system was incident upon the nailfold surface through the glass slide with an average optical power of $\sim 3.2\mathrm{mW}$ and was focused into the skin [Fig. 1(b)]. Such a sample preparation makes the working distance become 500 μm longer than the nominal one (25.1 mm in air) due to the presence of the glass (average RI: 1.5). The lateral resolution was measured to be 25 μm.

Fig. 1

(a) Photograph of the probe head in the sample arm of the swept source optical coherence tomography (OCT) system and (b) the schematic of the probe head-skin interface.

The A-scan rate of the system was fixed to $\mathrm{100,000}\text{lines}/\mathrm{s}$, with which the system sensitivity was estimated to be 103 dB at the focus spot of the probing beam. The OCT data acquisition was achieved by the X-Y galvo scanning in which the scanning probe beam was tracked with a red laser that was used as an aiming beam. The scan resulted in one volumetric OCT data over the scanned tissue area. From the obtained 3-D data cube, both the tissue structure and corresponding vasculature were extracted simultaneously.

2.2.

Data Acquisition and Processing

For 3-D vascular imaging of the skin tissue with the system, we adopted a scanning protocol of UHS-OMAG in which the 3-D tissue perfusion maps have been successfully achieved with a flow sensitivity down to a capillary perfusion level.¹⁸ In this protocol, the raster scanning was performed to record a B-frame (XZ cross-sectional image) with 256 A-lines for fast B-scan (in $X$-direction) and 2560 cross sections for slow C-scan (in $Y$-direction). This means that the completed scanning produces a 3-D volume data containing $256(X)\times 2560(Y)\times 4096(Z)\text{voxels}$, where 4096 is a sampling pixel number in one A-line. Using the B-frame imaging speed of $100\text{frames}/\mathrm{s}$ (fps), we were able to acquire the 3-D OCT data cube in $\sim 26\mathrm{s}$. Considering a scan region of $2\mathrm{mm}(X)\times 2\mathrm{mm}(Y)$, spatial gaps between adjacent A-lines and inter-B-frames were 7.8 and 0.78 μm, respectively. Because phase measurement is challenging to the SS-OCT system due to sampling or A-scan trigger jitter,²⁷ the flow image processing was conducted with amplitudes of the obtained OCT raw data, calculated after fast Fourier transform of the complex OCT signals. One cross-sectional flow intensity image was extracted from the absolute difference between the subsequent amplitude B-frames as follows:

Through the application of the high-pass filtering on the blood flow calculation, it was found that the scattering signals arising from the high reflective tissue structures still resulted in residuals on the vascular image, degrading the flow image quality. To reduce these artifacts, we carried out additional cmOCT imaging with the same OCT data set. The cmOCT image is based on amplitude correlation between inter-B-frames, which gives relative immunity to the artifacts from the high intensity structural regions compared to the UHS-OMAG.²⁸ Thus, a binary form (1: flow region, 0: flow-free region) of the cmOCT image was used as a mask on the UHS-OMAG image. By overlaying the cmOCT mask on the UHS-OMAG image, it can block the static signals from the flow distribution. Concept and process of this scheme are detailed in a recent paper.²⁸ As a result, the 256 cross-sectional UHS-OMAG images were masked with a corresponding binary cmOCT image, and the masked UHS-OMAG images (named mOMAG images) were displayed with a logarithmic scale. The post-data processing was performed with a laboratory-software coded by a MATLAB language (Ver. 7.6, MathWorks) on a personal laptop (2.9 GHz, 8 GB RAMS). Total processing time to obtain a resultant 3-D vascular map took about 15 min. One en face microangiogram was displayed by projecting voxels with maximum intensity values in the 3-D vasculature.

2.3.

Experimental Design

In order to monitor skin blood flow response to the applied pressure, we performed microvascular imaging of human nailfold with the following experimental protocol; first, a healthy volunteer was in a seated position and allowed to adapt to the environment for 15 min at room temperature (23°C) before the measurement. A baseline measurement was conducted without any stress on the nailfold tissue with the glass slide almost just in touch with the skin surface. The load was then applied through the glass slide by manually moving the probe head forward to the nailfold with a step increment of 50 μm, which was monitored in real time through B-mode OCT imaging at 100 fps. Displacement of the glass slide ranged from 0 to 200 μm relative to its initial position, where 200 μm was an empirically set minimal displacement for complete obstruction of the capillary flow in the nailfold, meaning a total disappearance of the dynamic scattering within B-mode OCT imaging. For each displacement (took about 5 s), 3-D OCT imaging of the pressed nailfold was carried out for 26 s and followed by a rest time of 10 s. After that, the pressure was suddenly released by quickly moving the probing head backward and one OCT measurement was taken right after the pressure release. Then, two postpressure measurements were acquired with 1-min intervals. Subsequently, a total of eight measurements consisting of four sessions (prepressure (baseline), pressure application, pressure release, and postpressure) were taken within a time period of $\sim 6\min$.

2.4.

Quantification of Vascular Perfusion and External Pressure

To characterize the pressure-flow relationship, it is necessary to quantify the vascular perfusion and the load distribution over the nailfold. In this work, the extent of vascular perfusion was represented as a flow index (FI), which can be defined as a ratio of perfusion area to total scanned area given by

Fig. 2

Flow diagram of the procedure for strain calculation. (a) Segmentation of the tissue structure (between red and yellow lines) to be deformed from the cross-sectional OCT structural image of the nailfold, (b) segmented three-dimensional (3-D) tissue structure obtained from all OCT structural images of the nailfold, (c) en face thickness map calculated from the segmented 3-D tissue structure (b), and (d) corresponding en face strain map calculated using Eq. (3) with en face thickness maps obtained from baseline and each measurement with external loading.

3.1.

Microvascular Imaging of Human Nailfold

Figure 3 shows in vivo microvascular imaging of the nailfold with no pressure (baseline) where the imaged area ($2\mathrm{mm}\times 2\mathrm{mm}$) is marked as a square box in a photograph of the fourth finger of the left hand [Fig. 3(a)]. In Fig. 3(b), a representative $XZ$ cross-sectional OCT image exhibits structural features in the nailfold such as epidermis (E), dermis (D), nail root (NR), and nail matrix (NM). In the corresponding cross-sectional mOMAG image [Fig. 3(c)], many bright signals are visible, representing blood perfusion in the capillaries. The structure image with the false colored (red) mOMAG overlay identifies the capillary flow in the dermal layer (D) as shown in Fig. 3(d). Maximum intensity projection (MIP) views were made with the structure and the vasculature, and the results are shown in Figs. 3(e) and 3(f), respectively. In Fig. 3(f), regular disposition of the hair pin-like capillary loops (arrows) along the cuticle (eponychium) is clearly seen, a typical characteristic pattern of the nailfold capillary for the healthy subject.²⁵

Fig. 3

In vivo microvascular imaging of a normal human nailfold with no pressure. (a) Photograph of a fourth finger of a left hand with an imaged area of the nailfold (a square box, $2\mathrm{mm}\times 2\mathrm{mm}$). (b) Representative cross-sectional (XZ) OCT structural image, EP: epidermis, D: dermis, NM: nail matrix, and NR: nail root. (c) Corresponding cross-sectional capillary flow image (mOMAG image). (d) The structural image of the nailfold with the color-coded (red) mOMAG image overlay. (e and f) En face (XY) maximum intensity projection (MIP) images of the nailfold structure and color-coded (red) vasculature from the same location, respectively. The horizontal white line on (e) and (f) indicates where the cross sections are shown in (b) and (c), respectively. Scale bars: 500 μm.

3.2.

Microvascular Imaging of Human Nailfold with Graded External Pressure

The capillary perfusion change against the displacements of the glass slide is presented in Fig. 4 with the en face angiograms captured from each measurement. At displacement of $Z=50\mu \mathrm{m}$ [Fig. 4(b)], the signal loss is observed in the middle of the proximal region from the capillary loops (indicated by the white circle) compared to the baseline [Fig. 4(a)], suggesting that the blood flow was ceased due to the obstruction of the vessel under the pressed tissue structure. Qualitatively, it is noted that there is an increase in capillary flow (arrow heads) and new appearance of capillaries (arrows) after loading was applied [Fig. 4(b)]. It can be explained that the blood flow was detoured to different intact vessels close to the obstructed vessels, leading to flow increase in the detoured vessels, meaning that the reserve capillaries are recruited. In normal tissue state, the reserve vessels are in existence but with no or minimum blood flow.³¹ However, the greater the displacements of the glass slide from 100 to 150 μm, the larger the vessel area is occluded. This is shown in Figs. 4(c) and 4(d), followed by complete occlusion across the whole capillary network at $Z=200\mu \mathrm{m}$ [Fig. 4(e)]. On releasing the occlusion, the mOMAG angiogram in Fig. 4(f) shows remarkable increase in the capillary reperfusion over the preocclusion (baseline), indicative of reactive hyperemia.⁴ This hemodynamic behavior may be due to vasodilation of the constricted capillary vessels after releasing pressure.⁴ The capillary perfusion is recovered but remained elevated to the baseline level 2 min after the pressure release [Figs. 4(g) and 4(h)]. For intuitive observation of this hemodynamic, we additionally performed real-time B-mode OCT imaging at the certain location in the nailfold that underwent consecutive pre/post loading. The B-mode imaging speed was 100 fps. OCT structural images and corresponding mOMAG images were obtained from the B-scanned OCT data set and made as a video. Figure 5 shows the representative OCT structural image and corresponding mOMAG image excerpted from the animation video, clearly visualizing time-lapsed change in the structure and the blood perfusion in the nailfold during the pressing.

Fig. 4

En face capillary angiograms of the nailfold against the displacements from the initial position of the glass slide in contact to the nailfold surface. (a) Baseline, (b) $Z=50\mu \mathrm{m}$, (c) $Z=100\mu \mathrm{m}$, (d) $Z=150\mu \mathrm{m}$, (e) $Z=200\mu \mathrm{m}$, (f) $Z=\text{back}$ to 0 μm, (g) $Z=0\mu \mathrm{m}$ (1 min after pressure releasing), and (h) $Z=0\mu \mathrm{m}$ (2 min after pressure releasing). In the initial loading (b), cessation of local capillary perfusion is observed (white circle). Perfusion in other vessels close to the ceased capillaries is increased (arrow heads) allowing reserved capillary recruitments (arrows). A scale bar: 500 μm.

Fig. 5

Representative OCT structural image and corresponding vascular image excerpted from the animation showing time-lapsed change in structure and vasculature of the human nailfold with consecutive pre/postloading process (MPEG, 4.5 MB) [URL: http://dx.doi.org/10.1117/1.JBO.19.5.056003.1].

Moreover, strain maps corresponding to the displacements of the glass slide are visualized in Fig. 6. In Fig. 6(b), distribution of strain is localized at the same site where the capillary occlusion occurred in Fig. 4(b) and enlarged as the displacement is increased from 50 and 200 μm. Since the nailfold is roughly circumferential, its strain magnitude becomes larger around the top surface of the nailfold. In Figs. 6(c) and 6(d), interestingly, the load distributions are caused by the shut-down of blood supply to the peripheral capillaries, leading to subsequent flow cessation in the capillary loops free from the external pressure [see Figs. 4(c) and 4(d)]. At the displacement of $Z=200\mu \mathrm{m}$ [Fig. 6(e)] where the complete occlusion occurred in Fig. 4(e), the strain distribution is the hottest in color, meaning the strongest pressure distribution over the tissue region of interest. Significant reduction of the strain is apparent right after the lift of the pressure due to immediate recovery of the deformed skin tissue [Fig. 6(f)] where, however, the residual strain is present. After the postpressure, the nailfold deformity is almost restored to the baseline level from the strain maps of Figs. 6(g) and 6(h), respectively. Combining Figs. 4 and 6, one can conclude that the blood flow response due to applied pressure is correlated very well with the resulted strain maps.

Fig. 6

En face strain maps of the nailfold corresponding to the displacements from the initial position of the glass slide in contact to the nailfold surface. (a) Baseline, (b) $Z=50\mu \mathrm{m}$, (c) $Z=100\mu \mathrm{m}$, (d) $Z=150\mu \mathrm{m}$, (e) $Z=200\mu \mathrm{m}$, (f) $Z=\text{back}$ to 0 μm, (g) $Z=0\mu \mathrm{m}$ (1 min after pressure releasing), (h) $Z=0\mu \mathrm{m}$ (2 min after pressure releasing). A scale bar: 500 μm.

Figure 7(a) shows the overlaid images of the mOMAG angiograms (Fig. 4) and the corresponding strain maps (Fig. 6) for each measurement, showing that the strain distributions are coregistered with the vessel occlusion regions. We calculated the FI and the averaged strain value for each image and the results are graphically represented in Fig. 7(b). In the graph, a progressive decrease in the FI is seen with the linear increase of the strain from 0 to 0.072. Especially, a quick drop of the FI is observed between the strains of 0.018 and 0.037. As the averaged strain reaches 0.072, the FI essentially becomes zero (complete capillary occlusion). After immediate release of the strain (0.012), the FI overshoots up to 0.602, an increase of $\sim 62\%$ above the baseline (0.372), and then slowly decreases by $\sim 34\%$ during a 2-min period of the postreleasing. This pressure-flow curve is quite similar to the results reported in the previous works using the LDI techniques.^3,4,8

Fig. 7

(a) Overlaid images of the mOMAG angiograms and corresponding strain maps. (b) Flow index versus averaged strain value for each measurement.