2.1.

Experimental Setup of SS-OCTA System for Dermatology

The schematic of SS-OCTA system for dermatology is illustrated in the top panel of Fig. 1. We employed a high-speed broadband VCSEL swept laser (SL1310V1-20048, Thorlabs, Inc.) based on microelectromechanical systems (MEMS) technology, which is capable of 200 kHz sweeping rate with a $\sim 100\mathrm{nm}$ wavelength tuning range centered at 1300 nm. This swept source also provided a precise A-line trigger and optical clock for linear $k$-space sampling, generated by a built-in Mach–Zehnder interferometer (MZI) with 48 mm delay. The light from the laser was delivered into a fiber-based Michelson interferometer via an optical fiber circulator. A broadband 50:50 fiber coupler split the beam into the reference arm and sample arm. In the reference arm, the light was focused on the reference mirror by the lens and then coupled back into the fiber through a collimator. The lens and mirror were mounted on a precision translation stage so that the optical path length difference can be fine-tuned between the reference and sample arms. A compact handheld scan probe was used to capture the OCT images of skin, which was convenient and flexible to access different parts of human body. The incident power on the sample was 5 mW, which is below the FDA safe limit for skin imaging. A guiding light with 635 nm wavelength was also coupled in the interferometer by a wavelength-division multiplexing (WDM) to provide the aiming beam for OCT scanning. The backreflected and backscattered light from the reference arm and sample arm were combined to generate spectral interferometric fringes and then detected by a balanced photodetector with 400 MHz bandwidth. A high-speed digitizer (ATS9360, AlazarTech, Inc.) was utilized to record the OCT signals, which provided 12-bit resolution with up to $1.8\mathrm{GSam}/\mathrm{s}$ sampling rate. The OCT interferogram data were transmitted to a host computer through eight-lane PCI-Express Gen2 interface.

Fig. 1

Experimental setup of the SS-OCTA system and the SD-OCTA system for dermatology imaging applications. BPD, balanced photodetector; WDM, wavelength-division multiplexing; and SLD, superluminescent diode.

In order to show the advantages of SS-OCTA in skin imaging, we also purposely established an SD-OCT system for imaging performance comparison. The bottom panel of Fig. 1 shows the setup of the SD-OCTA system used in this study. The light source was a broadband superluminescent diode (SLD) with 85-nm full-width at half-maximum bandwidth centered at 1310 nm, which is similar to that of the swept laser in the SS-OCTA system. To make a fair comparison, we also kept the incident power on the sample at 5 mW by controlling the current value of the SLD. The configuration of fiber-based Michelson interferometer was the same as the SS-OCTA system, except for the detection part as well as the light source. In the spectrometer, an InGaAs line-scan camera (GL2048R, Sensors Unlimited) with 2048 pixel-array was utilized to capture the OCT interferometric fringes, capable of running at different line rates up to 147 kHz. The raw OCT data were streamed to a host computer by a Camera Link^® interface. To ensure the consistent imaging region and imaging condition, the SD-OCTA system and SS-OCTA system shared the same handheld scan probe by simply switching the fiber in the sample arm. Both the SD-OCTA system and SS-OCTA system were automatically controlled on LabVIEW platform.

2.2.

Imaging Protocols and Algorithms

For all of the OCTA images in this paper, we adopted the ultrahigh sensitive optical microangiography (OMAG) scanning protocol as described in Refs. 33 and 34, where four repeated B-scans were taken at the same B-scan position to extract the dynamic change of blood flow from static tissue. The galvo scanner situated in the handheld probe was driven by a saw tooth waveform with 80% duty circle in the fast axis ($X$ direction) and a step function waveform in the slow axis ($Y$ direction). The B frame rate of both SD-OCTA system and SS-OCTA system was set at 160 Hz, providing the same time interval of 6.3 ms to achieve high flow sensitivity down to capillary level. After the imaging acquisition, we processed the OCT data offline on the platform of MATLAB^®. To construct the blood flow image, we applied the algorithm of complex OMAG,³³ which performed a differential operation of the complex OCT information between the repeated B frames. Since the algorithm of complex OMAG is phase sensitive and the MEMS-VCSEL swept source has severe phase noise due to unstable wavelength sweeping and the synchronization time jitter, an additional fully automatic and numerical technique based on phase gradient algorithm was utilized to compensate the phase error for the SS-OCTA data processing.³⁵ The imaging FOV, A-line speed of the SD-OCTA system and the objective lens in the handheld probe are variable for the requirements of different dermatological examinations.

The OCTA imaging of human skin by the home-built systems was reviewed and approved by the Institutional Review Board of University of Washington, and the informed consent was obtained from all subjects before imaging. This study followed the tenets of the Declaration of Helsinki and was conducted in compliance with the Health Insurance Portability and Accountability Act.

3.1.

Sensitivity Roll-Off Performance

By tuning the optical delay using the translation stage in the reference arms, we generated the system sensitivity roll-off curves for the SD-OCTA system and SS-OCTA system, respectively, the results of which are shown in Fig. 2. With the full spectral span of $\sim 120\mathrm{nm}$ centered at 1310 nm wavelength and 2048 pixel-array in the line-scan camera, the imaging range of the SD-OCTA system was 7.3 mm. The spectral resolution of the SD-OCTA system was $\sim 0.08\mathrm{nm}$, which is determined by the configuration of the spectrometer, such as the pixel width of the line-scan camera, the line density of the dispersive grating, and the waist and incident angel of the optical beam. With 5-mW optical power in the sample arm and 147 kHz A-line rate, the measured sensitivity of the SD-OCTA system was $\sim 102\mathrm{dB}$ at the position close to zero-delay line. However, as shown in Fig. 2(a), the increase of the depth position makes the OCT sensitivity to drop quickly, largely due to the relatively limited spectral resolution in the spectrometer. Although the A-line speed of SS-OCTA system was 200 kHz (higher than the SD-OCTA system), the measured sensitivity was 105 dB near the zero-delay line, which is still better than SD-OCTA system. This is largely attributed to the high efficiency of the dual balanced photodetector, fast acquisition speed of the A/D card, and the narrow line width of the laser. The digitizer was run at external clock mode using the optical clock from the swept laser to provide linear $k$ sampling. The optical clock was generated by the built-in MZI with 48-mm optical delay, resulting in 12 mm imaging range with 2560 sampling points per spectrum. Since the MEMS-VCSEL laser is operated in single longitudinal mode, the instantaneous linewidth is extremely narrow ($<3\mathrm{pm}$),³⁶ leading to an ultralong coherence length. As shown in Fig. 2(b), the sensitivity is maintained at $>100\mathrm{dB}$ within the depth range of 8 mm. The sudden sensitivity drop beyond the distance of 8 mm is due to the bandwidth limitation of the balanced photodetector that was only 400 MHz, which can be still further improved if a higher bandwidth of balanced detector is employed.^21,37 The comparison between the two systems as shown in Fig. 2 demonstrates that the SS-OCTA system has a longer imaging range with higher sensitivity and better roll-off performance, which would greatly benefit the wide-field imaging of human skin (see below).

Fig. 2

Comparison of system sensitivity roll-off between (a) SD-OCT system and (b) SS-OCT system.

3.2.

OCTA Imaging and Comparison on Normal Skin

After the assessment of the system sensitivity roll-off, we further compared the in vivo dermatological OCTA imaging between these two systems. We first performed the dermatological OCTA imaging on a normal skin sample. Here, the palm skin of a healthy volunteer’s right hand was selected. To enhance the optical penetration, a drop of glycerol solution covered with a thin glass slide was applied on the human skin for refractive-index matching.³⁸ Figure 3 shows the en face maximum intensity projections (MIPs) of the vascular networks, color coded by depth, captured by the SD-OCTA system and the SS-OCTA system, respectively. In the handheld probe, we employed an objective lens with 36-mm focal length, which provided $\sim 20\mu \mathrm{m}$ lateral resolution. The sampling spacing of these images was $10\mu \mathrm{m}$ in $X$ and $Y$ directions for a fine visualization of both the structure and vasculature of human skin. Since the line-scan camera is capable of running at different speed, we set two imaging A-line rates for the SD-OCTA system. One imaging speed was 76 kHz as shown in Fig. 3(a), which is a typical imaging speed for most SD-OCT systems for skin imaging. The other imaging speed was 147 kHz as shown in Fig. 3(b), which is the highest speed for a commercially available line-scan camera. With 160-Hz B frame rate and 80% fast scanning duty circle, the en face blood flow image [Fig. 3(a)] captured by 76 kHz A-line rate contains $380\times 380\text{pixels}$ in $X$ and $Y$ directions, covering an FOV of $3.8\times 3.8{\mathrm{mm}}^2$. While the 147-kHz SD-OCTA has more sampling pixels ($735\times 735$), permitting a wider FOV ($7.35\times 7.35{\mathrm{mm}}^2$) [Fig. 3(b)]. Although the lower A-line rate results in smaller FOV, it has longer integration time to collect more OCT photons emerging from the skin tissue. Thus, the 76-kHz SD-OCTA has higher sensitivity and better visualization of the deep blood vessels, compared with the 147-kHz SD-OCTA [comparing between Figs. 3(a) and 3(b)]. On the other hand, the SS-OCTA system was run at 200-kHz swept rate, which permitted the highest imaging speed in this study. With the same B frame rate and sampling spacing, the blood flow image [Fig. 3(c)] captured by this SS-OCTA system contained $1000\times 1000\text{pixels}$, providing a wide FOV of $10\times 10{\mathrm{mm}}^2$. Furthermore, due to the higher OCT sensitivity and better roll-off performance, the SS-OCTA system delivered impressive information about vascular networks, where not only the small blood vessels located at the superficial layer (in green color) are observed but also the big vessels permeating the deep dermal layers (in red color) are visualized.

Fig. 3

The en face OCTA images with color coded by depth in the region of human palm (right hand), scanned by SD-OCT and SS-OCT, respectively. (a) The vascular image taken by the SD-OCTA system running at 76 kHz with $3.8\times 3.8{\mathrm{mm}}^2$ FOV. (b) The vascular image taken by the SD-OCTA system running at 147 kHz with $7.3\times 7.3{\mathrm{mm}}^2$ FOV. (c) The image taken by the SS-OCTA system with 200 kHz A-line rate and $10\times 10{\mathrm{mm}}^2$ FOV. The white dashed box in (b) corresponds to imaging area in (a), while the white dashed box in (c) corresponds to the imaging area in (b). The white scale bar is 1 mm.

Figure 4 shows the representative cross sectional images of both the structure (gray color) and vasculature (red color) at the position indicated by the white dashed lines in Fig. 3. The top two images [Figs. 4(a) and 4(b)] were taken by the SD-OCTA system with 76- and 147-kHz A-line rate, respectively, whereas the bottom one [Fig. 4(c)] was from the SS-OCTA system. In these images, the physiological structures of stratum corneum, stratum lucidum, epidermal–dermal junction, and dermal tissue are clearly observed. The green dashed line separates the epidermis from dermis. We also draw a yellow dashed line, which is $\sim 0.5\mathrm{mm}$ deeper than the green dashed line. The structure and the blood flows under the yellow dashed line are almost invisible for the 147-kHz SD-OCTA system. With lower imaging speed, the 76-kHz SD-OCTA system has slightly better visualization in this deeper region. Due to the much higher system sensitivity in deep axial position (Fig. 2), the SS-OCTA system demonstrates remarkable penetration depth in the skin [Fig. 4(c)], giving more internal details where we are able to clearly identify the tissue structure as well as functional blood vessels innervating the deep dermal layer. For comparison, Fig. 4(d) is a schematic diagram of skin, illustrating the structure and vasculature at different depths of the skin. The cover glass in Fig. 4(c) looks blurred and thicker than that in SD-OCT images, which is due to the side-lobe artifact in the point spread function of the SS-OCT system, resulting from the top-hat-shaped spectrum of VCSEL laser.

Fig. 4

Typical cross sectional B-scans of OCT structure (gray) overlaid with blood flow signals (red). Shown are the B-scans obtained at the location indicated by the white dashed lines in Figs. 3(a) and 3(b) are from SD-OCT running at (a) 76 and (b) 147 kHz, respectively. (c) is from SS-OCT running at 200 kHz. (d) Schematic drawing of the structure and vasculature of skin. E, epidermis; D, dermis; SC, stratum corneum; SL, stratum lucidum; EDJ, epidermal–dermal junction; SD, superficial dermis; and DD, deep dermis. The white vertical and horizontal scale bars are 0.5 mm.

We further generated the blood perfusion maps of the human palm for different depth layers (Fig. 5). In doing so, we used a proprietary automatic segmentation software³⁹ to segment the 3-D images into depth resolved physiological layers. The images list from top to bottom in Fig. 5 correspond to the 76-kHz SD-OCTA, 147-kHz SD-OCTA, and 200-kHz SS-OCTA, respectively. They have different FOVs as described in Fig. 3, which are also indicated by the white dashed boxes here. The images of Fig. 5 listed from left to right demonstrate the vascular networks in three-depth layers. The first layer is from 0 to 0.5 mm measured from skin surface, which represents the epidermis layer above the green dashed line in Fig. 4. Both the SD-OCTA system and SS-OCTA system can show the rich epidermal vascular plexus, including capillary loops permeating this layer. The vessel diameter is relatively small and the vascular pattern coincides well with the palm print of the volunteer. The second layer is from 0.5 to 1 mm, which depicts the superficial dermis layer between the green dashed line and yellow dashed line in Fig. 4. The vessels in the superficial dermis layer are mainly the interconnecting vessels, which are clearly visualized by these OCTA systems. But the image brightness generated by the SD-OCTA system is weaker than that by the SS-OCTA system. The third layer is the deep dermal layer from 1 to 2 mm, which is under the yellow line in Fig. 4. The blood vessels in this deep region are termed as deep vascular plexus, which have relatively larger diameter to supply nutrition for skin metabolism. The image taken by 147-kHz SD-OCTA system [Fig. 5(b3)] is difficult to identify the deep vascular plexus. With slower imaging speed that improved the sensitivity, the 76-kHz SD-OCTA system has better image quality for these large vessels [Fig. 5(a3)] but still is not sufficient to provide good visualization for them. Using the 200-kHz SS-OCTA system, we are able to achieve superior OCT sensitivity in the deep axial position, resulting in the best visualization for the deep vascular plexus as shown in Fig. 5(c3). The large vessels as well as some interconnecting vessels are clearly observed, providing more comprehensive vascular information that may be useful to understand skin pathology for clinical applications. The comparison of the vascular networks in different skin layers demonstrates the advantages of the SS-OCTA system over the SD-OCTA system in imaging skin blood perfusion, providing a potential opportunity for dermatological diseases diagnosis and treatment monitoring.

Fig. 5

The en face OCTA images of the human palm displayed for different depth layers. The top images (a1–a3), middle images (b1–b3), and bottom images (c1–c3) correspond to Figs. 3(a)–3(c), respectively. The left images (a1–c1) are the blood perfusion maps of the depth layer from 0 to 0.5 mm, while the middle images (a2–c2) are from 0.5 to 1 mm and the right images (a3–c3) are from 1 to 2 mm, respectively. The white $\text{scale bar}=1\mathrm{mm}$.

3.3.

OCTA Imaging and Comparison on Benign Nevus

An examination of the abnormal skin site was also conducted by the two OCTA systems. Figure 6(a) shows the photograph of a black nevus with $\sim 3.5\mathrm{mm}$ diameter on the dorsal forearm of a volunteer. The white dashed box in this figure is the region of interest to perform the OCTA imaging, which has an FOV of $10\times 10{\mathrm{mm}}^2$. To cover this large area, the SD-OCTA system was running at 147 kHz. The imaging protocol including the B frame rate, repetition number, sampling points, refractive-index matching for skin, and scanning lens in the handheld probe was the same as described above. Figure 6(b) is the en face structural image taken by the SS-OCTA system, where we can clearly observe the position and boundary of the nevus, the curves of the epidermal ridges as well as the hair on the skin. The SD-OCTA system delivered very similar results for the en face structural image, thus not presented here. Figures 6(c) and 6(d) are the en face vascular networks at this skin site measured by both SD-OCTA and SS-OCTA. With the color coded by depth, the vascular imaging capabilities of the two systems are different. The SD-OCTA can identify the blood vessels in the superficial skin layer from 0 to $\sim 1\mathrm{mm}$, where the green color is dominated in Fig. 6(c). On the contrary, the SS-OCTA has better visualization of deep vasculature, where the small vessels in the superficial skin layer as well as the large vessels in deep dermal layer are delineated by the color from green to red as shown in Fig. 6(d). It is interesting to observe that the blood vessels in the nevus have less density and different vascular patterns compared to the surrounding normal skin region, indicating that the nevus is a living skin tissue but with less metabolic activities.

Fig. 6

OCTA images the microvascular networks permeating a benign nevus. (a) The photograph of the skin with a nevus. (b) The en face structural image by the SS-OCTA system for the skin area indicated by the white dashed box in (a). (c) and (d) are the en face vascular images with color coded in depth of the same skin area by the SD-OCTA and SS-OCTA, respectively. The $\text{scale bar}=1\mathrm{mm}$. The FOV of the OCTA images is $10\times 10{\mathrm{mm}}^2$.

Figure 7 shows the two-dimensional (2-D) B-scan images of the structure (gray) combined with the vasculature (red) measured by the two systems at the position indicated by the white dashed lines in Fig. 6. Compared to the skin of the human palm as shown in Fig. 4, the stratum corneum is thinner in the dorsal forearm. The bump in the middle part of the images is the black benign nevus. Although there are rich pigments in the nevus, both systems are able to visualize the dermis tissue within the nevus. As expected, the SD-OCTA has less penetration into the deep region of the skin [shown in Fig. 7(a)], while the SS-OCTA has much stronger OCT signals, resulting in better visualization in the deep dermal layer for both the normal skin part and the nevus area [shown in Fig. 7(b)]. The imaging capability of the blood vessels is also greatly affected by the strength of OCT signals. Thus, the blood flow signals in the nevus region are weaker and the vessels in the deep dermal layer are more difficult to be observed in the SD-OCTA, whereas the SS-OCTA provides much better imaging quality.

Fig. 7

Typical 2-D cross sectional B-scan images of the skin structure (gray) overlaid with blood vessels (red) captured by (a) SD-OCTA and (b) SS-OCTA, respectively. The vertical and horizontal scale bars are 1 mm.

We also generated the en face blood perfusion maps of the nevus by the two systems at different depth layers as shown in Fig. 8. The skin was separated into two layers by the green dashed lines in Fig. 7: the superficial skin layer from 0 to 0.8 mm depth and the deep dermal layer from 0.8 to 1.6 mm. The blood vessels in the first layer are given in green color, and the vessels in the second layer are shown in red color. Both SD-OCTA and SS-OCTA are capable of seeing rich superficial vascular plexus within the first skin layer albeit the SS-OCTA has better image contrast and clarity. As for the deep dermal vascular plexus, the SS-OCTA demonstrates better visualization than the SD-OCTA. The networks of large blood vessels are clearly identified by the SS-OCTA [shown in Fig. 8(b2)], whereas they are barely visualized with low contrast and SNR by the SD-OCTA system [shown in Fig. 8(a2)]. The segmentation of the blood vessels in different layers of the skin provides insightful information of the cutaneous circulation in nevus, where the SS-OCTA demonstrates obvious advantages in skin assessments over the SD-OCTA system.

Fig. 8

The en face vascular networks of the skin with a benign nevus for the superficial layer from 0 to 0.8 mm (a1 and b1) and the deep layer from 0.8 to 1.6 mm (a2 and b2), respectively. The top panel: SD-OCTA. Bottom panel: SS-OCTA. The $\text{scale bar}=1\mathrm{mm}$.

3.4.

OCTA Imaging and Comparison on Human Finger

To further illustrate the advantage of the SS-OCTA, we also conducted comparison between the two systems for a human finger, which is more complex in terms of its topological features, i.e., uneven surface. The depth locations of the finger skin could vary by several millimeters or even centimeter range that challenges OCTA imaging if the system ranging distance is limited. In this case, the SD-OCTA system was running at its maximum speed of 147 kHz to generate more sampling points. The B frame rate, number of repetition, and data size are the same as described above. A long focal length (110 mm) lens with wide diameter was used in the handheld scan probe, providing a lateral resolution of $\sim 50\mu \mathrm{m}$. The imaging FOV was $18\times 18{\mathrm{mm}}^2$ to cover the area of the whole fingertip. The top images in Fig. 9 are taken by the 147-kHz SD-OCTA system while the bottom images are captured by the 200-kHz SS-OCTA system. The en face structural images (left column) help us identify different parts of the fingertip including the nail plate, lunula, cuticle, eponychium, and proximal nail fold. Pointed by the yellow arrows, the lateral nail folds are clearly presented by the SS-OCTA system; however, they are invisible by the SD-OCTA system due to its rapid sensitivity roll-off along the depth and limited ranging distance. The brightness of finger structure in Fig. 9(d) is relatively uniform, but some parts in Fig. 9(a) are dark, which may result in misleading information when interpreting the fingertip morphology. The middle column images are the en face MIPs of the vascular networks of the fingertip. The SD-OCTA system is able to see the blood vessels in the area of proximal nail fold, where it is relatively flat and close to the zero-depth position. But the vessels under the nail plate are partly or barely observed due to the relatively poor OCT signals. Using the SS-OCTA system, the entire cutaneous circulation networks of the dorsal fingertip are visualized in a single scan [shown in Fig. 9(e)], including the vessels under thick nail plate, in the lateral nail fold, and the distal edge. The 2-D cross sectional images on the right column also verify the imaging capabilities of these two systems well. The lateral nail folds, as indicated by the white arrows, are located in deep axial position, where it is almost invisible by the SD-OCTA system. The nail plate has a thickness of $\sim 0.8\mathrm{mm}$ thickness, which also attenuates the OCT light, leading to relatively weak blood flow signal in the SD-OCTA system. As shown in Fig. 9(f), the profile of the fingertip is visualized from the top of the nail plate to the bottom of lateral nail folds, which has $\sim 5\mathrm{mm}$ depth range. The blood vessels in the skin dermis under the thick nail as well as in the lateral nail folds at deep positions are also well identified. This improvement of the visualization for both morphology and functional vasculature promises the SS-OCTA as a useful tool for clinical translation, since the human finger is an important health indicator that relates to many diseases, such as diabetes, anemia, thyroid, malnutrition, etc.^17,40

Fig. 9

The images of human fingertip captured by SD-OCTA system (top) and SS-OCTA system (bottom), respectively. (a) and (d) show the en face MIPs of the structure. The middle images (b and e) show the en face vascular networks. (c) and (f) are the cross sectional images of structure (gray) overlaid with blood vessels (red) at the position indicated by the white dashed lines in (a) and (d), respectively. The FOV is $18\times 18{\mathrm{mm}}^2$. The vertical and horizontal scale bars are 1 mm.