1.

Introduction

The microcirculation comprises microvascular networks of arterioles, capillaries, and venules, which are fundamental for thermoregulation and for transporting nutrients and gases to maintain the metabolism of cells.¹ However, under disease states, such as severe hemorrhage, cardiogenic shock, sepsis,^1,2 and systemic scleroderma,^3–5 the associated dysfunction of the microcirculation may cause heterogeneous hypoxia, impairing cell functioning in tissues and even causing multiple organ failures.^1,2 In other cases, hypertension and diabetes mellitus can cause microvascular complications, such as microvessel rarefaction and retinopathy, respectively.^6,7 Tumors will often induce angiogenesis of the microvascular system in their microenvironment.⁸ To better understand the fundamental mechanisms of these diseases, diagnose them in early stages, and evaluate the effectiveness of various therapies, it is essential to develop tools to monitor important microvascular parameters of blood perfusion. These parameters include function capillary density (which is defined as the total length of capillaries perfused by RBCs per observed area in units of ${\mathrm{cm}}^{-1}$),¹ total hemoglobin concentration (${\mathit{C}}_{\mathrm{Hb}}$), the oxygen saturation of blood (${\mathrm{sO}}_2$), the directional derivative of ${\mathrm{sO}}_2$ along the blood flow direction (${D\mathrm{sO}}_2$), and the speed of blood flow ($v_{\mathrm{Hb}}$).⁹

Primary medical imaging modalities, such as single-photon emission computed tomography, functional magnetic resonant imaging, positron emission tomography, ultrasonography, and diffuse optical tomography, have been used for years to image cardiovascular or cerebral blood flow.^10–14 Additionally, contrast-enhanced ultrasonography, which detects nonlinear oscillation of microbubbles (only a few microns in size) under low mechanical index conditions, has been applied to imaging blood perfusion around focal liver lesions and the renal cortex.^15–17 Even though these techniques are the best we have to date to image blood flows in organs deep in the body, they are limited by their millimeter-size resolution. Thus, these modalities are not efficient in monitoring microcirculation, which in general contains vessels smaller than $100\mu \mathrm{m}$.¹

As an alternative, the cutaneous and sublingual microcirculations have been proposed as a representative model for visceral microcirculation, because they are accessible by optical-based imaging techniques, which provide higher resolution than most other imaging modalities.^18–20 Patients with chronic diseases such as hypertension, renal disease, and coronary artery disease have been observed to have distinct cutaneous microvascular parameters.^6,21,22 Monitoring cutaneous microvascular functioning provides valuable information for evaluating peripheral microvascular diseases, such as Raynaud’s disease and peripheral arterial disease.^4,23,24 To this end, optical scattering-based techniques such as laser Doppler imaging,^22,25 near-infrared spectroscopy,^26,27 and reflectance spectroscopy^28,29 are used to detect scattered light from tissues. Over a submillimeter sampling volume, laser Doppler imaging can measure the average speed of flow, and near-infrared spectroscopy and reflectance spectroscopy can measure both the average flow speed and the oxygen saturation. On the wide-field scale, nailfold videomicroscopy,^4,30,31 orthogonal polarization spectral imaging,^32,33 sidestream dark field imaging,³⁴ and optical coherent tomography^35,36 can provide wide-field information about function capillary density and the speed of flow, with lateral resolutions ranging from submicrons to around $15\mu \mathrm{m}$, which covers from the thinnest capillaries to the wider arterioles and venules. The imaging depth can go as deep as $400\mu \mathrm{m}$ for nailfold videomicroscopy and around 1 to 3 mm for optical-based techniques. Combined with an endoscope, these modalities can image the gastric or intestinal microcirculation with a tolerable compromise of image quality. However, none of these imaging modalities can provide ${\mathrm{sO}}_2$ and $v_{\mathrm{Hb}}$ information at the same time.

In recent years, optical resolution photoacoustic microscopy (OR-PAM) has shown promise in in vivo microvascular imaging, with its ability to provide wide-field, capillary-resolving, and hemoglobin-sensitive images.^37–43 Combined with the flow speed imaging techniques reported previously,^44–46 OR-PAM has been demonstrated as a powerful tool to acquire such important parameters of the microcirculation as ${\mathrm{sO}}_2$, ${D\mathrm{sO}}_2$, ${\mathit{C}}_{\mathrm{Hb}}$, $v_{\mathrm{Hb}}$, and the metabolic rate of oxygen in tissues.^47,48 In this study, we implemented dual-wavelength in vivo OR-PAM for investigating oxygen release in cuticle capillaries. This is the first time that oxygen release dynamics in human cuticle capillaries have been monitored. The correlation between oxygen release and the speed of RBCs and between oxygen release and the first-order time-derivative of ${\mathrm{sO}}_2$ have also been analyzed in a cuticle capillary. The spatial- and time-resolved information acquired by OR-PAM may help in early-stage diagnosis of perivascular diseases, such as Raynaud’s syndrome, and in diagnosing heterogeneous microcirculation of interior organs.

2.

Methods

2.1.

Experimental Protocol

Nine healthy, consenting volunteers (with ages ranging from 23 to 30; seven males and two females) were recruited in this study. For each volunteer, we imaged the cuticle capillaries in the fourth finger (the ring finger) of the left hand.³⁶ Before each experiment, the volunteer rested in the temperature-controlled laboratory (at 20°C) for 15 min to adapt to the environmental temperature, since nailfold microcirculation is known to be sensitive to the surrounding temperature. The imaged area was then cleaned with alcohol swabs, and the hand was comfortably put on a homemade hand mount, without occlusion of blood flow, as shown in Fig. 1. During the data acquisition period, the photoacoustic (PA) scanning head was scanned over a single cuticle capillary at a time for three-dimensional imaging, with a 10-Hz C-scan rate (high-speed scanning mode) for about 40 s. At least three cuticle capillaries were recorded for each volunteer. The total experimental time spent on a volunteer was $<1.5\mathrm{h}$, including rests every 20 min to prevent numbness of the extremities. The human study was approved by the Institutional Review Board of Washington University in St. Louis, and the pulse energies of the excitation lasers used in each experiment were within the American National Standards Institute (ANSI) laser safety limit ($20\mathrm{mJ}·{\mathrm{cm}}^{-2}$).

Fig. 1

Schematic of the single-cell OR-PAM system. BS, beam splitter; CG, coverglass; LSM, linear step motor; PD, photodiode; PH, pin hole; PM, plastic membrane; UG, ultrasound gel; UT, ultrasound transducer; VC, voice coil motor; and WT, water.

3.

Results

3.1.

Monitoring of Oxygen Saturation of Blood Dynamics in Cuticles

Figures 2(a)–2(c) show a top view (C-scan) and a cross-sectional view (B-scan) of the typical morphology of finger cuticle capillary loops. The acquisition time of a C-scan image was 75 s. The cuticle capillary loops angle toward the distal nail bed and gradually toward the epidermis. Figure 2(d) shows the result of using the curvature calculated from the C-scan images to quantitatively describe the geometric profile at different positions along the cuticles. The full width at half maximum distance is around $40\mu \mathrm{m}$, which suggests that it is reasonable to define a region $\pm 20\mu \mathrm{m}$ from the position with maximal curvature as the tip region of a cuticle. It is also noticeable but not surprising to observe that in most cases the tip positions ($0\mu \mathrm{m}$) coincide with the uppermost ends of the cuticles in the B-scan images. In Fig. 2(b), the insets also show pixel-by-pixel calculation of ${\mathrm{sO}}_2$ distribution in different areas of the cuticle capillary network with different color bars. The ${\mathrm{sO}}_2$ reduction across the tip of a cuticle capillary is within 0.2. In high-speed scanning mode, the flow and the ${\mathrm{sO}}_2$ of single RBCs can be resolved, as shown in the snapshots in Fig. 3(a) and Video 1. Figure 3(b) shows the results of time-averaging over all the frames of the ${\mathrm{sO}}_2$ image. Around the cuticle tip (the most curved position along the cuticle), an abrupt drop in ${\mathrm{sO}}_2$ can be observed. Figure 3(c) shows ${\mathrm{sO}}_2$ versus $s$, where $s$ denotes the displacement along the central axis of a cuticle capillary loop (i.e., the trace of the blood flow). The origin of $s$ is coincident with the cuticle loop tip, and the RBCs flow from the negative coordinates (the upstream side of a cuticle vessel) to the positive coordinates (the downstream side). The ${\mathrm{sO}}_2$ change with distance can be revealed more clearly by plotting the derivative of ${\mathrm{sO}}_2$ with respect to $s$, which is defined as $D{\mathrm{sO}}_2\equiv \partial ({\mathrm{sO}}_2)/\partial s$, as shown in Fig. 3(d). The $D{\mathrm{sO}}_2$ values within $\sim 15\mu \mathrm{m}$ the cuticle loop tips are approximately twice as high as those in regions 25 to $40\mu \mathrm{m}$ away from the tip. The paired Student’s $t$-test between the tip region (yellow) and two sides (green) validates that the cuticle loop tips have significantly greater decreases in ${\mathrm{sO}}_2$ than the sides do.

Fig. 2

(a) Photograph of a finger with the imaged area boxed. (b) Wide-field PA image of cuticle capillaries shown with normalized PA amplitude. The insets show ${\mathrm{sO}}_2$ images of selected cuticles with different color bars. (c) B-scan image of cuticle capillary loops. (d) Curvature along the cuticles (fitting: sum of two Gaussians).

Fig. 3

(a) Selected time-lapse images of single-RBC ${\mathrm{sO}}_2$ (Video 1, MPEG, 207 KB) [URL: http://dx.doi.org/10.1117/1.JBO.21.5.056004.1]

. (b) Time-averaged images ($\sim 10\mathrm{s}$) of all time-lapse frames of ${\mathrm{sO}}_2$ imaging. (c) Time-averaged ${\mathrm{sO}}_2$ along the length of a cuticle capillary loop (i.e., a trace of the blood flow). (d) Time-averaged directional derivative of ${\mathrm{sO}}_2$ along the length of the loops. (e) Statistics of (d): paired Student’s $t$-test between the tip region (yellow) and the two side regions (green). NS: not significant ($P=0.48$), ***$P<0.001$, and $n=21$.

3.2.

Measurement of Red Blood Cell Flow Speed

By mapping the length of a curved cuticle loop $s$ into a straight line $l$, Fig. 4(a) shows the method we used to measure the speed of RBCs flowing in cuticle capillary loops.⁴⁵ Based on Fourier analysis of the frames of a specific segment of a capillary loop acquired at different times, the longitudinal flow speeds of different segments in a cuticle can be determined by $v_{\mathrm{Hb}}=(\mathrm{\Delta }s/\mathrm{\Delta }t)=(\mathrm{\Delta }l/\mathrm{\Delta }t)=(N_{\mathit{t}}/N_{\mathit{l}})(\mathrm{\Delta }{F}_{\mathit{t}}/\mathrm{\Delta }{F}_{\mathit{l}})$. Here, $N_t$ and $N_l$ are the sampled temporal and spatial lengths and $F_t$ and $F_l$ are the temporal and spatial frequencies. From Fig. 4(b), we can observe that the time-averaged RBC flow speed within the region of $15\mu \mathrm{m}$ around the cuticle tip is approximately one-third lower than that in the regions between 25 to $40\mu \mathrm{m}$ away from the cuticle tip. A paired Student’s $t$-test between the tip and side regions shows a significantly lower RBC flow speed around the tip region.

Fig. 4

(a) Image for speed measurement. (b) Time-averaged RBC flow speeds along the length of a cuticle loop. (c) Statistics of (b): paired Student’s $t$-test between the tip region (yellow) and the two side regions (green). NS: not significant ($P=0.45$), ***$P<0.001$, **$P<0.01$, and $n=18$.

3.3.

Measurement of Hemoglobin Flux in Red Blood Cell Flow and Time Derivative of Oxygen Saturation of Blood

As well as imaging ${\mathrm{sO}}_2$, we can also image the relative concentration of hemoglobin (${\mathit{C}}_{\mathrm{Hb}}$) by summing the calculated images for oxy- and deoxyhemoglobin. To calculate the time-averaged hemoglobin flux, we assume that the concentration and the speed are independent variables, which means that the time average of the product of the two variables is approximately equal to the product of the two time-averaged variables (in case the variances of $C_{\mathrm{Hb}}$ and $v_{\mathrm{Hb}}$ are small), so we have $\overline{{\mathrm{\Phi }}_{\mathrm{Hb}}}\approx \overline{C_{\mathrm{Hb}}}·v_{\mathrm{Hb}}$, where $v_{\mathrm{Hb}}$ is the time-averaged RBC flow speed around $x$. Figures 5(a) and 5(c) show a nearly flat trend, and the paired $t$-tests shown in Figs. 5(b) and 5(d) suggest that the hemoglobin flux in RBC flow is approximately the same along the cuticle capillary loops: the flow of RBCs is conserved. Under steady-state blood flow, $[\mathit{d}({\mathrm{sO}}_2)/\mathit{d}t]={D\mathrm{sO}}_2·(\mathit{d}s/\mathit{d}t)$. Similarly, we assume that both ${D\mathrm{sO}}_2$ and $\mathit{d}s/\mathit{d}t$ are independent variables; we have $[\mathit{d}({\mathrm{sO}}_2)/\mathit{d}t]\approx {D\mathrm{sO}}_2·v_{\mathrm{Hb}}$. The total time derivative of ${\mathrm{sO}}_2$ along cuticle capillary loops is shown in Fig. 5(e). The $p$ value between the upstream side and the tip is 0.03, and the $p$ value between the downstream side and the tip is 0.07, according to the paired $t$-test shown in Fig. 5(f).

Fig. 5

(a) Time-averaged relative flow rates along the length of cuticle capillary loops. (b) Statistics of (a): paired Student’s $t$-test between the tip region (yellow) and the two side regions (green). NS: not significant (up: $P=0.24$, left: $P=0.10$, right: $P=0.40$), $n=13$. (c) Time-averaging of hemoglobin concentration along the direction of the length of cuticles. (d) Statistics of (c): paired Student’s $t$-test between the tip region (yellow) and the two side regions (green). NS: not significant (up: $P=0.33$, left: $P=0.45$, right: $P=0.21$), $n=18$. (e) Time-averaged values of $d({\mathrm{sO}}_2)/dt$ along the length of cuticle capillary loops. (f) Statistics of (e): paired Student’s $t$-test between the tip region (yellow) and the two side regions (green). *$P=0.03$, NS: not significant (up: $P=0.07$, down: $P=0.25$), $n=15$.

4.

Discussion

In this study, we demonstrated the ability of single-cell resolution OR-PAM to monitor the microcirculation in cuticle capillaries with a temporal resolution of 0.1 s. Compared to nailfold videocapillaroscopy and optical computed tomography,³⁶ OR-PAM can not only image the morphology, dimensions, and vessel density of cuticle capillary loops, but also measure multiple hemodynamic parameters, such as ${\mathrm{sO}}_2$, ${D\mathrm{sO}}_2$, ${\mathit{C}}_{\mathrm{Hb}}$, and $v_{\mathrm{Hb}}$. Monitoring these functional parameters at the fundamental level of the physiology of oxygen transport can potentially help biologists and physicians to understand the mechanisms of oxygen transport in the skin and to define clinical standards for early-stage diagnosis and evaluation of perivascular diseases, such as Raynaud’s phenomenon and systemic scleroderma, before the capillaries undergo observable changes in morphology.

The time-averaged ${D\mathrm{sO}}_2$ results in Fig. 3(d) indicate that RBCs release more oxygen in the tip region over a length of around $30\mu \mathrm{m}$ then they do further down on the two sides. A similar result has been mentioned in one previous work, with no further investigation.⁵² It is interesting to note that the $30\text{-}\mu \mathrm{m}$ length is approximately equal to the length of a capillary loop in the dermal papillae in the skin outside of the cuticle area.⁵³ Capillary loops in dermal papillae are extensions of the subpapillary plexus in the reticular dermis, and they are responsible for oxygen and nutrient transport to living cells in the epidermis. Because nails are specialized structures of the skin,⁵⁴ cuticle capillaries and dermal capillaries should be functionally similar parts of the capillary loop system (except that cuticle capillaries extend toward the distal nail bed), it will not be surprising to discover that the tip region of a cuticle capillary releases more oxygen than the other regions.

In Figs. 4(b) and 4(c), the RBC flow speed is reduced in the tip region (around two-thirds of the speed in the side regions). In blood rheology, RBC flow in capillaries is treated as a non-Newtonian fluid because of the special viscoelasticity of erythrocytes, which complicates the RBC flow in a capillary.^55–57 The reduced RBC flow speed may result from deformation of RBCs and a consequent change of their viscoelasticity while passing through the highly curved pathway of the tip region. Another possibility is that RBCs partially accumulate in the tip region. In order to test this hypothesis, we examined the hemoglobin concentration and the hemoglobin flux along the cuticles. Further, we used a paired $t$-test to compare the effects of the straight part and the curved part of a cuticle on the hemoglobin flow and concentration. To improve the statistical accuracy, we excluded outlier data points that have large standard deviations ($>30\%$). It can be seen that this hypothesis is not supported by the results in Figs. 5(a)–5(d), which show that both the flux of hemoglobin and the time-averaged hemoglobin concentration do not significantly differ between the side regions and the tip. Therefore, RBC flow is shown to be conserved along a cuticle capillary loop. The slower RBC flow in the tip region seems to meet a functional demand which requires a longer transit time of RBCs to release enough oxygen for metabolism. To investigate the relation between $D{\mathrm{sO}}_2$ and RBC flow speed, we calculated the time derivative of ${\mathrm{sO}}_2$ along cuticle capillary loops. Without introducing physical cuffing and compression on the arm imaged, and without any extra physiological stimulation, we assumed that the RBC flow can be considered as in a steady or quasi-steady state, which means $[d({\mathrm{sO}}_2)/\mathit{dt}]=\mathit{D}{\mathrm{sO}}_2·(\mathit{ds}/\mathit{dt})+[\partial ({\mathrm{sO}}_2)/\partial t]\approx \mathit{D}{\mathrm{sO}}_2·(\mathit{ds}/\mathit{dt})$.⁵⁸ Figures 5(e) and 5(f) show that RBCs release more oxygen per unit time in the tip region than in the sides. Although the statistics are not strongly significant ($p>0.01$), this finding still suggests that there are factors other than RBC flow speed, such as partial oxygen pressure, that can drive RBCs to release more oxygen in the tip region.

In this investigation, our single-cell resolution OR-PAM system performed monitoring of several hemodynamic parameters on nine human volunteers. Cell-by-cell based statistics also provided insights. In the future, OR-PAM promises to help greatly in the early-stage diagnosis of perivascular diseases and to illuminate more fundamental mechanisms in hemodynamics.

5.

Conclusion

In this paper, the cuticle microcirculations of healthy volunteers were monitored by real-time single-cell resolution OR-PAM. Hemodynamic parameters such as ${\mathit{C}}_{\mathrm{Hb}}$, ${\mathrm{sO}}_2$, $D{\mathrm{sO}}_2$, $v_{\mathrm{Hb}}$, and relative blood flow rate were extracted from the images. A drop in $D{\mathrm{sO}}_2$ and slower RBC flow were observed in the tip region than in the side regions of a cuticle capillary loop. The conserved blood flow rate in a cuticle capillary loop and the drop in the time-derivative of ${\mathrm{sO}}_2$ in the tip region suggest that the heterogeneity of the RBC flow speed over a cuticle capillary loop is not the only factor that determines the heterogeneity of the oxygen release in the loop.