1.

Introduction

Hemodynamic monitoring after an ischemic event is an effective means of assessing and understanding recovery from numerous diseases, such as strokes.^1,2 Vascular occlusion tests (VOTs) are often employed for this type of monitoring and are based on microcirculation functions. Thus, VOTs can be used to examine blood flow capacity increases after ischemic events, such as temporary disruptions of arterial or venous flow, and to monitor postischemic reactive hyperemia, by investigating the responses of microvessels to a local stimulus.^3,4The use of high-resolution optical techniques to monitor vascular network changes can yield further insight into the nature of hemodynamic responses.^5,6

Some full-field optical imaging techniques are currently being developed with the objective of obtaining high-resolution spatial and temporal data for local hemodynamics monitoring.^7,8 Laser speckle imaging, which is based on temporal speckle contrast analysis (TSCA), has been widely used to study blood flow characteristics.^9–11 This method is based on the dynamic speckle pattern blurring, which results from the integration time being longer than the typical speckle pattern decorrelation time. The level of blurring is quantified by the speckle contrast to reconstruct blood flow images. However, TSCA imaging is also affected by the moving particles concentration and the flow velocity; thus, it is not easy to measure a specific physical quantity from a reconstructed image. Other disadvantages include that TSCA imaging is not immune to low-frequency background noise and that its spatial resolution cannot meet the requirements of optical microangiography. Recently, we developed a laser speckle imaging technique, namely, laser speckle imaging with intensity fluctuation modulation (LSI-IFM).^12,13 In this method, LSI-IFM is used to separate the dynamic and static speckle signals that originate from moving and nonmoving particles, respectively. This approach yields the moving particles distribution and strongly suppresses the low-frequency background noise. Thus, high-spatial-resolution vascular network images can be obtained. However, their temporal resolution, of about 3 s per LSI-IFM image, is not sufficiently high to capture blood microcirculation changes.¹⁴

In this study, we further developed this technique and achieved high-temporal-resolution, full-field optical angiography by using the short-time modulation depth (STMD) to perform imaging. This revised method can achieve a temporal resolution of 0.04 s per image, significantly better than the 3-s temporal revolution of the LSI-IFM method. Phantom and biological experiments were also performed and the results verify the feasibility of using the STMD imaging method to perform high-temporal-resolution, full-field VOTs in vivo.

2.

Materials and Methods

A schematic diagram of the experimental setup is shown in Fig. 1. An 85-mW laser beam with a central wavelength of 650 nm was emitted by a semiconductor laser and was expanded by a beam expander. A parallel beam with a diameter of 20 mm obliquely incidents on the sample from the same side of the camera. In this experiment, a custom varifocus bitelecentric system lens (depth of field: $\pm 2\mathrm{mm}$) and matrix CMOS camera (A504, Basler, with maximum resolution of $1024\times 1024\text{pixels}$) with an image acquisition rate of 500 fps were used to acquire the raw laser speckle images, and Labview^® and MATLAB^® software were employed for data acquisition and processing, respectively. The entire imaging apparatus was fixed on a vibration-isolating optical platform to reduce the effects from external shaking.

Fig. 1

Schematic diagram of experimental setup. Tube driven by double-channel syringe pump and rabbit ear were used to obtain STMD images.

In the proposed method, full-field optical angiography is realized by using an undersampling method to separate the dynamic and static speckle signals, which arise from moving and nonmoving particles, respectively. In general, dynamic speckle signals are generated by the Doppler effect of the moving particles, whose frequencies can be up to 10 kHz.⁷ According to the Nyquist sampling theorem, to extract a precise dynamic speckle signal, the camera’s frame rate should be at least twice the Doppler signal frequency. Consequently, the camera’s resolution must be low to obtain a sufficiently high acquisition speed because the rate of data transmission from the camera to the computer is finite.¹⁵ However, in our method, a high-speed digital camera is used to avoid this problem and can obtain high-spatial-resolution images since its exposure and acquisition periods can be independently set by software. As shown in Fig. 2(a), if the camera’s exposure period is very short, i.e., less than the speckle signal period, the high-frequency speckle signal variations can be captured. For example, if the exposure period is set to $50\mu \mathrm{s}$, as required by the Nyquist sampling theorem, dynamic speckle signal fluctuations with frequencies of up to 10 kHz can be captured. On the other hand, if the camera’s acquisition period is relatively low (such as that corresponding to the 500 fps image acquisition rate used in our system) and is less than what is demanded by the Nyquist sampling theorem, then the camera can only capture fragments of the high-frequency dynamic speckle signal. In this case, the recorded signals are completely random [Fig. 2(b)]. In contrast, the stationary speckle intensity from nonmoving particles is not affected by changes in the camera’s exposure and acquisition periods because their signal approaches a constant signal over time [Fig. 2(c)].Therefore, the dynamic and static speckle signals can be separated in the frequency domain, as shown in Fig. 2(d). In our approach, a highly undersampled Doppler signal is used to extract the dynamic speckle signal.

Fig. 2

(a) Sketch of speckle signal recorded by camera; (b) dynamic speckle signal; (c) static speckle signal captured using shorter exposure time; (d) dynamic signals ($P_{\mathrm{AC}}$) and static signals ($P_{\mathrm{DC}}$) in frequency domain. $u_0$ is cut-off frequency and $F_0$ is sample rate of camera.

In our previous work, the modulation depth (MD) was used as the image parameter to reconstruct blood flow images and was defined as¹⁴

Fig. 3

Schematic of STMD signal processing: (a) raw speckle image sequence; (b) example long-time window ($M=256$, red solid-line rectangle). Long-time windows were used to extract (c) dynamic and (d) static speckle signals, and short-time windows ($N=20$, red dashed-line rectangle) were used to compute STMD.

3.

Experiment and Results

The results of the following experiment verified that STMD can be used as an imaging parameter to perform VOTs. A ${\mathrm{TiO}}_2$ solution was used to simulate blood flow, and ${\mathrm{TiO}}_2$ particles with diameters between $\sim 2\text{-}10\mu \mathrm{m}$ were employed as the moving particles. A precision double-channel syringe (JZB-2800, JYM) was used to pump pure water [syringe A in (Fig. 1)] and ${\mathrm{TiO}}_2$ solution [syringe B in (Fig. 1)] into a single polyethene tube, which had an inner diameter of 1 mm. The ${\mathrm{TiO}}_2$ concentration in the polythene tube was precisely controlled by adjusting the speeds of the flows from syringes A and B. The imaging area was about $2\mathrm{mm}\times 2\mathrm{mm}$. The camera captured 9000 frames at an acquisition rate of 500 fps during the VOT processing. The experimental process was as follows.

When the acquisition began, syringes A and B pumped pure water and ${\mathrm{TiO}}_2$ solution ($5\mathrm{g}/\mathrm{L}$) at flow rates of 1 and $0.05\mathrm{mm}/\mathrm{s}$, respectively, into the polyethene tube. After 3.72 s, the ${\mathrm{TiO}}_2$ solution pump, i.e., the flow from syringe B, was stopped and begun again at 11.84 s. The entire VOT process took 8.12 s. Video 1 shows the dynamic changes of the STMD image in detail and Fig. 4 presents selected STMD images taken before, during, and after stopping the ${\mathrm{TiO}}_2$ solution flow. A $0.2\mathrm{mm}\times 0.2\mathrm{mm}$ area [shown in Fig. 4(a)] was used to calculate the average STMD value. Figure 4(g) presents the STMD trace resulting from the VOT. Points (a)–(f) that are marked in Fig. 4(g) correspond to the STMD images shown in Figs. 4(a)–4(f). The decreasing region in Fig. 4(g) shows the gradual decline in ${\mathrm{TiO}}_2$ concentration that occurs after the occlusion at 3.72 s. The reason that this decrease is gradual rather than immediate is that the ${\mathrm{TiO}}_2$ particles diffusion in the pure water delays the concentration drop. When the ${\mathrm{TiO}}_2$ flow begins again at 11.84 s, STMD rapidly returns to its initial value, demonstrating the high-temporal-resolution nature of our method.

Fig. 4

(a)–(f) Sequential STMD images obtained during phantom VOT; (g) average temporal STMD profile of area indicated by rectangle in (a). Simulated occlusion begins at 3.72 s in (b) and is removed at 11.84 s in (e) [Video 1, MOV, 9.42 MB (URL: http://dx.doi.org/10.1117/1.JBO.21.8.086002.1)].

A biomedical VOT experiment was performed and demonstrated that our method can be used to monitor vascular network changes in vivo. The specimen was a rabbit with a mass of 200 g that was anesthetized with 10% chloral hydrate (0.4 mL). Its ear was depilated with a hair remover lotion and the $20\mathrm{mm}\times 16.7\mathrm{mm}$ area used for imaging was pressed by a glass rod to cause blood vessel occlusion. The animal was treated according to Regulations for the Administration of Affairs Concerning Experimental Animals, which were approved by the State Science and Technology Commission of China, and the Guangdong Province regulations for the care and handling of laboratory animals. The high-speed camera (500 fps) recorded 40116 raw speckle images during the 80.23-s acquisition time to reconstruct a 1994 frame STMD image. The occlusion began at 3.48 s and was released at 47.20 s. The resulting raw laser speckle and STMD image sequences are presented in Figs. 5 and 6, respectively. The snapshots were obtained before, during, and after occlusion. In the raw laser speckle image sequences, the blood flow cannot be distinguished from the background tissue. Thus, the tracking profile in Fig. 5(g) shows no changes during the VOT, which is also verified by Video 2. However, as shown by the STMD images in Fig. 6, the vascular network can be reconstructed by using our STMD-parameter-based, full-field optical angiography method. During the occlusion, the vessels and background brighten at first; then, the vessels disappear gradually, and the background becomes dark. A sufficient time after releasing the occlusion, the vessel network returns to its normal state. From the full-field optical angiography image (Fig. 6 and Video 3), an important congestion phenomenon can be directly observed, where Figs. 6(g) and 6(h) present VOT time-trace profiles at the vessel position indicated by rectangles “A” and “B” ($0.1\mathrm{mm}\times 0.1\mathrm{mm}$), respectively, in Fig. 6(a). When the occlusion starts, the first congestion is mainly caused by venous congestion (point “b” at 4.24 s); when the occlusion stops, arterial congestion occurs at the dominant position (point “e” at 49.16 s) and the network gradually returns to the normal condition. The difference between the two congestions occurs because veins and arteries have different reactions to vascular occlusion. A similar phenomenon has been observed in other experiments performed using ultrasound.¹⁶ As shown in Video 3 and Figs. 6(b) and 6(e), the background signal abruptly increases when the congestion occurs. A possible explanation is that the interstitial fluid around the adipose tissue undergoes an emergency reaction to response to the sudden global congestion. However, at some instances [point “1” at 8.26 s and point “2” at 13.47 s that are marked in Figs. 6(g) and 6(h)], the two vessels show different dynamic responses. This result demonstrates that STMD images can also reveal the local dynamic changes in blood vessels. Obviously, an additional benefit of the high-temporal-resolution and full-field character of the STMD method is that rapid global and local changes in vascular networks can be monitored.

Fig. 5

Time series of raw images (a) before, (b)–(d) during, and (e), (f) after occlusion of rabbit ear artery (Video 2). Imaging area is $20\mathrm{mm}\times 16.7\mathrm{mm}$. (g) Time trace of blood flow, where intensity is average value in area indicated by rectangle in (a) [Video 2, MOV, 10.5 MB (URL: http://dx.doi.org/10.1117/1.JBO.21.8.086002.2)].

Fig. 6

Time series of STMD images (a) before, (b)–(d) during, and (e), (f) after occlusion of rabbit ear artery (Video 3). Imaging area is $20\mathrm{mm}\times 16.7\mathrm{mm}$. (g) and (h) are time traces of blood flow, where STMDs are average values in areas indicated by rectangles A and B in (a), “1” and “2” are at 8.26 and 13.47 s, respectively [Video 3, MOV, 8.09 MB (URL: http://dx.doi.org/10.1117/1.JBO.21.8.086002.3)].

4.

Discussion and Conclusion

Although the results of both the phantom and biological experiments demonstrate the abilities high-temporal-resolution of our method, the STMD image display is not in real time. For this purpose, graphics processing unit techniques could be employed in our system, since these could simultaneously acquire raw data and display STMD images. The exposure time of the camera on the STMD images is obtained significantly. When the flow speed is sufficiently fast, a substantial part of its energy will be aliased back into the direct current (DC)-band, which presumably could affect the MD measurement. In the future, a pulsed laser could be used as the illumination source to trigger the synchronization acquisition. For a pulsed laser with short pulse width and high peak power, an ultrashort integration time and a high illumination intensity could be obtained simultaneously. The above-mentioned experimental results indicate that our method could be applied to develop a microcirculation function imaging tool suitable for studying small animal brain models or eye retinas, e.g., to investigate ischemic thrombotic strokes or neurological responses to sensory simulations.¹⁷ It can also be used clinically to study cutaneous angiogenesis during wound healing. Since STMD images are mainly sensitive to move red cell concentrations, a dual-wavelength STMD imaging technique could be developed to perform simultaneous imaging of blood flow and hemoglobin oxygenation changes at high spatiotemporal resolutions.¹⁸ Another possible application is in the study of microfluidic systems, since this method could be used to control and monitor the reaction time and chemical concentration in each droplet precisely.¹⁹ However, the most distinct advantage of our method is that it enables high-temporal-resolution, full-field imaging for moving particle concentration change monitoring in transparent microfluidic systems.

In summary, we proposed a high-temporal-resolution, full-field optical angiography method that employs STMD as the imaging parameter. Phantom and biological VOTs were conducted and demonstrated that STMD images can be used to monitor vascular network changes with high temporal and spatial resolution. Our method could be developed into a functional imaging tool to study tissue hemodynamics in vivo or reaction kinetics in microfluidic systems.