1.

Introduction

Noninvasive blood flow imaging can provide critical information on the state of biological tissue and the efficacy of approaches to treat disease. Fluorescein angiography is used routinely to study blood flow dynamics in the retina.¹ Laser Doppler flowmetry and laser Doppler imaging have been applied in numerous preclinical and clinical studies on the brain,^{2, 3} retina,^{4, 5} skin,^{6, 7} and joints.⁸ Principles of Doppler shift imaging have been combined with optical coherence tomography to provide high-resolution cross-sectional images of blood flow. ^{9, 10, 11, 12, 13, 14, 15, 16, 17}

In 1981, Fercher and Briers¹⁸ proposed a laser speckle imaging (LSI) approach as an alternative to laser Doppler imaging. On the basis of this study, it was concluded that variations in speckle contrast can be used to provide directly a wide-field velocity distribution map. LSI has since been employed for wide-field blood flow imaging of the brain, ^{2, 19, 20, 21, 22} retina,^{18, 23} and skin, ^{6, 24, 25, 26} with high temporal resolution. Briers²⁷ has demonstrated that LSI and laser Doppler imaging provide equivalent assessments of blood flow.

With LSI, relative changes in blood flow are typically reported, with the assumption that the measured values are on a linear scale. A linear relationship between the measured and actual flow rate values has been suggested.^{2, 19} The actual flow rate range over which this linear relationship is valid is unknown. Knowledge of the relationship between measured and actual blood flow values is important because researchers would then know to which in vitro and in vivo flow models LSI could be applied to quantify relative flow changes.

The primary objective of this paper is to provide insight into the relationship between LSI measurements of speckle flow index with actual flow rate. The secondary objective is to demonstrate application of LSI to an in vivo rodent dorsal skinfold model and to show its potential as a wide-field microvascular imaging modality in a widely used animal model. Based on the analysis of Yuan, ²¹ the central hypothesis was that image integration time would affect the overall velocity dynamic range and signal-to-noise ratio of LSI.

2.

Materials and Methods

2.2.

In Vivo Application of LSI

The objective of this study was to demonstrate in vivo application of LSI. To accomplish this objective, we applied chemical agents to the microvascular network of the rodent dorsal skinfold chamber model. ^{25, 30, 31, 32, 33, 34, 35, 36, 37}

2.2.1.

Experimental Design

Rodent dorsal skinfold chamber model. Surgical installation of a dorsal skinfold window permitted observation of full-thickness skin from both the epidermal and subdermal sides (Fig. 1 ). The animal was anesthetized with a combination of Ketamine and Xylazine (4:3 ratio, $0.1\mathrm{g}∕100\mathrm{g}$ body weight) administered by intraperitoneal injection. A section for window placement on the back of the animal was selected, surgically scrubbed, shaved, and depilated. Sutures attached to a temporary mount retracted the dorsal skin away from the animal’s body. A circular section with an approximate diameter of $1\mathrm{cm}$ was cut from one side of the symmetrical skinfold, thus exposing blood vessels in the underlying skin. An aluminum chamber was sutured to both sides of the skin and the sutures cut to release the skin from the mount. To prevent dehydration, saline was applied periodically to the subdermal skin. The surgery was performed as defined in a protocol approved by the University of California, Irvine, Animal Use Committee.

Fig. 1

Representative subdermal image of a dorsal skinfold window chamber preparation.

Chemical agents

We were interested in studying microvascular flow during the application of a recently developed polyurethane agent designed to reduce temporarily optical scattering in biological tissues such as skin.³⁸ The agent was applied directly to the subdermal microvascular network and LSI images acquired over a $15\text{-}\min$ period from the subdermal side. Due to lateral diffusion of the agent away from the imaged microvascular network after application, the agent was reapplied on a regular basis during the image acquistion period. In separate experiments, each on a different animal, phosphate buffered saline and $13.6\mathrm{M}$ glycerol were applied as negative and positive controls, respectively, for inducing microvascular flow changes.

Image acquisition and processing

Raw speckle images were acquired and SFI images computed as described above. White light reflectance images were also acquired to serve as anatomical reference images. To visualize changes in blood flow induced by agent application, the SFI images were converted to binary images using the iterative selection method.³⁹ This automated method was used to determine a threshold SFI value based on an initial estimate ${\mathrm{SFI}}_0$ . The mean SFI values of all pixels below $\left({\mathrm{SFI}}_{below}\right)$ and above $\left({\mathrm{SFI}}_{above}\right)$ ${\mathrm{SFI}}_0$ were calculated and a new estimate ${\mathrm{SFI}}_{new}$ computed as the arithmetic mean of ${\mathrm{SFI}}_{below}$ and ${\mathrm{SFI}}_{above}$ . This process was repeated with each new estimate until ${\mathrm{SFI}}_{below}$ was equal to ${\mathrm{SFI}}_{above}$ .

With each experiment, a threshold SFI value was determined for the SFI image acquired immediately after agent application and was applied to binarize SFI images acquired at later time points. To visualize better arteriolar and venular blood flow dynamics, the binary SFI images were superimposed on the corresponding white light anatomical images.

3.

Results

Camera integration time impacts the relationship between measured SFIs and actual flow rate (Fig. 2 )

With an increase in integration time from 1 to $10\mathrm{ms}$ , the following trends were observed with the in vitro flow phantom: (1) the linear response range decreased from 2 to $20\mathrm{mm}∕\mathrm{s}$ , to 0 to $5\mathrm{mm}∕\mathrm{s}$ ; (2) the best case velocity resolution, defined as the reciprocal of the slope of a linear fit to the data, decreased from 2 to $5\mathrm{\mu }\mathrm{m}∕\mathrm{s}$ ; and (3) as shown by the correlation coefficient of the linear fits, the amount of scatter in the data decreased. Collectively, these data provide us with knowledge on the actual flow rate range over which our LSI instrument has a linear response.

Fig. 2

Effect of camera integration time on the linear response range between SFI and actual flow rate. Data acquired with integration times of (a) 1 and (b) $10\mathrm{ms}$ are presented. The data show that an increase in integration time decreases the linear response velocity range and velocity resolution and increases the linear regression fit to the discrete data points. In (a), the full set of acquired data is shown, and the linear fit is restricted to actual flow rate values of 2 to $20\mathrm{mm}∕\mathrm{s}$ .

Wide-field, arteriolar, and venular blood flow dynamics are readily monitored during exogenous chemical agent application

Representative white light, SFI, and superimposed white light–binary SFI images of the rodent dorsal skinfold window chamber are shown in Fig. 3 . A good correspondence between visible blood vessels is evident from a comparison of the white light and SFI images and in the superimposed image. By applying the same threshold SFI value to all images acquired in a time-resolved sequence, blood flow dynamics are readily visualized in a qualitative fashion. With direct application of saline (i.e., negative control) to the microvascular network for $16\min$ , a slight increase in blood flow was observed, primarily in the smaller arteriolar and venular branches (Fig. 4 ). With direct application of glycerol (i.e., positive control) to the microvascular network, a gradual decrease in blood flow was observed (Fig. 5 ). From the superimposed white light–binary SFI image, we observed an immediate decrease in venular flow and a delayed decrease in arteriolar flow (between 12 and $16\min$ after glycerol application). With direct application of the polyurethane agent to the microvascular network, a slight increase in blood flow was observed in the smaller vessel branches (Fig. 6 ), similar to that observed with saline application. Collectively, these data demonstrate the potential of LSI to allow wide-field monitoring of heterogeneous blood flow dynamics.

Fig. 3

Representative (a) white light reflectance, (b) SFI, and (c) superimposed white light–binary SFI images of the microvascular network in the rodent dorsal skinfold chamber. Image dimensions are $9\times 6.8{\mathrm{mm}}^2$ .

Fig. 4

Representative (a) white light reflectance and (b, c) superimposed white light–binary SFI images at 0 and $16\min$ , respectively, after direct saline application to the microvascular network of the rodent dorsal skinfold chamber. The images show a slight increase in blood flow in the smaller blood vessel branches [(arrows in (c)]. Image dimensions are $9\times 6.8{\mathrm{mm}}^2$ .

Fig. 5

Representative (a) white light reflectance and (b to f) superimposed white light–binary SFI images at various time points after direct glycerol application to the microvascular network of the rodent dorsal skinfold chamber. The images show a heterogeneous decrease in blood flow, with an initial decrease in venular flow [“V” in (a)] and a delayed decrease in arteriolar flow [“A” in (a)]. Image dimensions are $9\times 6.8{\mathrm{mm}}^2$ .

Fig. 6

Representative (a) white light reflectance and (b to f) superimposed white light–binary SFI images at various time points after direct polyurethane application to the microvascular network of the rodent dorsal skinfold chamber. The images show a slight increase in blood flow in the smaller blood vessel branches. Image dimensions are $9\times 6.8{\mathrm{mm}}^2$ .

4.

Discussion

We are primarily interested in studying hemodynamics in in vivo microvascular network animal models, such as the dorsal skinfold window chamber preparation. Before we could use LSI reliably as a quantitative tool to assess blood flow in these models, we first needed to study the relationship between SFI and actual flow rate. In this study, we identified a range over which the relationship is linear, for two commonly used camera integration times. Specifically, we have shown that, for integration times of 1 and $10\mathrm{ms}$ , the linear relationship holds for flow rates of 2 to 20 and 0 to $5\mathrm{mm}∕\mathrm{s}$ , respectively (Fig. 2). There is slightly more scatter in the 1-ms integration time data than that at $10\mathrm{ms}$ and hence more deviation from the linear regression fit. A similar trend in improved speckle contrast-to-noise ratio was observed by Yuan ²¹

Our data suggest that camera integration time should be selected with the anticipated velocity range to be evaluated. Vargas ¹⁵ used optical Doppler tomography to measure blood flow velocities of $\sim 4\mathrm{mm}∕\mathrm{s}$ in arterioles and venules in the rodent dorsal skinfold model. From blood flow rates measured in arterioles and venules of the rodent cranial window model, velocities of less than $5\mathrm{mm}∕\mathrm{s}$ were anticipated.³² To ensure that a linear relationship is maintained between measured speckle flow index values and actual flow rate, our data suggest an integration time of approximately $10\mathrm{ms}$ should be used to monitor flow rate dynamics directly in arterioles and venules. From the analysis of Yuan, ²¹ the speckle contrast to noise ratio is only $\sim 8\%$ less at an integration time of $10\mathrm{ms}$ than the maximum ratio at $5\mathrm{ms}$ , which we believe is an acceptable compromise to ensure measurements are acquired over a linear response range.

The data obtained in this study can be used to estimate a velocity dynamic range (VDR) for our LSI instrument¹⁰

We performed a series of experiments (Figs. 4 to 6) to demonstrate the potential of LSI for wide-field monitoring of blood flow dynamics in an in vivo microvascular network model. We observed a heterogeneous decrease in blood flow after glycerol application (Fig. 5), with an initial decrease in venules and a delayed decrease in arterioles. A similar trend was observed by Vargas ¹⁵ with optical Doppler tomography monitoring of a single arteriole-venule pair. A potential advantage of LSI over such a technique is the ability to interrogate multiple vessels and associated branches simultaneously.

A surprising finding was that the blood flow rate increased slightly with application of phosphate buffered saline (Fig. 4) and the polyurethane agent (Fig. 6). The thermoregulatory aspect of cutaneous blood flow may have led to this increase, because the agent temperature $(\sim 25°\mathrm{C})$ was slightly lower than the in vivo skinfold temperature $(\sim 28°\mathrm{C})$ . Another possibility is that the reduced access of the subdermal tissue to ambient oxygen due to the addition of saline may have led to a compensatory increase in blood flow to reestablish steady-state oxygenation conditions. The strong flow reduction effect of glycerol was probably sufficient to overcome any compensatory increase in blood flow due to either effect.

Our current LSI instrument is capable of raw speckle image acquisition rates of $\sim 100$ frames per s, which corresponds to a temporal resolution of $\sim 10\mathrm{ms}$ . The intent of the current in vivo imaging study was to compare our data (Figs. 4 to 6) with previously published data,¹⁵ so we acquired image sequences during agent application at a considerably slower frame rate of 0.5 frames per min. Previously, we have acquired raw speckle images within seconds after pulsed dye laser irradiation of select blood vessels,²⁵ demonstrating the potential of LSI to study rapid transients in blood flow.

With our optical imaging configuration, we acquired images at unity magnification to provide a raw speckle image lateral resolution of $6.45\mathrm{\mu }\mathrm{m}$ . To achieve this spatial resolution, we selected the f-number of the macro lens so that the speckle size was equivalent to that of a single pixel²

Theoretical²⁷ and experimental² studies have established that LSI and laser Doppler imaging provide equivalent relative flow values. Thus, if the same laser wavelength is used, both methods interrogate perfusion within equivalent volumes of tissue. The primary advantages of LSI include its relative simplicity (i.e., no need for scanning mechanical components) and its image acquisition rate. LSI images can be acquired at the maximum camera frame rate (typically $30+$ frames per s for $300+$ kilopixel images) and processed offline. With laser Doppler imaging, acquisition times of several minutes are required to achieve a 65-kilopixel image.⁶ Such prolonged acquisition times are associated with a considerably higher risk of motion artifact. Laser Doppler imaging does provide information on the direction of blood flow, a capability that is not present with conventional LSI. These considerations should be taken into account during selection of the appropriate wide-field blood flow imaging method.

The LSI velocity dynamic range and signal-to-noise ratio reported here provide an imaging platform that can be used to study normal and tumor microvasculature in the rodent dorsal skinfold model. Angiogenesis associated with tumor growth has been studied extensively with this animal model.^{31, 32, 37} In such studies, blood flow dynamics are typically monitored with point laser Doppler probes or direct red blood cell counting algorithms. Coupling of LSI with other optical imaging modalities, such as reflectance,²² fluorescence, and multiphoton microscopy, may enhance knowledge of critical factors such as heterogeneities in tumor blood flow and local oxygen consumption dynamics.

In summary, we have shown that SFI and actual flow rate have a linear relationship at typically used LSI integration times and at blood flow rate values typical of arterioles, venules, and capillaries. With an increase in integration time, the VDR decreases and signal-to-noise ratio increases. These findings imply that LSI can provide accurate wide-field maps of microvascular blood flow rate dynamics and highlight heterogeneities in flow response to the application of exogenous agents.