2.1.

Surgical Procedures

All animal experiments were performed with a study protocol approved by and conducted in accordance with the guidelines of the White Oak Institutional Animal Care and Use Committee. Optical measurements were performed on 13 male C57BL/6J mice (12- to 24-weeks old). Anesthesia was induced by an initial exposure to 4% isoflurane (vaporized in medical grade compressed oxygen) for $<30\mathrm{s}$. Animals were additionally administered an injection of xylazine ($18\mathrm{mg}/\mathrm{kg}$, IP) to provide a stable plane of anesthesia at low isoflurane concentrations (0.1% to 0.2%) for the remainder of the experimental session, which typically lasted 3 to 6 h. Maintenance doses of xylazine ($6\mathrm{mg}/\mathrm{kg}$) were administered once every $\sim 2.5\mathrm{h}$. This anesthesia regimen mitigated the vasodilation effects known to occur with isoflurane administration. Following initial anesthesia induction, animals were positioned in a stereotaxic apparatus (David Kopf Instruments, California). Their body temperature was measured and maintained at 37°C with a closed-loop temperature-controlled heating pad (Model TC-1000, CWE). Respiratory rate was also monitored and maintained at $\sim 100\text{breaths}/\min$ during the surgical and experimental procedures. Skin incisions were infused with lidocaine, and the eyes were covered with ointment (Lacri-Lube) to prevent drying. A midline sagittal incision was made in the skin, which exposed the coronal and lambdoid sutures on the skull. The intersection of these sutures with the midline (i.e., bregma and lambda) served as a landmark for recording locations and was also used as a guide for surgery.

A craniotomy ($\sim 2.5\times 3\mathrm{mm}$) was performed using a handheld dental drill (Osaka) and a $2\times 2.5\mathrm{mm}$ window (sterilized #0 cover glass) was placed over the somatosensory cortex [Fig. 1(a)]. The gap between the outer edge of the glass window and the inner edge of the craniotomy was filled with surgical silicone adhesive (Kwik-Sil, World Precision Instruments, Florida), and the coverslip was then sealed to the skull with instant adhesive gel (Loctite 454, HenKel, Wisconsin). A metal head-bar was attached to the skull contralateral to the craniotomy by dental cement (C&B Metabond, Parkell, New York). Animals recovered overnight. The transparent region of the cortex beneath the window was $\sim 2\times 2\mathrm{mm}$.

Fig. 1

Overview of imaging region and approach for identifying regions of sensory activation. (a) The location of the recording window (blue square) on the head of a mouse relative to the underlying brain regions. The blue and red dots adjacent to the window correspond to the emitter and detector fibers from the DCS apparatus. (b) A representative en face OCT-A image of the cortical vasculature. $\text{Scalebar}=200\mu \mathrm{m}$. (c) The custom 3D-printed probe for simultaneous DCS and OCA imaging. (d) A photograph of a completed probe assembly.

2.2.

Evoking Sensory Responses

The median nerve contralateral to the recording location was stimulated via a pair of 27-gauge stainless-steel needles inserted subcutaneously into the forelimb of the animal. The electrical stimulus consisted of a train of 12 current pulses generated by a constant current stimulator (DS7A, Digitimer) with the following characteristics: amplitude 1-4 mA, frequency 3 Hz, pulse duration $200\mu \mathrm{s}$, and total pulse-train duration 4 s. Individual measurements of sensory-evoked changes in cerebral blood flow were separated by 30 s, which we empirically found to be the minimum intertrial duration that did not elicit monotonic alterations in the steady-state flow. At this temporal separation between trials, we did not observe habituation effects over the full duration of any preparation. To further avoid any potential habituation effects, the stimulus current was randomly selected between consecutive trials.

2.3.

DCS System

The basic theory underlying DCS has been extensively described in previous publications.^16,31 Briefly, when laser light migrates through tissue, the emerging intensity pattern, called a speckle pattern, is composed of bright and dark spots, which are caused by constructive and destructive interference of photons that traverse different path lengths. The intensity fluctuations of a single region, which are caused by interactions of scattered light with moving particles (i.e., red blood cells), can be used to extract information about blood flow. In particular, it is the autocorrelation decay that correlates with blood flow. Our DCS system has been previously described.³⁶ Briefly, it consisted of a long coherence length continuous-wave near-infrared laser (785 nm, CrystaLaser, Nevada), a photon-counting avalanche photodiode (APD) (SPCM-AQRH-12-FC, Excelitas, Quebec, Canada), and an autocorrelator signal processing board (Correlator.com, New Jersey). NIR excitation light was delivered to the brain with a multimode optical fiber ($200\text{-}\mu \mathrm{m}$ core diameter, Thorlabs, New Jersey), and the scattered light was detected with a single-mode fiber ($5\text{-}\mu \mathrm{m}$ core diameter) connected to the APD. The sampling rate of the system was 4.4 Hz.

2.4.

Imaging Probe Design for Simultaneous DCS and OCT-A

To permit simultaneous DCS and OCT-A angiographic portraits of blood flow, we fabricated a custom probe from semiflexible acrylate polymer using a three-dimensional printer (Objet 260 Connex3 printer, Stratasys, Minnesota) [Figs. 1(c) and 1(d)]. For the DCS modality, embedded microprisms (2 mm, Edmund Optics Inc., New Jersey) directed illumination light from the source fiber down to the head and backscattered light to the detector fiber. Midway between the source and detector fiber ports, which were separated by 5-mm, a $2\mathrm{mm}\times 2\mathrm{mm}$ aperture provided access for the OCT-A beam. All OCT-A images collected during stimulation were acquired through the aperture. During experiments, the probe was positioned such that the central clearing was directly above the implanted cranial window and sealed with silicone (Kwik-Cast, Silicone Elastomer, World Precision Instruments, Florida).

The design of the probe was guided by several factors. At the most basic level, we needed to ensure that the centroid of the forelimb representation on the cortex could be fully observed within the imaging window.³⁷ This region of activation has a diameter of $\sim 2\mathrm{mm}$; given preparation-to-preparation variability, we wanted to ensure that errors in window placement would not cause the probe to entirely occlude the response centroid. Another constraint, however, was that we wanted the OCT-A scan depth to match to the depth of DCS maximum sensitivity, which is influenced by the source–detector separation. Additional constraints on the probe geometry included the dimensions of the fibers themselves as well as the fact that the probe assembly needed to accommodate the window for OCT imaging. As a compromise, a source–detector separation of 5 mm was selected, which, based on previous studies, likely keeps relative blood flow measurement error below 12%.³⁸

2.5.

Modeling the DCS and OCT-A Sensory Evoked Response Waveforms

To extract quantitative features from the sensory evoked optical signals, we fit the observed waveforms to a canonical hemodynamic response function that uses two gamma density functions to approximate the hemodynamic response,³⁹ i.e.,

2.6.

OCT-A System and Image Processing

The OCT-A device (Fig. 2) used herein has been described previously,^34,40 apart from several key modifications described below for synchronization to the DCS instrument. The system uses a 1315 nm (center wavelength) superluminescent diode light source (Exalos, Schlieren, Switzerland) with a 3-dB bandwidth of 85 nm, which produces a theoretical axial resolution in air of $9\mu \mathrm{m}$. A custom-designed spectrometer (Wasatch Photonics Inc., Durham, North Carolina) with a 76-Hz line rate InGaAs linear detector (Sensors Unlimited Inc., Princeton, New Jersey) achieves a depth roll-off of $1.8\mathrm{dB}/\mathrm{mm}$ (in the range 0 to 2.5 mm). In configuring the system optics, we had to balance the lateral spot size required to resolve capillaries with a moderate depth-of-focus to match the volume from which the majority of the DCS signal derives. Therefore, we used a 19-mm focal length source collimator to produce a $\sim 3\text{-}\mathrm{mm}$ input beam diameter and a $10\times$ telecentric objective (LSM02, 7.5 mm working distance, Thorlabs Inc., Newton, New Jersey). The lateral resolution measured with a nanoparticle-embedded point spread function phantom was $\sim 9\mu \mathrm{m}$ (FWHM) and the depth-of-focus was $\sim 500\mu \mathrm{m}$.

Fig. 2

Hybrid optical apparatus (simplified) for simultaneous DCS and OCT-A. DC, dispersion compensation cube; FI, fiber interferometer; ${\mathrm{G}}_{\mathrm{x},\mathrm{y}}$, galvanometer scanners; M, mirror; ND, neutral density filter; SLD, superluminescent diode. The two modalities are combined through a custom 3D-printed probe placed on the region of the skull above the primary somatosensory cortex at a location coinciding with the forepaw representation. The sample is here represented by a depiction of a capillary network.

Custom OCT system software written in LabVIEW (National Instruments, Austin, Texas) was developed to control hardware, including source power control, galvanometers for scanning, data acquisition boards for waveform generation, and cameras and frame grabbers for image acquisition. The acquired OCT images were processed in near real-time on the system computer using custom CUDA-based code running on the graphics processing unit (GPU) of a high-performance video card (GeForce GTX-760, NVIDIA). For this study, a real-time axial motion correction algorithm was implemented in the GPU software, based upon subpixel alignment of adjacent A-scans using cross-correlation.

Two OCT videos were created for each acquisition: a reflectance video and an angiography video. Each reflectance video frame is a simple average of all reflectance B-scans collected at a given lateral position and time point, whereas each angiography video frame was created using the average absolute difference of the magnitude of the complex OCT signal for all reflectance B-scans from the same set.⁴¹ The angiography processing (average absolute difference), similar to other OCT speckle variance approaches,^31,34,42,43 separates the static DC OCT reflectance signal from the higher frequency OCT signal that arises from flowing blood. Note that the resulting measurements represent relative blood flow rather than an absolute measurement of blood velocity. The angiography frames were cropped to a 0.4-mm deep region-of-interest (ROI) within the system’s depth of field. We then used the Otsu method⁴⁴ (implemented with the multithresh function in Matlab) to segment blood vessels in each angiography frame into three sizes. A representative angiography frame is shown in Fig. 3(b). Once masks were created for each vessel size, the gray-scale values in the angiography frame for all pixels in each segmented region were summed. This value is a measure of the total blood flow in the segmented region, though not a direct measure of flow velocity without further processing.^35,43–47 Therefore, the flow values are normalized to the prestimulation period to generate the temporal response curves in relative units.

Fig. 3

Method for dissecting OCT-A signal contributions with Otsu thresholding. (a) A sample en face OCT-A image. $\text{Scalebar}=200\mu \mathrm{m}$. The red line indicates the location selected for collection of high-resolution B-scans. (b) The B-scan corresponding to the red line drawn in (a). The bright punctae correspond to cross-sections of small blood vessels. Superimposed on this B-scan image, (c) two populations of vessel cross-sections selected from among the regions identified using Otsu’s method. In particular, vessels of diameter 10 to $20\mu \mathrm{m}$ (in red) and $>20\mu \mathrm{m}$ (in blue) are shown.

2.7.

Multimodal Synchronization and Data Acquisition Protocol

The OCT system was configured to generate two additional waveforms to synchronize the OCT B-scan acquisition with the median nerve stimulation and the DCS acquisition. The total acquisition time for a single stimulation was 30 s. The first waveform drove the current stimulator with 12 TTL pulses at 0.2 ms duration and 3 Hz (4 s total stimulation duration), starting after a delay of 5 s. The second waveform triggered the DCS system acquisition synchronous to the first (stimulus) waveform. The OCT acquisition was set up to acquire B-scans with 250 A-scans (not including fly-back) at a rate of 240 B-scans/s (4.17 ms interval). In order to collect angiography images of sufficient contrast, 32 B-scans at each lateral location were acquired in 133 ms, where the average absolute difference between B-scans was used to generate the angiography images.⁴¹ The total acquisition and GPU processing time for each composite B-scan was 460 ms, so the OCT acquisition software was set up to acquire 65 composite B-scans during the 30-s stimulation duration, where the stimulus was on from the $\sim 11$ to 20 B-scans of the acquisition. OCT and DCS acquisition and stimulus waveforms were triggered synchronously.

Each animal was first prepared with anesthesia, coarsely aligned on the stage beneath the objective, and stimulus electrodes were inserted. The focus was set $\sim 200\mu \mathrm{m}$ beneath the window and the window precisely aligned to the OCT beam using the real-time display. Any angle in the window was also adjusted using the two-axis goniometer stage to which the head-bar was affixed. Once aligned, a wide-field OCT scan was acquired to guide probe placement over the somatosensory ROI and further image/data collection. The wide-field $2\times 2\mathrm{mm}$ OCT scan ($4\mu \mathrm{m}/\text{pixel}$) covered the entire cortical region beneath the window and used 10 B-scans at each location to generate the angiography image. The DCS probe was not used and stimulation was not activated during this scan, which took $\sim 2.5\min$ to acquire. After the wide-field scan was acquired, the DCS probe was placed over the somatosensory region using the $2\times 2\mathrm{mm}$ OCT scan and the surgical photos as a guide. The OCT beam was then aligned through the center channel.

2.8.

OCT-A En Face Video Survey of Evoked Blood Flow

To select the optimal target location for simultaneous acquisition of OCT B-scans and DCS signals during sensory-evoked stimulation, we created an en face OCT video survey of blood flow activation during somatosensory stimulation over an ROI spanning $1\times 1\mathrm{mm}$ ($4\mu \mathrm{m}/\text{pixel}$), limited by vignetting of the probe aperture. This region was probed densely in one lateral direction (250 A-scans/B-scan over 1 mm), coarsely in the perpendicular lateral direction (10 B-scan locations over 1 mm), and averaged temporally where 10 somatosensory stimulus presentations were collected at each B-scan location. The en face OCT video survey took 50 min to acquire. Data were then processed offline using custom Matlab code (Mathworks Inc., Natick, Massachusetts) to generate an en face response survey video as follows. First, at each lateral location, the B-scans collected for the 10 stimuli were averaged. Second, the B-scan were reduced in the dense lateral and axial dimensions by accumulating all pixel values in a $10(\text{width})\times 100(\text{depth})$ pixel region. This reduced each B-scan region from 250 A-scans ($100\text{pixels}/\mathrm{A}\text{-}\text{scan}$) to a 25-pixel vector for each location. Third, each of the 65 vectors over time was normalized to the average of the first nine vectors (representing the first 4 s prior to the stimulus onset 5 s after the acquisition start). Finally, the 10 locations in the coarse lateral dimension were expanded by interpolation to 25. This resulted in the final en face OCT video where each frame is $25\times 25\text{pixels}$ with 65 frames over the 30-s image collection duration. This procedure is depicted in more detail in Fig. S1 in the Supplementary Materials, and an example of the response video is depicted in Fig. 4 and shown in Video 1. It should be noted that the en face surveys were used exclusively for selecting the primary recording location, and the actual quantification of vessel diameter-dependent signal was based on the segmented B-scans.

Fig. 4

OCT-A en face temporal survey of cerebral blood flow patterns following somatosensory stimulation. $\text{Scalebar}=200\mu \mathrm{m}$. (a) The size of the survey field-of-view relative to the full-field en face scan. (b) The general scanning strategy and a representative individual movie frame. (c) Sample extracted OCT-A frames of blood flow patterns during stimulation. The OCT-A survey videos, an example of which is Video 1, have a temporal resolution of $0.46\mathrm{s}/\text{frame}$. The red bar below the frames indicates the period during which the forelimb was stimulated. Color bar represents changes in blood flow relative to the average of 4 s prior to the stimulus (Video 1, mp4, 645 KB [URL: https://doi.org/10.1117/1.NPh.8.2.025006.1]).

2.9.

Simultaneous OCT-A and DCS of Evoked Blood Flow

After the response video was examined, the best single lateral location was selected as the ROI, and the animal stages were moved to collect the remainder of the scans during the experimental session. Nine additional OCT B-scan sets spanning 1 mm in one axis (250 A-scans/B-scan) were collected with simultaneous DSC signal acquisition: three trials each at three different evoked stimulus currents (1, 3, and 4 mA, collected in random order), where each trial included 10 stimulus presentations and took 5 min to acquire. The resultant B-scans were averaged over the 10 stimuli, segmented according to vessel diameter with a binary mask using Otsu’s thresholding method over the B-scan focal region ($1\mathrm{mm}\text{width}\times 0.4\mathrm{mm}\text{depth}$), and each timepoint normalized to the average of the corresponding baseline B-scan values (first 4 s) to produce a temporal response curve for each trial. The entire experimental procedure took about 3 to 4 h to complete, including time for setup, alignment, and reconfiguration between scans.

2.10.

Data Analysis

Experimental data were included in analysis if the animal survived and maintained steady physiological state during the full course of the recording session. This criterion excluded one animal, and an additional animal was excluded from our analysis because of an error during data acquisition. In all, 11 out of 13 animals were included in our data analysis. Significance was tested using two-tailed Student’s $t$-tests. Normality of the data distribution for any group average was tested using a one-sample Kolmogorov–Smirnov test prior to significance analysis.

3.1.

Locating the Centroid of the Hemodynamic Response Using OCT-A

While the mouse somatosensory cortex has a relatively stereotypical representation of the forelimbs,⁴⁶ the exact centroid location of the functional hemodynamic response varies among animals. Given that the OCT-A-measured hemodynamic response waveform can differ significantly depending on the relative distance to the response centroid,⁴⁷ it was important to ensure that our relative recording location was as consistent as possible among all experiments. This presented a challenge given that the peak of the functional response was rapid, on the order of 6 s following stimulus, and obtaining a full en face image such as the one shown in Fig. 2(b) required a scan time of 2.5 min (for a single stimulus and with reduced OCT-A averaging). As a compromise, we subdivided the field of view into a grid of subregions, in each of which we performed more dense B-scans, as described in Sec. 2. Because this sparser, more distributed sampling could be achieved relatively rapidly ($\sim 50\min$), we were able to effectively create a topological survey of activity (Fig. 4). We used this approach to identify the regions of maximal response and thus select the primary location to acquire higher pixel density B-scans.

3.2.

Sensory-Evoked Blood Flow in the Cerebral Cortex Varies as a Function of Vessel Diameter

As shown in Fig. 3, OCT-A imaging resolved cross-sections of blood vessels of different diameters. Using Otsu’s thresholding method,⁴⁴ we created binary masks that were then used to dissect flow measurements from different vessel populations, including diameters of 10 to $20\mu \mathrm{m}$ and $>20\mu \mathrm{m}$. Because the axial and lateral resolutions of the OCT-A device were $\sim 9\mu \mathrm{m}$, smaller features were not well resolved using thresholding. Consequently, the regions of the B-scans that yielded non-zero flow but did not overlap with other vessel populations were attributed to vessels of diameter $<10\mu \mathrm{m}$.

Integrating the sensory-evoked blood flow selectively in the three vessel diameter ranges within the OCT-A recordings yielded hemodynamic responses with qualitatively different waveform attributes (Fig. 5). Averaged among all animals ($n=11$), the most prominent features for both DCS and OCT-A measurements were positive and negative peaks at $\sim 8$ and $\sim 15\mathrm{s}$, respectively. The relative flow amplitude of the positive peak, resolved by OCT-A at a stimulation current of 1 mA, was $2.19\pm 0.8\%$ ($\text{mean}\pm \mathrm{SD}$) in the areas putatively representing vessels of diameter $<10\mu \mathrm{m}$ (Fig. 6). This was more than double the response summed among regions putatively representing the largest vessel diameters, $0.92\pm 0.8\%$. Meanwhile, for the negative peak, the amplitude magnitude in the same large vessel population $1.96\pm 0.86\%$ was $\sim 2.3$ times greater than the flow in the putatively $<10\mu \mathrm{m}$ diameter vessels ($0.86\pm 1.26\%$). Overall, in our experiments, the maximum positive peak amplitudes in OCT-A were found for the $<10\mu \mathrm{m}$ diameter group at a stimulus current of 4 mA, whereas the maximum amplitude for the undershoot was found in the 10- to $20\text{-}\mu \mathrm{m}$ diameter vessel group at a stimulus current of 3 mA.

Fig. 5

Image segmentation-based dissection of OCT-A signals reveals that hemodynamic response features depend on vessel diameter. Traces on the left-most column depict DCS measurements of somatosensory-evoked hemodynamic responses from the grand average of all experiments. Each trace represents the average of 11 animals, and the three traces aligned vertically depict the responses to increasing stimulation current intensities. The three columns of traces on the right depict spatially segmented components of the OCT-A response. The traces at each stimulus current for both DCS and OCT-A were recorded simultaneously.

Fig. 6

Side by side comparison of simultaneously measured DCS and OCT-A hemodynamic responses, averaged across all animals ($n=11$). The top waveform depiction illustrates how waveforms were quantified; the red dashed line is a dual gamma function fit (see Sec. 2) to the evoked DCS blood flow response (blue trace) in one experiment, which was selected for illustration because of its prominent biphasic features. The blue and green arrows show the peak locations at which amplitudes were assessed. Note that the exact time points varied from experiment to experiment depending on the fit. Analyzed using image segmentation, putative vessels of $<10\mu \mathrm{m}$ diameter in OCT-A imaging results featured the most prominent positive initial peaks, which occurred $7.3\pm 0.3\mathrm{s}$ after the stimulus ($\text{mean}\pm \text{standard deviation}$). Putative vessels of larger diameter were relatively unresponsive during the initial positive phase and did not demonstrate significant stimulus intensity dependence. The amplitude of the subsequent undershoot, however, was greater in these putative vessels with larger diameters. In DCS signals, the initial positive peaks demonstrated a greater stimulus intensity dependence between 1 and 4 mA compared with the responses during the undershoot period. Error bars represent $\text{mean}\pm \text{standard deviation}$. The brackets and asterisks above the data points depict the statistical significance of the difference in the grand averaged response compared with the response to 1 mA stimulation. Asterisks indicate the $P$-value from a two-tailed Student’s t-test as follows: * indicates $P<0.05$, ** indicates $P<0.01$, *** indicates $P<0.001$.

Compared with the blood flow waveforms measured in small vessels with OCT-A, the blood flow waveforms obtained through DCS had a more pronounced undershoot. For example, at 1-mA stimulus current in OCT-A measurements, the positive peak amplitude of the putatively $<10\mu \mathrm{m}$ diameter vessels was $\sim 2.5$ times greater than the vessel population’s average undershoot amplitude. At the same stimulus current, the positive peak magnitude of the DCS signal was only $\sim 10\%$ greater than the measurement undershoot ($3.37\pm 1.67\%$ and $-3.04\pm 1.83\%$, respectively). As an additional contrast, while the magnitudes of the DCS signal were within an order of magnitude of the OCT-A counterparts, the DCS peaks were broader. This can be visualized in terms of the integrated signal, as seen in Fig. 7. The difference was most prominent for the undershoot, where the integrated DCS values were larger than their OCT-A counterparts at any stimulus current and vessel diameter population.

Fig. 7

Integrated waveform features accentuate the diameter-dependence of response amplitude during positive and negative peaks post stimulus. As in Fig. 5, the top DCS waveform depiction illustrates how waveforms were quantified. The dashed red trace superimposed on the blue trace is a double gamma function fit. Superimposed on the curve are red and blue shaded regions that correspond to the positive and negative peaks, respectively. These regions are defined by the points at which the blood flow waveform crosses above/below the baseline value. The exact time points that defined these positive and negative phases varied from experiment to experiment depending on the fit. Bar charts represent the average over all animals ($n=11$). Error bars represent mean ± standard deviation. The use of brackets and asterisks is as in Fig. 5.

3.3.

Correlation Between Evoked Blood Flow Waveforms in DCS versus OCT-A-Derived Vessel Subpopulations

As shown in Fig. 5, the flow waveform varied considerably between different vessel populations. As one strategy for assessing underlying vascular architectural substrates for the DCS signal, we calculated the Pearson’s correlation coefficient of DCS waveforms with the concurrently recorded OCT-A signals. Figure 8 depicts correlation values, averaged over all animals, of DCS versus OCT-A flow waveforms integrated among putative vessel subpopulations of differing diameter. Overall, the mean correlation coefficients for all three vessel diameter ranges increased when the stimulus current was increased from 1 mA. The highest correlation value, $0.68\pm 0.49$, was observed in the putatively $<10\mu \mathrm{m}$ diameter population at a stimulus current of 4 mA. Within this small diameter vessel population, the correlation value measured at a stimulus current of 3 mA represented a significant increase in correlation coefficient relative to the waveform elicited by 1 mA stimulus current ($P=0.04$ using a two-tailed Student’s $t$-test). At 4-mA stimulus current, the correlation values of the 10 to $20\mu \mathrm{m}$ and $>20\mu \mathrm{m}$ groups were each statistically different from that of the $<10\mu \mathrm{m}$ diameter group ($P<0.05$ using a two-tailed Student’s $t$-test).

Fig. 8

Correlation values, averaged over all animals, of DCS versus OCT-A flow waveforms integrated among putative vessel subpopulations of differing diameter. The values represent Pearson’s correlation coefficient comparing full, 30-s waveforms. Correlation coefficient values were moderate for all vessel populations, and highest among the small vessels. The putatively $<10\mu \mathrm{m}$ diameter vessels demonstrated a significant increase in correlation coefficient when forelimb stimulus intensity was increased from 1 to 3 mA. Values are $\text{mean}\pm \text{standard deviation}$, and the asterisk* indicates that $P<0.05$ when using a two-tailed Student’s $t$-test to assess statistical differences between the correlation coefficient $\text{mean}\pm \text{standard deviation}$ of the 3 mA response versus that of the 1 mA response. ★ indicates that the correlation coefficient ($\pm \text{standard deviation}$) differs significantly from that of the $<10\mu \mathrm{m}$ diameter vessels at 4 mA stimulus current, also assessed using a criterion of $P<0.05$ obtained through a two-tailed Student’s t-test.