2.1.

Tissue Phantoms

To model biological tissue, 10 phantoms were made by heating 1% TCBS agar (Sigma-Aldrich, Missouri) and 2% intralipid diluted from 20% stock (Sigma-Aldrich, St. Louis, Missouri) in water. Intralipid mimics the scattering properties of biological tissue,^22,23 and agar solidifies the resultant suspension upon cooling. The heated mixture was cooled in chambers containing two microtubes to simulate blood vessels. Five prepared phantom chambers contained microtubes with internal diameters of $600\mu \mathrm{m}$, and five more with internal diameters of $300\mu \mathrm{m}$. A double-syringe high-precision pump (Model NE-4002X, New Era Pump Systems, Farmingdale, New York) was used to fill the two microtubes simultaneously; OCT measurements under both no-flow and flow conditions were then possible. The average blood flow velocities in both microtubes were 0, 0.5, 1, 2, 3, 5, 8, and $12\mathrm{mm}/\mathrm{s}$ corresponding to venules, capillaries, and arterioles.^24–26 One syringe was filled with blood, the other with the mixture of blood and MBs. Rotating magnets inside both syringes were used to avoid red blood cells (RBC) settling and MBs floating during all experiments.

2.2.

Human Blood

Ethics approval was granted by the Research Ethics Boards of Ryerson University and the Canadian Blood Services. All experiments were conducted using whole blood donated by healthy volunteers recruited by the Canadian Blood Services’ Network Center for Applied Development (Vancouver, BC, Canada). The whole blood was 1-week old and contained 14% citrate phosphate dextrose anticoagulant.

2.3.

Microbubbles

Commercially available Definity^® microbubbles (perflutren lipid microspheres, Lantheus Medical Imaging, Billerica, Massachusetts), FDA approved for diagnostic ultrasound contrast, were used in this study. The mean diameter of MBs was $1.2\pm 0.5\mu \mathrm{m}$. A 2.5% volume fraction of a stock suspension (average $\sim 1.2\times {10}^{10}\mathrm{MBs}/\mathrm{mL}$) was injected into the blood microtube. The resultant MB concentration was comparable to that used in ex-vivo ultrasound tests.²⁷

2.4.

Optical Coherence Tomography System

A home-built and previously described swept source OCT system²⁸ was used in this study. Briefly, a 36 kHz short cavity laser source with a polygon-based tuneable filter and wavelength centered at 1310 nm had a sweep range of 110 nm and an average output power of 40 mW. The system contained a $50/50$ Mach-Zehnder interferometer and an imaging probe mounted on three-axis micropositioning stage in the sample arm. A dual-balanced photo detector was used in the output arm of the interferometer for DC signal component removal. A fiber Bragg grating provided the A-scan trigger at the beginning of each laser sweep. An OCT signal recalibration was performed with a standard Mach–Zehnder recalibration interferometer to generate linear k-space data from the raw data sampled with a $125\mathrm{MS}/\mathrm{s}$ data acquisition card. After the data were equally spaced in k-space, it was inversely Fourier transformed for further analysis. The axial resolution (in air) of the system was $8\mu \mathrm{m}$. The experimental setup is shown in Fig. 1.

Fig. 1

Experimental setup.

2.5.

Data Acquisition and Processing

OCT imaging was performed for both no-flow and flow conditions for blood (left microtube on subsequent images) and blood + MB mixture (right microtube). Data were recorded using Thorlabs software (Thorlabs Inc., Newton, New Jersey) in one-dimensional (1-D), two-dimensional (2-D), and 3-D modes.

In 1-D, A-scans were acquired from both microtubes to quantify MB-induced differences in OCT signal intensities. The autocorrelation function (ACF) was calculated from 18 pixels along the chosen A-line (shown in Fig. 2 as a vertical white dashed line) within the region of interest (ROI) of the corresponding microtube. The average of speckle decorrelation times ($1/e$ level) of these 18 pixels is displayed in Fig. 3. Autocorrelation was considered as the correlation between intensity values $I_i$ of the same pixel at times $t_i$ and $t_{i+\tau }$ with increasing time delay $\tau$:

Fig. 2

Typical structural B-scan of the phantom. The left microtube was filled with blood, and the right with blood + MBs [2.5% volume fraction of a stock suspension (average $\sim 1.2\times {10}^{10}\mathrm{MBs}/\mathrm{mL}$) were injected into the blood microtube]. ROIs of blood and blood + MB mixture are labeled with half-circles at the top circular section of microtube lumens. Background ROI was chosen at the same depth as shown. The vertical dashed line indicates the location of the acquired A-scans for ACF calculation (Fig. 3).

Fig. 3

Decorrelation time plots for blood and blood + MB mixture under eight flow conditions in $300\text{-}\mu \mathrm{m}$ tube.

2-D images (B-scans) of each phantom were acquired with 4 mm of lateral scanning. MATLAB software (MATLAB, Mathworks, Natick, Massachusetts) was used for data processing. Images of both blood and blood + MB microtubes were used to quantify the difference in resultant phantom contrasts. Since the majority of the signal from the bottom of the microtube was low due to high scattering of blood, the ROI was chosen as the top-half circular section of the microtube lumen (see Fig. 2). The averaged ROI intensity was used to quantify the contrast via

3-D OCT images were acquired in sv mode in order to obtain sv average intensity projections for $3\times 1\times 1{\mathrm{mm}}^3$ data volume as previously described,²⁸ with the number of sequential same location B-scans equal to 8, and 25 ms step size in slow-axis ($Y$) direction. svOCT calculates the interframe intensity variance of structural images, where contrast is based on the difference in time-varying properties. Specifically,

3.1.

One-Dimensional Decorrelation Analysis

The ACFs of OCT signals with and without MBs at no-flow and varying flow conditions are plotted in Fig. 3. The results from the $300\text{-}\mu \mathrm{m}$-diameter microtube are shown; those from the larger microtube show similar trends with slightly higher values at low flows. As expected, the decorrelation time decreases with blood flow velocity. The decorrelation times were typically $\sim 15\mathrm{ms}$ at no-flow condition, decreasing to $\sim 1\mathrm{ms}$ for velocities $>5\mathrm{mm}/\mathrm{s}$. As seen, the presence of MB causes faster decorrelation, and this differential effect diminishes with increasing blood velocities. Thus the potential utility of these MB contrast agents, at least as suggested by this 1-D analysis, appears enhanced in the slower flow regimes. This may be relevant for improved visualization of smaller vessels with slower blood velocities with svOCT imaging.

The reasons for the observed decrease in decorrelation times may be twofold: (i) in the no-flow regime, larger particles (RBC) undergo slower Brownian motion than the scatterers in the blood + MB mixture, with the latter’s smaller average diameter of $\sim 1.2\mu \mathrm{m}$ increasing the rapidity of temporal variation and signal intensity fluctuations; (ii) in the flow regime, a given MB ensemble is continuously being replaced by another, resulting in a higher temporal variation of the scattered OCT signal and higher intensity fluctuations for a given applied velocity (flow rate). As seen from Fig. 3, the former effect is considerably stronger than the latter.

3.2.

Contrast Enhancement of Two-Dimensional Structural Images

Structural OCT images, corresponding image contrasts, and ratios of blood + MB to blood contrasts acquired for different flow conditions are shown in Fig. 4. Although structural snapshot images show little visual evidence of MB contrast enhancement in Figs. 4(a) and 4(d), their averaged ROI intensities used for $C_{\mathrm{W}}$ calculation via Eq. (2) clearly demonstrate the contrast difference for both microtube diameters in Figs. 4(b) and 4(e). Dynamic OCT signal changes better reflect the contrast difference as shown in standard deviation image [Fig. 4(c)], calculated from eight subsequent structural images at $0.5\mathrm{mm}/\mathrm{s}$ blood flow velocity.

Fig. 4

Structural images, corresponding contrast of blood and blood + MB mixture and ratios of blood + MB to blood contrasts under eight flow conditions for 600- and $300\text{-}\mu \mathrm{m}$ microtubes. (a) Structural image of the phantom with $600\text{-}\mu \mathrm{m}$ microtubes at $0.5\mathrm{mm}/\mathrm{s}$ blood flow velocity; (b) structural image contrast for $600\text{-}\mu \mathrm{m}$ microtubes; (c) standard deviation image calculated from eight sequential structural images at $0.5\mathrm{mm}/\mathrm{s}$ blood flow velocity; (d) structural image of the phantom with $300\text{-}\mu \mathrm{m}$ microtubes at $0.5\mathrm{mm}/\mathrm{s}$ blood flow velocity; (e) structural image contrast for $300\text{-}\mu \mathrm{m}$ microtubes. Error bars in (b) and (e) were evaluated from data at 200 different locations (B-scans) within each phantom. Note the different vertical scales in (b) and (e), see text for explanation; (f) ratio of blood + MB/blood structural contrasts for both microtube diameters.

Similar to the temporal 1-D analysis (Fig. 3), the level of enhancement in Figs. 4(b) and 4(e) diminishes with increasing flow, but does not disappear (asymptote to unity) even for fast flows; as seen, there are still differences in the blue versus red histogram bars. Interestingly, despite the more than twofold difference in contrast values between 300- and $600\text{-}\mu \mathrm{m}$ diameter microtubes [note the different vertical scales in Figs. 4(b) and 4(e)], ratios of blood + MB to blood contrasts (height of red histogram bars divided by height of blue ones) for both of them are similar [Fig. 4(f)]. This likely stems from the fact that contrast in OCT structural images is based on the refractive index mismatches in the interrogated material. By introducing MB to the blood, the refractive index mismatches increase which increases scattering. This phenomenon is likely independent of the flow rate and yields additional structural image contrast, compared to blood-only conditions.

The two fold difference in contrast values between 300- and $600\text{-}\mu \mathrm{m}$ microtubes in Figs. 4(b) and 4(e) can be explained by the chosen contrast evaluation method. The OCT signal is lower at the bottom of the chosen half-circles of microtube lumens [see Figs. 2, 4(a), and 4(d)] for $600\text{-}\mu \mathrm{m}$ diameter microtubes in comparison to $300\text{-}\mu \mathrm{m}$ diameter ones. As contrast evaluation is based on averaged ROI intensity, more low intensity pixels in the larger microtube’s center will result in lower contrast value. The reason for the observed difference (especially at low flows) between contrast ratio curves in Fig. 4(f) may be in the faster RBC settling at low flows due to less intensive mixing/laminarity within the microtube of bigger diameter. We consider these effects not significant in the context of real in-vivo blood flow velocities reported in the literature for comparable vessel diameters ($\sim 8\mathrm{mm}/\mathrm{s}$ or faster). What is important here is a higher structural image contrast in the presence of MB in all experiments, more pronounced at low flows. Findings in Secs. 3.1 and 3.2 now set the stage for dynamic OCT image contrast investigation, as per sv studies discussed as follows.

3.3.

Contrast Enhancement of Two-Dimensional and Three-Dimensional Speckle Variance Optical Coherence Tomography Images

As pointed out in Sec. 2, in the svOCT mode, the B-scan acquisition time was set to 25 ms between the $N=8$ frames. This was (1) fast enough that signals from stationary agar did not decorrelate between frames for svOCT signal to approach zero, and (2) sufficiently slow to ensure complete interframe decorrelation (high svOCT signal) for pixels representing blood (see Fig. 3 for the decorrelation time plots).

Figures 5(a) and 5(c) show the svOCT images corresponding to the structural images in Figs. 4(a) and 4(d) at $0.5\mathrm{mm}/\mathrm{s}$ blood flow velocity. As seen, there are more bright pixels in the lumen area of $300\text{-}\mu \mathrm{m}$ diameter microtubes in comparison to $600\mu \mathrm{m}$ ones, likely due to OCT signal wash-out in bigger microtubes. Similar to structural contrast case, this also caused ∼ twofold difference between corresponding sv contrasts in the two microtubes [Figs. 5(b) and 5(d)]; note the different vertical scales.

Fig. 5

(a) svOCT image of the phantom with $600\text{-}\mu \mathrm{m}$ microtubes calculated from eight subsequent structural images at $0.5\mathrm{mm}/\mathrm{s}$ blood flow velocity; (b) svOCT image contrast for $600\text{-}\mu \mathrm{m}$ microtubes; (c) svOCT image of the phantom with $300\text{-}\mu \mathrm{m}$ microtubes at $0.5\mathrm{mm}/\mathrm{s}$ blood flow velocity; (d) svOCT image contrast for $300\text{-}\mu \mathrm{m}$ microtubes. Error bars in (b) and (d) were evaluated from data at 200 different locations within each phantom. Note the different vertical scales in (b) and (d), see text for explanation; (e) ratio of blood + MB/blood sv contrasts for both microtube diameters.

Assessment of the ratio between blood + MB and blood contrasts [Fig. 5(e)] revealed a slightly higher contrast for smaller diameter microtube at low flows (similar trends to the structural contrast ratio metric above). Despite this similarity, it is important to emphasize that svOCT contrast values were found to be (i) 2 orders of magnitude higher than structural contrast values [see Figs. 4(b), 4(e), 5(b), and 5(d) for comparison], and (ii) 1.3 to 2 times higher for blood + MB mixture than for blood for flows up to $3\mathrm{mm}/\mathrm{s}$ [see Fig. 5(e)].

Moving on potential practical implementation of svOCT vascular contrast enhancement with MB added to the blood stream, we present en-face ($x-y$ lateral dimensions) average intensity projections of volumetric phantom scans in Fig. 6. There are eight projections for different flow conditions of a phantom with embedded $600\text{-}\mu \mathrm{m}$ microtubes. The average velocities are indicated in each panel, and the flow direction is labeled with black arrows.

Fig. 6

En-face ($x-y$ lateral dimensions) svOCT average intensity projections of volumetric phantom scans. Eight projections corresponding to eight flow conditions in $600\text{-}\mu \mathrm{m}$ diameter microtubes with blood (left microtube) and blood + MB mixture (right microtube). Image intensities are normalized to the maximum value. The flow direction is labeled with black arrows. The OCT slow scanning axis is antiparallel to the flow direction.

The average contrast values for these panels with deviations from phantom to phantom (error bars) can be found in Fig. 5(b). It is noticeable that at 0 to $1\mathrm{mm}/\mathrm{s}$ flows there is a slight gradient of svOCT signal in $Y$-direction along microtubes. That is because svOCT imaging takes time (tens of seconds for each panel) and images reflect RBC gradual settlement at low flows. As other parameters (blood flow, MB concentration, hematocrit) remain unchanged, from these low flow images, where brighter svOCT signal is seen at right microtubes (filled with blood + MB mixture), we can assume that gradual RBC settlement occurs slower than in left microtubes (filled with blood). The possible reason for that may be MB particles exhibiting faster Brownian motion relative to RBC at low flows due to their smaller size (as mentioned in Sec. 3.1). One of the recent reports²⁹ on OCT study of diffusive versus flow contributions to backscattered OCT signal at low flows for small particle sizes supports this hypothesis. As a result, faster motion of surrounding smaller particles contributes to slower RBC settlement, leading to more rapid intensity fluctuations and higher svOCT contrast.

However, at higher velocities the flow contribution to OCT scattering dominated over diffusive (Brownian motion) effects.²⁹ In our experiments, this leads to increased role of RBC, dominating over MB contribution and therefore diminishing the difference between blood and blood + MB mixture contrasts in both snapshot structural OCT and dynamic svOCT images. Indeed, the common trends observed here from different analysis methods [Figs. 3, 4(b), 4(e), 5(b), and 5(d), as well as faster flow panels in Fig. 6] support these suggested mechanisms.

For potential in-vivo applications of MB-enhanced OCT, assessment of microvascular response to treatment (ultrasound ablation, photodynamic, chemical and radiation therapies of tissues) appears promising. For example, in the last scenario,³⁰ smallest microvessels with low blood flow are known to be most sensitive to irradiation; however, their detection with svOCT is difficult due to low SNR especially at lower depths. Adding MB to the blood stream before OCT imaging may increase SNR for these vessels and improve their detection and thus help visualize their response to irradiation. Further, vascular-targeted treatments (e.g., angiogenic or antiangiogenic therapies) generally affect the vascular functionality. Preimaging the vasculature, then adding MB and imaging vasculature again may reveal “compromised” stagnant/blocked vessels with unchanged contrast (no MB) versus “functioning” increased-contrast vessels (with MB).