2.1.

Experimental System

In VE-PAT, the major device is a linear-array-based photoacoustic imaging probe.^19,20 The probe used here consisted of a linear array ultrasonic transducer (Visualsonics Inc. LZ250, 21 MHz center frequency, 78% one-way bandwidth, 256 elements, $23\times 3\mathrm{mm}$ array size) with two optical fiber bundle strips ($20\times 1.25\mathrm{mm}$) mounted on each side [Fig. 1(a)]. The laser beams coming out of the two optical fiber bundle strips had an angle of incidence of 30 deg with respect to the imaging plane. The two optical fiber bundles were bifurcated from a single optical bundle that was incorporated into the probe together with an ultrasound signal cable. A tunable optical parametric oscillator laser (680 to 970 nm, 20-Hz pulse repetition rate) was coupled into the single fiber optical bundle for photoacoustic excitation. The wavelength was set to 850 nm to achieve deep penetration for vascular elasticity imaging. In our experiments, the fluence on the sample surface was $10\mathrm{mJ}/{\mathrm{cm}}^2$, well within the safety limit set by the American National Standards Institute ($20\mathrm{mJ}/{\mathrm{cm}}^2$). Each element of the transducer array was cylindrically focused with a focal length of 15 mm. For each of four laser pulses, photoacoustic signals were captured sequentially on one quarter (i.e., 64 elements) of the transducer array elements. After all data were acquired from the four segments, we reconstructed a two-dimensional (2-D) photoacoustic image using the filtered back-projection algorithm.²¹ The reconstructed 2-D photoacoustic image is referred to hereafter as the B-scan photoacoustic image. An imaging station (Vevo LAZR, Visualsonics Inc.) displayed the reconstructed B-scan photoacoustic images at $5\text{frames}/\mathrm{s}$, as determined by the 20 Hz laser repetition rate and the four-to-one multiplexing in the image acquisition system. The VE-PAT system has spatial resolutions of $119\mu \mathrm{m}$ in the lateral direction, $86\mu \mathrm{m}$ in the axial direction, and $1237\mu \mathrm{m}$ in the elevational direction.²⁰ The minimum displacement that can be resolved is $18.3\mu \mathrm{m}$, which is determined by the data acquisition sampling rate of 84 MHz and the average speed of sound in biological tissue of $1540\mathrm{m}/\mathrm{s}$. The typical signal-to-noise ratios of the VE-PAT system were around 25 dB in the following phantom experiments and 20 dB in the in vivo experiments in humans.

Fig. 1

Schematic of vascular elastic photoacoustic tomography (VE-PAT): (a) photoacoustic imaging probe and (b) VE-PAT setup for a blood vessel phantom imaging. The imaging probe is incorporated with a customized compression stage.

To implement VE-PAT, a customized compression stage was developed and incorporated into the probe. The compression stage had an aluminum plate controlled by a three-dimensional translation stage, which could position the plate in the $x-y$ plane and induce precise displacements along the $z$ axis for accurate sample compression [Fig. 1(b)]. Above the compression plate, a water tank held water for acoustic coupling. An imaging window slightly larger than the probe was machined in the center of the compression plate. To ensure the compression force applied to the sample was normal and uniaxial, a piece of fully stretched polymethylpentene plastic membrane was attached to the bottom of the compression plate, which allowed the transmission of both the optical excitation light and the generated photoacoustic wave. An aluminum block on an optical table held the sample against compression.

2.2.

Phantom Preparation

Silicone microtubes with 0.3 mm (60985-700, VWR) and 3.4 mm inner diameters (60985-720, VWR) perfused with bovine blood (905, Quad-Five) mimicked small and large blood vessels, respectively. Blood was pumped into the microtubes through a syringe, and the flow speed was controlled by a syringe pump (BSP-99M, Braintree Scientific). The thrombosis was simulated by injecting a small drop of glue into the microtube downstream from the measurement point, which hardened on the tube wall and partially blocked the flow. The microtube was then embedded in tissue-mimicking gelatin phantoms 3 mm deep [Fig. 2(a)]. To achieve stiffness similar to that of soft tissue, the gelatin concentration in the phantoms was $100\mathrm{g}/\mathrm{L}$.²² To achieve similar optical scattering as in biological tissue, 1% intralipid was added to the gelatin phantoms.

Fig. 2

Characterization of a large blood vessel phantom: (a) diagram of the large blood vessel phantom embedded 3 mm deep in gelatin and (b) stress–strain response measured for the large blood vessel phantom with and without simulated thrombosis. Stress–strain curves with strain smaller than 10% were fitted to linear functions to calculate the compliances as shown by the dashed lines.

2.3.

Method

3.1.

Standard Compression Test of the Large Blood Vessel Phantom

Before the microtubes were embedded into tissue-mimicking gelatin phantoms, the compliances of the perfused large blood vessel phantoms with and without simulated thrombosis were measured with a standard compression test. In the standard compression test, the large blood vessel phantom was placed on a high-precision digital weighing scale (S200, Ohaus) and compressed using the customized compression stage. The compression stress was calculated based on the difference in the scale readings before and after compression. The axial displacement of the top boundary was read from the translation stage that moved the compression plate along the $z$ axis for precise sample compression. Because the bottom boundary did not move, the strain was calculated as the deformation over the original diameter of the blood vessel phantom [Eq. (1)]. Consistent with the previous elasticity imaging studies,²³ the strain–stress relationship appeared linear with the coefficient of determination ($R^2$) of 0.993 and 0.987 when the strain was smaller than 10% [Fig. 2(b)]. The ratio of the compliances between cases with and without simulated thrombosis was 0.64.

3.2.

Vascular Elastic Photoacoustic Tomography of Large Blood Vessel Phantom

Photoacoustic images of a cross section of the large blood vessel phantom before and after 10 small steps of compression with a step size of $50\mu \mathrm{m}$ are shown in Fig. 3. The diameter of the large blood vessel phantom before compression was measured to be $3513.6\mu \mathrm{m}$. The axial displacements of the top and bottom boundaries and the strain values of the large blood vessel phantom are summarized in Table 1. Under the same stress, the blood vessel phantom without simulated thrombosis [Figs. 3(a) and 3(b)] underwent larger deformation than that with simulated thrombosis [Figs. 3(c) and 3(d)]. The strain values of the large blood vessel phantom with and without simulated thrombosis were plotted at each compression step [Fig. 3(i)]. During the 10 small steps of compression, the strain–stress response still appeared linear, as shown in Fig. 3(i). We quantified the relative compliance by fitting the two data sets to linear functions. The compliance ratio between the two states was estimated to be 0.67, which is close to the ratio of 0.64 measured in the standard compression test.

Fig. 3

VE-PAT of a large blood vessel phantom. Photoacoustic images of a cross section in the large blood vessel phantom without thrombosis (a) before and (b) after 10 small steps of compression, and (e)–(f) after two large steps of compression. Photoacoustic images of the cross section in the large blood vessel phantom with simulated thrombosis (c) before, (d) after 10 small steps of compression, and (g)–(h) after two large steps of compression. (i) Strain curves under 12 steps of compression for the large blood vessel phantom with and without simulated thrombosis. Dashed lines: linear fits with the coefficient of determination ($R^2$) of 0.998 and 0.995 for $\le 13\%$ strain. Error bars: standard deviation.

Table 1

Vascular elastic photoacoustic tomography (VE-PAT) of axial displacements and strains of the large blood vessel phantom. The linear compression is the result from 10 total compression steps.

Figures 3(e)–3(h) show B-scan photoacoustic images of a cross section of the large blood vessel phantom before and after two large steps of compression with a step size of $500\mu \mathrm{m}$, during which the strain reached the apparent nonlinear strain–stress response regime. The blood vessel phantom became harder to compress in the presence of simulated thrombosis [Figs. 3(g)–3(h)], akin to the linear strain–stress response result. The strain values at the first compression state for the blood vessel phantom with and without simulated thrombosis were $12.9\%\pm 1.3\%$ and $16.3\%\pm 1.1\%$, respectively. At the second compression state, they were $15.8\%\pm 1.4\%$ and $22.4\%\pm 1.7\%$, respectively [Fig. 3(i)]. Since the same compression forces were applied to the blood vessel phantom with and without simulated thrombosis at each state, the smaller strains indicated that a decrease in compliance of the blood vessel phantom was induced by the simulated thrombosis.

3.3.

Vascular Elastic Photoacoustic Tomography of Small Blood Vessel Phantom

Photoacoustic images of a cross section of the small blood vessel were acquired before and after compression with and without simulated thrombosis [Figs. 4(a)–4(d)]. The diameter of the small blood vessel phantom before compression was measured to be $329.4\mu \mathrm{m}$. The axial displacements of the top and bottom boundaries and the strain values of the small blood vessel phantom are summarized in Table 2. Similar to the large blood vessel phantom, the small blood vessel phantom was harder to compress with the simulated thrombosis. The strain for the small blood vessel phantom without simulated thrombosis was $13.3\pm 0.9\%$ while the strain with simulated thrombosis was only $6.7\%\pm 0.4\%$ [Fig. 4(e)]. Under the same compression force, the decrease in strain indicated a decrease in compliance due to the simulated thrombosis.

Fig. 4

VE-PAT of a small blood vessel phantom. Photoacoustic images of a cross section of the small blood vessel phantom without simulated thrombosis (a) before and (b) after compression, with simulated thrombosis (c) before and (d) after compression. (e) Strain values for the small blood vessel phantom with and without thrombosis. Statistics: paired Student’s $t$-test. $P$ values: **$<0.05$.

Table 2

VE-PAT of axial displacements and strains of the small blood vessel phantom.

3.4.

Vascular Elastic Photoacoustic Tomography of Blood Vessels in a Human

Figure 5 shows the VE-PAT results of the middle finger blood vessels of the human subject with and without vessel occlusion. The diameter of the blood vessel in the human finger before compression was measured to be $786.9\mu \mathrm{m}$. The axial displacements of the top and bottom boundaries and the strain values of the finger blood vessel are summarized in Table 3. During five small steps of compression with a step size of $50\mu \mathrm{m}$, the blood vessel in the finger underwent smaller deformation with vessel occlusion downstream [Figs. 5(a)–5(d)]. Strain values for the blood vessels at each step of compression were calculated for both normal and vessel occlusion conditions [Fig. 5(e)]. Since the largest deformation was still within the linear strain–stress response regime, we estimated a relative decrease of compliance as 30.0% when the vessel was occluded.

Fig. 5

VE-PAT of a human finger in the linear strain–stress response regime. Photoacoustic images of a cross section of the human finger without vessel occlusion (a) before and (b) after compression and with vessel occlusion (c) before and (d) after compression. (e) Strain curves under five steps of compression for the human finger with and without vessel occlusion. Dashed lines: linear fits with the coefficient of determination ($R^2$) of 0.993 and 0.987. Error bars: standard deviation.

Table 3

VE-PAT of axial displacements and strains of the human finger blood vessel in vivo. The linear compression is the result from five total compression steps.

The finger was then compressed with two $300\mu \mathrm{m}$ steps, during which the blood vessels underwent larger deformation so that the strain–stress response became nonlinear [Figs. 6(a)–6(f)]. Again, the blood vessels with occlusion underwent smaller deformation than that without occlusion under the same load. Strain values for blood vessels without occlusion at the two steps were $37.5\%\pm 5.8\%$ and $47.9\%\pm 6.1\%$, while they were $16.7\%\pm 3.4\%$ and $22.9\%\pm 4.5\%$ with vessel occlusion [Fig. 6(g)]. These results demonstrated that VE-PAT can detect the decrease of vessel compliance in the human finger due to vessel occlusion in both the linear and nonlinear strain–stress response regimes.

Fig. 6

VE-PAT of a human finger in the nonlinear strain–stress response regime. Photoacoustic images of a cross section of the human finger without vessel occlusion (a) before compression, (b) after compression step 1, and (c) after compression step 2. Photoacoustic images of the same cross section of the human finger with vessel occlusion (d) before compression, (e) after compression step 1, and (f) after compression step 2. (g) Strain values for the blood vessels with and without thrombosis in two compression steps. Statistics: paired Student’s $t$-test. $P$ values: **$<0.05$.