2.1.

PA/TA/US System

Figure 1(b) shows a block diagram of the trimodality imaging system that we developed by modifying a clinical US imaging scanner (iU22, Philips Healthcare). The system had four main components: a laser system, a microwave system, the modified clinical ultrasound scanner, and a custom-made data acquisition system. The commercial US scanner was modified to allow access to raw per-channel RF acoustic data, whereas all imaging capabilities of the scanner were retained.¹⁹

The laser system consisted of a tunable dye laser (PrecisionScan-P, Sirah) pumped by a Q-switched Nd:YAG laser (PRO-350-10, Newport). The chosen laser wavelength was 650 nm, where our laser system provides the highest power. The laser pulse width was 6.5 ns and the repetition rate was 10 Hz. Laser pulses were delivered by free-space optics to the opening of a horn antenna through a small hole on the narrow side wall and optically expanded to form a beam with a 1.6-cm radius at the antenna opening.

The microwave system generated 3.0-GHz microwave pulses with different pulse widths (0.16 to 1.2 μs) and repetition rates ($<100\mathrm{Hz}$) and directed the pulses toward the target through the horn antenna. For this study, the microwave pulse width was 0.6 μs and the repetition rate was 10 Hz. The peak power with 0.6-μs pulse width was 58 kW, which generated an average power density of $4.5\mathrm{mW}/{\mathrm{cm}}^2$ at the opening of the antenna, which is below the maximum permissible exposure limit set by the IEEE standard.²⁵ The size of the antenna opening was 11 by 7 cm. The hole in the antenna that allowed the laser beam to pass through was small compared to the microwave wavelength in air. The current distribution on the antenna, and therefore the electromagnetic field, was not significantly disturbed.

An US phased array probe (S5-1, Philips Healthcare) with 80 elements and a nominal frequency band of 1 to 5 MHz was used to acquire US, PA, and TA signals. Although this probe had a lower center frequency than typical breast imaging probes, it was well suited for detecting low-MHz-frequency, microwave-induced thermoacoustic signals. The custom-made data acquisition system controlled the triggering of all three modalities and collected raw data for image display and postprocessing. PA and TA images were reconstructed using a delay-and-sum or Fourier beam-forming algorithm implemented in Matlab, generating cross-sectional B-mode images.²⁶ To improve the SNR, PA and TA acquisitions were repeated 200 times and raw data were averaged before image reconstruction.

2.2.

PAT Experiments

Low-density polyethylene (LDPE) tubes filled with either oxygenated bovine blood or methylene blue dye (30 mM) were embedded in layers of chicken breast tissue and imaged using PA/US. In PA penetration-depth experiments, additional layers of chicken breast tissue were sequentially added to assess the PA detection limits for the trimodality system. Methylene blue dye was used as a target for PAT because it is routinely used clinically during sentinel lymph node biopsy for axillary staging of breast cancer patients. In methylene blue detection-sensitivity experiments, an LDPE tube filled with methylene blue solution was buried within layers of chicken breast tissue at a fixed depth of $\sim 3.5\mathrm{cm}$, and the concentration of the solution was varied from 2 μM to 1 mM. In spatial resolution experiments, the PA resolutions were evaluated using black human hairs mounted on a plastic holder placed in a water tank.

2.3.

TAT Experiments

LDPE tubes filled with saline solution (0.9%) diluted from phosphate saline buffer and embedded in porcine fat were imaged using TA/US. Saline solution was used for TAT because previous studies showed that TAT contrast between normal and tumor tissues was mainly due to the higher sodium and water content in tumors.^4,27 In TA penetration-depth experiments, additional layers of porcine fat were sequentially added to assess the TA detection limits for the trimodality system. The TA spatial resolutions were evaluated using a thin saline tube placed in a tank filled with mineral oil for TAT.

3.1.

Multimodality Phantom Imaging

A phantom test was conducted to validate and show the benefits of the trimodality system. The experiment setup is illustrated in Fig. 2(a). Three LDPE tubes filled with mineral oil, methylene blue (30 mM), or 0.9% saline solution were placed at the opening of the antenna. The US image, overlaid PA/US image, and overlaid TA/US image are shown in Fig. 2(b), 2(c), and 2(d), respectively. As expected, all three tubes were observed in the US image. Owing to the strong optical absorption of the dye at 650 nm, only the methylene blue tube was observed in the PA image. Both the methylene blue and saline solution tubes were observed in the TA image, but the signal from methylene blue was weaker than that from saline solution, as saline has stronger microwave absorption than methylene blue. Accurate coregistration of the tube locations was maintained in all three modalities, which was facilitated by using the same probe for detection of US, PA, and TA signals.

Fig. 2

PA/TA/US phantom images. (a) Diagram of the experimental setup imaging three tubes filled with different solutions. (b) US image showing all three tubes. MB, methylene blue; Sa, saline; MO, mineral oil. (c) Overlaid PA and US image. The PA image shows only the tube filled with MB. (d) Overlaid TA and US image. Both tubes filled with MB and Sa are shown in the TA image, and the Sa tube has stronger signal. (e) PA image showing the cross-section of three black human hairs. (f) MAP image of one hair in the en face view. (g) TA image showing the cross-section of a saline tube.

The tubes were identified in the images by the boundaries facing the ultrasound transducer. Because the low-frequency information in the acoustic signal was filtered by the transducer bandwidth, only the boundaries of the tubes were visible. Moreover, the transducer position was fixed and had a limited aperture for detection, leading to incomplete boundaries of the tubes in the images.²⁸

The PA spatial resolutions of the trimodality imaging system were estimated using black human hairs immersed in a water tank. The PA axial and lateral resolutions were calculated from the FWHM widths in the cross-sectional hair images that were reconstructed using a Fourier transform algorithm. Figure 2(e) shows an image of the hairs placed at a depth of 2.7 cm. For PA elevational resolution, one hair was imaged in a longitudinal view (i.e., along the long axis of the transducer), and a 2-D maximum amplitude projection (MAP) image was created by scanning the transducer in the elevational direction. The PA elevational resolution was then measured from the FWHM of the hair signal in the resulting MAP image. Figure 2(f) shows one such MAP image after interpolation in the width direction at a depth of 6 cm. The PA axial resolution, 640 μm, was nearly constant over the full range of depths. The PA lateral resolution ranged from 720 μm to 2 mm, and it increased with depth. The PA elevational resolution also varied with depth, and the best resolution was 3.5 mm. It is worth noting that in our PAT study, we used diffuse light to illuminate the sample, so the photoacoustic resolution of the system was determined by the ultrasound transducer. Scattering tissues such as human breast tissue will affect the SNR but not the resolution.

A similar resolution measurement was done for TAT using a thin tube (inner diameter $\sim 0.3\mathrm{mm}$) filled with 0.9% saline. TAT resolution depends on the microwave pulse width.^29,30 To minimize the effect of the microwave pulse width, we measured the temporal profile of the microwave pulse and used the Fourier deconvolution method described in Ref. 31. Figure 2(g) shows a TA image of the tube placed at a depth of 2.8 cm after deconvolution. The TA axial and lateral resolutions were $\sim 0.5$ and $\sim 0.8\mathrm{mm}$, respectively. In general, the spatial resolutions of this system are adequate for imaging clinically significant breast tumors, which range in size from several millimeters to several centimeters.

3.2.

PAT Penetration Depth in Chicken Breast Tissue

To measure the PAT penetration depth, an LDPE tube (7 mm inner diameter, 25 mm length) filled with oxygenated bovine blood was embedded in layers of chicken breast tissue. Chicken breast tissue has optical properties comparable to human breast tissue. The optical $1/e$ penetration depth in chicken breast tissue at 650 nm is between 0.9 and 1.6 cm,³² and that of human breast tissue at 656 nm is about 0.78 cm.³³ During the experiment, the whole sample was placed inside a plastic container to support the layers of chicken breast tissue, as shown in Fig. 3(a). The same thicknesses of tissue were used above and below the tube to ensure that both the laser pulses and the generated acoustic signals traveled the same distance. With this arrangement, penetration-depth measurements acquired in transmission mode are analogous to reflection mode acquisitions, where both the source and detector are placed on the same side of the tissue. The laser beam had a diameter of $\sim 3.2\mathrm{cm}$ at the surface of the tissue, and the fluence was $\sim 19\mathrm{mJ}/{\mathrm{cm}}^2$. With these settings, a maximum penetration depth of 6.6 cm was obtained. Figure 3(b) shows a US image of the full sample captured on the US machine, when the blood tube boundary was 7.0 and 6.6 cm from the top and bottom surfaces, respectively. Figure 3(c) shows the corresponding overlaid PA/US image. PA signals from the blood at different depths were normalized to the signal at the smallest depth and are plotted in Fig. 3(d). By fitting the data according to the exponential decay, the $1/e$ penetration depth was calculated to be 0.84 cm, which was influenced by both laser and acoustic attenuations. The $R^2$ value of the fitting was 0.99. The 6.6-cm depth is approximately 7.9 times the $1/e$ penetration depth, corresponding to nearly 34 dB attenuation of the incident laser power density. The SNR at 6.6 cm in the reconstructed image was $\sim 24\mathrm{dB}$.

Fig. 3

PA imaging of blood tube in chicken tissue. (a) Photograph showing the blood tube embedded in chicken breast tissue. The blood tube was covered by additional layers of chicken breast tissue during PA imaging. (b) US image of the whole sample when the depth is 6.6 cm. (c) PA and US overlaid image at 6.6 cm. (d) PA signal (normalized by the signal at the smallest depth) as a function of depth. The maximum depth is 6.6 cm. SNR at this depth is 16 (24 dB). The $1/e$ penetration-depth based on linear curve-fitting is 0.84 cm.

The experiment was repeated using a tube filled with methylene blue solution (30 mM) as the target. The laser beam size was 2.5 cm in diameter and the laser fluence was $\sim 17\mathrm{mJ}/{\mathrm{cm}}^2$. The maximum depth obtained was 8.4 cm, with an SNR of $\sim 15\mathrm{dB}$. Figure 4(a) shows a photograph of the sample setup. Figure 4(b) and 4(c) shows the overlaid PA/US images at depths of 3.4 and 8.4 cm, respectively. To view the tube signal better, PA signals below a threshold value were suppressed, and only a small portion of the PA image containing the tube was shown, indicated by the dotted frame. PA signals at different depths were normalized to the signal at the smallest depth and are plotted in Fig. 4(d).

Fig. 4

PA imaging of a MB-containing tube in chicken tissue. MB, methylene blue. (a) Photograph of the sample setup. (b) Overlaid PA and US image of the tube at 3.4 cm depth. (c) Overlaid PA and US image of the tube at 8.4 cm depth. (d) PA signal (normalized by the signal at the smallest depth) as a function of depth. The maximum depth is 8.4 cm. SNR at this depth is 5.6 (15 dB). The $1/e$ penetration-depth based on linear curve-fitting is 1.11 cm.

3.3.

TAT Penetration Depth in Porcine Fat Tissue

A similar penetration-depth experiment for TAT was performed using porcine fat. The tube filled with 0.9% saline solution had an inner diameter of 13 mm and was 15 mm in length. This tube had comparable dimensions in all directions so that the polarity effect of the electric field was minimized. Figure 5(a) shows the tube embedded in layers of porcine fat. Figure 5(b) and 5(c) shows overlaid TA/US images at depths of 1.1 and 4.4 cm, respectively. Deconvolution was used to process the TA data. The maximum penetration depth obtained was 4.4 cm, where the SNR was $\sim 15.1\mathrm{dB}$. Figure 5(d) shows the normalized TA signal in dB from the tube as a function of depth. Linear fitting of the measurement results gave an $R^2$ value of 0.99.

Fig. 5

TA imaging of saline tube in pork fat tissue. (a) Photograph showing the saline tube embedded in fat tissue. The saline tube was buried under additional layers of fat tissue during TA imaging. (b) TA and US overlaid image at a depth of 1.1 cm. (c) TA and US overlaid image of the tube at a depth of 4.4 cm (in the dotted frame). (d) TA signal (normalized by the signal at the smallest depth) as a function of depth. The maximum depth is 4.4 cm. SNR at this depth is 5.7 (15.1 dB).

3.4.

PAT Sensitivity for Methylene Blue

An LDPE tube was buried in layers of chicken breast tissue, positioned as shown in Fig. 6(a). The LDPE tube was tested to ensure that it was not stained by methylene blue during the experiment. The ends of the LDPE tube were cut open and connected to soft tygon tubes. Methylene blue solution of different concentrations was injected by syringe through the LDPE tube via the tygon tubes, without disrupting the sample position. The thicknesses of the chicken tissue above and below the tube were 3.5 and 3.4 cm, respectively. The laser fluence at the illuminated surface was $17\mathrm{mJ}/{\mathrm{cm}}^2$. First, the tube was filled with distilled water and the PA signal was recorded as a baseline. Next, methylene blue solution was injected into the tube at concentrations ranging from 2 μM to 1 mM. The baseline PA signal, defined as the signal from the tube filled with water, was subtracted from the PA signal from each concentration of methylene blue. The contrast-to-noise ratio (CNR) in dB, defined as the ratio between the PA signal after background subtraction and the noise, is plotted in Fig. 6(b) versus the concentration of methylene blue on a log scale. The linear curve-fitting of these data gave an $R^2$ value of 0.996, which means the PA signal changed linearly as the concentration changed. At concentrations higher than 1 mM, the PA signal varied in a nonlinear fashion because of the depth-dependent distribution of optical energy deposition inside the methylene blue.³⁴ The SNR from the tube filled with 2 μM methylene blue was 7.8, or 17.8 dB. Therefore, the noise-equivalent sensitivity, defined as the ratio of the methylene blue concentration to the SNR, was $\sim 260\mathrm{nM}$ at a depth of 3.4 cm in chicken breast tissue.

Fig. 6

PA signal strength as a function of methylene blue concentration at a depth of 3.4 cm measured using the S5-1 probe. (a) Photograph showing the methylene blue tube embedded in chicken breast tissue. The methylene blue tube was covered by 3.5 cm of chicken breast tissue during PA imaging. (b) CNR versus concentration after subtraction of the baseline signal.