2.1.

Experimental Systems

2.1.1.

Photoacoustic imaging system

Figure 1 shows the experimental configuration of the PAT-guided DOT study. A Ti:Sapphire (Symphotics TII, LS-2134) laser optically pumped with a Q-switched Nd:YAG laser (Symphotics TII, LS-2122) delivered 10 ns pulses at 15 Hz. The laser output was coupled into two optical fibers using a convex lens and circular beam splitter arrangement as shown in Fig. 1. The convex lens, which has 99% transmittance at 700 nm wavelength, has a focal length of 20 cm and focuses the light into the pair of fibers that are placed at its focal point. The beam splitter was used to split the incoming beam into two beams, one for each fiber. It was measured to have a 60% transmitting and 36% reflecting split ratio for horizontally polarized light at the 700 nm that was used for the experiment. The overall coupling efficiency of the setup (including the losses in the lens and beam splitter) was about 85%. The two output ends of the fibers were then mounted on opposite (longitudinal) sides of a low-frequency linear ultrasound transducer described below and used for photoacoustic imaging.

Fig. 1

Experimental configuration of the PAT setup coregistered with the DOT system. The DOT system consists of four laser diodes, 14 detectors, and data processing electronics integrated into a box, and interfaced with a computer for performing image reconstruction.

The electronics of the photoacoustic imaging system consists of 64-channel receiving circuits. Signals from 64-array elements are individually amplified by 20 dB and multiplexed to two parallel channels for further amplification, low pass filtering, and analog-to-digital (A/D) conversion. The second stage electronics consists of dual-channel two-stage variable gain amplifiers (Analog Devices AD604), 10 MHz low pass filters (Mini-Circuits SCLF-10), and 12-bit analog-to-digital converters running at 50 MS/s. The A/Ds are connected to a National Instruments PCI-DIO-32HS high-speed digital I/O card, which controls and receives the digitized data. The system is controlled with a custom C-language software on the host computer through the two digital I/O cards. Due to the multiplexing, 32 laser firings are required to generate a single 64-channel capture. Because the Ti:Sapphire laser has 15 Hz pulse repetition rate, the acquisition rate is about one frame every 5 s. Images were reconstructed using the delay-and-sum beamforming algorithm, which provided similar imaging quality as the model-based backprojection algorithm mainly because of the limited receiving aperture.

Clinically, DOT has demonstrated the capability to detect tumors of 1 cm or larger in extent. Because of the large target size, a low-frequency ultrasound transducer was employed to maximize the sensitivity of the system. The piezocomposite transducer, produced by Vermon (France), consisted of 64 elements with 0.85 mm pitch. The center frequency of the transducer is 1.3 MHz with a 6 dB response from 500–1800 kHz. An integrated acoustic lens with 25 mm focal length increases sensitivity at imaging depths for which the optical fluences are low. Figure 2 depicts the amplitude frequency response of the photoacoustic system along with the theoretical spectra for uniformly absorbing spheres with radii of 1, 5, and 10 mm. The spectra were obtained from FFT of the expressions derived by Diebold ³⁴ The center frequency and bandwidth of the mainlobe of the 1-mm-diameter sphere match the transducer response well. In contrast, the 10-mm-diameter sphere possesses a narrower spectrum located near the low-frequency edge of the transducer response. As a result, the system exhibits sensitivity only to the edges for objects with dimensions greater than 1 mm.

Fig. 2

Impulse response of the photoacoustic system along with the theoretical spectra for uniformly absorbing spheres with radii of 1, 5, and 10 mm.

2.1.2.

Frequency domain DOT system and imaging reconstruction

The DOT system used for the experiment was a frequency-domain imager consisting of four laser diodes emitting at 740, 780, 808, and 830 nm wavelengths. Since phantom targets were used in the study and the multiwavelength imaging was not a concern, only 780 nm wavelength was used to demonstrate the PAT guided DOT approach. The outputs of these laser diodes were coupled into nine fibers via a 4×1 and a 1×9 optical switches.⁴ The nine source fibers were then deployed around the PAT/DOT probe in an arrangement as shown in Fig. 3. On the receiving side, fourteen 3 mm-diameter fiber bundles were used to couple reflected light from the target to 14 photomultiplier tubes (PMTs). Light from the source fibers was sequentially delivered to the target, but the received light was detected in parallel by the PMT detectors. Semi-infinite partial reflection boundary condition with an effective refraction coefficient of 0.6 was used for the DOT image reconstruction. The measured scattered field used for the optical imaging reconstruction was the difference between measurements obtained from a 0.5–0.7% intralipid solution and a target immersed inside and located at various depths. The calibrated optical absorption (μ_a) and reduced scattering coefficient ( [TeX:] $\mu _s^\prime $ ${\mu }_s^{\prime }$ ) of the intralipid solution used in the experiment were in the ranges of 0.01–0.02 cm⁻¹ and 5–7 cm⁻¹. To compare the improvement of PAT-guided reconstruction to reconstruction with no a priori target information and depth-only guidance, we reconstructed targets using three procedures, namely: 1. no a priori target location information, 2. target depth only, and 3. PAT-guided reconstruction using target depth and twice the target size as measured from PAT images. In PAT-guided reconstruction, the dual-zone mesh optical imaging algorithm that we developed earlier was modified for reconstructing the target optical properties.⁶ Briefly, the entire imaging volume was segmented into a target region or two-target regions (ROIs), and a background region; this segmentation was based on the target absorption map provided by the high-resolution PAT. The center region of each target was then precisely determined from the reconstructed photoacoustic image in the lateral (x) dimension, depth (z), as well as a region twice the size of the target diameter. Because the photoacoustic images provided only 2-D information, the target center in the y dimension was assumed to be on the same ultrasound transducer plane at y = 0. The center locations in the x and y directions measured by PAT were slightly perturbed in the DOT reconstruction in order to obtain an improved convergence as measured by the object function. Because of the intense light scattering, the target size seen by DOT was much larger than its actual size. A ROI twice the target size was necessary for avoiding boundary distortion in reconstructed DOT images. A finer grid of 0.2×0.2×0.5 (cm³) was chosen for the target region and a coarse grid of 1.5×1.5×1.0 (cm³) was used for the background region. The total imaging volume was chosen to be 12×12×4 (cm³). This multiple-ROIs scheme significantly reduced the total number of voxels with unknown optical properties and dramatically improved the convergence of inverse mapping of target optical properties.

Fig. 3

Photograph of DOT/PAT hybrid probe covered at the bottom with aluminum foil.

2.2.

Hybrid Probe

Figure 3 presents a close-up photograph of the hybrid PAT/DOT probe. The transducer occupies the central slot with the two PAT optical fibers straddling the transducer on both sides. The transducer was painted white to reduce the front-face light absorption that may generate unwanted photoacoustic signals. The nine DOT source fibers and 14 detector fiber bundles are arranged nearly symmetrically on both sides of the probe in a pattern that provides a distribution of source-detector pair distances from 1.5–7 cm. The probe, made from transparent plexiglas, measured 9 cm in diameter and 1 cm in height and was covered with aluminum foil at the bottom. This organization, coupled with the partial reflection probe boundary, enabled optimized imaging for targets ranging from 1–3 cm in depth suitable for breast imaging in reflection geometry.

2.3.

Phantoms

One centimeter-cubical soft absorbers of higher (μ_a = 0.25 cm⁻¹) and lower (μ_a = 0.08 cm⁻¹) optical contrast representative of typical malignant and benign lesion optical properties served as more realistic soft tissue targets. The soft absorbers were made of polyvinyl chloride plastisol³⁵ and the calibration was performed on a larger piece of 10 cm × 10 cm × 5 cm phantom using the DOT system.

3.1.

Single Target Imaging

A 1 cm-cubical absorber of higher (μ_a = 0.25 cm⁻¹) and another of lower (μ_a = 0.08 cm⁻¹) optical contrast were used as soft tissue targets and placed inside intralipid of calibrated optical properties of μ_a = 0.02 cm⁻¹ and [TeX:] $\mu _s^\prime $ ${\mu }_s^{\prime }$ = 6.9 cm⁻¹, and imaged. Figures 4a1, 4a1, 4a2, 4a3, 4a4, 4b1, 4b2, 4b3, 4b4, 4c1, 4c2, 4c3, 4c4, 4d1 show the PAT images of the higher contrast target with its center located at 1.5, 2.0, 2.5, and 3 cm depth, respectively, from the transducer. The high-resolution PAT images clearly delineate the front and back face of the target. Without any a priori information about the target location, the reconstructed DOT absorption maps shown in Figs. 4a2, 4a3, 4a4, 4b1, 4b2, 4b3, 4b4, 4c1, 4c2, 4c3, 4c4, 4d1, 4d2 fail to provide target location and contrast. However, if the target depth is available, the reconstructed absorption maps, given in Figs. 4a3, 4a4, 4b1, 4b2, 4b3, 4b4, 4c1, 4c2, 4c3, clearly show the target location and the quantification accuracy has reached 39–48% for a target located from 1.5–2.5 cm depth. The maximum reconstructed absorption coefficient is used to quantitatively evaluate the reconstruction accuracy. Although the quantification accuracy is only 25% when the target is located deeper at 3 cm as shown in Fig. 4d3, the target is reconstructed at the correct depth and is visible. With the precise guidance of the PAT target location and size, the reconstructed DOT absorption maps shown in corresponding Figs. 4a4, 4c1, 4d1, 4d2, 4d3, 4d4 further improved target quantification by 22% on average. The reconstructed maximum target absorption coefficient and quantification accuracy are given in Fig. 5. Figures 6a1, 6a2, 6a3, 6a4, 6d1 show the PAT images of the lower-contrast target whose center is located at 1.5, 2.0, 2.5, and 3 cm depth, respectively, from the probe. The contrast of the PAT images was not as good as that of the higher contrast one. However, the front and back faces of the target were fairly well resolved. Without any a priori target information, the reconstructed DOT absorption maps shown in Figs. 6a2, 6b1, 6c1, 6d1, 6d2, 6d3, 4d4 cannot resolve the target location and provide a poor contrast from the background. With the target depth information, the reconstructed absorption maps, given in Figs. 6a1, 6a2, 6a3, 6a4, 6d1, 6d2, 6d3, clearly show the target location and the quantification accuracy has reached 51–60% for a target located from 1.5–2.5 cm depth and 27% for a target located at 3 cm depth. With the precise PAT guidance on target location and size, the reconstruction accuracy has reached 76–99% for a target located from 1.5–2.5 cm, and 65% for a target located at 3 cm depth [Figs. 6a4, 6d1, 6d2, 6d3, 6d4]. The reconstructed maximum target absorption coefficient quantification accuracy is shown in Fig. 7. Clearly, a priori spatial information from PAT provides a significant improvement in the absorption coefficient quantification accuracy.

Fig. 4

Experimental results of the higher-contrast 1 cm cubical phantom target. (a1)–(d1) PAT images of the target (shown inside the circled regions) with center located at 1.5, 2, 2.5, and 3 cm depths, respectively. The x-axis of the PAT images is the lateral dimension in x (mm), and vertical axis is the depth in z (mm). (a2)–(d2) Corresponding absorption maps reconstructed without using any a priori information. (a3)–(d3) Corresponding absorption maps reconstructed using only depth information. (a4)–(d4) Corresponding absorption maps reconstructed using PAT location and target size guidance. Each absorption map has six subimages. Each subimage is an x-y image of 12 cm × 12 cm. The depth range is from 0.5–3 cm in 0.5 cm increments.

Fig. 5

Absorption coefficient quantification accuracy for higher-contrast phantom (with calibrated μ_a = 0.25 cm⁻¹) using DOT reconstruction with no a priori information, depth-only information, and PAT spatial information.

Fig. 6

Experimental results of the lower-contrast 1 cm-cubical phantom target. (a1)–(d1) PAT images of the target (shown inside the circled regions) with center located at 1.5, 2, 2.5, and 3 cm depths, respectively. The x-axis of the PAT images is lateral dimension in x (mm), and vertical axis is the depth in z (mm). (a2)–(d2) Corresponding absorption maps reconstructed without using any a priori information. (a3)–(d3) Corresponding absorption maps reconstructed using only depth information. (a4)–(d4) Corresponding absorption maps reconstructed using the PAT location and target size guidance. Each absorption map has six subimages. Each subimage is an x-y image of 12 cm × 12 cm. The depth range from 0.5–3 cm in 0.5 cm increments.

Fig. 7

Absorption coefficient quantification accuracy for lower-contrast phantom (with calibrated μ_a = 0.08 cm⁻¹) using DOT reconstruction with no a priori information, depth-only information, and PAT spatial information.

3.2.

Dual Target Imaging

The 1 cm-cubical absorbers immersed inside the intralipid were again used to mimic two lesions. The first set of experiments was conducted using a pair of the higher contrast absorbers to represent typical optical properties of clustered malignant lesions. Both targets were located at the same depth from the probe and were again varied from 1.5–3.0 cm as before. For each depth, the center-to-center separation of the two targets was also varied from 1.5–3.0 cm in 0.5 cm increments. The images shown in Figs. 8a1, 8a2, 8a3, 8d1 are the PAT images of the two targets located at the four different depths as mentioned above and 3.0 cm center-to-center separation. The PAT images clearly delineate the front and back face of each target; however, the contrast is reduced at a depth of 3 cm. In addition, the target size and center-to-center distance can be accurately measured. Figures 8a2, 8a3, 8d1, 8d2 are the reconstructed absorption maps of the two targets using target depth information without target size and spatial location guidance from PAT. The two targets are recognizable at 1.5 cm depth as shown in Fig. 8a2 and merge together at deeper depths. The reconstructed values were 0.13 cm⁻¹ (52%) and 0.12 cm⁻¹ (48%). The reconstructed absorption maps using two ROIs estimated from coregistered PAT images are shown in Figs. 8a3, 8d1, 8d2, 8d3. It is seen that this time, the two targets are well separated, and the reconstructed values are 0.18 cm⁻¹ (71%) and 0.18 cm⁻¹ (71%). The results of the accuracies for other depths and center-to-center separations are given in Table 1. In summary, when the center-to-center (C-C) separation of targets was 2.0 cm and targets located at 1.5 cm, or C-C was 2.5 cm and targets located at 2.0 cm and shallower, or C-C was 3 cm and targets located at 3.5 cm and shallower, the reconstructed absorption maps without and with PAT guidance can separate the two targets. However, the target quantification can be improved by using two ROIs provided by precise PAT guidance. The average improvement ranges from 11–26% with an average of 20%. For other C-C target separations and locations without and with PAT, guidance cannot separate the two targets. However, the target quantification has improved from 8–20% with an average improvement of 13% by using precise PAT guidance on target spatial location and size.

Fig. 8

Phantom experiments of two higher-contrast targets, with a center-to-center separation of 3.0 cm. The images in the left column (a1)–(d1) are the PAT images (shown inside the circled regions). The reconstructed absorption maps using a priori target depth information are only given in the middle column (a2)–(d2). The reconstructed absorption maps using PAT guidance are shown in (a3)–(d3).

Fig. 11

Experiment results for a higher contrast cube submerged in the chicken breast. (a) PAT image, circled region is the cube; (b) Reconstructed target absorption map, max μ_a = 0.152 cm⁻¹.

Table 1

Reconstruction results of phantom experiments of two higher contrast targets (calibrated μa = 0.25 cm−1).

A second set of experiments were performed on a pair of 1 cm-cubical soft absorbers, one of higher contrast and the other of lower contrast. This set of experiments emulates the condi-tion of imaging one malignant and one benign lesion next to each other. As before, both targets were located at the same depth from the probe and the center depth was varied from 1.5–3.0 cm in the intralipid. For each target depth, the C-C separation between two targets was varied from 1.5–3.0 cm in 0.5 cm increments. An example of C-C = 3 cm and target located from 1.5–3 cm is given in Fig. 9. As shown in Figs. 9a2, 9a3, 9b1, 9b2, 9b3, 9c1, 9c2, 9c3, 9d1, 9d2, the reconstructed absorption maps using the single ROI with the a priori target depth information can barely resolve the low contrast target at 1.5 and 2 cm depth and both targets are merged at 2.5 and 3 cm depth. Using the two ROIs estimated from PAT images, the two targets can be resolved at 1.5 and 2 cm, but barely resolved at 2.5 cm depth with an average 17% improvement on target quantification for high contrast target and 15% for low contrast one at depths of 1.5 and 2 cm. The reconstruction accuracy results are given in Table 2 and show a similar trend as in Table 1. When the center-to-center separation of the two targets was 2.0 cm at a depth of 1.5 cm, or C-C was 2.5 cm at a depth of 2.0 cm or shallower, or C-C was 3 cm at a depth of 3.5 cm or shallower, the reconstructed absorption maps without and with PAT guidance can separate the two targets. However, the target quantification can be improved by using two ROIs provided by precise PAT guidance. On average, both high and low contrast quantification accuracy was improved by 18% using PAT guidance compared with depth only. For other C-C target separations and locations, neither PAT method was able to separate the two targets, and only the higher contrast target was able to be reconstructed. The PAT guidance improved the quantification of the higher contrast target by 13%.

Fig. 9

Phantom experiments of two targets with different contrasts, with a center-to-center separation of 3.0 cm. The images in the left column (a1)–(d1) are PAT images (shown inside the circled regions), with the left one being the higher-contrast and the right one the lower-contrast target. The reconstructed absorption maps using a priori target depth information are only given in the middle column (a2)–(d2). The reconstructed absorption maps using PAT guidance are shown in (a3)–(d3).

Table 2

Reconstruction results of phantom experiments of one higher (calibrated μa = 0.25 cm−1) and one lower contrast targets (calibrated μa = 0.08 cm−1).

3.3.

Absorber Embedded in Chicken Breast

Apart from imaging target in intralipid medium, the PAT/DOT system was also used in imaging target through a chicken breast of 1.5 cm thickness. For this experiment, the higher-contrast cube was inserted into a hole made in a chicken breast as illustrated in Fig. 10. The size of the hole was made a little bit smaller than the cube to allow the cube to fit tightly inside the hole. A thin layer of transparent ultrasound gel was used over the top surface of the cube and the chicken breast. A second layer of chicken breast of about 1.7 cm thickness was then placed on top of the cube so that the latter was covered on sides with chicken breast. This setup mimicked a tumor that is located at a depth of 2.2 mm inside breast tissue. The PAT/DOT probe was placed over and in contact with the top layer chicken breast. Coregistered PAT and DOT images were then taken and the results are shown in Fig. 11. The value of the reconstructed absorption coefficient (0.152 cm⁻¹) of the cube obtained is 67% of the calibrated value (0.25 cm⁻¹).

Fig. 10

Diagram showing cube inside chicken breast tissue and the PAT/DOT probe used for taking coregistered images.