Fan-shaped scanning approach for miniaturized photoacoustic tomography

Abstract

We describe a novel scanning approach for miniaturized photoacoustic tomography (PAT), based on fan-shaped scanning of a single transducer at one or two discrete positions. This approach is tested and evaluated using several phantom and animal experiments. The results obtained show that this new scanning approach provides high image quality in the configuration of miniaturized handheld or endoscopic PAT with improved effective field of view and penetration depth.

1 |. INTRODUCTION

Photoacoustic imaging (PAI) is one of the fastest growing biomedical optical imaging techniques capable of providing nonionizing and noninvasive imaging of tissue absorption with high spatial resolution, and has been demonstrated in biological and clinical applications [1]. In general, PAI is classified into photoacoustic microscopy (PAM) and photoacoustic tomography (PAT), according to different implementation strategies and image reconstruction algorithms. While PAM enables high-resolution (several micrometers) imaging of biological structures like blood vessels at a superficial depth of up to ~1 to 2 mm in tissue, PAT provides deep tissue imaging ability with a penetration depth of up to several centimeters and a good spatial resolution (~100 μm), and is more suitable for a wide range of clinical applications such as breast cancer detection [2–5], brain investigation [6, 7], arthritis diagnosis [8, 9] and pathological analysis [10].

In the past decade, increased effort has been made in the development of handheld and endoscopic PAI devices through miniaturization in hardware [11–19]. Compared with a handheld PAT device with a diameter of several centimeters, an endoscopic PAT always has limited size, typically ~1 cm for gastrointestinal imaging, and ~1 mm in diameter for intravascular imaging. While a large number of endoscopic PAI devices have been reported based on PAM to date [13–15], the design of a miniaturized PAT especially endoscopic PAT remains largely as a challenge. There are mainly two factors limiting the development of endoscopic PAT. The first factor is that the conventional linear or circularly scanning based endoscopic PAT provides a very small effective imaging region or field of view (FOV). The second one is that an effective image reconstruction algorithm is not readily available. For example, delay and sum (DAS), the most commonly used reconstruction algorithm, is typically implemented for the configuration of linear or circular scanning PAT, and not suitable for an endoscopic PAT.

To overcome the aforementioned two limitations, in this work, we present a novel scanning approach based on fan-shaped scanning of a single transducer at one or two discrete positions along with an appropriately modified DAS algorithm specifically tailored for such novel scanning that will allow us to realize endoscopic PAT. We demonstrate this approach using phantom and animal experiments.

2 |. METHODS

2.1 |. Fan-shaped scanning

The fan-shaped scanning approach is schematically shown in Figure 1. To demonstrate this approach, a single element acoustic transducer (Valpey Fisher, Hopkinton, Massachusetts), having an effective diameter of 6 mm with a center frequency of 1 MHz, was used as the detector to receive photoacoustic (PA) signal from a sample/target immersed in water upon absorption of light emitted by an OPO laser at 780 nm (Phocus Mobile, Opoteck, Carlsbad, California). The transducer placed on a rotation stage (Newmark, California) revolved around the center of the transducer surface to realize the fan-shaped scanning and to detect the PA signals from different directions. The scanning covered an angle of 120° with an interval of 0.1°, providing a large of FOV. It took 20 minutes to complete the scanning, which was limited by the rotating speed of the rotator used in this system.

FIGURE 1.

Schematic of fan-shaped scanning approach for PAT. L stands for the distance between the transducer and the target

We hypothesize that the number of detector positions will affect the resulting PAT image quality. Therefore, two scanning schemes (A and B) were performed and compared, in which scheme A had only one detector position while scheme B had two discrete detector positions with a separation of 4 mm. The PA signal was acquired by a data acquisition (DAQ) board (NI, 5152) controlled by a LabVIEW program.

2.2 |. Image reconstruction

In the conventional DAS algorithm, the initial pressure rise can be solved in time domain by back-projection of the received pressure wave. The transducer is considered as a point detector located at the center of the transducer surface. To reconstruct an image, the center of the transducer and the object of interest (OOI) must have relative displacement to provide difference in delay time. However, in our fan-shaped scanning scheme for PAT, as the center of the transducer is fixed, there is no relative displacement between the transducer center and the OOI. Thus, the traditional DAS approach is not suitable for the fan-shaped scanning configuration.

It is known that photoacoustic signal collected by a transducer is the integral of the acoustic wave pressure over the entire active surface of the transducer [20]. In our approach, as shown in the schematic Figure 2, we divide the signal received by a single transducer into many subsignals received by a set of virtual elements over the transducer surface. The virtual signal for each virtual element can be expressed as

$$s_{ki}\left(t_{ki}\right)=s_k\left(t_{ki}\right)\frac{S_i}{S},$$ (1)

where s_ki(t) is the collected acoustic signal at virtual element i for the kth scanning; $t_{ki}=\frac{\left|\overrightarrow{r}-{\overrightarrow{r}}_{ki}\right|}{c}$ is the time for the initial pressure wave at position $\overrightarrow{r}$ to reach virtual element i at position ${\overrightarrow{r}}_{ki}$ at scanning position k; s_k is the signal collected by the physical element (ie, the transducer); S_i is the size of the virtual element i; S is the active size of the physical element.

FIGURE 2.

Schematic of conventional DAS vs. virtual signal reconstruction approach

The distance between two adjacent virtual elements is defined as λ/4, where λ is the acoustic wavelength in water at the center frequency of transducer. The initial pressure $p_0\left(\overrightarrow{r}\right)$ at position $\overrightarrow{r}$ is reconstructed by back-projection of the virtual signals received by all the virtual elements, which can be expressed as $p_0\left(\overrightarrow{r}\right)={\sum }_{k=0}^{n-1}{\sum }_{i=0}^{N-1}{s}_{ki}\left(\overrightarrow{r},t_{ki}\right)$. As the transducer rotates, it provides the difference in delay time between the virtual elements, which allows the reconstruction of a photoacoustic image.

To further improve the focusing quality, a coherence factor (CF) is applied in the reconstruction approach. The signal coherence is defined as

$$\text{CF}=\frac{1}{N}\frac{{\left|{\sum }_{i=0}^{N-1}{s}_i\right|}^2}{{\sum }_{i=0}^{N-1}{\left|s_i\right|}^2},$$ (2)

where s_i is the received signal of virtual element i after proper time delay; N is the total number of virtual elements on a transducer.

3 |. RESULTS AND DISCUSSION

To validate the fan-shaped scanning approach, two sets of phantom experiments were performed. Figure 3A shows the photograph of the phantom with three targets (pencil leads, 0.5 mm in diameter each), in which the three pencil leads were embedded in phantom in a vertical line with a separation of 2 mm in depth for any two adjacent targets. The distance between the transducer and the top target was 7 mm. Figure 3B presents the PAT images recovered using the conventional DAS, and Figure 3C,D presents the virtual signal approach with scanning schemes A and B, respectively, where we see that the targets are recovered with entirely incorrect dimensions using the convent DAS, that the targets can be detected with appreciable dimensions using the virtual signal approach with scanning scheme A, and that the targets can be reconstructed with accurate dimensions using the virtual signal approach with scanning scheme B. Figure 3D shows the intensity profiles across the center of each target along the white dashed lines shown in Figure 3C. The full width at half maximum (FWHM) of these profiles was calculated to be 1.78, 1.57, and 2.42 mm for targets 1, 2 and 3, respectively, using scheme A, and 0.5638, 0.5428, and 0.5448 mm for targets 1, 2 and 3, respectively, using scheme B.

FIGURE 3.

Photoacoustic imaging of three targets embedded in the background phantom along a vertical line in depth direction. A, Photography of the phantom with three vertically arranged pencil leads (0.5 mm diameter). Recovered PA images by using conventional DAS (B) and the virtual signal reconstruction approach (C), respectively; D, Intensity profiles of the recovered three targets along the white dashed line across the center of each target shown in C. T1-T3 indicate targets 1–3

In the second phantom experiment, three pencil leads were embedded in the phantom in a triangular shape as shown in Figure 4A. The distance between the transducer and the closest target is 14 mm. Figure 4 gives the PAT images for this phantom experiment. Again, we see that the virtual signal approach using scanning B provides the best quality of image (right, Figure 4C), while the targets are incorrectly recovered using the conventional DAS (Figure 4B). Figure 3D shows the intensity profiles along the white dashed line across each target center. The FWHM of each recovered target was calculated to be 1.74, 2.21, and 2.64 mm for targets 1, 2 and 3, respectively, using scheme A, and 1.13, 1.56, 1.78 mm for targets 1, 2 and 3, respectively, using scheme B. We notice that the resolution is lower than that for the first experiment. This is because the targets were located farther away from the detector in the case of the second experiment, resulting in lower signal-to-noise ratio. In other words, the spatial resolution based on this new fan-shaped scanning approach is related to the distance between the target and transducer, that is, the larger this distance and the poorer the spatial resolution. The reason is that when the target is closer to the transducer the signal from the target can be detected at more scanning angles/positions for more effective image reconstruction, and that the signal-to-noise ratio is larger in this case.

FIGURE 4.

Photoacoustic imaging of three targets embedded in the phantom in a triangular shape. A, Photography of the phantom with three pencil leads (0.5 mm diameter). Recovered PAT images using conventional DAS (B) and the virtual signal reconstruction approach (C), respectively; D, Intensity profiles of the three targets along the white line across the center of each target shown in C. T1-T3 indicate targets 1–3

We also notice that for both the experiments, the conventional DAS cannot deliver a well-reconstructed image (Figures 3B and 4B). The new approach can provide a much better lateral resolution in both cases.

To further demonstrate the ability of the fan-shaped scanning approach, we conducted experiments using two mice bearing patient tissue-derived xenograft (PDX) tumor administrated with near-infrared (NIR) 830-ATF-ironoxidate nanoparticles (IONP) solution as contrast agent (for detail on this contrast agent, see our previous publications [21, 22]). The PAT images obtained for the two animals using scheme B are shown in Figure 5. We can see that the tumors are recovered clearly for both cases. The intensity profiles along the dashed lines across the tumor center are given in Figure 5C,D,G,H. From these profiles, the sizes of the tumors were photoacoustically measured to be 6.06 and 4.65 mm, which are in good agreement with the actual tumor sizes.

FIGURE 5.

Photoacoustic imaging of two mice bearing PDX tumor. Photographs of tumor-bearing mice (A, E); PAT images of the tumors for the two mice (B, F); Intensity profiles of the tumors along the dashed lines 1 and 2 shown in B (C, D), and along the dashed lines 1 and 2 in F (G, H), respectively

The results from both the phantom and animal experiments have demonstrated the advantages of this novel fan-shaped scanning approach for miniaturized PAT with a large FOV. Compared to the miniaturized transducer array-based PAT system [19, 23], our new approach would provide an inexpensive alternative as it avoids the use of costly transducer array and the associated multi-channel data acquisition system. Meanwhile, the use of a larger transducer element in our approach would provide significantly better sensitivity and imaging depth. We note that the fan-shaped scanning of a single transducer certainly is relatively slow and offers lower temporal resolution. This, however, can be significantly improved by the combination of a single transducer with a fast scanning mechanism such as MEMS mirror scanning of the ultrasound beam.

4 |. CONCLUSION

In summary, we have described and explored a new scanning approach to realize miniaturized PAT along with a virtual signal reconstruction algorithm. Several phantom and animal experiments were conducted, and the results obtained indicate that this fan-shaped scanning approach is capable of imaging small targets located in deep tissue. This work would produce at least the following three significant impacts: (a) a highly miniaturized handheld or endoscopic PAT system could be achieved using the fan-shaped scanning configuration; (b) A large effective imaging region or FOV could be realized using the approach described; and (c) A high detector sensitivity and deep penetration depth could be obtained by utilizing one or two transducers having a relatively large size. We plan to develop an endoscopic PAT system based on the fan-shaped scanning for applications such as image-guided cancer surgery and GI tract imaging in the near future.