Significance Photoacoustic tomography (PACT) is a novel medical-imaging modality. During the imaging process, biological tissues are irradiated by nanosecond ultrashort light pulse. Fast energy deposition in tissues causes thermoelastic expansion and generates ultrasound emission. Such emissions can be detected by ultrasonic detectors. The penetration depth of PACT (>5 cm) is much higher than that of most optical-imaging modalities, approaching that of ultrasonic imaging. Therefore, PACT has wide potential applications in areas such as blood pressure monitoring, cancer detection, and small-animal studies.

PACT reconstruction requires knowledge of distribution of speed-of-sound (SoS). In practice, acoustic properties of biological tissues are inhomogeneous and unknown, resulting in image degradation in PACT. In its simplest implementation, PACT reconstruction assumes single SoS for both biological tissues and surrounding acoustic-coupling medium (i.e., water in most cases). However, considering that the SoS inside soft tissues varies from 1350 m/s (fat) to 1700 m/s (skin), such an assumption can sometimes be too idealized, resulting in splitting, blurring, and distortion of structural features. Furthermore, the SoS of a range of tumors is much higher than that of normal tissues. Therefore, imaging degradation is much severe in tumor imaging.

To solve this problem, a more rigorous acoustic model is required during the reconstruction process. Some studies have employed ultrasound devices for directly imaging the distribution of SoS. However, the most effective method is the joint reconstruction of initial sound pressure (IP) and SoS. Presently, there are several methods for joint reconstruction. Each method has advantages and disadvantages and is effective in different scenarios. Herein, we introduce several methods developed by our teams and summarize their advantages and disadvantages, supporting the biomedical application of PACT.

Progress The ring-array PACT system (Fig. 1) is widely used in IP-SoS joint reconstruction. Most joint-reconstruction methods are based on delay-and-sum (DAS) or back-projection (BP) reconstruction.

Several conventional joint-reconstruction methods have been developed, including dual-SoS, passive element, and model-based methods. Each of the methods has disadvantages and cannot meet practical application demand.

We have developed four joint-reconstruction methods, including feature coupling (FC), multisegmented feature coupling (MSFC), adaptive PACT (APACT), and signal compensating (SC) methods. The FC method, separates ring arrays into two halves. Two images are reconstructed using signals detected by the two halves separately. The correlation coefficient between the two images is then calculated to measure their similarities. The aim of optimization is to maximize the correlation coefficient. In vivo experiments have shown promising results (Fig. 2). The MSFC method further separates the ring array into eight subarrays. In numerical studies, the MSFC method has shown better reconstruction results for both IP and SoS distributions, with lower computational complexity (Fig. 3). The APACT method, inspired by adaptive optics methods widely used in optical imaging, estimates the inhomogeneity-induced wavefront distortions of photoacoustic signals in the frequency domain. The SoS distribution is calculated based on the estimates of the wavefront distortion. In vivo experiments have also shown promising results (Fig. 5). Numerical studies have further shown the advantages and disadvantages of the APACT method compared with the FC method. In summary, when using the FC method, prior knowledge of the distribution of SoS is vital. However, currently, the APACT method works well in the absence of prior knowledge (Fig. 6). The SC method utilizes the characteristics of electronic impulse response (EIR). In photoacoustic imaging, the EIR of ultrasonic detectors has positive and negative peaks. For the reconstructed IP images, when the positive peak detected by one detector is superposed by the negative peak detected by the opposite detector, the intensity of the reconstructed image is the lowest. Therefore, by minimizing the intensity of the reconstructed IP image, the inhomogeneity-induced wavefront distortion can be estimated. Fourier transform is applied to the reconstructed image to further analyze the relationship between delay time and image intensity in different directions. Numerical studies have shown promising results (Fig. 7).

Conclusion and Prospect Although PACT is a promising imaging modality for various biomedical applications, a better solution to the joint-reconstruction problem is needed. In summary, there are several methods for joint reconstruction, and each is effective in different scenarios. However, all these methods are designed for ring- or hemisphere-array systems, which limits their applications. Therefore, further research on joint reconstruction in linear array and multimodal systems is required to promote the development of this imaging modality.