Research On Reconstruction Algorithms For Photoacoustic Imaging Based On Simulated Data

In photoacoustic imaging, pulsed laser is been used as the radiation source to generate ultrasound waves, and this signal is been collected by ultrasonic transducers to reconstruct the distribution of optic absorption within biological tissues. As the novel technique of biological tissue imaging, it combines the advantages of pure optical imaging and ultrasound imaging. Since photoacoustic imaging has good properties like, high contrast, good spatial resolution and noninvasive to human body, it has been getting more and more attentions in recent years.Image reconstruction algorithm is one of the key technologies of photoacoustic imaging. Nowadays, most research about reconstruction algorithm is based on some ideal assumptions, such as full scan data, which limits their practicability. Although photoacoustic imaging achieved rapid development in recent years, but due to the limitations of imaging method and the reconstruction model, much effort is still needed to further improve the quality of the reconstruction results. To improve the above problems, this paper is focusing on the research and implement of image reconstruction algorithms. The major work of this dissertation is described as follows.1. At first, a simulation system for the photoacoustic imaging is build. This system simulates the working principles of photoacoustic tomography (PAT) based on focused ultrasound transducer. The output of this system is photoacoustic signals collected by transducer, which is used as the input in the following reconstruction algorithms.2. Several classic photoacoustic imaging reconstruction algorithms are implemented, including filter back projection algorithm, ML-EM algorithm, ISTA and ISRA. After studying the theory of reconstruction algorithm, it is found that computation of system matrix is the most time consuming part in these algorithms. To speed up the algorithms a new reconstruction idea is been proposed and implemented on ISRA. Experiments have been done, and it proved that the reconstruction speed is accelerated about four times.3. To improve the reconstruction accuracy of the limited-view data, ADMM algorithm which is implemented in two different models is proposed.4. Also, an idea based on auxiliary function is proposed in this paper. The idea is used in five different models. The algorithms automatically satisfied the non-negative constraints of pixels without step size factor required, then ensuring the monotonous decreasing of objective function with global convergence provided.5. Three types of experiments are designed to testify the algorithms mentioned in this paper.