Research On Iterative Reconstruction Methods For Limited-view Photoacoustic Tomography

Photoacoustic tomography (PAT) is a novel technique of noninvasive medieal imaging. It can combine the high contrast in optical imaging and the good spatial resolution in ultrasound imaging, and it has been attracting more and more attention. In PAT, the pulsed laser is employed as the radiation source to generate ultrasound waves, the ultrasonic transducers are used to receive photoacoustic signals, and the captured signals are utilized to obtain the distribution of optic absorption within biological tissues based on a certain reconstruction algorithm. Image reconstruction is a key technology of PAT. Most of the proposed algrithms need some ideal conditions, such as full-view scan, and is therefore restrieted in practicability. Although much progress has been reported in recent years, problems still remain in PAT. Because of the limitations of the imaging methods and reconstruction models, much effort is still needed to further improve the quality of the reconstruction results.This thesis focuses on the limited-view PAT, and researches on further improving the accuracy and the convergence rates of the limited-view PAT reconstructions are conducted. Three new iterative imaging methods, which incorporate the incomplete data of limited-view, the effective scanning angel and the sparse priori knowledge of the absorption sources in the reconstruction, are proposed to improve the accuracy of PAT. Therefore, the quality of the limited-view PAT images can be better. On the other hand, the preconditioning technology and total variation regularization is employed to optimize the reconstruction models, which further improves the convergence properties and stability of the above methods. The tasks are summarized as follows:1. The unit detection photoacoustic imaging system was builded. PAT experiments including gelatin phantoms, biological tissue in-vitro and vascular structure of rat brain in-vivo are conducted with satisfactory results obtained. And a computer simulation platform of PAT is establised. Both of them provide data and experiments basis for limited-view PAT.2. A combined iterative reconstruction method is used for the limited-view PAT. Such a combination of simultaneous algebraic reconstruction technique and the filtered back projection algorithm can effectively reduce artifacts and distortions caused by the incomplete data, and greatly improve the quality of the limited-view PAT. The preconditioned conjugate gradient method is applied to improve the ill-posedness of the limited-view PAT problem. By converting the high-dimensional optimization problem into a one-dimensional optimization problem, the number of iterations and reconstruction time can be effectively reduced.3. An iterative adaptive weighted filter backprojection approach is applied to the limited-view PAT. Based on the effective scanning angle weighted back projection model, an image-based adaptive weighted value is emplied to update the correction residual image at each iteration step. Therefore, the image residuals gradually become smaller. To improve the convergence properties and stability of the iterative loop, a regularization is proposed. By adding the residual image and the regularized image to the reconstruction, this method is expected to deliver a good result in less than five iterations.4. An alternating direction reconstruction method is proposed for the limited-view PAT. Based on the compressed sensing theory and L1 norm optimization technique, this method incorporates the sparse prior information into the reconstruction. It eliminates the artifacts caused by back projection and reconstructs a high-resolution image with limited-view signals. During the process of iterative, the total variation is utilized in the reconstruction model. The results of simulation show that this method is more effective for noise suppression and can accurately obtain a photoacoustic image with a small amount of data.