Comparative Studies of Reconstruction Algorithms for Sparse-View Photoacoustic Tomography

Inverse source reconstruction is one of the most challenging problems in photoacoustic tomography (PAT) due to its ill-posed nature. Despite the extensive work done by researchers on this problem, there is currently no universally accepted solution. Regularization methods assume a significant role in the sparse-view PAT inverse problem. This study compares six inverse source reconstruction methods based on Lp (0≤p≤2) regularization and investigates the effects of signal sampling quantity, measurement noise, and sparsity on the performance of the reconstruction algorithms through a series of numerical simulations. The experimental results indicate that the average peak signal-to-noise ratio of the iterative hard threshold (IHT) method is twice that of the Tikhonov method and 3 dB higher than that of the L1magic method. The average running time of the IHT method is half that of the Tikhonov method and one-seventh of the L1magic method. To further assess the performance of these six reconstruction methods, we conducted agar phantom experiments to examine their ability to resolve details. The aim is to provide valuable guidance for the development and application of algorithms in relevant fields.