Comparison of reconstruction algorithms for sparse-array detection photoacoustic tomography

A photoacoustic tomography (PAT) imaging system based on a sparse 2D array of detector elements and an iterative image reconstruction algorithm has been proposed, which opens the possibility for high frame-rate 3D PAT. The efficacy of this PAT implementation is highly influenced by the choice of the reconstruction algorithm. In recent years, a variety of new reconstruction algorithms have been proposed for medical image reconstruction that have been motivated by the emerging theory of compressed sensing. These algorithms have the potential to accurately reconstruct sparse objects from highly incomplete measurement data, and therefore may be highly suited for sparse array PAT. In this context, a sparse object is one that is described by a relatively small number of voxel elements, such as typically arises in blood vessel imaging. In this work, we investigate the use of a gradient projection-based iterative reconstruction algorithm for image reconstruction in sparse-array PAT. The algorithm seeks to minimize an 1-norm penalized least-squares cost function. By use of computer-simulation studies, we demonstrate that the gradient projection algorithm may further improve the efficacy of sparse-array PAT.