Purpose: Optoacoustic tomography (OAT) is inherently a three-dimensional (3D)inverse problem. However, most studies of OAT image reconstruction still employtwo-dimensional (2D) imaging models. One important reason is because 3D imagereconstruction is computationally burdensome. The aim of this work is toaccelerate existing image reconstruction algorithms for 3D OAT by use ofparallel programming techniques. Methods: Parallelization strategies are proposed to accelerate a filteredbackprojection (FBP) algorithm and two different pairs ofprojection/backprojection operations that correspond to two different numericalimaging models. The algorithms are designed to fully exploit the parallelcomputing power of graphic processing units (GPUs). In order to evaluate theparallelization strategies for the projection/backprojection pairs, aniterative image reconstruction algorithm is implemented. Computer-simulationand experimental studies are conducted to investigate the computationalefficiency and numerical accuracy of the developed algorithms. Results: The GPU implementations improve the computational efficiency byfactors of 1, 000, 125, and 250 for the FBP algorithm and the two pairs ofprojection/backprojection operators, respectively. Accurate images arereconstructed by use of the FBP and iterative image reconstruction algorithmsfrom both computer-simulated and experimental data. Conclusions: Parallelization strategies for 3D OAT image reconstruction areproposed for the first time. These GPU-based implementations significantlyreduce the computational time for 3D image reconstruction, complementing ourearlier work on 3D OAT iterative image reconstruction.
展开▼
