Comparison of iterative reconstruction approaches for photoacoustic tomography

In many iterative algorithms for photoacoustic tomography (PAT), images are reconstructed from an integrated data function g(r&vec;, t) rather than from the measured acoustic pressure data function p(r&vec;, t). The integrated data function is related to the object by a spherical Radon transform, which can be inverted by use of standard reconstruction algorithms. In this work, we investigate a different reconstruction approach that utilizes the measured pressure data function p(r&vec;, t) to directly invert the PAT imaging model. We reveal that these two reconstruction approaches, which are preconditioned versions of each other, have distinct statistical and numerical properties. Each is demonstrated to have characteristics that are advantageous to certain types of data inconsistencies. Numerical results are presented to corroborate our analysis.