Block Coordinate Descent Based Algorithm for Image Reconstruction in Photoacoustic Tomography

Abstract

Photoacoustic (PA) imaging is a rising, congenital, in vivo biomedical imaging configuration combining equally optics and ultrasonics used for tumor angiogenesis monitoring. A nanosecond laser pulse is accustomed to illuminate biological tissue at a light wavelength typically in the near-infrared (NIR) window when deep light entrance into tissue is desired. The emission pressure rise at the origin is reciprocal to the immersed power and the force wave drives within smooth organic tissues as an audible wave also recognized as Photoacoustic wave. Several photoacoustic image reconstruction algorithms have been developed which includes analytic methods in terms of filtered back-projection (FBP) rather algorithms founded on Fourier transform but these methods generally give suboptimal images. In this paper, we developed an algorithm based on block coordinate descent method for image reconstruction in photoacoustic tomography (PAT). Block coordinate descent (BCD) algorithms optimize the target function over individual segment, at every sub-repetition, whereas keeping all the other segments fixed. This scheme is used to compute pseudoinverse of system matrix and reconstruct image for pressure distribution in photoacoustic tomography. The proposed BCD method is compared with back-projection (BP) method, direct regularized pseudoinverse computation and conjugate gradient based method using simulated phantom data sets.