Photoacoustic reconstruction for linear scanning geometry includes the delay-and-sum method, the
spectral-domain method and the time-domain based method. In practice, the data collection using the planar detection
geometry is not full-view, causing the details of the reconstructed object to be blurred and distorted. In addition to the
exact formulation, we adopt a heuristic reconstruction method. In this paper, we demonstrate photoacoustic
reconstruction for linear scanning geometry by formulating the image reconstruction into an optimization problem, and
solve the problem with the particle swarm optimization (PSO) method. In this method, first we guess the initial optical
energy distribution. According to photoacoustic model, described by the Helmholtz equation, the generated
photoacoustic wave can be collected with planar detection geometry. The spherical Radon transform is adopted for the
simulation of arbitrarily guessed optical energy distribution. Next we compare the collected signals generated from the
guessed optical energy distribution with the measured signals by the sum of squared differences. By minimizing the error
sum among various guesses, the initial optical energy distribution is obtained. In this regard, no limited-view is
encountered. To guess the initial distribution efficiently such that the sum of the squared differences is minimized is an
optimization problem with the dimension of unknowns being the size of the initial optical energy distribution. PSO is a
derivative-free and population-based stochastic method that has been used to solve various optimization problems due to
its simplicity and efficiency. High computational costs aroused from a large number of particles required can be
alleviated with the use of the graphic processing units (GPUs). The proposed reconstruction method based on the PSO
algorithm along with the spherical Radon transform is implemented on a NVIDIA Telsa C1060 GPU.
|