Data Structure Assisted Accelerated Reconstruction Strategy for Handheld Photoacoustic Imaging

Abstract

In photoacoustic computed tomography (PACT), advanced model based iterative image reconstruction (MoBIIR) offers several advantages over analytical methods such as back-projection, time reversal, Fourier transform, delay and sum algorithms. However, MoBIIR also shows some disadvantages such as requirement of large storage memory, higher matrix computation time and necessity of selecting optimum parameters for the right solution. When using model based reconstruction methods for high resolution photoacoustic and ultrasound tomography, large matrix computation time is an important concern. In this chapter, we will discuss about filtered back-projection, time reversal methods, F-K migration and a specific model based iterative photoacoustic image reconstruction scheme where the direct non-symmetric photoacoustic system matrix of form \(Hx=z\) (where H is m by n matrix and \(m>n\)) has been analyzed in detail using Least Squared Conjugate Gradient (LSCG) method where the computation of the \(H^TH\) and thereafter regularization are explicitly avoided. Apart from this, a unique pseudo-dynamical systems approach based iterative algorithm is also discussed to demonstrate the insensitivity of tikhonov type physical regularization (\(\lambda \)), which is used frequently in normal equation of form \(H^THx = H^Tz\). However, to implement the algorithms, the photoacoustic equation is usually discretized over the spatial and temporal domain to form spatial-temporal interpolated model photoacoustic system matrix (H), where the data structure for sparsity is considered for accelerating the computation and hence the reconstruction. Finally, the applications of algorithms in photoacoustic imaging modality are shown. The computational requirements of different reconstruction strategies suitable for handheld photoacoustic imaging are also analyzed and discussed in detail.