A Quantitative Assessment of PET Brain Image Reconstruction using MAP and Neural Network based Segmentation of CG Algorithm

Abstract

This paper addresses a comparative analysis of PET Brain image reconstruction based on iterative and weighted least-square (WLS) algorithms. In previous years, the analytical approach was used to reconstruct the Positron Emission Tomography (PET). This approach requires a minimization of a convex cost function and accompanied by many problems related to the computational complexity. The poles apart iteration methods are Conjugate Gradient (CG), Coordinate Descent (CD) and Image Space Reconstruction Algorithm (ISRA). It has many advantages compared to conventional approach. The functional protocol used here is neural network based segmentation of PET brain image using CG method to improve the CG algorithm. In this step of process, the image was segmented using Neural Network and reconstructed with CG algorithm. This statistical fashion can provide better and high PSNR along with lowest noise in the PET Brain image. An assortment of image quality parameters is considered to analyze the PET brain image in this algorithm. The PET brain image is constructed and simulated in MATLAB /Simulink package.