Performance evaluation of PET image reconstruction using radial basis function networks

Abstract

In this paper, for the reconstruction of the positron emission tomography (PET) images, Artificial Neural Network (ANN) method and Artificial Neural Network-Radial Basis Function (ANN-RBF) method are pursued. ANN is a dominant tool for demonstrating, exclusively when the essential data relationship is unfamiliar. ANN imitates the learning process of the human brain and can process problems involving nonlinear and complex data even if the data are imprecise and noisy. But, ANN calls for high processing time and its architecture needs to be emulated. So, ANN-RBF method is implemented which is a two-layer feed-forward network in which the hidden nodes implement a set of radial basis functions. Thus, the learning process is very fast. By the image quality parameter of peak signal-tonoise ratio (PSNR) value, the ANN method and the ANN-RBF method are compared and it was clinched that better results are obtained from ANN with RBF method.