Parametrically Optimized Pulse Coupled Neural Network for Analysis of Multimodal Medical Images

The present work proposes an effective way to integrate multimodal medical images to analyze the prognosis of brain lesion and Alzheimer’s disease (AD). The shift invariant non-subsampled shearlet transform (NSST) based decomposition is utilized to obtain multidirectional detail information. The salient features of low frequency (LF) components are extracted to highlight every fine detail in the fused images. Besides this, a parametrically optimized pulse coupled neural network (POPCNN) model is employed to merge high frequency (HF) components. The dynamic parameters of the proposed POPCNN model are determined by the static property of the input HF sub images. Hence it is named so. The superiority of proposed approach is reflected in the experimental result section.