Multi-modality Anatomical and Functional Medical Image Fusion Based on Simplified-spatial Frequency-pulse Coupled Neural Networks And Region Energy-weighted Average Strategy In Non-sub Sampled Contourlet Transform Domain

Spatial frequency pulse coupled neural networks (SF-PCNN)'s parameters need to be set manually and have great influence on multi-modality medical image fusion, and unsuitable parameters setting will lead to low-quality fusion result. In order to solve this problem, pulse generator of SF-PCNN was simplified and automatic setting strategy of linking strength was designed, Simplified-SF-PCNN model suitable for multi-modality medical image fusion was proposed. This study proposed a multi-modality medical image fusion framework using Simplified-SFPCNN and region energy-weighted average strategy in non-sub sampled contourlet transform (NSCT) domain. Source images were decomposed into low and high-pass bands by NSCT, then region energy-weighted average strategy was adopted as low-pass bands fusion rule and Simplified-SF-PCNN was adopted as high-pass bands fusion rule, finally the inverse NSCT was used to generate fusion result. MRI-PET, MRI-SPECT and CT-SPECT medical image fusion was performed, and evaluation metrics of fusion performance were adopted to verify the proposed method's superiority. Compared to five mainstream medical image fusion solutions, the proposed method has better performance, and it can solve the manual setting problem of SF-PCNN's parameters.