Medical image fusion with convolutional neural network in multiscale transform domain

Multimodal medical image fusion approaches have been commonly used to diagnose diseases and involve merging multiple images of different modes achieve superior quality reduce uncertainty redundancy in order increase the clinical applicability. In this paper, we proposed a new algorithm based on convolutional neural network (CNN) obtain weight map for multiscale transform (curvelet/ non-subsampled shearlet transform) domains that enhance textual edge property. The aim method is achieving best visualization highest details single fused without losing spectral anatomical details. method, firstly, (NSST) curvelet (CvT) were decompose source into low-frequency high-frequency coefficients. Secondly, coefficients by generated Siamese Convolutional Neural Network (SCNN), where get series feature maps fuses pixel activity information from sources. Finally, was reconstructed inverse multi-scale (MST). For testing standard gray-scaled magnetic resonance (MR) colored positron emission tomography (PET) taken Brain Atlas Datasets used. can effectively preserve detailed structure performs well terms both visual objective assessment. experimental results evaluated (according metrics) with quantitative qualitative criteria.