A Multi-module Medical Image Fusion Method Based on Non-subsampled Shear Wave Transformation and Convolutional Neural Network

Abstract

In order to solve the fusion limitation of single module medical image, a multi-module medical image fusion method based on the non-subsampled shear wave transform (NSST) and convolutional neural network (CNN) is proposed. Firstly, NSST is used to decompose the source image into multi-scale and multi-direction layers to obtain the sub-band coefficients of low frequency and high frequency. Secondly, the low-frequency sub-band coefficients are derived from region energy and variance, and the fusion method based on region feature weight is adopted. The high frequency inner layer sub-band coefficient is firstly calculated by CNN, and then selected by average gradient weighting. The high frequency outer layer sub-band coefficient adopts the larger region absolute value fusion rule. Finally, the image is reconstructed by inverse NSST. The experiment results show that the proposed method achieves better fusion effect on both subjective and objective evaluation indexes.