Medical image fusion model using CT and MRI images based on dual scale weighted fusion based residual attention network with encoder-decoder architecture

Deep learning strategies have evolved as powerful tools to rectify the complexities related to medical image integration. These strategies utilize effective models to learn and draw out the related features automatically from distinct images, allowing the generation of high-resolution fused images. Therefore, in this paper, we developed an innovative deep learning-based medical image fusion model to solve traditional problems. Initially, the required Computed Tomography (CT), and Magnetic Resonance Imaging (MRI) images are collected manually. The medical image fusion process is performed using the Dual-scale Weighted Fusion-based Residual Attention Network with Encoder-Decoder Architecture (DWF-RANED). In the developed model, the CT and MRI images are given to the first and second scales of the encoder module. Further, weighted fusion is performed using the features obtained from the encoder module. Here, the weights are tuned by the Modified Good and Bad groups-based Optimizer (MGBO) to enhance the effectiveness. Finally, the attained features are carried to the decoder component, where the images are reconstructed into individual images. The designed model's performance is verified by performing various performance investigations with recent models. The final solutions illustrated the better functionality of the suggested model.