Brain MRI Scans Super-Resolution With Wavelet and Attention Mechanisms

High-resolution magnetic resonance imaging has become more popular in the medical field because it produces detailed structural information. However, clinical acquisition of the high-resolution (HR) magnetic resonance imaging (MRI) from a low-resolution (LR) image remains intricate task due to hardware limitations, long scan times, and high system costs. Therefore, promising super-resolution techniques have been investigated, notably those based on deep learning techniques, to overcome these challenges and improve the LR images into HR images. In this paper, we propose a Deep Extract Features with Wavelet Transform and Attention Connection (DEFWAC) model to enhance MRI scans resolution. In our SR model, the wavelet function is employed as a dilatation filter technique to enhance feature extraction. This integration of convolutional layers with wavelet processing offers a significant advantage. By combining the local information obtained from convolutional neural networks (CNN) with the global information derived from wavelet analysis, our model effectively extracts additional features for the purpose of super-resolution (SR) image generation. The proposed method has been compared with SR baselines techniques. Experimental results show that our model performs noticeably than other methods as well as being beneficial to the MRI SR reconstruction.