Residual Edge Attention in U-Net for Brain Tumour Segmentation

Abstract

Identification and delineation of the tumour area in images of the brain constitute the crucial job of brain tumour segmentation in medical imaging. This task is crucial for diagnosis, treatment organizing, and keeping a track of brain tumours. Medical imaging methods like magnetic resonance imaging (MRI) or computed tomography scans are frequently used to divide brain tumours in real time. (CT). These imaging techniques provides high-resolution images for the brain that allows doctor to identify and locate tumours. There are several approaches to brain tumour segmentation, including manual segmentation by a radiologist, semi-automated segmentation using software tools that require some manual intervention, and fully automated segmentation using artificial intelligence (AI) algorithms. In this probing work, For segmenting brain tumours, we had anticipated Residual Edge Attention in U-Net design (ResEA-U-Net). Residual Edge Attention (ResEA) is a novel approach that enhances the performance of the U-Net architecture for brain tumour segmentation. The U-Net is often used in deep learning architecture for medical MRI brain image segmentation tasks, but it suffers from limited receptive field and feature reuse. To address this limitation, ResEA is expected to capture wide-range dependencies and enable network to focus on important regions of the image. The ResEA block contains of a residual block and an attention block that are connected in series. The residual block is created to improve the gradient flow and feature reuse, while the attention block focuses on important regions of the image by assigning higher weights to informative edges. The expected approach to evaluated on the BraTS data, which contain images of magnetic resonance of brain tumours. Experimental outcomes demonstrate that the ResEA-U-Net outperforms the baseline U-Net and other state-of-the-art methods. Overall, the suggested ResEA-U-Net architecture is a promising approach for brain tumour segmentation because it improves segmentation accuracy and lowers segmentation false positive rate, which can be essential for precise detection and therapy planning.