JoWUA

Abstract

Image segmentation and classification is a critical and difficult problem in many computer-aided diagnostic and surgical systems, attracting considerable research interest in the computer vision and medical image processing domains. Deep learning-based medical image segmentation has recently been thoroughly researched and demonstrated state-of-the-art performance for several modalities of medical information. In this paper, we introduce the mechanism of spatial and channel wise attention that incorporates in U-Shaped Neural Network. In the task of image segmentation, SCA-U-Net dynamically modulates the multi-layer feature map extraction from the foetal cardiac ultrasound image. SCA utilizes attention method to highlight discriminant regions and acquire vital channel maps to enhance the performance of the network. Extensive studies on two benchmark medical picture datasets demonstrate that our suggested network design outperforms the basic U-Net and its modifications. The efficacy of the results is assessed using metrics such as accuracy, average precision, mean average precision, F-measure, and error rate.