Title: MGU-Net: a multiscale gate attention encoder-decoder network for medical image segmentation

Authors: Le Liu; Qi Chen; Jian Su; Xiao Gang Du; Tao Lei; Yong Wan

Addresses: Shaanxi Joint Laboratory of Artificial Intelligence, Shaanxi University of Science and Technology, Shaanxi, Xian, China; The College of Electronic Information and Artificial Intelligence, Shaanxi University of Science and Technology, Shaanxi, Xian, China ' Shaanxi Joint Laboratory of Artificial Intelligence, Shaanxi University of Science and Technology, Shaanxi, Xian, China; The College of Electronic Information and Artificial Intelligence, Shaanxi University of Science and Technology, Shaanxi, Xian, China ' Shaanxi Joint Laboratory of Artificial Intelligence, Shaanxi University of Science and Technology, Shaanxi, Xian, China; The College of Electronic Information and Artificial Intelligence, Shaanxi University of Science and Technology, Shaanxi, Xian, China ' Shaanxi Joint Laboratory of Artificial Intelligence, Shaanxi University of Science and Technology, Shaanxi, Xian, China; The College of Electronic Information and Artificial Intelligence, Shaanxi University of Science and Technology, Shaanxi, Xian, China ' Shaanxi Joint Laboratory of Artificial Intelligence, Shaanxi University of Science and Technology, Shaanxi, Xian, China; The Department of Geriatric Surgery, The First Affiliated Hospital of Xi'an Jiaotong University, Shaanxi, Xian, China ' The Department of Geriatric Surgery, The First Affiliated Hospital of Xi'an Jiaotong University, Shaanxi, Xian, China

Abstract: Medical image segmentation, the prerequisite of numerous clinical needs, has been significantly prospered from recent advances in encoder-decoder networks. However, uneven reflection of human organs and the subject's tremor and movement cause blurred edges in the image, which is difficult to segment. Hence and more details and context information are needed to resolve this problem. Most of the existing Unet-like architectures do not take into account the multiscale characteristics of medical images and do not make full use of the spatial information and channel information of feature maps, resulting in the loss of detail information. This paper proposes a Multiscale Gate Attention (MGU-Net) encoder-decoder network. Firstly, we use multiscale blocks to focus on the fusion of contextual information. Besides, we use two gate attention to deploy more detailed information. On three different public datasets, compared with other State-of-the-Art (SOTA) methods, the proposed method achieves an improvement.

Keywords: medical image segmentation; gate mechanism; multiscale feature fusion.

DOI: 10.1504/IJCAT.2023.132397

International Journal of Computer Applications in Technology, 2023 Vol.71 No.4, pp.275 - 285

Received: 19 May 2022
Received in revised form: 04 Jul 2022
Accepted: 09 Jul 2022
Published online: 19 Jul 2023 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article