MADE-TransUNet Induced Brain Tumor Detection for Smart Medicare Using Internet of Medical Things

Abstract

The traditional medical diagnosis using magnetic resonance imaging (MRI) is tedious and seriously depends on doctor's experience. In order to handle this issue, this paper proposes an internet of medical things (IoMT) based intelligent MRI medical diagnosis architecture for brain tumor detection, which adopts cloud computing and deep learning technologies. First, the MRI data is obtained through the scanning using MRI machine and to send to a cloud server in which an intelligent image segmentation model is deployed. Second, the cloud server uses deployed intelligent image segmentation model to implement automatically brain tumor detection and sends the results to clients. The key of proposed medical diagnosis architecture is the intelligent image segmentation model in cloud server. This paper proposes MADE-TransUNet in which bilinear fusion multimodal feature module (BFMF) is applied in the encoder stage for better fusion of multimodal features, adaptive response fusion (ARF) is applied in the bottleneck stage for fusion of features with different resolutions to improve feature expression, and the edge-sensitive enhancement and learning (ESEL) module are applied in the encoder to enhance the edge information. The results of experiments and simulation demonstrate MADE-TransUNet outperforms existing networks for brain tumor segmentation tasks. The code is available at https://github.com/zhzhuac/MADE.