Convolutional Gated Recurrent Unit with Spectral Attention for Medical Image Recognition in the Internet of Medical Things

Medical image recognition in Internet of Medical Things (IoMT) environment presents to the process of automatic analysis and inference of medical images, such as X-rays or CT scans, within a related healthcare system. The IoMT enables data transmission between smart medical devices, allowing effective remote analysis and monitoring. This integration enhances analytical accuracy, minimizes manual effort, and supports faster decision making in healthcare. However, in IoMT systems, medical images are collected from diverse connected devices of varying sizes and differing quality, and processing these data precisely remains challenging. This minimizes the effectiveness of medical image recognition and delays diagnosis. To overcome these challenges, the proposed Spectral Attention-Convolutional Gated Recurrent Unit (SA-CGRU) is suggested for enhanced medical image recognition performance and improved model diagnostic ability. The proposed SA-CGRU model offers improved classification accuracy by focusing on the most relevant spectral data, enabling accurate image recognition and enhancing diagnostic precision while efficiently managing heterogeneous information from IoMT devices. The proposed SA-CGRU achieves an accuracy of 93.5%, recall of 90.8%, 91.2% of F1-score and precision of 91.7%, outperforming conventional models.