An Enhanced Multi-Scale Collaborative Learning Network for Medical Image Segmentation in Internet of Medical Things

Multi-organ segmentation in the abdomen is a key area in medical image segmentation and is essential for accurate diagnosis and treatment of diseases. The diversity of abdominal organs, differences in size, and ambiguous boundaries pose significant challenges to segmentation tasks, making it difficult for existing models to meet clinical demands in complex scenarios. To address this, this paper proposes a Multi-Scale Collaborative Learning Network (MSCL-Net) in conjunction with a medical Internet of Things (IoMT) platform, aiming to enhance the accuracy of abdominal organ segmentation. The introduction of IoMT technology dramatically expands the channels for obtaining medical image data. By collecting massive amounts of data in real-time from various scenarios, such as outpatient clinics and hospitals, doctors can quickly use the segmentation model to obtain preliminary segmentation results of CT images, which specialists can rapidly correct. This workflow improves clinical work efficiency and provides a large volume of highquality data for training the model, further enhancing its accuracy. Regarding model design, the Feature Space Module is used for global feature extraction, the Feature Extraction Module captures local semantic information at different scales, and the Feature Aggregation Module performs deep integration of local and global information. A considerable number of experimental results based on the Synapse dataset demonstrate that MSCL-Net outperforms existing advanced algorithms and achieves 81.71% in the Dice Similarity Coefficient (DSC) and 17.81 in the Hausdorff Distance at 95% (HD95) metrics, demonstrating its outstanding ability in medical image segmentation.