RcsANet: Segmentation of Bladder MRI tumors Based on Improved U-Net Network

Bladder cancer is a common and aggressive malignancy of the urinary system, characterized by high morbidity, mortality, and treatment costs. Tumor segmentation plays a critical role in diagnosis, and the use of neural networks in medical image segmentation has shown significant promise. While U-Net remains widely used in this domain, its inability to effectively capture long-range dependencies limits its capacity to integrate global information. This paper introduces RcsANet, which incorporates residual structures and multiple attention mechanisms to address these limitations. Compared to the traditional U-Net, RcsANet improves the average Dice coefficient for MRI bladder tumor segmentation by 5%.