Application of Improved nnU-Net in MRI Image Segmentation of Breast Tumors

Breast tumors, as a highly prevalent disease, can cause significant harm to patients' physical and mental health. Accurate and effective segmentation techniques for preoperative image analysis play a decisive role in subsequent surgical diagnosis and treatment. This study utilizes the nuU-Net network as the base architecture and enhances it by incorporating the Convolutional Block Attention Module (CBAM). By combining channel and spatial attention, CBAM enhances focus on tumor regions while suppressing background noise. Simultaneously, the Atrous Spatial Pyramid Pooling (ASPP) module is introduced to capture multi-scale contextual information through dilated convolutions with different rates, addressing the challenges of varying tumor sizes and blurred boundaries. Experimental results demonstrate that the improved nou-Net network outperforms other models in segmenting breast tumors in MRI images.