Multimodal MRI Brain Tumor Segmentation: Enhancing Detailed Features and Multimodal Information

Brain tumor segmentation in multimodal MRI images is crucial for clinical diagnosis and treatment. However, the location of the lesion area is uncertain and the edge blur is very prominent in the image performance, so automated segmentation faces huge challenges. Currently, most brain tumor segmentation methods make insufficient use of multi-modal information and do not describe edges well, resulting in low segmentation accuracy. To this end, this paper proposes a multi-modal MRI brain tumor segmentation method based on deep learning. This method uses a deep neural network for training, making full use of the complementarity and difference of multi-modal information, paying special attention to the edges and details of the tumor, and providing a global receptive field through the attention mechanism to focus on the location information of the tumor. This network model enhances tumor localization, extraction of edge detail features, utilization of multi-modal information, and filtering of redundant information. Our method is validated on the dataset of the Brain Tumor Segmentation Challenge, and experimental results show that our method has superior performance compared to many advanced brain tumor segmentation methods.