Improved Brain Tumor Segmentation Framework Based on Multimodal MRI and Cascaded Segmentation Strategy

Accurate segmentation of brain tumors is crucial for their early diagnosis. Multi-modal MRI images can provide complementary information, which is essential for improving segmentation accuracy. This paper presents a brain tumor segmentation method based on multi-modal MRI images and a cascade segmentation strategy, taking into account the characteristics of brain tumor MRI images and the hierarchical nature of brain tumor anatomy. The method employs superpixels to generate labels for model training and utilizes a self-supervised One-Shot brain tumor segmentation network. Experimental results demonstrate that the proposed method achieves higher segmentation accuracy compared to existing small-sample brain tumor segmentation models on the brain tumor segmentation competition dataset.