U-CCNet: Brain Tumor MRI Image Segmentation Model with Broader Global Context Semantic Information Abstraction

Since brain tumor detection can help in the diagnosis of diseases such as Alzheimer's disease, segmentation and classification of brain tumors are crucial for diagnosis and treatment. Although radiologists can find the tumor area using magnetic resonance imaging (MRI), it is more time-consuming. The high-precision segmentation results provided by deep learning models can assist radiologists in better-identifying brain tumors. In this paper, we propose a U-Net-based approach to detect brain tumors in MRI. In this paper, the cross-attention mechanism is added to the U-Net benchmark architecture and trained and tested on the LGG Segmentation dataset (TCGA-LGG) dataset. The test results show that the proposed model performs well in brain tumor MRI segmentation tasks. Finally, the proposed U-CCNet model is compared with other studies. Experimental results show that this study is superior to other models in the Dice similarity coefficient and IoU index. The proposed method is tested on the LGG brain MRI segmentation dataset. There is a significant improvement in segmentation accuracy in terms of dice score and mIoU, reducing the use of computationally expensive 3D networks.