Research on Multimodal MRI Glioma Segmentation Based on Attention Mechanism

Glioma is the most common primary malignant intracranial tumor, accounting for about 70 % of primary brain tumors. However, the complex tissue structure of brain gliomas, their small proportion in MRI brain images, and the imbalance in MRI image grayscale hinder the development of glioma segmentation techniques and their practical application in clinical settings. This study focuses on glioma segmentation algorithms using deep learning with a lightweight attention mechanism to optimize MRI image preprocessing and model design. The research includes using data augmentation to prevent overfitting, adding attention mechanisms to the skip connection part of the UNet model to improve segmentation accuracy and conducting experimental analysis. The experimental results show that our proposed algorithm achieves an accuracy of 86.57 %, sensitivity of 83.11 %, and Dice coefficient of 85.20 % on the BraTS2020 dataset. In conclusion, our innovative model outperforms the traditional UNet model.