Session 5

[BISC5-03] The Modified SE(2) CNN: Geometric Deep Learning for Classifying Magnetic Resonance Imaging of Brain Tumor

*Clara Lavita Angelina^1,2, Sunil Vyas³, Fu Ren Xiao³, Yuan Luo^3,4,5, Hsuan Ting Chang^1,2 (1. Department of Electrical Engineering, National Yunlin University of Science and Technology, 2. Graduate School of Engineering Science and Technology, National Yunlin University of Science and Technology, 3. Institution of Medical Device and Imaging, National Taiwan University, 4. YougLin Institute of Health, National Taiwan University, 5. Program for Precision Health and Intelligent Medicine, National Taiwan University)

Early-stage brain tumor detection is crucial for survival. This research uses a modified SE(2) CNN for MRI brain tumor categorization. Trained on 2611 scans in five classes, the model excels in binary classification, especially in AVM vs. non-AVM, with high accuracy and F1 score. It also identifies Meningioma, Pituitary, Metastases, and Schwannoma, with metrics over 0.5.