Publication

Abstract : Analyzing MRI scans is complex due to the often blurry boundaries and inconsistent sizes of tumors, leading to variability in results. This study underscores the critical role of hyperparameter tuning in optimizing 3D U-Net models for accurate brain MRI segmentation, using the BraTS 2020 dataset. Key parameters, including learning rate, optimizer, and batch size, were systematically adjusted to assess their impact on model performance. Identifying the optimal configuration proved challenging, addressed effectively using MARCOS and MAIRCA multicriteria decision methods. The optimal configuration, featuring a learning rate of 0.0001 and the Adam optimizer, achieved an accuracy of 0.966 and a precision of 0.978. This finely tuned 3D U-Net model demonstrated superior performance compared to other models like Hybrid DL and SLF-UNet, emphasizing the importance of meticulous hyperparameter tuning for improving model accuracy and robustness in brain MRI segmentation.