Automated Brain Tumor Detection in MRI Using Enhanced U-NetArchitecture: A Comparative Analysis of Segmentation Methods

Accurate segmentation of brain tumors in MRI scans is critical for early diagnosis and treatment planning,but manual segmentation is time-consuming and highly dependent on the expertise of the operator. Toaddress this, automated systems leveraging deep learning have gained attention for their potential tostreamline the process. This study focuses on optimizing the U-Net architecture to enhance thesegmentation accuracy of brain tumors in two-dimensional MRI images. Various training parameters,such as dropout rates, data preprocessing techniques, and loss function configurations, were exploredacross six different experimental setups. The performance of these configurations was evaluated basedon segmentation accuracy using the BraTS datasets. Our findings demonstrate that fine-tuning theseparameters leads to significant improvements in tumor segmentation, providing a robust foundation forfully automated, computer-aided diagnosis systems.

Keywords:

Deep Learning for Brain Tumor Segmentation , Enhanced U-Net Architecture , AutomatedTumor Detection , MRI Image Analysis , Medical Image Segmentation , Dropout and Batch Normalization