U-Net Based approach for Brain Tumor Segmentation

Abstract

Brain tumor segmentation plays a vital role in medical image analysis, offering crucial insights for diagnosis, treatment planning, and surgical guidance. However, manual segmentation by radiologists is often time-intensive, subjective, and susceptible to variability between observers. In this study, an automated segmentation approach is proposed using a U-Net-based convolutional neural network (CNN), which is specifically tailored for biomedical image segmentation tasks. The model is trained and tested on MRI images, with preprocessing and data augmentation techniques applied to improve its generalization performance. To evaluate the effectiveness of the segmentation, commonly used metrics such as dice coefficient, Intersection over Union (IoU), accuracy, and sensitivity are employed. These metrics collectively assess the model’s precision in identifying tumor boundaries, ensuring high overlap with tumor regions while minimizing errors like false positives and false negatives. The used model achieved an accuracy of 99.44%, a Dice score of 83.76%, and an IoU of 72.70%. These results demonstrate the U-Net-based framework's robustness and reliability, highlighting its potential to assist radiologists in achieving faster and more consistent brain tumor segmentation