BRAIN TUMOR SEGMENTATION AND CLASSIFICATION FROM MRI USING A U-NET, GAN AND EFFICIENTNET HYBRID MODEL

Accurately interpreting MRI scans plays a crucial role in diagnosing brain tumors. However, doing this manually can be both time-consuming and prone to variation between different readers. To address these challenges, this study introduces an automated deep learning framework that improves MRI image quality, precisely identifies tumor regions at the pixel level, and classifies tumor types.The system combines three key components: a Generative Adversarial Network (GAN) for enhancing image quality, a U-Net model for segmentation, and an EfficientNet- based classifier for tumor identification. The overall workflow includes preprocessing MRI scans, enhancing contrast, segmenting tumor regions, refining masks, estimating tumor volume, and finally determining the tumor type.When tested on the BraTS 2020 and BraTS 2021 benchmark datasets, the proposed approach achieved a Dice Score of 92% and an Intersection over Union (IoU) of 85%. These results outperform several existing baseline methods, while also offering faster and more consistent support for clinical decision-making.