Brain Tumor Classification Using Modified ResNet50

Brain tumors pose serious healthcare difficulties due to their varied biological activities and effects on patient outcomes. For treatment planning to be successful, accurate diagnosis and classification are essential. The gold standard for detecting brain tumors is still magnetic resonance imaging (MRI), but manual interpretation is labor-intensive and prone to error. Medical imaging has changed as a result of deep learning, especially convolutional neural networks (CNNs), which allow for automatic and extremely precise tumor categorization. This study proposes an ensemble-based deep learning framework integrating ResNet50 and EfficientNet-B7 for classification and U-Net for segmentation, leveraging the crystal clean brain MRI Dataset containing approximately 22,000 high-resolution images. Comprehensive preprocessing techniques, including data augmentation and normalization, enhance model performance. Additionally, the integration of the Segment Anything Model (SAM) improves tumor boundary delineation. Experimental results demonstrate superior classification accuracy (98.66%) and robust segmentation, underscoring the potential of AI-driven solutions in advancing brain tumor diagnostics and optimizing clinical workflows.