An Optimized hybrid Learning Approach for Brain Tumor Identification and Segmentation

Brain tumor is life-threatening disease and one of the most critical challenges in medical imaging detecting the tumor and early stage. Magnetic resonance imaging (MRI) offers detailed visualization of the tumor, but it requires complex models for reliable interpretation. In this work, we introduced an optimized hybrid learning framework built upon MobileNetV2 for lightweight feature extraction and XGBoost for precise tumor classification and segmentation. This hybrid approach enhances both feature quality and decision accuracy while maintaining the low computational cost. The proposed method is trained and tested on the Brain Tumor MRI dataset, achieving more accuracy compared to conventional CNN-based systems. This model reduces model parameters and training time while maintaining performance, making it suitable for development in clinical and even in resource limited environments. In general, this method demonstrate that hybrid deep and machine learning integration can provide an efficient and automated brain tumor identification and segmentation.