Enhancing Brain Tumor Diagnosis through Adaptive Feature Aggregation based Transfer Learning

In recent years, computer-aided diagnosis systems have emerged as valuable assets for assisting in the detection and classification of brain tumors using magnetic resonance imaging (MRI). This study addresses the growing challenges in brain tumor diagnosis by introducing an automated system that utilizes MRI as a diagnostic tool. Our proposed approach employs a deep learning architecture utilizing pretrained Resnet18 and the Adaptive Feature Aggregation technique for classifying tumors into four categories (Normal, Pituitary, Meningioma, and Glioma). Achieving a validation set accuracy of 98.28% and a test set accuracy of 97.46%, our system outperforms existing brain tumor classification systems. With enhanced performance, flexibility, and interpretability, our suggested strategy is a viable resource for practical applications in brain tumor classification.