Comparing Explainable Artificial Intelligence and Deep Learning Models for MRI-Based Brain Tumor Diagnosis

Abstract

Brain tumors represent one of the most life-threatening diseases, and their early and accurate detection is critical for improving patient outcomes. Magnetic Resonance Imaging (MRI) is the most reliable imaging technique for identifying brain tumors, yet manual interpretation by radiologists is time-consuming and prone to errors. To address these challenges, this study investigates the application of deep learning architectures, including Convolutional Neural Network (CNN), VGG16, VGG19, ResNet50, and MobileNet, for brain tumor detection and classification. A publicly available MRI dataset consisting of glioma, meningioma, pituitary, and no-tumor cases was used. The models were trained and evaluated using accuracy, precision, recall, F1-score, ROC curves, and confusion matrices, while interpretability was assessed using Local Interpretable Model-Agnostic Explanations (LIME). Experimental results demonstrate that ResNet50 achieved the highest performance with 96.9% accuracy, followed closely by MobileNet at 96.6%, whereas CNN performed less effectively at 87.9%. The findings confirm that advanced deep learning architectures not only achieve high classification accuracy but also improve interpretability, thereby offering reliable and clinically applicable solutions for automated brain tumor diagnosis.