Toward Clinically Trustworthy Alzheimer’s Diagnosis: Combining EfficientNetV2B0 and XAI Techniques for MRI Analysis

Alzheimer’s disease (AD) is a progressive neurodegenerative disorder where early diagnosis is essential for effective intervention. Recent advancements in neuroimaging and deep learning have led to promising automated diagnostic systems, yet their “black box” nature limits clinical adoption. In this study, we present a novel framework that combines a fine-tuned EfficientNetV2B0-based model with state-of-the-art explainable AI (XAI) techniques, including Grad-CAM and LIME, to classify brain MRI images from the OASIS-1 dataset into four distinct stages of Alzheimer’s progression. The model achieves over 99% validation accuracy while generating visual explanations that reliably highlight disease-relevant brain regions, thereby enhancing interpretability and trust for clinical decision-making. These results not only validate the effectiveness of our approach but also underscore the importance of integrating transparency into AI-driven diagnostics. Future research will explore multimodal imaging data and additional XAI methods to further improve diagnostic reliability and support personalized treatment strategies.