Alzheimer's Disease (AD) is a long-term neurodegenerative condition that demands precise and timely diagnosis to enable early intervention. This study proves that combining deep learning with fuzzy logic along with explainable AI (XAI) can significantly enhance diagnostic accuracy for AD detection and grading. This research presents a new deep learning (DL) methodology for grading the severity of AD based on brain MRI images. The technique combines Swish-activated ResNet for extracting features, fuzzy logic for enhanced decision-making, and XAI methods to increase model transparency. For increased interpretability, Grad-CAM and SHAP methods are used with consistency and clinical applicability. The model is rigorously evaluated on multiple benchmark datasets, including OASIS, ADNI, PDAD, and Alzheimer's Dataset. Experimental results on several datasets show the better performance of the proposed model. It reports 99.2% accuracy on OASIS's and other datasets, outperforming DenseNet-121 (92.8%) and XGBoost (89.8%). This research contributes a hybrid CNN–fuzzy–XAI framework that not only improves diagnostic performance but also enhances explainability, making it clinically relevant and scalable for real-world deployment. This research emphasizes the diagnostic accuracy improvement and early detection support provided by CNN–fuzzy–XAI.
Mohanraj et al. (Tue,) studied this question.