An Explainable AI Approach Using Lightweight-CNN Model for Alzheimer’s Disease Detection

Alzheimer's disease (AD) is a degenerative brain disorder that causes brain tissue degradation, reducing the affected person's memory and cognitive function. While the exact remedy for the disease remains unknown, early detection may slow its progression. The disease's best treatment and therapeutic interventions require early detection so patients can take preventive steps before permanent neuronal loss occurs. Our research aims to improve the early diagnosis of Alzheimer's disease (AD) by employing a compact and powerful lightweight Convolutional Neural Network (CNN). Our suggested model was trained using a KAGGLE dataset of 11,519 images demonstrating four distinct kinds of Alzheimer's disease. This model demonstrates exceptional performance, with an accuracy of 99.06%, an average Area Under the Curve (AUC) score of 99.98%, and precision, recall, and F1-score of 99%, respectively. Being lightweight, our model uses fewer parameters (0.661 M) than prior works. Additionally, we employed Shapley Additive Explanations (SHAP) for clear and understandable prediction explanations and Local Interpretable Model-Agnostic Explanation (LIME) to improve interpretability. A conceptually transparent artificial intelligence technology, Gradient Weighted Class Activation Mapping (Grad-CAM), was used to represent the model's identified condition. A heatmap was generated to highlight regions of concern for easier classification interpretation.