Early Prediction of Dementia Diseases using Explainable Learning Frameworks

Alzheimer’s disease (AD) is a progressive neurodegenerative disorder and dominant cause of dementia worldwide. Early and accurate prediction of dementia symptoms in AD patients is essential for timely intervention and personalized care planning. Recent progress in Machine Learning (ML) have demonstrated the potential of speech-based analysis in detecting early signs of cognitive decline. However, many of these models operate as opaque black-box systems, limiting their acceptance in clinical environments due to a lack of interpretability and transparency. This study presents a novel explainable learning framework that combines Gated Attention Networks (GANs) with SHAP (SHapley Additive exPlanations) to enable both high-performance prediction and interpretable decision-making. The Gated Attention mechanism enables the approaches to focus dynamically on relevant temporal and acoustic features while suppressing noise and irrelevant segments, thereby improving the feature representation and classification performance. It uses the ADReSS-O Challenge dataset, which contains speech and linguistic data from Alzheimer’s patients and cognitively normal individuals, to capture subtle cues associated with early dementia. In parallel, SHAP analysis provides a comprehensive explanation of the approach’s predictions by quantifying the contribution of every input feature, thus offering clinicians insight into the reasoning behind each decision. Experimental evaluations utilized on the ADReSS-O dataset show that the approach attains outstanding performance, with an accuracy of 99.4%, precision of 99.2%, recall of 99.1%, and an F 1 -score of 99.15%. These results significantly surpass traditional deep learning models and highlight the efficacy of combining attention-based learning with explainable AI techniques for real-world clinical deployment. This investigation underscores the importance of integrating interpretability into predictive models, particularly in sensitive domains such as healthcare, where understanding the rationale behind AI decisions is as important as the predictions themselves.