Published Paper Details:

Alzheimer's disease (AD) is a progressive neurodegenerative disorder in which early and accurate prediction of disease progression remains a clinical challenge. This study proposes an attention- based multimodal deep learning framework that integrates structural magnetic resonance imaging (MRI) and clinical features for multi-class progression prediction. MRI volumes are processed using a three- dimensional convolutional neural network to extract spatial biomarkers, while demographic and cognitive variables are encoded through a clinical feature encoder. A modality-level attention fusion mechanism is introduced to adaptively weight imaging and non-imaging information, enabling the model to learn clinically meaningful representations. The system is trained and evaluated on the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset with class imbalance handled through weighted sampling and cost-sensitive learning. The framework also supports explainable artificial intelligence using gradient- based localization and feature attribution for clinical transparency. Experimental results demonstrate improved macro-F1 performance compared with unimodal approaches and provide interpretable attention distributions indicating modality contribution.