HiMAL: Multimodal Hierarchical Multi-task Auxiliary Learning framework for predicting Alzheimer's disease...

Area of Science:

• Neuroscience
• Artificial Intelligence
• Medical Informatics

Background:

• Mild Cognitive Impairment (MCI) is a precursor to Alzheimer's Disease (AD).
• Accurate prediction of MCI to AD transition is crucial for timely intervention.
• Existing predictive models often lack multimodal data integration and longitudinal analysis.

Purpose of the Study:

• To develop and validate a novel multimodal framework, Hierarchical Multi-task Auxiliary Learning (HiMAL), for predicting cognitive composite functions.
• To estimate the longitudinal risk of transition from MCI to AD.
• To provide clinically informative explanations for disease progression prediction.

Main Methods:

• Utilized multimodal longitudinal data (imaging, cognitive, clinical) from the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset.
• Developed the HiMAL framework for predicting AD conversion within 6 months.
• Compared HiMAL performance against state-of-the-art baselines using AUROC and AUPRC metrics.
• Conducted an ablation study to determine modality contributions and provided longitudinal explanations.

Main Results:

• HiMAL demonstrated superior prediction performance compared to single-task, single-modality baselines (AUROC=0.923, AUPRC=0.623).
• Ablation analysis identified imaging and cognitive scores as key contributors to prediction accuracy.
• The model successfully predicted disease progression with high accuracy (P < .05).

Conclusions:

• The HiMAL framework offers a robust and accurate method for predicting MCI to AD progression.
• Model explanations provide clinically relevant insights into cognitive decline.
• HiMAL's reliance on EHR data indicates significant translational potential for point-of-care monitoring and management of high-risk patients.