Longitudinal Alzheimer's Disease Progression Prediction With Modality Uncertainty and Optimization of Information...

Area of Science:

• Neuroimaging
• Computational Biology
• Biostatistics

Background:

• Alzheimer's disease (AD) is a major global neurodegenerative disorder.
• Existing longitudinal models struggle with imaging modality uncertainty and information retention.
• Current recurrent models face challenges in robustly managing information flow over time.

Purpose of the Study:

• To develop a novel model for predicting Alzheimer's disease progression.
• To address the challenge of modality uncertainty in multimodal longitudinal data.
• To enhance the robustness of information flow control in recurrent models.

Main Methods:

• Propose a model to constrain multimodal data into a common representation space.
• Capture intermodality interactions and address modality uncertainty.
• Introduce an auxiliary function to improve recurrent gate performance for longitudinal data.
• Utilize data from the Alzheimer's Disease Neuroimaging Initiative (ADNI).

Main Results:

• The proposed model effectively extracts and constrains information from different modalities.
• The auxiliary function enhances the robust control of information flow over time.
• Comparative analysis demonstrated superior performance across all evaluation metrics compared to existing methods.
• The model successfully addresses modality uncertainty in predicting AD progression.

Conclusions:

• The developed model offers a promising approach for multimodal longitudinal Alzheimer's disease progression prediction.
• The method effectively handles modality uncertainty, a significant challenge in AD research.
• This work advances the capability of predictive modeling for neurodegenerative diseases using complex imaging datasets.