A physics-informed geometric learning model for pathological tau spread in Alzheimer's disease

Area of Science:

• Neuroscience
• Artificial Intelligence
• Medical Imaging

Background:

• Tau tangles are key indicators of Alzheimer's disease (AD) progression and cognitive decline.
• Tau spread is linked to neuronal connectivity, not just physical proximity.

Purpose of the Study:

• To develop a novel physics-informed geometric learning model for predicting tau buildup and spread in Alzheimer's disease.
• To leverage longitudinal tau imaging and structural connectivity data for improved prediction accuracy.

Main Methods:

• Implemented a graph neural network with physics-based regularization for effective training on smaller datasets.
• Utilized longitudinal tau positron emission tomography (PET) and diffusion tensor imaging (DTI) data from the Harvard Aging Brain Study.
• Validated the model using two- and three-timepoint tau PET measures and cross-validation.

Main Results:

• The proposed model achieved higher peak signal-to-noise ratio and lower mean squared error compared to unregularized graph neural networks and differential equation solvers.
• Demonstrated effective prediction of tau accumulation and spread patterns.
• Confirmed the model's robustness and effectiveness through rigorous validation.

Conclusions:

• The physics-informed geometric learning model offers a promising advancement in predicting tau pathology spread in Alzheimer's disease.
• This approach enhances the accuracy and efficiency of Alzheimer's disease progression modeling.
• The findings support the use of AI and neuroimaging for understanding and tracking neurodegenerative diseases.