Fully individualized models for cross-sectional and longitudinal network-based tau spread.

Area of Science:

• Neuroimaging
• Neurology
• Biophysics

Background:

• Regional tau burden exhibits heterogeneity, complicating disease progression evaluation.
• Existing tau positron emission tomography (PET) studies link connectivity to tau pathology epicenters but lack individualization.
• Population-based connectomes and epicenters are insufficient for precise tau burden prediction.

Purpose of the Study:

• To develop and validate fully individualized models for predicting regional tau burden using structural connectomes and individualized epicenters.
• To assess the cross-sectional and longitudinal predictive power of these individualized models.

Main Methods:

• Utilized diffusion MRI-derived structural connectomes and individualized tau pathology epicenters.
• Modeled tau burden prediction based on distance along individual structural connectomes from individualized epicenters.
• Evaluated models cross-sectionally and longitudinally, including validation datasets.

Main Results:

• Fully individualized models significantly outperformed population-based models in explaining regional tau burden.
• Individualized models demonstrated improved prediction accuracy in validation datasets.
• Achieved stronger single-subject level prediction of tau burden.

Conclusions:

• A fully individualized approach effectively explains regional tau heterogeneity.
• Findings provide strong in vivo evidence for network-based tau pathology spread.
• This method enhances understanding of Alzheimer's disease progression and network dynamics.