UCSF

Alzheimer's Disease (AD) is characterized by the accumulation of tau neurofibrillary tangles, which spread along the brain's white matter tracts and correlate with cognitive de-cline. Modeling and predicting tau progression is crucial for developing of therapeutic treatments, but current approaches provide limited insight into individual variation. This study combines statistical and biophysical modeling approaches to capture individuals' longitudinal tau spread. We employed an event-based model (EBM) to identify time-evolving pathology patterns in 650 Alzheimer's Disease Neuroimaging Initiative (ADNI) subjects. We then fit a biophysical network diffusion model (NDM) to the resulting tau trajectories. By optimizing the initial tau pattern of the NDM, we achieved a strong fit to individuals' empirical data, enabling predictions of past and future tau. Our findings reveal a phenomenon of convergence of inter-region and inter-subject tau over time, as well as two underlying seed archetypes. Our novel integration of event-based and biophysical modeling may prove clinically valuable and further an understanding of heterogeneity in AD pathophysiology.