Network Modelling of Multimodal Dynamics in Alzheimer’s Disease and Dementia

University Of California, San Francisco, San Francisco CA

Investigators

Linked publications, trials & patents

Abstract

PROJECT SUMMARY / ABSTRACT Alzheimer's Disease (AD) is caused by misfolded proteins that march through brain circuits trans-neuronally, causing stereotyped patterns of damage to the brain over decades of progression, and increasing clinical and cognitive impairments. Using new imaging techniques, spatiotemporal mapping of the biomarkers of AD, including of atrophy, metabolism, and pathology deposition, are becoming possible. However, the precise relationship of these biomarkers to each other is not known. These factors, coupled with insidious onset, clinical heterogeneity, overlap with other dementias and variability in progression, make a rigorous characterization and prognosis difficult. Although the ?trans-neuronal? mechanism of pathology naturally suggests that pathology spread must follow the brain's fiber connectivity, existing methods of predicting progression and cognitive decline do not currently exploit the network information, relying instead on phenomenological or statistical approaches unanchored in the biophysics of networked spread. These gaps hinder understanding of the biophysical mechanism underlying dementias, and preclude accurate quantitative predictors of patients' future trajectory. The objective of this application is to learn, test and apply biophysical models of networked spread in AD. Our central hypothesis is that once a patient's baseline disease status is known, all subsequent disease-related processes are enacted on the brain's fiber connectivity network, i.e. the ?connectome?, in a fully predictable manner. Influence of genetic and environmental actors is already factored in the baseline data. This project will build on and extend our recent novel graph theoretic Network Diffusion model, which mathematically captures the process of trans-neuronal network spread, and is ideally suited for investigating these issues. With this network model as a foundation, we will bring together all key elements of the causal AD progression chain. Then, using human imaging data (atrophy from MRI, A? from AV45-PET, tau from T807-PET and metabolism from FDG-PET) from the public ADNI study, we will mathematically characterize 1) network-based spread of tau and amyloid-beta, 2) the relationship between tau deposition and regional atrophy, and amyloid deposition and regional metabolism. Next, these validated models will be entered in a state-space generative mode of progression that will predict future spatial patterns of the biomarkers. An alternative deep learning approach will also be developed as a comparison with the proposed biophysical modeling approach. Success of this proposal could have wide implications in treatment, care, planning and monitoring of dementia in susceptible populations. Our long-term goal is to develop a common connectome-based biophysics model underlying all dementias, forming the core of novel computational diagnostic and prognostic biomarkers.