Correlating tau pathology to brain atrophy using a physics-based Bayesian model

• Published in: Engineering with computers Vol. 38; no. 5; pp. 3867 - 3877
• Main Authors: Schäfer, Amelie, Chaggar, Pavanjit, Goriely, Alain, Kuhl, Ellen
• Format: Journal Article
• Language: English
• Published: London Springer London 01.10.2022; Springer Nature B.V
• Subjects: Alzheimer's disease; Atrophy; Bayesian analysis; Brain; CAE) and Design; Calculus of Variations and Optimal Control; Optimization; Classical Mechanics; Computer Science; Computer-Aided Engineering (CAD; Control; Correlation; Math. Applications in Chemistry; Mathematical and Computational Engineering; Mathematical models; Medical imaging; Neuroimaging; Original Article; Parameters; Pathology; S.I. : Image-Based Methods in Computational Medicine; Statistical inference; Systems Theory; Alzheimer’s disease; Tau PET; Network model; Brain atrophy
• ISSN: 0177-0667; 1435-5663
• DOI: 10.1007/s00366-022-01660-3

Misfolded tau proteins are a classical hallmark of Alzheimer’s disease. Increasing evidence indicates that tau—and not amyloid—is the main agent in driving neurodegeneration and tissue atrophy in Alzheimer’s brains. However, the precise correlation between tau and atrophy remains insufficiently understood. Here we explore tau-atrophy interactions by integrating a multiphysics brain network model and longitudinal neuroimaging data for n = 61 subjects from the Alzheimer’s Disease Neuroimaging Initiative. Using Bayesian inference with a hierarchical prior structure, we personalize subject-level parameter distributions for each individual subject and infer group-level parameter distributions for amyloid positive and negative groups. Our results show that the group-level tau growth for amyloid positive subjects of 0.0161/year is significantly larger ( p = 0.0036 ) than for amyloid negative subjects of - 0.2042 /year. Similarly, the group-level tau-induced atrophy for amyloid positive subjects of 0.0165/year is significantly larger ( p = 0.0048 ) than for amyloid negative subjects of 0.0111/year. These findings support the hypothesis that amyloid pathology has a magnifying effect on tau pathology and tissue atrophy. Our model may serve as a descriptive tool to quantify the correlation between tau and atrophy, as well as a predictive tool to estimate personalized tau pathology, atrophy, and cognitive impairment timelines from a sequence of medical images.