Friday, May 27, 2022 • 11:15 am–12:30 pm (CT) • Light Hall, Room 214
Organizer: Eardi Lila, University of Washington
Chair: Elizabeth Sweeney, University of Pennsylvania
Analyzing brain structural connectivity as continuous functions
Zhengwu Zhang, University of North Carolina at Chapel Hill
Will Consagra, University of Rochester Medical Center
Martin Cole, University of Rochester Medical Center
Xing Qiu, University of Rochester Medical Center
Brain networks are often analyzed as discrete adjacency matrices, containing elements describing relationships between regions of interest (ROIs). A pre-specified brain atlas is typically used to define ROIs. The choice of atlas can be arbitrary, resulting in large ROIs and leading to a loss of connectivity information within and between ROIs. This talk introduces a continuous brain connectivity (ConCon) representation that can overcome these challenges and develops a fast dimension reduction technique for inferring relationships between ConCon networks and human traits. In ConCon, we consider the cortical band of brain as a non-linear manifold and define its connectivity as smooth functional data describing the relationship between any pair of points on the manifold. ConCon is resolution independent and prevents information loss caused by large ROIs. We develop an efficient functional principal components analysis tailored for ConCons to conduct network embedding and link it to human traits and behaviors. Using real data from the Human Connectome Project, we show that our proposed framework is superior compared with the discrete brain connectivity analysis.
Gaining power and reliability through cortical surface-based spatial Bayesian modeling
Daniel Spencer, Indiana University
David Bolin, KAUST
Yu (Ryan) Yue, Baruch College
Sarah Ryan, University of Pennsylvania
Amanda Mejia, Indiana University
The general linear model (GLM) is a widely popular and convenient tool for estimating the functional brain response and identifying areas of significant activation during a task or stimulus. However, the classical GLM is based on a massive univariate approach that does not explicitly leverage the similarity of activation patterns among neighboring brain locations. As a result, it tends to produce noisy estimates and be underpowered to detect significant activations, particularly in individual subjects and small groups. Our spatial Bayesian GLM leverages spatial dependencies among neighboring cortical or subcortical vertices to produce more accurate estimates and areas of functional activation. The spatial Bayesian GLM can be applied to individual and group-level analysis. In this talk, we first define the spatial Bayesian GLM for both single-subject and group analysis and the advantages of the stochastic partial differential equation prior in fMRI data. A validation of the method is shown using high-quality cortical surface data from the Human Connectome Project. Finally, a computationally efficient EM algorithm is outlined to facilitate analysis. All analyses are performed using the open-source BayesfMRI package in R.
A functional data approach to identifying Alzheimer's disease from cortical surface data
Eardi Lila, University of Washington
Wenbo Zhang, UC Irvine
Swati Rane Levendovszky, University of Washington
We introduce a novel statistical framework for the classification of functional data supported on non-linear, and possibly random, manifold domains. The motivating application is the identification of subjects with Alzheimer's disease from their cortical surface geometry and associated cortical thickness map. The proposed model is based upon a reformulation of the classification problem into a regularized multivariate functional linear regression model. This allows us to adopt a direct approach to the estimation of the most discriminant direction while controlling for its complexity with appropriate differential regularization. We apply the proposed method to a pooled dataset from the Alzheimer's Disease Neuroimaging Initiative and the Parkinson's Progression Markers Initiative, and are able to estimate discriminant directions that capture both cortical geometric and thickness predictive features of Alzheimer's disease, which are consistent with the existing neuroscience literature.