18. Statistical Approaches to Addressing Challenges in Neuroimaging Research
The video recording of this talk can be viewed here.
Organizer: Dayu Sun
Department of Biostatistics and Bioinformatics Emory University
Email: dayu.sun@emory.edu
Chair: Dayu Sun
Department of Biostatistics and Bioinformatics Emory University
Email: dayu.sun@emory.edu
Speakers:
1. Lexin Li
University of California, Berkeley
Email: lexinli@berkeley.edu
Title: Testing Mediation Effects Using Logic of Boolean Matrices with Applications in Neuroimaging Mediation Analysis
Time: 11:45am-12:05pm
Abstract:
A central question in high-dimensional mediation analysis is to infer the significance of individual mediators. The main challenge is that the total number of potential paths that go through any mediator is super-exponential in the number of mediators. Most existing mediation inference solutions either explicitly impose that the mediators are conditionally independent given the exposure, or ignore any potential directed paths among the mediators. In this talk, we present a new hypothesis testing procedure to evaluate individual mediation effects, while taking into account potential interactions among the mediators. Our key idea is to construct the test statistic using the logic of Boolean matrices, which enables us to establish the proper limiting distribution under the null hypothesis. We further employ screening, data splitting, and decorrelated estimation to reduce the bias and increase the power of the test. We show that our test can control both the size and false discovery rate asymptotically, and the power of the test approaches one, while allowing the number of mediators to diverge to infinity with the sample size. We illustrate our method with two applications in neuroimaging-based mediation analysis for Alzheimer's disease.
2. Sean L. Simpson
Wake Forest School of Medicine
Email: slsimpso@wakehealth.edu
Title: Mixed modeling frameworks for analyzing whole-brain network data
Time: 12:05pm-12:25pm
Abstract:
Brain network analyses have exploded in recent years, and hold great potential in helping us understand normal and abnormal brain function. Network science approaches have facilitated these analyses and our understanding of how the brain is structurally and functionally organized. However, the development of statistical methods that allow relating this organization to health outcomes has lagged behind. We have attempted to address this need by developing mixed-modeling frameworks that allow relating system-level properties of brain networks to outcomes of interest. These frameworks serve as a synergistic fusion of multivariate statistical approaches with network science, providing a needed analytic foundation for whole-brain network data. Here we delineate these approaches that have been developed for single-task, multitask, and dynamic brain network data.
3. Xin "Henry" Zhang
Florida State University
Email: henry@stat.fsu.edu
Title: Generalizing liquid association for multimodal neuroimaging
Time: 12:25pm-12:45pm
Abstract:
Alzheimer’s disease (AD) is the leading form of dementia, and the number of affected people is drastically increasing along with aging of the worldwide population. A key question of AD research is to understand the spatial associative patterns between two pathological proteins, amyloid-beta and tau, as the subject’s age varies. The problem can be formulated as studying the associations of two sets of random variables conditional on the third set of random variables, a topic that has received relatively little attention in the statistics literature, but is crucial for multimodal neuroimaging analysis in general. In this article, motivated by a multimodal positron emission tomography (PET) study for AD, we extend the notion of liquid association of K.C. Li (2002) from the univariate setting to the multivariate and high-dimensional setting. We propose a novel generalized liquid association analysis approach, which offers a new and unique angle to study associations among three sets of random variables. We establish a population dimension reduction model, transform the problem to sparse Tucker decomposition of a three-way tensor, and develop a higher-order singular value decomposition estimation algorithm. We derive the non-asymptotic error bound and asymptotic consistency of the proposed estimator, while allowing the variable dimensions to be larger than and diverge with the sample size. We analyze the motivating multimodal PET dataset, and identify important brain regions that exhibit the most contrastive associations as age varies. We further complement the data analysis with additional simulations to demonstrate the efficacy of the proposed method.