Abstract:
Statistical methods play a crucial role in brain imaging, enablingresearchers to uncover the complex patterns of brain function andconnectivity.In this talk, we will begin by highlighting the critical rolethat statistical approaches play in the analysis of imaging data,particularly in the context of functional magnetic resonance imaging(fMRl). We will discuss how appropriate statistical methods arenecessary to handle the complexity of spatial and temporalcorrelations typical of brain data. Building on this foundation, we wilthen discuss approaches to studying dynamic brain connectivity, whichseeks to understand the changing interactions between different brainregions over time. We will present two novel Bayesian approachesdesigned to capture these dynamic relationships within multivariatetime series data. First, we will present a scalable Bayesian timevarying tensor vector autoregressive(TVVAR)model, aimed atefficiently capturing evolving connectivity patterns. This modelleverages a tensor decomposition of the VAR coefficient matrices atdifferent lags and sparsity-inducing priors to capture dynamicconnectivity patterns. Next, we will introduce a Bayesian framework forsparse Gaussian graphical modeling,which employs discreteautoregressive switching processes. This method improves theestimation of dynamic connectivity by modeling state-specific precisionmatrices, using innovative prior structures to account for temporal andspatial dependencies. Throughout the talk, we will illustrate the powerand flexibility of these Bayesian methods with examples fromsimulation studies and real-world fMRl data. Our discussion willemphasize the importance of these innovative statistical tools inadvancing our understanding of brain connectivity and their potentialfor applications in neuroscience research and clinical practice.
About the Speaker:
Michele Guindani, Ph.D.,is a Professor in the Department ofBiostatistics at the Fielding School of Public Health, University ofCalifornia, Los Angeles (UCLA). Professor Guindani's research spansBiostatistics, Data Science, Machine Learning, Statistical decision-making under Uncertainty, Multiple comparison problems, Statisticallmaging, Clinical Trials, Study Design, Clustering, Bayesian modeling,and Nonparametric Bayesian models. Professor Guindani is Fellow ofthe American Statistical Association (ASA)and the InternationaSociety for Bayesian Analysis (lSBA). He is the current President ofthe International Society for Bayesian Analysis (ISBA).

Your participation is warmly welcomed!

欢迎扫码关注北大统计科学中心公众号,了解更多讲座信息!