Presented by Heather Shappell
Assistant Professor
Department of Biostatistics & Data Science
Wake Forest University
The human brain is a complex network whose patterns of communication change continuously over time. Rather than remaining in a single, fixed configuration, functional connections among brain regions appear to organize into recurring network states that the brain transitions between throughout a scan. Understanding these dynamics may provide insight that is missed by traditional approaches that average connectivity over time. This is especially compelling in the study of eating behavior and weight regulation, where moment-to-moment changes in networks involved in reward, salience, cognitive control, and motivation may help explain why individuals differ in their responses to food cues and in their ability to lose or maintain weight.
Hidden semi-Markov models (HSMMs) provide a natural framework for studying these dynamic brain networks. In this framework, observed neuroimaging data are modeled as arising from a sequence of latent brain states, while the duration of time spent in each state is modeled explicitly. This allows us to characterize not only which network configurations occur, but also how frequently they occur, how individuals transition between them, and how long the brain remains in each state. I will introduce the HSMM framework for dynamic functional connectivity analysis and discuss how it can be used to identify and characterize recurring patterns of whole-brain network organization.
I will then present a recent extension that allows state-specific dwell times to depend on subject-level covariates. We introduce a regression structure for the dwell-time distributions, considering both Poisson and Gamma models. Parameters are estimated using an expectation–maximization algorithm, and participant-level bootstrap resampling is used to obtain standard errors and carry out inference on the regression coefficients.
Finally, I will illustrate the framework using functional MRI data collected during a food-cue task in older adults participating in a weight-loss intervention. We examine how Power of Food Scale scores, which reflect psychological responsiveness to the food environment, relate to the amount of time individuals spend in distinct brain network states. Together, these methods provide a flexible approach for studying how the temporal organization of brain networks varies across individuals and how those differences may relate to meaningful behavioral and clinical outcomes.
A seminar tea will be held at 2:45 p.m. in University Office Plaza, Room 240. All are Welcome.


