Search Website
QUICK LINKS:

Dr. Michelle Miranda's Talk
Location: Kresge Building 106 (K106)
Speaker: Dr. Michelle Miranda - Department of Mathematics and Statistics, University of Victoria
Title: A Bayesian Framework for Estimating Long Memory in Resting-State fMRI
Abstract: Resting-state functional MRI (fMRI) captures spontaneous brain activity in the absence of tasks and is widely used to study intrinsic brain function. However, commonly used autoregressive or independence models in neuroscience fail to capture the long-range temporal dependence observed in fMRI time series. In this talk, I introduce a statistical pipeline that models these dependencies within a long-memory (LM) framework, where autocorrelations decay according to a power law. LM parameters are estimated voxelwise using a wavelet-based Bayesian method, which efficiently handles the multi-scale nature of long-range dependence and provides full uncertainty quantification. We use a composite basis to obtain a lower-dimensional representation of the maps, which are then linked to subject-level covariates through regression. Whereas functional connectivity emphasizes inter-regional relationships, the LM parameter reflects persistence of temporal dynamics within regions, offering complementary insights. Applying this framework to the ADHD-200 dataset, we found that age was positively associated with the LM parameter in the hippocampus, after adjusting for ADHD symptoms and medication status. This work highlights the potential of long-memory modeling for studying developmental changes in resting-state brain activity and provides a flexible framework for large-scale neuroimaging analysis.