Poster E121, Monday, March 26, 2:30-4:30 pm, Exhibit Hall C
Data-driven subgrouping of task-based and resting state fMRI timeseries
Jonathan T. Parsons1, Kathleen M. Gates1, Joseph B. Hopfinger1; 1University of North Carolina at Chapel Hill
Analysis of network activity has become an increasingly important method for interpreting fMRI data. Group Iterative Multiple Model Estimation (GIMME) is a method for detecting both the presence and direction of functional connections between brain regions (Gates et al., 2012). Previously, GIMME has been used to define subgroups of individuals, based on patterns of functional connectivity, that map onto groupings based on behavioral phenotype. Here, we investigate the robustness of this algorithm to recover differences in patterns of functional connectivity in a well-controlled experimental paradigm. Healthy young adults performed a series of separate tasks intended to stimulate processing in visual, motor, and auditory regions, in addition to a separate resting state scan. Time-series data from regions of interest in primary sensory and motor areas, as well as the default mode network and frontal-parietal control networks, were extracted. GIMME analyses revealed a robust ability to recover subgroups that mapped on to the tasks being performed. We discuss why these subgroupings corresponded more closely to task in some cases than others, including a comparison of task-based paradigms that followed the typical block design versus a “continuous task” paradigm that mimics typical resting state designs.
Topic Area: METHODS: Neuroimaging