Characterizing Motion-Related Contamination of Group Independent Component Analysis Networks in Resting State fMRI

Motivation: Resting-state network quality is a trade-off involving the amount of data used (data quantity) to form the network and confounds such as motion, yet these trade-offs remain largely unexplored. Goal(s): Establish the relationship between motion and group network spatial specificity across varying data quantities. Approach: We analyzed group resting-state networks with differing levels of motion contamination, modelled the relationship and identified data quantities required for specific spatial specificity levels. Results: The relationship between quantity of data and motion level changes based on the desired spatial specificity. We present a methodology to produce calibration curves that reveal this relationship. Impact: This contribution informs researchers who want to perform power calculations or group-level analyses following dual regression of network components with a methodology to approximate the quantity of data necessary to achieve quality results given an estimate of the motion levels.