Evaluating Acquisition Time of rfMRI in the Human Connectome Project for Early Psychosis. How Much Is Enough?

Abstract

Resting-state functional MRI (rfMRI) correlates activity across brain regions to identify functional connectivity networks. The Human Connectome Project (HCP) for Early Psychosis has adopted the protocol of the HCP Lifespan Project, which collects 20 min of rfMRI data. However, because it is difficult for psychotic patients to remain in the scanner for long durations, we investigate here the reliability of collecting less than 20 min of rfMRI data. Varying durations of data were taken from the full datasets of 11 subjects. Correlation matrices derived from varying amounts of data were compared using the Bhattacharyya distance, and the reliability of functional network ranks was assessed using the Friedman test. We found that correlation matrix reliability improves steeply with longer windows of data up to 11–12 min, and ≥14 min of data produces correlation matrices within the variability of those produced by 18 min of data. The reliability of network connectivity rank increases with increasing durations of data, and qualitatively similar connectivity ranks for ≥10 min of data indicates that 10 min of data can still capture robust information about network connectivities.