Test-retest reliability of edge-level resting-state functional connectivity in people with aphasia

There is a growing interest in using resting-state functional connectivity (RSFC) to investigate language processing and recovery in post-stroke aphasia due to its limited dependence on an individual's ability to follow directions and perform tasks, or the severity of their aphasia. However, the test-retest reliability of RSFC in people with aphasia has not been established, raising questions about the strength and validity of inferences based on this technique. In this study, we examined the reliability of RSFC at the level of individual edges (i.e., connections) in 14 adults with chronic aphasia due to left-hemisphere stroke. Intraclass correlations (ICCs) between two resting-state scans obtained over a few days were computed for every edge in a whole-brain network and several cognitive and language subnetworks. Based on median ICCs, reliability was fair at longer scan durations (10-12 min) and better in most subnetworks than the whole brain. Reliability was also positively associated with connectivity strength and had a weak negative relationship with inter-node distance (i.e., the distance between the regions that form an edge). Edges in the right hemisphere were more reliable than those in the left hemisphere and between hemispheres, though all three sets of edges were fairly reliable. The results indicate that edge-level RSFC is acceptably reliable for continued use in aphasia research but highlight the need for strategies to ensure that inferences are based on valid results, such as using sufficiently long scans and focusing analyses on established subnetworks, especially in longitudinal contexts.