Impact of motion correction and temporal domain preprocessing strategies on test-retest reliability of resting-state functional connectivity of language and sensorimotor networks.
Abstract
Background
Temporal domain choices and motion correction strategies are critical for rs-fMRI analysis. NEW
Method
This study compared the combined effect of temporal domain and motion correction choices of two widely-used and methodologically distinct preprocessing strategies implemented within FSL and CONN pipelines on test-retest reliability of the language and sensorimotor networks. METHODOLOGY 27 healthy volunteers underwent two rs-fMRI sessions. Data were analyzed using seed-based analysis for seven regions of interest within the language and sensorimotor networks. All preprocessing parameters were unified between FSL and CONN, except for their default temporal domain preprocessing and motion correction strategies to enable direct comparison. Test-retest reliability was primarily assessed using Intraclass Correlation Coefficient (ICC 3,1), with secondary analyses of group-level intersession FC variations.
Results
The FSL pipeline yielded significantly higher voxel-level reliability (p = 0.031), particularly for the sensorimotor network. At the ROI level, FC map reliability was comparable between pipelines. Group-level analyses revealed that FSL captured changes in basal ganglia connectivity patterns not detected by CONN. Furthermore, inter-network FC stability varied markedly between both pipelines. COMPARISON WITH OTHER
Methods
This method evaluates the impact of complete end to end rs-fMRI preprocessing strategies as implemented in clinical practice instead of segregating their impact in previous research work.
Conclusion
Preprocessing strategy has a network-dependent impact on the test-retest reliability of the rs-fMRI functional connectivity of the language and sensorimotor networks. The language network benefits from CONN's aggressive artifact removal. Conversely, the sensorimotor network, specifically subcortical components, requires the preservation of higher-frequency signals, favoring FSL approach.