Abstract
PURPOSE: With growing interest in modeling neurobehavior, there is increased interest in understanding patterns of functional connectivity (FC) during language processing. Previous research has suggested that static resting-state functional magnetic resonance imaging (rs-fMRI) and task-based functional magnetic resonance imaging (tb-fMRI) may be interchangeable in determining FC in language-related regions of interest. Authors have argued for the elimination of using tb-fMRI assessments in preoperative clinical workup of language mapping. However, given that language exhibits not only 3D spatial attributes but also temporal components, understanding the temporal dynamics is essential in developing adaptive computational models. Thus, the stability of language neural networks during rs-fMRI and tb-fMRI during auditory comprehension was examined in healthy participants. METHOD: Twenty-three participants underwent rs-fMRI and 12 participants underwent tb-fMRI while listening to an auditory description task. Sliding scale time series correlation was used to generate estimates of dynamic FC. We used Spearman correlation with Bonferroni correction to compute statistically significant whole-brain dynamic networks. A total of 59,831 data points (42,159 in rs-fMRI, 10,152 in tb-fMRI, and 7,520 in overlap) were analyzed. RESULTS: There was more notable fluctuation in linear correlation over time for dynamic FC networks activated during the task relative to baseline than during rs-fMRI. Differences in dynamic FC were noted in only specific common networks in the left hemisphere during tb-fMRI. CONCLUSIONS: The temporal dynamics of identical neural networks during rs-fMRI and tb-fMRI are markedly different, especially within the left hemisphere. This may be an important computational model feature that may improve model prediction of clinical outcomes following central nervous system injury.