(325) fcMRI maps during pain tasks vary based on the inclusion of paradigm modeling in analysis

Abstract

As an alternative to resting state functional connectivity (fcMRI) analysis, block-design fcMRI studies of pain have been increasingly performed. However, the effect of the stimulus-correlated MRI signal changes on fcMRI studies is still unknown. We compared analyses with and without modeling of the stimulation paradigm during a 30 s block-design experiment of electric nerve stimulation (ENS). Our primary hypothesis was that modeling the stimulation paradigm would result in reduced correlation, compared to an un-modeled analysis. fcMRI data was acquired in 14 healthy adults in an IRB-approved 3 T whole brain functional MRI study of ENS-generated pain in the right index finger (intensity self-adjusted to 7/10). A seed region was chosen in the left (contralateral) insula and modeled for each subject as the primary regressor with and without the stimulation paradigm included as an effect of no interest. These were compared to the fcMRI maps of a similar tonic pain task (without cyclic changes). Group average functional correlation maps were generated. Mean and maximum (max) Z-scores were tabulated from regions of interest (ROIs) with significant correlation. The left insula seed time course was found to be correlated to areas of the pain matrix (anterior cingulate cortex, bilateral primary and secondary somatosensory cortices, and right insula), as expected. Modeling the paradigm in this cyclic pain task decreased the strength of correlation in all significant brain areas, giving a map more similar to that of the tonic pain task. Thus, modeling the stimulus timing removed superimposed task-induced activation, giving more accurate fcMRI maps. As an alternative to resting state functional connectivity (fcMRI) analysis, block-design fcMRI studies of pain have been increasingly performed. However, the effect of the stimulus-correlated MRI signal changes on fcMRI studies is still unknown. We compared analyses with and without modeling of the stimulation paradigm during a 30 s block-design experiment of electric nerve stimulation (ENS). Our primary hypothesis was that modeling the stimulation paradigm would result in reduced correlation, compared to an un-modeled analysis. fcMRI data was acquired in 14 healthy adults in an IRB-approved 3 T whole brain functional MRI study of ENS-generated pain in the right index finger (intensity self-adjusted to 7/10). A seed region was chosen in the left (contralateral) insula and modeled for each subject as the primary regressor with and without the stimulation paradigm included as an effect of no interest. These were compared to the fcMRI maps of a similar tonic pain task (without cyclic changes). Group average functional correlation maps were generated. Mean and maximum (max) Z-scores were tabulated from regions of interest (ROIs) with significant correlation. The left insula seed time course was found to be correlated to areas of the pain matrix (anterior cingulate cortex, bilateral primary and secondary somatosensory cortices, and right insula), as expected. Modeling the paradigm in this cyclic pain task decreased the strength of correlation in all significant brain areas, giving a map more similar to that of the tonic pain task. Thus, modeling the stimulus timing removed superimposed task-induced activation, giving more accurate fcMRI maps.