Abstract
Enhanced correlation in BOLD fMRI signals under rest conditions has been found between visual areas with hierarchical (e.g., V1d -> V2d) and homo-topic (V3v -> V3d) relations. Separately, saccades, blinks, and changes in pupil diameter are associated with visual cortex activity. Here we tested if between visual area resting state correlations are explained by shared eye-related signals. We collected 22 minutes of rest-state data in total darkness from 20 subjects at 3T. Infrared eye tracking was performed. Following preprocessing with the HCP pipeline, the median time series was extracted from each of six visual regions of interest per hemisphere (V1–V3, dorsal and ventral). We removed variance related to head motion, respiration, heartbeat, and ventricular and white matter signal. We created covariates that modeled pupil radius, change in eye position, and blinks. These were convolved with a hemodynamic response function. We modeled and removed these signals (and their first derivatives) using linear regression from the cleaned time series. Across visual areas, covariates derived from eye signals accounted for 15.6% of variation in the BOLD fMRI signal (±1.3% SEM by bootstrap), as compared to 10.0% (±0.7%) when the subject pairing between eye and BOLD signals were permuted. Prior to removing eye-related variance from the BOLD fMRI time-series, the correlation between hierarchical and homotopic visual area pairings was greater than that observed for other (background) pairings (z’ of 0.756 vs 0.562), recapitulating prior work. Removal of eye signals from the BOLD data slightly reduced hierarchical and homotopic correlations as compared to background pairings (z’ of 0.722 vs 0.550). Eye signals appeared to explain different variance components in striate and extrastriate regions. BOLD fMRI signals in V1 and extrastriate cortex reflect eye and pupil dynamics. These signals are not entirely responsible for the correlation in spontaneous activity across visual areas.
Acknowledgement: U01EY025864, R01EY024681