Detection of eye movements from fMRI data.

Abstract

Awake humans make eye movements with amplitudes and frequencies that depend on behavioral state and task. This poses two problems for functional magnetic resonance imaging (fMRI) studies that compare brain activity across tasks. First, motion of the eye in the orbit increases the variance of the MR signal in adjacent regions of the orbitofrontal cortex, hampering activation detection. Second, eye movements are associated with activity in a distributed network of brain areas, confounding comparisons of task activation. Direct measurement of eye movements in the scanner bore is possible with expensive and technically demanding equipment. A method is described that detects eye movements directly from MR data without the use of additional equipment. Changes in the MR time series from the vitreous of the eye were observed that correlated with eye movements, as measured directly with an infrared pupil tracking system. In each of 10 subjects, the variance of the MR time series from the eye vitreous was greater when the subject made eye movements than when the subject fixated centrally (average standard deviation (SD) 99.7 vs. 75.6, P = 0.001). The assessment of eye movements directly from fMRI data may be especially useful for retrospective and meta-analyses.