Test-Retest Reliability of fMRI Activation Generated by Different Saccade Tasks

Abstract

Purpose: To assess the reproducibility of brain-activation and eye-movement patterns in a saccade paradigm when comparing subjects, tasks, and magnetic resonance (MR) systems. Materials and Methods: Forty-five healthy adults at two different sites (n = 45) performed saccade tasks with varying levels of target predictability: predictable (PRED), position predictable (pPRED), time predictable (tPRED), and prosaccade (SAC). Eye-movement pattern was tested with a repeated-measures analysis of variance. Activation maps reproducibility were estimated with the cluster overlap Jaccard index and signal variance coefficient of determination for within-subjects test-retest data, and for between-subjects data from the same and different sites. Results: In all groups latencies increased with decreasing target predictability: PRED < pPRED < tPRED < SAC (P < 0,001). Activation overlap was good to fair (>0.40) in all tasks in the within-subjects test-retest comparisons and poor (<0.40) in the tPRED for different subjects. The overlap of the different tasks for within-groups data was higher (0.40-0.68) than for the between-groups data (0.30-0.50). Activation consistency was 60-85% in the same subjects, 50-79% in different subjects, and 50-80% in different sites. In SAC, the activation found in the same and in different subjects was more consistent than in other tasks (50-80%). Conclusion: The predictive saccade tasks produced evidence for brain-activation and eye-movement reproducibility.