Effects of EPI readout bandwidth on measured activation map and BOLD response in fMRI experiments

The purpose of this study was to evaluate the effects of echo planar imaging (EPI) readout bandwidth and its interaction with data processing procedures on the measured blood oxygenation level dependent (BOLD) response and activation in fMRI experiments. Seventeen healthy subjects were scanned during a brief visual stimulation paradigm with two EPI pulse sequences having 'high' (1953 Hz/pixel) and 'low' (780 Hz/pixel) readout bandwidth. Functional data were analyzed with a general linear model including temporal filtering and a basic correlation model following (1) no preprocessing, (2) realignment, or (3) realignment and spatial smoothing. A range of statistical thresholds were used to generate activation maps. Despite slightly higher BOLD signal detected with the high bandwidth sequence from matched ROIs in the primary visual cortex, results showed that the low bandwidth pulse sequence was more sensitive under most conditions evaluated. That is, the low bandwidth sequence detected greater numbers of activated voxels with lower cluster average BOLD signal (e.g., low bandwidth detected 1.4 times more voxels, with average BOLD signal 30% lower compared to high bandwidth for P = 0.05 (corrected) with the 3rd preprocessing procedure using the general linear model). However, there was significant interaction between bandwidth and data preprocessing procedures. Of particular interest, the sensitivity advantage of the low bandwidth pulse sequence decreased for the smoothed data as the activation threshold became less conservative. For the frequently used threshold of P = 0.001 (uncorrected) and cluster size of at least 5 voxels, the bandwidth advantage became insignificant. These findings demonstrate that the effects of bandwidth should be considered carefully in the design, analysis, and interpretation of BOLD fMRI studies.