k-Space based summary motion detection for functional magnetic resonance imaging

Abstract

Functional MRI studies are very sensitive to motion; head movements of as little as 1-mm translations or 1 degrees rotations may cause spurious signals. An algorithm was developed that uses k-space MRI data to monitor subject motion during functional MRI time series. A k-space weighted average of squared difference between the initial scan and subsequent scans is calculated, which summarizes subject motion in a single quality parameter; however, the quality parameter cannot be used for motion correction. The evolution of this quality parameter throughout a time series indicates whether head motion is within a predetermined limit. Fifty functional MRI studies were used to calibrate the sensitivity of the algorithm, using the six rigid-body registration parameters (three translations and three rotations) from the statistical parametric mapping (SPM99) package as a reference. The average correlation coefficient between the new quality parameter and the reference value from SPM was 0.84. The simple algorithm correctly classified acceptable or excessive motion with 90% accuracy, with the remaining 10% being borderline cases. This method makes it possible to evaluate brain motion within seconds after a scan and to decide whether a study needs to be repeated.