Inertial Navigation Assisted Motion Correction for Brain MRI

To assess the utility of inertial sensors for estimating parameters describing rigid body motion of the head, and to reduce motion related artifact through a hybrid prospective steering of the rotation via the sensor and retrospective realignment of translation via the MRI data.

Inertial sensors are widely used for Earth navigation problems. However, drift accumulation from integrating noisy signals makes them impractical for MRI motion correction at millimeter resolution over minute-long scans. Prior work has shown that the MRI system's $B_0$ and gravity are approximately orthogonal and can be exploited as reference vectors for 3-DOF orientation estimation. We introduce a hybrid inertial measurement-based method, which uses these fields to estimate 3-DOF orientation directly, without integration, thereby eliminating random walk. The remaining 3-DOF translation is recovered via phase correlation from the MRI data. We validate the efficacy of this hybrid method retrospectively using a 3D radial koosh-ball sequence and prospectively using 2D EPI time-series acquisition during volunteer motions.

IMU-based orientation estimation followed by phase correlation achieved a mean accuracy of 0.6° and 0.4 pixels across all experiments. Image quality improved when motion correction was applied in all volunteer scans for both retrospective and prospective correction cases.

The proposed method measured head motion in the MRI scanner with high accuracy at unprecedented sampling rates and enabled both retrospective and prospective correction, improving image quality by aligning k-space data and reducing motion artifacts.