Murat Aksoy1, Christoph Forman1,2, Matus Straka1, Tolga ukur3, Samantha Jane Holdsworth1, Stefan Tor Skare1,4, Juan Manuel Santos3, Joachim Hornegger2, Roland Bammer1
1Department of Radiology, Stanford University, Stanford, CA, United States; 2Computer Science, Friedrich-Alexander-University Erlangen-Nuremberg, Erlangen, Germany; 3Electrical Engineering, Stanford University, Stanford, CA, United States; 4Karolinska Institute, Stockholm, Sweden
Correction of motion artifacts in MRI is essential to assure diagnostic image quality. In case where external pose information is used for motion-correction, cross-calibration errors may impair image quality. In this study, we propose a combined prospective & retrospective approach to prospectively correct for motion and to mitigate residual image distortions which emanate from subtle cross-calibration errors. Specifically, a single camera mounted on the head coil was used to measure and correct patient motion in real-time. Resulting data inconsistencies emanating primarily from cross-calibration errors were removed by a retrospective autofocusing algorithm wherein k-space was divided into segments. The relative rotation and translation needed to realign these segments were determined by means of entropy-based autofocusing. Phantom and in-vivo results show that in the presence of inaccuracies in cross-calibration, the current method provides improved image quality over prospective motion correction only.