磁気共鳴フィンガープリンティングの0.5秒間隔の動き補正

Motion correction in magnetic resonance fingerprinting (MRF) helps preserve the accuracy of quantitative maps; however, existing approaches provide motion updates only every 7-8 seconds. We propose a navigation framework that integrates compact k-space navigators throughout the MRF acquisition, enabling sub-second motion estimation at minimal sequence overhead. A 3D spiral-projection MRF sequence was augmented with three orthogonal spiral navigators inserted every 0.5 seconds, enabling motion estimation by comparing navigator signals with quantitative scout (Q-Scout) data, i.e., motion-free low-resolution k-space with matching contrast evolution. The Q-Scout is obtained via a rapid calibration during the dummy preparation period, incurring no additional scan time. Motion estimation is formulated as dictionary matching in a discriminant subspace with optimization refinement. The method was evaluated in simulation and in vivo for 1 mm isotropic brain 3D MRF at 3 T. Across 35 motion-corrupted acquisitions with motion-free references available, the proposed motion correction reduced MRF reconstruction normalized root-mean-square error (NRMSE) by 7.1% and increased the structural similarity index measure (SSIM) by 0.085. Motion estimates aligned with 8 second temporal-rate image-based navigation (mean absolute difference of 0.15 mm and 0.23 degrees), while the proposed method provided higher temporal resolution and improved motion correction. The proposed framework enables robust motion navigation in MRF at 0.5 second temporal resolution with minimal sequence overhead. By using contrast-consistent modeling and efficient inference, it improves the reliability of quantitative MRI under rapid, unpredictable motion.