| description abstract | Abstract. Magnetic resonance imaging (MRI) is a useful method to noninvasively measure vertebral kinematics (rotations and translations). Measurement of vertebral kinematics should be both fast and accurate, a need potentially satisfied by automatic registration of reference–deformed image pairs. So far, MRI registration has not been systematically optimized for this application. The objective of this study therefore was to apply automatic 3D image registration methods to the measurement of vertebral kinematics from MRI: first, to systematically optimize all registration parameters to minimize registration error across a representative dataset; second, to reanalyze a separate, previously published, MRI dataset of diurnal, flexion, and extension vertebral body (VB) mechanics using the optimized registration to reduce the dataset's measurement error and clarify its interpretation. Validation against manual registrations indicated that midsagittal vertebral body marker position error in the sagittal plane was 0.10±0.08 mm, well below the pixel size of 0.5 mm, with corresponding negligible errors in change in wedge angle (Δ wedge angle), change in disc height (Δ disc height), and A–P translation. Reanalysis of diurnal mechanics data revealed that diurnal Δ wedge angle, with subjects scanned supine, is essentially zero despite significant A–P translation and disc height loss. Distinct kinematics at the L5–S1 disc level were also observed. Relative to manual marker-based methods, use of this image registration method in future work would allow sample size to be halved with no change in statistical power. This optimized registration method will increase the efficiency of future research and may allow detection of effects that would otherwise be overlooked. | |
