Introduction Biomechanical modelling of the human body requires measurement of the relative positions of skeletal elements. This information provides data on joint kinematics directly affects muscle attachment site locations and hence determines muscle moment calculations. Spinal orientation is particularly difficult to measure due to small joint movements and relative inaccessibility of the bones to direct measurement. This study presents a novel method of accurately determining relative bone position in vivo using magnetic resonance imaging (MRI).
Methods A process incorporating both positional and conventional MRI was used to determine the skeletal positions of the lumbar spine and pelvis. The method uses higher quality conventional MRI to determine bone geometries and then registers these with lower resolution, positional MRI images of various postures.
Using the positional scanner four postures were investigated: Neutral Standing, Neutral Sitting, Flexed Sitting and Extended Sitting. These scans comprised simultaneous sagittal and coronal non-contiguous slices to facilitate three-dimensional registration and reduce acquisition time. Conventional MRI was then used to scan the subject at higher resolution contiguous slices. After segmentation and surface extraction of all bones from all scans, each bone geometry was registered with each of the positional scans to produce high quality in vivo skeletal position data.
For 2 subjects, each of the 5 lumbar vertebrae and the pelvis were registered 5 times in the 4 postures to investigate intra-tester reliability. This resulted in 48 sets of 5 registrations. Each bone surface was represented by surface points and a local coordinate system. Angular and translational differences between coordinate axes were examined for each set of five registrations.
Results The results indicate good intra-measurer reliability with a maximum rotational difference for all vertebral registrations of less than 1 degree and a maximum origin translation of less than 3mm. The pelvic registrations demonstrated larger discrepancies. Flexion/extension, lateral bend and axial twist rotations were measured for each joint. While there did not appear to be patterns between the two subjects, there were obvious trends within each subject and in particular trends of lateral bending throughout sagittal plane motion were identified.
Discussion The results showed that the technique was able to register the surfaces reliably. The intervertebral movements between postures were within normal ranges of motion and demonstrated kinematic trends within an individual. At present, the greatest disadvantage of the method described lies in its large data processing times. The data collected are three dimensional and represent the anatomy and movement of a specific individual. The method can be used to examine joint mechanics and centres of rotation in three dimensions, validate the predictions of finite element models and investigate the effects of medical interventions.
