We have developed a method for the noninvasive measurement of three-dimensional patellar kinematics in loaded flexion using magnetic resonance imaging (MRI). Knee models obtained by segmenting and reconstructing one high-resolution MRI scan of the knee are registered to bone outlines obtained by segmenting fast, low-resolution MRI scans of the knee in loaded flexion.

Accuracy of our MRI-based technique was assessed by comparing patellar tracking measurements made using the new method to measurements made using Roentgen Stereophotogrammetric Analysis (RSA) in three cadaver knee specimens loaded through a range of flexion in a test rig. The error in patellar spin and tilt measurements was less than 1.02° and the error in lateral patellar shift was 0.88 mm.

Intra-subject variability was measured as the variation in the patellar tracking patterns of 3 subjects across four different loading cycles, using the same high-resolution MRI scan to derive the 3D surface models for all four cycles. The intra-subject variability in patellar orientation measurements was less than 1.5° and the intra-subject variability in patellar translation measurements was less than 1 mm.

Inter-experimenter repeatability of our technique was measured as the variation in the patellar tracking patterns resulting from the same set of images (i.e. one high-resolution scan and one loading cycle of low-resolution scans) independently processed and analyzed by three experimenters. The inter-examiner repeatability was less than 1.3° for measurements of patellar flexion and tilt, and less than 0.7 mm for patellar position.

The method is particularly useful for multiple measurements on subjects because coordinate axes must only be defined once, and has the additional advantage that no hardware or software modifications to clinical MRI scanners are required to implement it.