Quantitative musculoskeletal magnetic resonance imaging at 3.0T: Developing new techniques to detect morphological changes in human cartilage and bone in osteoarthritis and osteoporosis.

摘要

This thesis concentrates on the ongoing work in musculoskeletal imaging using MRI. The research focuses on quantitative musculoskeletal MRI at 3.0 Tesla and uses imaging features derived from deformation based morphometry for better discrimination between normal and abnormal, and gender-based population studies. The goal is to develop a completely automated voxel based technique, which can be applied to the knee joint for the analysis of the femur and tibia bones and the femoral cartilage. An atlas is created using well documented and tested, 12 body rigid affine and elastic registration techniques with high accuracy and reproducibility. Using the atlas as a reference, a voxel-based comparison is performed between populations to study shape changes. The jacobian values calculated from the deformation fields obtained at the end of the registration, provides point-by-point information of the amount of enlargement and shrinkage between subjects, when compared to the common reference frame. Calculating the effective size, as well as doing voxel based statistical analysis, determines the difference in cartilage between populations. The ability to automatically characterize cartilage in the knee joint, as diseased or non-diseased would greatly aid the early detection, diagnosis and prognosis of Osteoarthritis. Similar analysis for the femur and tibia bones provides insight on the variations existing within a well-defined population. This information aids in automated characterization of the localized shape differences and size differences of the bone and cartilage in subject cohorts with different disease conditions. Another aspect of this research is analyzing the integrity of trabecular bone in Osteoporosis. The possibility of using texture features from low-resolution images as a surrogate to structural indices measured from high-resolution images is explored. The analysis of the structural properties of trabecular bone of the wrist from high resolution MRI, show that these indices correlate strongly with texture features calculated from low resolution images. Hence there is a possibility of using these texture features to characterize the bone as normal or osteoporotic.