Bone microarchitecture is subject to a continuous remodeling process that reflects the skeletal sequelae of mechanical loading, aging and chronic diseases. High resolution peripheral quantitative computed tomography (HR-pQCT) allows the non-invasive assessment of skeletal dynamics over time. However, current options for longitudinal image analysis do not include measures of structural directionality and information on the local distribution of microarchitectural changes. Optical flow is a comparative image-analysis tool primarily utilized for the detection of minimal motion in automated video surveillance. This technique is characterized by a particular strength in capturing minimal gradient changes between two images, and therefore bears the potential of being applicable to longitudinal HR-pQCT imaging and the quantification of subtle changes in bone microarchitecture.
The aims of this work are to establish a joined reference space to be able to conduct cross sectional studies, to demonstrate the feasibility of optical flow and structure tensor analysis in longitudinal HR-pQCT imaging and to find quantitative measures of changes in trabecular microarchitecture. In order to compare local microarchitecture and to establish correspondences across patients, all imaging data has to be registered to a unified reference coordinate system. This unified reference space allows voxel to voxel comparison of structural as well as remodeling information throughout a whole patient population. Optical flow is utilized to capture trabecular volume change between two successive HR-pQCT acquisitions within the reference space. Trabecular direction is represented by structure tensors. Quantitative measures for direction dependent trabecular reorganization are derived by linking trabecular volume change and direction. These combined measures are introduced as spatio-temporal morphometry (STM).
The conducted experiments evaluate the stability and feasibility of the proposed STM measures as well as their applicability on a clinical patient cohort. Extensive parameter tuning is performed to optimize the registration steps needed to establish the unified reference space. Measure stability evaluation is done on cadaver data with and without repositioning in between scans and on healthy control patients. To research STM in a clinical setting, HRpQCT scans of patients shortly after lung transplantation (LuTX) and subsequent scans within 1-2 years are evaluated. Furthermore the HR-pQCT data of the LuTX patients at baseline is investigated to characterize bone microarchitecture and bone strength (micro finite element analysis) shortly after lung transplantation.