Hip fracture is the leading cause of acute orthopaedic hospital admission amongst the elderly, with around a third of patients not surviving one-year post-fracture. Current risk assessment tools ignore cortical bone thinning, a focal structural defect characterizing hip fragility. Cortical thickness can be measured using computed tomography, but this is expensive and involves a significant radiation dose. Dual-energy X-ray absorptiometry (DXA) is the preferred imaging modality for assessing fracture risk, and is used routinely in clinical practice. This thesis proposes two novel methods which measure the cortical thickness of the proximal femur from multi-view DXA scans. First, a data-driven algorithm is designed, implemented and evaluated. It relies on a femoral B-spline template which can be deformed to fit an individual?s scans. In a series of experiments on the trochanteric regions of 120 proximal femurs, the algorithm?s performance limits were established using twenty views in the range 0? ? 171?: estimation errors were 0.00 ? 0.50 mm. In a clinically viable protocol using four views in the range ?20? to 40?, measurement errors were ?0.05 ? 0.54 mm. The second algorithm accomplishes the same task by deforming statistical shape and thickness models, both trained using Principal Component Analysis (PCA). Three training cohorts are used to investigate (a) the estimation efficacy as a function of the diversity in the training set and (b) the possibility of improving performance by building tailored models for different populations. In a series of cross-validation experiments involving 120 femurs, minimum estimation errors were 0.00 ? 0.59 mm and ?0.01 ? 0.61 mm for the twenty- and four-view experiments respectively, when fitting the tailored models. Statistical significance tests reveal that the template algorithm is more precise than the statistical, and that both are superior to a blind estimator which naively assumes the population mean, but only in regions of thicker cortex. It is concluded that cortical thickness measured from DXA is unlikely to assist fracture prediction in the femoral neck and trochanters, but might have applicability in the sub-trochanteric region.
Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:675907 |
Date | January 2015 |
Creators | Tsaousis, Nikolaos |
Publisher | University of Cambridge |
Source Sets | Ethos UK |
Detected Language | English |
Type | Electronic Thesis or Dissertation |
Source | https://www.repository.cam.ac.uk/handle/1810/252878 |
Page generated in 0.0014 seconds