Bone quality influences bone strength with important consequences for osteoporosis, fracture risk and dental implant success. Whilst imaging that is capable of capturing bone structure in 3D is becoming more common, quantitative clinical measures of bone quality rely on bone quantity, not structure. If bone quality could be more accurately measured, and the influence of bone architecture better understood, strength may be better predicted.This thesis presents methods for making structural comparisons between successive micro-CT images of loaded bone and explores the limitations of these. I present a novel method to detect where damage occurs in loaded rat vertebrae based on multiscale rigid registration and difference measures. Together these methods represent a quantitative approach to image guided failure analysis.Time-lapsed micro-CT images of 14 successively loaded rat vertebrae were acquired and damaged regions found using these. Using a random forest classifier I tested whether the damaged regions could be predicted by several commonly used structural measures (bone area and volume), three-dimensional texture measures (co-occurrence matrices and fractal dimension) and a more novel type of architectural measure (based on the structure tensor). A combination of parameters was able to predict damage regions with specificities in the range 70-90% and sensitivities of 60-70%.Using ovariectomised rats as a model of osteoporosis I have performed a pilot experiment to investigate how changes in bone quality might effect our results. The wider applicability of my methods are demonstrated by applying them to dental cone beam images of healthy and osteoporotic patients.
|Date of Award
|31 Dec 2012
- The University of Manchester
|James Graham (Supervisor) & Hugh Devlin (Supervisor)
- Trabecular Architecture Measures
- Bone Quality
- Image Guided Failure Analysis