Cortical Bone Thickness Estimation in CT Images: A Model-Based Approach Without Profile Fitting

Abstract

Structure of cortical bone is decisive for its strength, and quantification of the structure is crucial for early diagnosis of osteoporosis and monitoring of therapy effect. In three-dimensional computed tomography (CT) images, typically cortical thickness in proximal femur, lumbar vertebrae, and sometimes in distal forearm is estimated. However, resolution of clinical quantitative CT (QCT) scanners is comparable to the cortical thickness, especially for osteoporotic patients, leading to significant partial volume artefacts. A recent model-based approach recovers the cortical bone thickness by numerically deconvolving the image (profile fitting) using an estimated scanner point spread function (PSF) and a hypothesized uniform cortical bone mineralization level (reference density). In this work we provide an essentially analytical unique solution to the model-based cortex recovery problem using few characteristics of the measured profile and thus eliminate the non-linear optimization step for deconvolution. The proposed approach allowed to get rid of the PSF in the model and reduce the sensitivity to errors in the reference density value. Also, run-time and memory effective implementation of the proposed method can be done with the help of a lookup table. The method was compared to an existing approach and to the 50% relative threshold technique by evaluating performance of these three algorithms in a simulated environment with noise and various error levels in the reference density parameter. Finally, accuracy of the proposed algorithm was validated using CT acquisitions of European Forearm Phantom II, a widely used anthropomorphic standard of cortical and trabecular bone compartments that was scanned with various protocols.