Robustness evaluation of Bone Mineral Density measurements in spectral CT using Three Material Decomposition

Abstract

Osteoporosis is a disease that can have lethal consequences, which is why its early detection is crucial. New methods for osteoporosis diagnosis based on bone mineral density (BMD) measurement have been proposed, namely Three Material Decomposition (3MD). In this new study, we evaluate the performance of the 3MD against one of the current clinical standards, Quantitative Computed Tomography (QCT). We studied the regional variability of the 3MD method and determined the optimum region of the L1 vertebra for the automatic placement of a region of interest (ROI). With this goal, we developed an automatic ROI placement algorithm capable of avoiding unwanted tissues, such as, cortical bone, venous plexus and lesions. We evaluated 11 different positionings of circular ROIs, together with two spherical volumes of interest (VOI) placed around different levels of the vertebral height and proposed by Philips Healthcare. The result of this analysis indicated that the placement of a cleaned VOI provides the closest BMD values between 3MD and QCT methods, with a mean difference of 2.25%. However, we recorded large standard deviation values, meaning the precision of the tests must be improved to definitely identify an optimum region for VOI placement within the L1 vertebra. Overall, we believe accounting for the cortical tissue, venous plexus and lesions is critical in order to obtain accurate readings of the BMD value.