PROJECT SUMMARY / ABSTRACT
Osteoporosis (OP) and osteoarthritis (OA) cumulatively affect more than 40 million Americans. Both OP and OA
are underdiagnosed and undertreated because of the limited accuracy of existing tools for diagnosis and
treatment monitoring. The need for improved biomarkers of OP and OA spurred interest in quantitative evaluation
of texture features of cancellous bone derived from radiography, CT, and MRI. In such “bone radiomics”, image
texture provides an indirect assessment of the trabecular geometry (=100 µm detail size) that is better suited to
the limited resolution of diagnostic imaging modalities than the direct measurements used in e.g. micro-CT. Initial
clinical validation of textural bone biomarkers showed promising performance in prediction of vertebral failure
and progression of OA. However, rigorous investigation of how the image formation process affects textural
biomarkers is essential to establish standardized protocols for imaging and analysis in bone radiomics –
especially in light of emerging technologies for high-resolution imaging. Recently, new CT scanners with ~2x
improved spatial resolution compared to conventional CT have been introduced by major manufacturers,
including the Canon Precision system that will be used in this project. This new generation of ultra-high resolution
CT (UHR CT) is capable of visualizing ~150 µm details, approaching the trabecular thickness and thus potentially
enabling a breakthrough in in-vivo evaluation of bone micorarchitecture. We hypothesize that the improved
spatial resolution of UHR CT will lead to better quantitative performance of bone radiomics than normal resolution
CT (NR CT) or x-ray absorptiometry (DXA). To establish the clinical utility of bone radiomics using UHR-CT, the
following Aims will be pursued: 1) Perform the first comprehensive assessment of the sensitivity of CT-based
texture features of bone to key components of the CT imaging chain (e.g., scan and reconstruction protocol)
using a high-fidelity CT simulator and experimental studies in bone core samples. We will establish UHR and
NR CT features that are correlated to trabecular geometry and reproducible with respect to body size and dose.
2) Demonstrate improved prediction of trabecular stiffness using UHR CT texture features. Multivariate
regression between stiffness and texture bone features investigated in Aim 1 will be performed for ~300 bone
cores using UHR CT and NR CT. We will demonstrate improved stiffness estimates with UHR CT compared to
NR CT. 3) Perform a clinical pilot of UHR CT-based texture features in longitudinal monitoring of OP treatment.
We will acquire longitudinal UHR CT and DXA of 20 spine fusion patients being treated with OP drug to optimize
their bone quality. We will demonstrate that radiomic features from UHR CT detect changes in bone quality
earlier than DXA. We will also investigate the feasibility of bone radiomics in prediction of fusion outcomes.
Successful completion of the Aims will establish quantitative UHR CT-based bone radiomics as a novel tool for
in-vivo assessment of bone health in OA and OP, with downstream reduction of patient morbidity and mortality.