Combining universal access with high precision for revolutionary osteoporosis screening: CT-based 3D-HU volumetric bone density.
OBJECTIVE This study aimed to develop a novel 3D Hounsfield unit (HU) technique for assessing volumetric bone mineral density (vBMD) using routine preoperative CT scans. It was designed to enable opportunistic osteoporosis screening by leveraging existing imaging data. The diagnostic accuracy of this approach was rigorously evaluated against quantitative CT (QCT). METHODS The authors retrospectively analyzed the records of spine surgery patients who underwent both spinal CT and QCT within a 1-week period. Patients were divided into 3 groups (osteoporosis, osteopenia, and normal bone mineral density [BMD]) based on QCT-vBMD. Baseline characteristics and radiological parameters were compared among the 3 groups. Three-dimensional HU measurements were obtained from lumbar CT scans. Multivariable logistic regression, Pearson correlation analysis, and receiver operating characteristic (ROC) analysis were used to assess the relationships between vBMD and 3D-HU. To evaluate interscanner reliability, a subset of patients received additional CT scanning using both GE and Philips systems. Bland-Altman analysis was used to assess agreement between 3D-HU measurements from different scanner platforms. RESULTS Among 264 enrolled patients, 94 were classified as osteoporotic, 114 as osteopenic, and 56 as having normal BMD. Multivariable logistic regression analyses identified lower 3D-HU as an independent predictor of both osteoporosis (OR 0.723, 95% CI 0.618-0.806; p < 0.0001) and osteopenia (OR 0.698, 95% CI 0.564-0.799; p < 0.0001). Three-dimensional HU demonstrated a strong correlation with QCT-vBMD (T12, r = 0.979; L1, r = 0.968; L2, r = 0.962). ROC analysis demonstrated that 3D-HU achieved diagnostic accuracy comparable to QCT for both osteoporosis and osteopenia. Specifically, for osteoporosis, the area under the ROC curve (AUC) was 99.1% (sensitivity 94.7%, specificity 98.3%); for osteopenia, the AUC was 98.5% (sensitivity 93.9%, specificity 94.6%). Bland-Altman analysis demonstrated excellent interscanner agreement (mean difference 0.13; 95% limits of agreement -8.23 to 8.50), with only 2% of measurements falling outside the agreement limits. CONCLUSIONS Three-dimensional HU represents a novel vBMD assessment methodology that achieves diagnostic parity with QCT. Crucially, this precision is attained using conventional CT scanners without extra radiation exposure or specialized equipment. This study demonstrated that precise, universally accessible osteoporosis screening is now achievable, promising a transformative advance in public bone health.