Accuracy of Phantomless Calibration of Routine Computed Tomography Scans for Opportunistic Osteoporosis Screening in the Spine Clinic

Study Design. Diagnostic accuracy study Objective. To establish a simple method of phantomless bone mineral density (BMD) measurement by using preoperative lumbar Computed Tomography (CT) scans, and compare the accuracy of reference tissue combinations to diagnose low BMD against uncalibrated Hounsfield units (HUs). Summary of Background Data. HUs are used as a measure of BMD; however, associations between HU and T-scores vary widely. Quantitative CT (qCT) scans are more accurate, but they require density calibration with an object of known density (phantom), which limits feasibility. As an emerging technique, phantomless (internal) calibration of routine CT scans may provide a good opportunity for screening. Methods. Patients who were scheduled to undergo lumbar surgery, with a preoperative CT scan, and a dual-energy x-ray absorptiometry (DXA) scan within six months were included. Four tissues were selected for calibration: subcutaneous adipose (A), erector spinae (ES), psoas (P) and aortic blood (AB). The HUs of these tissues were used in linear regression against ground-truth values. Calibrations were performed by using two different internal tissues at a time to maintain simplicity and in-office applicability. Volumetric bone mineral densities (vBMD) derived from internally calibrated CT scans were analyzed for new threshold values for low bone density. Areas under the curve (AUC) were calculated with 95% confidence intervals (CI). Results. 45 patients were included (M/F=10/35, mean age:63.3). Calibrated vBMDs had stronger correlations with DXA T-scores when compared with HUs, with L2 exhibiting the highest coefficients. Calibration by using A and ES with the threshold of 162 mg/cm 3 had a sensitivity of 90% in detecting low BMD (AUC=0.671). Conclusions. This novel method allows simple, in-office calibration of routine preoperative CT scans without the use of a phantom. Calibration using adipose and erector spinae with a threshold of 162 mg/cm 3 is proposed for low bone density screening with high sensitivity (90%). Level of Evidence. Level III