Prediction of osteoporosis at the sacrum using opportunistic CT of the abdomen and pelvis: a retrospective feasibility study in 277 patients comparing CT and QCT data

Objectives To examine the distribution pattern of bone density in the L1-S3 vertebrae using opportunistic abdominopelvic imaging.QCT was employed as a reference to establish HU thresholds for the sacral vertebrae facilitating the prediction of osteoporosis and the exclusion of bone abnormalities. Methods A total of 277 subjects aged 19 to 81 years who underwent abdominopelvic CT were evaluated. Bone mineral density (BMD) measurements for the L1-S3 vertebrae and HU values for the S1-S3 vertebrae were collected. The study analyzed the correlation between sacral spine HU values and sacral spine BMD, along with the clinically utilized mean BMD for L1-L2, was analyzed. Receiver operating characteristic (ROC) curves were generated to identify the optimal diagnostic thresholds. Results The BMD of the lumbosacral vertebrae displayed a gradual decrease from L1 to L3, followed by an increase from L4 to S1, and a subsequent decline from S1 to S3. HU values of the sacral vertebrae across all planes were strongly correlated with both sacral spine BMD and the mean BMD values for L1-L2( r＝0.830 to 0.905, P ＜0.05）. For individual vertebrae, the area under the curve（AUC） of HU values for predicting osteoporosis ranged from 0.909 to 0.977, while the AUC for excluding bone abnormalities ranged from 0.933 to 0.950, with S1 demonstrating the highest predictive efficacy. The optimal threshold for S1 was >165.17 HU, yielding a specificity of 91.5% and a sensitivity of 83.0% for excluding bone abnormalities. Conversely, an S1 threshold of <130.50 HU resulted in a diagnostic specificity of 90.0% and a sensitivity of 96.6% for osteoporosis. Additionally, a predictive model that incorporated sex, age, and vertebral cancellous bone HU values achieved an AUC of 0.981. Conclusions Our data demonstrate a strong correlation between the HU values of the sacral spine and the clinically used BMD values for L1-L2, supporting the prediction of osteoporosis based on sacral spine HU values. Moreover, a predictive model that includes sex, age, and vertebral measurements offers improved diagnostic accuracy.