Radiomics based on lumbar CT to identify high-risk patients for OVCF in postmenopausal women.

Objective: Osteoporosis vertebral compressive fracture (OVCF) is a severe complication in patients with osteoporosis. There were limitations in finding the risk factor of OVCF in previous evaluation techniques. In this study, we developed a radiomics model (R-model) based on a lumbar CT scan to identify vertebrae at high risk of OVCF in postmenopausal women. Method: Radiographic data of postmenopausal patients in our hospital from January 2021 to August 2022 were collected. All the patients received both dual-energy X-ray absorptiometry (DEXA) and lumbar CT scan. Images in a dataset of 329 vertebral bodies without compressive fracture in lumbar 1 to 4 were extracted by a 3D slicer and randomly divided into a training group (n = 230) and a test group (n = 99). A number of radiomics features (129) were automatically calculated by the pyradiomics module, and minimum-redundancy maximum-relevancy (mRMR), least absolute shrinkage, and selection operator (LASSO) were used to shrink features for R-model construction. The sensitivity, specificity, accuracy, and area under the receiver operating characteristic curve (AUC) of the T-scores and R-scores were calculated. The AUCs of the two models were compared using the DeLong test. Decision curve analysis (DCA) shows the clinical usefulness of the R-model. Results: Eight features were chosen to construct the R-model. The AUCs of the T-score and R-score in the training group were 0.845 and 0.945, respectively, and 0.818 and 0.914, respectively, in the test group. There was a significant difference (p < 0.001) between the AUCs of the two models, and the decision curve analysis (DCA) shows the R-model has a better overall net benefit than the T-score model. Conclusion: The radiomics model based on lumbar CT scans in postmenopausal women can identify and predict patients at high risk of OVCF with better sensitivity and accuracy than DEXA, even in patients with the same T-scores.