Integration of bone turnover markers and radiomic features from chest CT imaging for osteoporosis diagnosis in patients with hip joint injury

Purpose: The objective of this study is to investigate whether the combination of bone turnover marker and radiomic features extracted from chest CT images can improve the diagnostic accuracy for predicting osteoporosis in patients with hip injury. Method: A group of 249 patients who suffered hip injury participated in this study. They underwent various diagnostic tests, including chest CT scans, bone density measurements, and bone turnover marker. The clinical data collected from these patients was used to create clinical models. To extract radiomics features, one thoracic vertebra was selected from each patient using ITK-SNAP. The feature extraction was conducted using a pyramid library tool that runs on the Python programming language. By randomly selecting patients, a training set consisting of 201 individuals and a testing set comprising of 48 participants were formed. A radiomics model was established after extracting radiomics features. The nomogram model was developed by combining radiomics models with clinical models. Finally, the sensitivity, specificity, accuracy, and area under the curve (AUC) of the three models were calculated separately. Result: The clinical model developed by our research team incorporated variables such as gender, age, vitamin D, and β B-Cross Laps. In the training group, the accuracy of Clinic model recognition was 0.731, AUC was 0.735. In the test group, the accuracy of Clinic model recognition was 0.708, AUC was 0.769. In the training group, the accuracy of nomogram model recognition was 0.891, AUC was 0.943. In the test group, the accuracy of nomogram model recognition was 0.812, AUC was 0.862. Conclusion: By combining the radiological features extracted from thoracic CT scans with the clinical features of patients, it is possible to distinguish hip joint injury patients with or without osteoporosis before surgery