Search for a command to run...
Background Diabetes mellitus has emerged as a global public health concern, boasting a high prevalence rate worldwide. Given this situation, accurately identifying type 2 diabetes mellitus (T2DM) holds great significance as it plays a pivotal role in enabling early intervention and facilitating more effective management of the disease. Against this backdrop, it becomes essential to explore the diagnostic value of radiomics features extracted from CT images of vertebral bodies (VB) and paravertebral muscles (PVM) in relation to type 2 diabetes mellitus (T2DM). Methods A total of 160 cases of clinical and imaging data were retrospectively collected, including 80 patients with T2DM and 80 non-diabetic patients. Regions of interest (ROIs) of VB and PVM were delineated for all subjects, and radiomics features were extracted. Patients were divided into a training group (n=112) and a validation group (n=48) at a 7:3 ratio. Key radiomics features of VB and PVM were screened using independent samples t-test and least absolute shrinkage and selection operator (LASSO) algorithm. A k-nearest neighbor (KNN) classifier was used to establish radiomics models based on VB and PVM, and radiomics scores (Rad-scores) were calculated by weighting the coefficients of the selected features. Clinical risk factors were identified via univariate and multivariate logistic regression to construct a clinical model. A nomogram was then developed by integrating the Rad-score with the clinical model using multivariate logistic regression. The diagnostic performance of the models was evaluated using the area under the receiver operating characteristic curve (AUC), calibration curves, and clinical decision curves, with the Delong’s test applied to compare performance among models. Results In the training set, the AUCs of the VB radiomics model, PVM radiomics model, VB-PVM combined radiomics model, clinical model, and radiomics-clinical combined model were 0.902, 0.948, 0.952, 0.857, and 0.956, respectively; in the validation set, the corresponding AUCs were 0.873, 0.880, 0.894, 0.758, and 0.926. The radiomics-clinical combined model showed the best diagnostic performance. Calibration and decision curves indicated that the radiomics nomogram had good consistency and clinical applicability. Conclusion The combined radiomics and clinical model based on CT images of VB and PVM has good diagnostic value for the differential diagnosis of T2DM.