Intelligent prediction of cervical instability based on MRI radiomics combined with clinical features

OBJECTIVE: To build machine learning models based on radiomics features from cervical MRI and clinical features to enable early and intelligent prediction of cervical instability (CI). METHODS: We included 196 subjects with CI and 112 healthy controls from January 2022 to May 2024, divided into training and test sets based on the chronological order. Cervical MRIs were collected, and radiomics features were extracted from 5 regions of interest (ROI) based on cervical anatomical structures. Clinical features were collected through questionnaires. Radiomics models, clinical models, and radiomics-clinical combined models for CI were built using 6 algorithms. The area under the curve (AUC) from the receiver operating characteristic (ROC) curve, along with the F1 score, precision, and recall, were calculated. The DeLong test was applied to compare the AUC of the models. An application for predicting CI was to develop based on the study results to enhance its use in clinical settings. RESULTS: In the training set, the highest AUCs of the clinical model, radiomics model, and radiomics-clinical combined model were 0.73, 0.97, and 0.97, respectively. The AUC of the radiomics model (P < .001) and the radiomics-clinical combined model (P < .001) was significantly higher than that of the clinical model. There was no significant difference in AUC between the radiomics model and the radiomics-clinical combined model (P = .66). In the test set, the AUCs of the clinical model, radiomics model, and radiomics-clinical combined model were 0.64, 0.74, and 0.74, respectively. The AUC of the radiomics model (P < .001) and the radiomics-clinical combined model (P < .001) was significantly higher than that of the clinical model. There was no significant difference in AUC between the radiomics model and the radiomics-clinical combined model (P = .79). CONCLUSION: The radiomics-clinical combined model, developed based on cervical MRI and clinical features, enables early and intelligent prediction of CI and demonstrates excellent classification performance. TRIAL REGISTRATION: The study protocol is registered with Chinese Clinical Trial Registry (registration number ChiCTR2100053525).