Prediction of Preterm Birth Based on Cervical Ultrasound Radiomics Combined with Clinical Features

Objective: To evaluate the feasibility of predicting preterm birth using an ultrasound radiomics model combined with clinical features.Methods: We retrospectively analyzed 521 pregnant women who underwent prenatal care at Yichang Central People’s Hospital between January 2018 and August 2024. Patients were randomly assigned to a training set (n = 417) and a validation set (n = 104) at an 8:2 ratio. Radiomic features were extracted from the region of interest (ROI) of the cervix on 2D ultrasound images, and combined with clinical high-risk factors. All features were standardized and normalized to a (0,1) range. Feature selection was performed using variance thresholding, optimal feature selection (by number and percentage), and significance-based filtering. Logistic regression, random forest, and support vector machine models were constructed for preterm birth prediction. Model performance and clinical utility were evaluated using ROC curves, AUC, Hosmer-Lemeshow test, and decision curve analysis (DCA).Results: The combined model achieved the highest predictive performance for preterm birth, with AUCs of 0.874 and 0.841 in the training and validation sets, respectively, indicating good consistency. Hosmer-Lemeshow test and decision curves demonstrated good model calibration and high clinical net benefit.Conclusion: Ultrasound radiomics combined with clinical features can effectively predict preterm birth and may support early, non-invasive clinical interventions.