Ultrasound radiomics model shows promise for predicting spontaneous preterm birth

Researchers have developed a prediction model that combines radiomic features extracted from cervical ultrasound images with clinical indicators to estimate the risk of spontaneous preterm birth in late pregnancy. The study, published in BMC Pregnancy and Childbirth, was designed as a nested case-control study and included 272 singleton pregnant women who underwent transvaginal ultrasound at 28 to 32 weeks of gestation. Premature birth remains a major cause of perinatal death and long-term health complications, and the authors note that existing prediction methods have limitations during the third trimester, which motivated the search for an improved approach.

The team then used logistic regression to build three competing models: a clinical model, a radiomics-only model and a combined model. The clinical variables incorporated into the combined model were the anterior uterocervical angle, cervical length and maternal age. Model performance was evaluated using the area under the receiver operating characteristic curve, calibration curves and decision curve analysis.

The combined model achieved the best overall performance, with an area under the receiver operating characteristic curve of 0.907 in the training set and 0.830 in the validation set. However, the authors report an important nuance: in the validation set, the combined model performed similarly to the radiomics-only model, suggesting that the clinical variables may have contributed less to the combined model’s efficacy than expected. The anterior uterocervical angle, cervical length, maternal age and the radiomics score were all identified as independent predictors. The combined model also demonstrated good calibration and delivered a positive net benefit across a reasonably broad range of threshold probabilities, from 0.01 to 0.99 in the training set and 0.02 to 0.99 in the validation set, according to decision curve analysis.

The authors conclude that ultrasound radiomics shows promising value for predicting spontaneous preterm birth in the third trimester and that incorporating clinical indicators may improve predictive performance in clinical practice. They also caution that these findings require validation in large multicenter prospective cohort studies before any clinical translation. The study’s design and scope should be kept in mind when interpreting the results: it was a nested case-control study of 272 women with singleton pregnancies assessed at a specific gestational window, and the reported performance figures come from internal training and validation splits rather than an independent external cohort. The shared authorship note accompanying the article indicates that this version was shared early to provide faster access to peer-reviewed, accepted research, is citable and carries a permanent DOI, and is subject to further edits before the final Version of Record replaces it. Until larger prospective studies confirm the results, the model should be regarded as a research tool under development rather than an established clinical test, with the radiomics component appearing to carry most of the predictive signal observed in the validation data.

Subject of Research: Predicting spontaneous preterm birth using cervical ultrasound radiomics combined with clinical indicators

Article Title: The predictive value of ultrasound radiomics combined with clinical indicators for spontaneous preterm birth: a nested case-control study

Article References: Mei, Y., Li, X. M., Zeng, B. H., Zhou, H. E., Huang, Y. Y., & Guo, X. M. (2026). The predictive value of ultrasound radiomics combined with clinical indicators for spontaneous preterm birth: a nested case-control study. BMC Pregnancy and Childbirth. https://doi.org/10.1186/s12884-026-10139-8

Image Credits: AI Generated

DOI: 10.1186/s12884-026-10139-8

Keywords: preterm birth, ultrasound radiomics, cervical length, uterocervical angle, LASSO regression, nested case-control, predictive model, third trimester, BMC Pregnancy and Childbirth, machine learning, transvaginal ultrasound, decision curve analysis