Predicting Preterm Birth Using Multimodal Fetal Imaging.

Preterm birth (PTB) (<37 weeks' gestational age (GA)) is associated with increased risk of short- and long-term sequelae. Accurate predictive tools allow to improve the outcomes of those born preterm by offering early obstetric interventions to mothers at high-risk of PTB. Methods: This study combines a wide range of structural and functional MRI parameters, from the fetal head, lung, placenta with clinically available Ultrasound and outcome data. A preprocessing pipeline adapted to the special requirements of the often incomplete and highly GA dependant data and a supervised machine learning model based on these derived markers derived is proposed. Data from 58 preterm and 217 term-born neonates were analysed. Results: The best SVR model achieved an R-2 value of 0.67 and correctly predicted 92% of true preterm cases using a combination of two maternal and four fetal features. Conclusion: The significance of this study is uncovering the potential of markers derived from multi-modal imaging data in the prediction of PTB using large-scale fetal studies. This study paves the way for future studies focusing on at-risk women to further enhance the data set and thus predictive power.