Non-invasive prediction of uterine contractions vies electronic nose and oral microbiota profiling

Abstract
Background: Accurate and non-invasive prediction of uterine contractions is essential for optimizing obstetric care and improving perinatal outcomes. We aimed to develop and validate a diagnostic model that combines exhaled volatile organic compounds (VOCs) patterns and oral microbiota profiles to predict contraction status in term pregnancy.
Methods: We prospectively enrolled 84 third-trimester pregnant women and analyzed exhaled breath samples by electronic nose (E-nose). A convolutional neural network (CNN) was used to develop a predictive model of uterine contraction status. Subsequently paired saliva samples from a subset of 15 participants underwent 16S rRNA gene sequencing to explore potential microbial sources of breath VOCs patterns.
Results: The CNN-based E-nose model achieved an area under the curve of 0.63 (95% CI: 0.56-0.71) for differentiating non-contractions from irregular contractions, with performance improving to 0.79 (95% CI: 0.72-0.89) for latent-phase contractions and 0.82 (95% CI: 0.77-0.87) during first-stage labor. Concordant with these findings, 16S rRNA analysis revealed significant enrichment of specific oral taxa in the contraction group, including Streptococcus sanguinis (P=0.0016), Lactobacillus (P=0.0263), Lautropia (P=0.0453), and Lachnospiraceae NK4A136 group (P=0.0408). Bacterial interaction network analysis revealed enhanced synergistic relationships among microbial taxa.
Conclusion: This study demonstrates that E-nose profiling of exhaled breath can non-invasively predict uterine contraction status in late pregnancy. The integration of microbial data provides plausible biological insights into the breath VOCs changes, positioning breath analysis as a novel tool for personalized obstetric monitoring.&#xD.