Preterm Birth: A Global Challenge
Preterm birth (PTB) remains one of the most pressing global health issues, affecting up to 16% of pregnancies worldwide and accounting for a significant share of neonatal mortality. Despite advancements in healthcare, predicting PTB risks has been challenging due to the complex interplay of factors influencing preterm labor. Traditional methods relying on electronic health records (EHRs) or biomedical signals often fall short in providing continuous, real-life monitoring.
A recent study is changing this narrative by leveraging wearable technology to monitor maternal autonomic nervous system (ANS) activity and predict PTB risks. By integrating smartwatches with advanced machine learning models, researchers have unlocked the potential for early PTB risk estimation through continuous, noninvasive monitoring.
The Study: Harnessing Wearables for Maternal Health
This groundbreaking longitudinal study monitored 58 pregnant women, including seven preterm cases, from gestational weeks 12–15 to three months postpartum. Using smartwatches to collect long-term photoplethysmogram (PPG) signals, the study extracted heart rate (HR) and heart rate variability (HRV) data, key indicators of ANS activity.
Key Findings
- Distinct Patterns of Abnormality Scores:
Abnormality scores predictive of PTB risks emerged as early as the second trimester, highlighting the feasibility of early intervention. - HRV Metrics as Significant Predictors:
Key features like average interbeat intervals (AVNN), SD1SD2 ratio, and SDNN were identified as critical indicators of PTB risks. - Machine Learning with Explainable AI:
An autoencoder model with SHAP analysis provided interpretable insights into the impact of HR and HRV features on PTB predictions.
Authors: Mohammad Feli, Iman Azimi, Fatemeh Sarhaddi, Zahra Sharifi-Heris, Hannakaisa Niela-Vilen, Pasi Liljeberg, Anna Axelin & Amir M. Rahmani
