Application of Artificial Intelligence in Predicting Preterm Birth and Promoting Women’s Health

Introduction
Preterm birth remains a major challenge in obstetrics and is considered the leading cause of neonatal mortality worldwide. Most current clinical interventions are based on traditional observations, basic laboratory tests, or physicians’ experience, which often lack sufficient precision. Due to the complexity of factors influencing preterm birth—genetic, physiological, environmental, and psychological—the need for tools capable of simultaneous analysis of multiple variables is evident. Machine learning algorithms can identify hidden patterns within massive medical datasets and play a critical role in identifying women at high risk for preterm delivery.
Methods
This study utilized a hypothetical dataset comprising clinical information from 1,000 pregnant women in their second trimester. Variables included demographic factors (age, BMI, race), obstetric history, laboratory markers (CRP, HbA1c), ultrasound findings (cervical length, amniotic fluid volume), clinical symptoms (pelvic pain, uterine contractions), and uterine electrical signals (EHG). Data were analyzed using two approaches:
- XGBoost model: optimized parameters included max_depth = 6, learning_rate = 0.15, and n_estimators = 250.
- CNN-LSTM hybrid model: applied to time-series EHG data.
Data were split into training and testing sets at an 80/20 ratio. Evaluation metrics included accuracy, sensitivity, specificity, and AUC.
Results
The XGBoost model achieved 91% accuracy, 88% sensitivity, 93% specificity, and an AUC of 0.95 in classifying patients at risk of preterm birth. The CNN-LSTM model, analyzing EHG data, achieved 93% accuracy and 90% sensitivity. Statistical analysis identified cervical length (p<0.001), history of abortion (p=0.03), elevated systolic blood pressure (p=0.02), and increased CRP (p<0.01) as key predictive variables.
Conclusion
This study concludes that artificial intelligence can be an efficient complementary tool for diagnosing and predicting pregnancy complications, especially preterm birth. The use of algorithms such as XGBoost and LSTM can enhance clinical care accuracy and reduce complications. These findings demonstrate AI’s high potential in developing personalized medical services for women’s health.