A new model based on artificial intelligence to screening preterm birth
Valter Lacerda de Andrade Júnior1, Marcelo Santucci França2, Roberto Angelo Fernandes Santos1
1Graduate and Postgraduate Department, Impacta School of Technology, São Paulo, Brazil.
Summary
Artificial intelligence (AI) offers a new method for screening spontaneous preterm birth (sPTB) before 35 weeks. This AI model significantly improves prediction accuracy compared to traditional cervical length measurements.
Area of Science:
- Perinatology and Obstetrics
- Medical Artificial Intelligence
- Biomedical Data Science
Background:
- Spontaneous preterm birth (sPTB) before 35 weeks gestation is a significant concern in singleton pregnancies.
- Current screening methods, such as cervical length (CL) measurement via transvaginal ultrasound, have limitations in predictive accuracy.
- There is a need for improved screening tools to identify pregnancies at high risk for sPTB.
Purpose of the Study:
- To develop and evaluate a novel artificial intelligence (AI) based screening tool for spontaneous preterm birth (sPTB) before 35 weeks gestation.
- To compare the performance of the AI model against traditional methods like cervical length (CL) measurement and logistic regression (LR).
- To assess the clinical utility of AI in improving sPTB prediction accuracy and reducing false positives.
Main Methods:
- A cohort of 524 singleton pregnancies underwent transvaginal ultrasound for CL measurement between 18-24 weeks gestation.
- An AI model was constructed using a stacking-based ensemble learning method (SBELM), incorporating CL, logistic regression variables, and neural network algorithms.
- Performance metrics including Area Under the Curve (AUC), sensitivity, specificity, and predictive values were calculated for CL, LR, and the SBELM AI model.
Main Results:
- The AI model (SBELM) demonstrated superior performance compared to CL < 25 mm and logistic regression.
- At a 10% false positive rate, SBELM achieved an AUC of 0.808, sensitivity of 47.3%, and specificity of 92.8%.
- The AI model showed statistically significant improvement (p < .00001) over CL < 25 mm in predicting sPTB < 35 weeks.
Conclusions:
- Artificial intelligence applied to clinical and ultrasonographic data presents a viable strategy for screening sPTB before 35 weeks.
- The AI model significantly enhances the predictive performance associated with a short cervix, offering improved accuracy with a low false-positive rate.
- This AI-driven approach holds promise for more effective early identification of pregnancies at risk for preterm birth.
