From Data to Prediction: Development and Clinical Validation of a Machine Learning-Based Preterm Birth Risk Assessment Tool
PetrSU has developed an AI model, CatBoost, for early assessment of preterm birth risks based on medical records. The accuracy rate is 81%, and case detection is up to 87%.
Specialists from Petrozavodsk State University have developed an AI-based system for early assessment of preterm birth risks based on electronic medical records. The results were published in the journal "Obstetrics, Gynecology, and Reproduction." The researchers note that the best risk prediction results among machine learning algorithms were achieved by the gradient boosting-based algorithm, CatBoost Classifier (Categorical Boosting Classifier).
Gradient Boosting is a machine learning method that sequentially creates a set of weak predictive models (usually decision trees), combining them into a single strong model. Each new model seeks to correct the errors made by previous models.
CatBoost demonstrated 81% accuracy, 87% sensitivity, and 99.8% text extraction recall among 14 algorithms. The model was trained on 10,000 anonymized charts with 54 parameters, including clinical and laboratory records. The system analyzes combinations of factors: placental insufficiency, infections, cervical insufficiency, multiple pregnancies, and IVF. This will allow doctors to identify risks earlier and prescribe preventive measures, reducing neonatal mortality.
The authors of the development note that preterm birth is the result of a complex interaction of multiple factors unique to each patient, requiring a shift from universal approaches to personalized ones based on the integration of a wide range of clinical data. They also emphasize that AI is a support tool, not a replacement for a physician; multicenter tests and biomarkers are needed.
More details https://www.gynecology.su/jour/article/view/2618
Also read here https://medvestnik.ru/content/news/rossiiskie-uchenye-sozdali-ii-model-dlya-ocenki-riskov-prejdevremennyh-rodov.html
