MEDCYB — ML recognition of gestational diabetes predictors in pregnant women
ML recognition of gestational diabetes predictors in pregnant women
Challenge
Gestational diabetes mellitus (GDM) is hard to predict in advance: dozens of factors, no reliable predictor model on clinical data.
How we built it
Together with the OB/GYN department of the St. Petersburg Pediatric University we built an ML model on a dataset of 5000+ women and ~85 parameters: data cleaning, feature selection, training and validation of a GDM predictor classifier.
How we built it
5000+ women, ~85 parameters
Cleaning, missing values, encoding
Significant GDM predictors
Train & validate classifier
Result
The model surfaces significant GDM predictors and estimates risk — a clinical decision-support and early-screening tool.
How it helps you
Picture an antenatal visit: gestational diabetes is easier to prevent than to treat, but dozens of risk factors are impossible to keep in your head, and clear symptoms appear late. The MEDCYB model analyses ~85 parameters across 5000+ patients and flags GDM risk in advance — giving the doctor time for prevention instead of managing complications.
Early GDM screening: the model flags risk in advance — leaving time for prevention.
Clinical decision support on 5000+ patients’ data, not on intuition.
Clear significant predictors — the doctor sees exactly what drives the risk.
Ready to discuss your project?
We scope the task and propose a solution. Leave a request — we will get back to you shortly.