Interactive research prototype

Test the maternal risk model.

Enter six routine maternal measurements to explore the study’s low, mid, and high-risk classification. Inference runs locally in your browser.

Loading model…

Patient measurements

Use a sample profile or enter a new observation.

6 inputs

Age training: 10–70 years Age during pregnancy Systolic pressure training: 70–160 mmHg Upper blood-pressure value Diastolic pressure training: 49–100 mmHg Lower blood-pressure value Blood glucose training: 6–19 mmol/L Dataset field: BS Body temperature training: 98–103 °F Temperature in Fahrenheit Heart rate training: 60–90 bpm Resting beats per minute

Ensemble output Local inference

— confidence

Ready to assess

Complete the measurements to run the research model.

Low risk —

Mid risk —

High risk —

What-if influence

Available after assessment

Proof-of-concept screening output only. It is not a diagnosis, treatment recommendation, or substitute for qualified clinical assessment.

74.2% Accuracy on the leakage-resistant 93-row holdout.

5 models SVC, Extra Trees, XGBoost, CatBoost, and LightGBM.

Device only Measurements stay in this browser and are not uploaded.

Scenario comparison

Save up to four assessments to compare how the predicted pattern changes.

No saved cases yet. Run an assessment and select Compare.

Scenario	Age	BP	Glucose	Temp	Heart rate	Prediction	Confidence

Read the result carefully

Designed for exploration, with the limits visible.

The interface keeps research performance and clinical meaning separate.

Probability, not certainty

The largest of three averaged model probabilities determines the displayed class.

Local what-if view

Influence shows how confidence changes when one input is replaced by its dataset median; it is not causal.

Mid-risk caution

The champion recalled only 15.8% of mid-risk holdout cases, so intermediate patterns need particular caution.