Fetal Monitoring with Cardiotocography (CTG)

Inspiration

We were inspired by the fact that complications in pregnancy are often detected too late, even with continuous monitoring tools like CTG. Many healthcare providers struggle with interpreter variability and false positives/negatives, leading to unnecessary interventions or missed warning signs. We wanted to build something that supports clinical teams with real-time, explainable insights to improve maternal-fetal outcomes.

What it does

Our solution processes CTG traces (fetal heart rate and uterine contractions) using machine learning to detect risk patterns. It provides:

Automated risk classification (e.g., normal, suspicious, pathological).
Explainable AI outputs, linking predictions to physiological features and academic evidence.
Clinical decision support, so obstetricians, nurses, and midwives can act earlier and with more confidence.

How we built it

Machine Learning Model: Trained using proprietary model for fetal risk classification. RAG-enabled LLM: Retrieves and summarizes PubMed/O&G journal papers to justify predictions. Prototype App: Built wireframes and dashboards for clinicians to visualize real-time CTG with alerts and explanations.

Challenges we ran into

Data Access: Getting high-quality, annotated CTG datasets for training and validation.
Interpretability Gap: Making sure explanations are clinically meaningful, not just statistical.
Resource Constraints: Building a reliable pipeline with limited compute and startup-stage budget.

Accomplishments that we're proud of

Built a working pipeline that integrates ML and RAG-based literature grounding.
Created clinically interpretable outputs with structured explanations (e.g., SHAP-linked reasoning + references to published studies).
Developed a scalable, bootstrap-friendly setup that can support pilot deployments in hospitals.

What we learned

Clinicians prioritize trust and interpretability over raw accuracy in AI tools.
Academic grounding via PubMed retrieval increases adoption willingness among doctors.
Balancing research R&D with practical deployment requires lean but robust infra choices.

What's next for Fetal Monitoring with Cardiotocography (CTG)

Clinical Trials: Run pilot studies with hospitals to validate accuracy and usability.
Expanded Dataset: Partner with OBGYN departments to enrich training data across demographics.
Integration: Seamlessly embed into existing hospital monitoring systems and EHRs.
Regulatory Pathway: Continue HSA Class B medical device clearance
Long-Term Vision: Build a comprehensive prenatal AI platform, expanding beyond CTG to wearable maternal health monitoring.