FetalCare AI: Early Risk Detection

• 

  FetalCare Agent detects abnormal CTG pattern. Status: Pathologic. Immediate medical attention advised. Confidence: 0.71

• 

  FetalCare Agent dashboard showing AI-based fetal health prediction with CTG inputs. Status: Normal. Confidence: 0.93

• 

  FetalCare AI architecture

Inspiration During a visit to a special child centre in Bangalore, I met children who couldn’t walk or crawl. When I asked the physiotherapist what had caused it, she said quietly:

“Most of them didn’t get enough oxygen during birth.”

That moment changed me. These children weren’t born with limitations — they were failed by delayed detection. No mother should lose her child to preventable hypoxia. That day, I resolved to build FetalCare AI — a system designed to help doctors detect fetal distress early, giving every unborn child a fair start in life.

What I Learned Working with the UCI Cardiotocography (CTG) dataset, I explored how features like fetal heart rate, uterine contractions, accelerations, and decelerations correlate with oxygen supply and fetal health.

I realized that AI is not just algorithms and numbers — it is compassion translated into computation. Every prediction, every recommendation can directly impact a life.

⚙️ How I Built It

Dataset: UCI CTG dataset (2126 samples, expertly labeled)

Algorithm: CatBoost (chosen for interpretability and robust performance)

Backend: FastAPI service providing real-time predictions

Frontend: Streamlit / React-based dashboard for clinicians

Deployment: AWS Bedrock and serverless services for scalable, secure inference

The system classifies fetal states into Normal, Suspect, and Pathologic, empowering healthcare professionals with actionable insights before complications arise.

Challenges Overcome

Data Imbalance: Medical data was skewed toward normal outcomes; addressed via resampling and careful validation.

Explainability: Integrated SHAP values to visualize model reasoning, building clinical trust.

Time Constraints: Developed and deployed rapidly for the AWS AI Agent Global Hackathon.

Emotional Balance: Maintaining empathy-driven motivation while ensuring scientific precision.

Impact FetalCare AI aims to assist doctors and healthcare workers in early detection of fetal hypoxia — especially in low-resource clinics and rural areas where advanced fetal monitors are not available.

Even if this project helps one mother save her child, or guides one doctor to make a timely intervention, it has fulfilled its purpose.

Vision To make AI-powered fetal monitoring accessible, interpretable, and affordable for every healthcare center — ensuring that no child’s first battle for oxygen goes unheard.

“Every heartbeat matters — and every life deserves a fair beginning.” ❤️

Built With

• amazon
• amazon-web-services
• api)
• backend
• bedrock
• catboost
• cloud
• engineering)
• feature
• flask
• javascript
• lambda
• languages:
• learning
• machine
• model
• numpy
• pandas
• preprocessing
• programming
• python
• react
• s3
• training
• xgboost