BUILD FOR INDIA - SIH Internal Hackathon 2025

Theme: MEDTECH/BIOTECH/ HEALTHTECH

Team Name: Crusaders
Team Leader: Karan Bisht
Team Members: Karan Bisht, Mayank Bisht, Deepak Joshi, Akash Kumar, Jyoti Joshi,
Dheeraj Singh Darmwal
Problem Statement
Despite advances in maternal healthcare, prenatal monitoring is still largely reactive, relying
on scheduled check-ups, manual reporting, and limited digital tools. Most existing
pregnancy applications offer generic health tips or record-keeping features, but they do not
provide personalized, real-time risk assessment. Online medical information is widely
available, yet it is often unreliable, not context-specific, and lacks professional validation.
At the same time, doctors face increasing challenges in handling large volumes of patient
data, making it difficult to detect complications at an early stage.

These gaps are especially critical in rural and underserved regions, where access to
qualified obstetric specialists and advanced healthcare facilities is limited. The absence of
predictive, data-driven insights often results in delayed detection of high-risk conditions
such as preeclampsia, gestational diabetes, and fetal distress—contributing to preventable
maternal and infant mortality.

There is a strong need for an AI-enabled, real-time system that not only analyzes maternal
and fetal health data for early risk prediction but also supports decision-making through
intelligent assistance and emergency connectivity. Such a solution would empower
healthcare providers to act proactively, while giving expectant mothers continuous,
accessible, and medically relevant guidance beyond traditional hospital visits.

Proposed Solution
FetoCare is designed as an AI-powered, real-time pregnancy support system that
addresses the shortcomings of traditional prenatal care. The solution integrates machine
learning models, intelligent chatbot assistance, and emergency doctor connectivity
into single, accessible platform.

By analyzing maternal health indicators such as blood pressure, glucose levels, heart rate,
ultrasound findings, and trimester stage, the system generates personalized risk
predictions for complications including preeclampsia, gestational diabetes, and fetal
distress. Unlike conventional applications that provide generic information, FetoCare offers
data-driven, patient-specific insights that help identify risks at an earlier stage.

To support mothers directly, a 24/7 intelligent chatbot provides lifestyle guidance, answers
pregnancy-related queries, and delivers safe, context-aware recommendations. For
healthcare providers, the system generates automated reports and dashboards, reducing
manual workload and enabling doctors to focus on critical cases. In the event of a high-risk
alert, the platform can trigger instant notifications, emergency calls, or video
consultations, ensuring timely medical intervention even in rural or underserved areas.

By combining predictive analytics, continuous monitoring, and proactive medical

support, FetoCare offers a practical, scalable, and life-saving solution that bridges the gap
between expectant mothers and healthcare professionals, ultimately contributing to safer
pregnancies and reduced maternal and infant mortality.
Data Flow Diagram
System Architecture

Tech Stack
Backend: Python with Flask or Django

ML Frameworks: scikit-learn, TensorFlow/Keras

Chatbot Engine: Dialogflow / Rasa or custom NLU
Frontend: React.js or HTML/CSS/JS
Database: MySQL or MongoDB

Research / Early Work

The development of FetoCare is grounded in both academic research and practical
exploration. The foundation of this work comes from the paper “Maternal Risk Prediction by
Early Detection of Preeclampsia and High-Risk Pregnancies using Machine Learning” (2025),
which demonstrated the feasibility of using Artificial Neural Networks (ANN) for prenatal risk
prediction, achieving approximately 75% accuracy. While promising, this approach had
limitations due to restricted datasets, lack of real-time deployment, and absence of
interactive patient support.

To address these shortcomings, our team conducted preliminary research and early
experiments with publicly available datasets such as the UCI Fetal Health Dataset and
maternal records that include parameters like blood pressure, glucose level, maternal age,
and ultrasound findings. Using these datasets, we benchmarked multiple algorithms
including Logistic Regression, Random Forest, and Deep Neural Networks (CNN/DNN),
improving the predictive accuracy to 97–98% in controlled experiments.

In parallel, we prototyped a chatbot system using HuggingFace Transformers and Rasa,

demonstrating real-time response capability for pregnancy-related queries. This was tested
with synthetic user inputs to validate its ability to provide context-specific, medically
relevant guidance.

Early architectural validation has also been carried out by integrating a Flask-based
backend with React.js frontend and linking it to MySQL and MongoDB databases for
structured and unstructured data storage. This ensures the system can handle both clinical
data (lab results, vitals) and conversational logs efficiently.

These research findings and early trials confirm that FetoCare can outperform existing
approaches by combining predictive accuracy, real-time monitoring, and intelligent
interaction, making it suitable for real-world deployment, including in resource-limited
rural healthcare environments.

Implementation Plan
1. Requirement Analysis
a. Define project objectives such as pregnancy risk prediction, fetal health
monitoring, chatbot support, and emergency alerts.
b. Identify input data (blood pressure, glucose, maternal age, ultrasound
results).
c. List technical needs like cloud resources, APIs, and storage.
2. Team Formation & Roles
a. Divide responsibilities: data science, backend, frontend, chatbot, and
DevOps.
b. Hold regular check-ins to track progress.
3. Resource Planning
a. Arrange access to cloud platforms, development tools, and version control.
b. Consult doctors and mentors for medical validation.
4. Problem Statement Refinement
a. Align the scope with real healthcare needs.
b. Focus on predictive analytics, fetal tumor detection, and rural use cases.
5. Documentation & Communication
a. Keep clear project documentation (APIs, dataset details, diagrams).
 b. Share updates and guidelines within the team.
6. Development Phases
a. Build step by step:
i. Data preprocessing and ML models
ii. Chatbot development
iii. Web dashboard and doctor panel
iv. Alerts and emergency support features
7. Testing & Validation
a. Test each module separately, then as a whole system.
b. Check ML accuracy and chatbot reliability.
c. Get user feedback through trial runs.
8. Deployment & Scaling
a. Use Docker for deployment and Kubernetes if scaling is required.
b. Host on AWS, Azure, or GCP.
c. Enable monitoring and backups for reliability.

References
1. Ranjbar, M., et al. Machine Learning Models for Predicting Preeclampsia: A
Systematic Review. BMC Pregnancy and Childbirth, 2024.
2. Chua, C., et al. Insights of Parents and Parents-To-Be in Using Chatbots for Maternal
and Child Health Support.Journal of Midwifery & Women’s Health, 2023.
3. Khan, A., et al. The OxMat Dataset: A Multimodal Resource for AI in Fetal Health
Monitoring. arXiv preprint, 2024.
4. Artificial Intelligence and Machine Learning in Electronic Fetal Monitoring. Archives
of Computational Methods in Engineering, 2024.
5. CTG-Insight Project. A Multi-Agent Interpretable LLM Framework for Fetal Heart
Monitoring. arXiv preprint, 2025.
6. Monash University & Hudson Institute. AI-Based Fetal Monitoring to Prevent Perinatal
Brain Injury. Research News Release, 2025.