Inspiration
In India, thousands of children are born with Thalassemia every year, many of whom require lifelong blood transfusions. We were deeply moved after learning that:
Patients often suffer adverse immune reactions due to poorly matched blood.
Donors rarely know the impact of their contribution.
There’s no mechanism to track post-transfusion health outcomes or give real-time feedback.
This inspired us to build ThalMatch+ — a solution that not only ensures biological compatibility between donor and patient but also closes the loop with feedback and learning.
What We Learned During this hackathon, we learned:
The importance of cross-matching blood beyond blood groups, including tissue compatibility and graft rejection risks in Thalassemia patients.
How AI can assist in predicting compatibility using donor/patient medical data.
How to design a feedback-driven loop to refine future predictions and engage both patients and donors meaningfully.
We also explored:
Concepts around Hb level estimation, SpO₂ monitoring, and hematologic markers.
Best practices in patient data privacy, encryption, and secure medical storage.
What it does
ThalMatch+ is an AI-powered blood compatibility and feedback system designed specifically for Thalassemia patients. It goes beyond basic blood type matching to ensure safer, smarter, and more personalized transfusions. Here's how it works:
Donor Blood Screening
Donors provide a small blood sample.
The system analyzes Hemoglobin (Hb) levels, Oxygen saturation (SpO₂), and other vital markers.
Health metrics are logged and verified for transfusion suitability.
Patient Matching
Thalassemia patients upload their tissue compatibility data and medical history.
AI performs a compatibility check and predicts graft rejection risk.
A match score is calculated based on multiple biological parameters.
Transfusion Approval
If compatibility is confirmed, the system greenlights the transfusion.
Doctors review match reports and approve the process.
Feedback Collection
After transfusion, patients report symptoms and outcomes.
Doctors add post-transfusion observations.
The AI uses this data to refine future match predictions, creating a smart feedback loop.
Donor Engagement
Donors receive feedback about the impact of their donation.
A trust-based ecosystem is built between donors, patients, and doctors.
How we built it
Frontend: Flutter for cross-platform mobile app.
Backend: Node.js and Express.
Database: Firebase Firestore for real-time updates + PostgreSQL for structured clinical data.
Machine Learning:
Python + scikit-learn for donor match scoring
Initial model trained on mock datasets with features like: MatchScore = w1.HbDiff + w2.SpO₂Diff + w3.TissueMatchScore Feedback System:
Patients rate outcomes post-transfusion via chatbot form
Doctors input clinical observations
AI uses this to adjust matching algorithm (active learning approach)
Challenges we ran into
Data Availability: Real-world compatibility datasets for Thalassemia were difficult to access, so we worked with simulated data.
Biometric Simulation: We had to simulate blood biosensor input like Hb and SpO₂ readings due to lack of real hardware.
Privacy Compliance: Designing secure storage for sensitive health information pushed us to learn about encryption techniques and legal frameworks like HIPAA.
AI Feedback Loop: Building a reliable and explainable loop for match improvement based on feedback was complex — especially in designing meaningful weights and thresholds.
Accomplishments that we're proud of
Built an end-to-end prototype in record time — from donor blood analysis to AI-powered match scoring and post-transfusion feedback collection.
Successfully simulated Hb and SpO₂ health data to test compatibility models, even without access to physical biosensor hardware.
Designed a secure, privacy-compliant flow for sensitive medical data, using encryption and role-based access control.
Implemented a feedback loop where patient outcomes are used to refine AI predictions — turning each transfusion into a learning opportunity.
Created a user experience that supports not just patients, but also empowers donors and doctors with clarity and transparency.
Proposed a solution that could realistically scale across India, especially through platforms like e-RaktKosh and Blood Warriors’ Blood Bridge.
What we learned
During this hackathon, we learned:
The importance of cross-matching blood beyond blood groups, including tissue compatibility and graft rejection risks in Thalassemia patients.
How AI can assist in predicting compatibility using donor/patient medical data.
How to design a feedback-driven loop to refine future predictions and engage both patients and donors meaningfully.
We also explored:
Concepts around Hb level estimation, SpO₂ monitoring, and hematologic markers.
Best practices in patient data privacy, encryption, and secure medical storage.
What's next for SmartBlood: AI for Safe Transfusions in Thalassemia
We aim to:
Integrate with live biosensor hardware for real-time blood scanning.
Partner with Blood Warriors and hospitals for real-world testing.
Expand the model to include antibody profiling, iron overload risk, and auto-scheduling for repeat transfusions.
Publish the feedback-learning AI model as an open-source healthcare tool.
Log in or sign up for Devpost to join the conversation.