TransfuSense

🧬 Inspiration

Thalassemia patients undergo regular blood transfusions every 3–4 weeks. For families—especially in rural or low-resource areas—this process involves deciphering complex lab reports, tracking hemoglobin levels, predicting transfusion dates, and finding compatible donors, often without medical guidance.

We asked ourselves: In the age of AI, why are caregivers still relying on paperwork and guesswork for life-saving decisions?

🛠 What We Built

TransfuSense is an AI-powered multilingual health assistant built specifically for Thalassemia care. It combines OCR, NLP, predictive modeling, and translation to simplify and automate key parts of the transfusion process.

• Upload a lab report (image or CSV)
  
• Automatically extract important biomarkers using OCR
  
• Summarize and translate the report into any language
  
• Predict the patient’s next transfusion date
  
• Provide voice-enabled interaction (optional)
  

💡 How We Built It

• OCR powered by PaddleOCR to read medical reports
  
• Summarization using DistilBART, fine-tuned for medical text
  
• Transfusion prediction using a Random Forest Regressor trained on synthetic and real (anonymized) data
  
• Language translation via Meta’s NLLB and Argos Translate
  
• Modular Python agent for routing user input based on intent
  
• Frontend powered by Streamlit for demo simplicity
  
• Optional voice interface via Web Speech API
  

🚧 Challenges We Faced

• Handling noisy, handwritten or scanned medical reports during OCR
  
• Generalizing prediction for diverse patient histories with limited data
  
• Providing meaningful summaries while preserving critical medical info
  
• Building a multilingual interface for users with low digital literacy
  

🌟 What We Learned

• Medical AI tools need to be human-first and context-aware
  
• Language inclusion is essential for real-world health tech adoption
  
• Simplicity, modularity, and explainability in AI design make a huge difference
  

🚀 The Future

We envision TransfuSense becoming a full-fledged health assistant for Thalassemia and other chronic blood disorders—with integrations into:

• Blood bank APIs
  
• WhatsApp-based carebots
  
• Offline mobile deployments
  

Built With

• biobert)-meta-nllb-+-argos-translate-for-multilingual-support-randomforestregressor-for-transfusion-prediction-streamlit-+-html/css-for-frontend-ui-web-speech-api-(browser-based-voice-input)-git
• canva
• development
• for
• github
• pandas)-paddleocr-for-report-extraction-hugging-face-transformers-(distilbart
• python-(streamlit
• scikit-learn