Medical Report Analyzer & Healthy Habits Recommender

• 

  Main Page

• 

  Suggestions

Inspiration

Every year during my annual hospital checkups, I receive multiple test reports. Some values are normal, while others need improvement. But after a few days, I tend to forget about them. I wanted a solution that could analyze my reports, remind me of areas to improve, and suggest actionable health habits so I can stay on track year-round.

This project was inspired by that need—to have an intelligent assistant that turns my medical data into daily, personalized health advice.

What it does

• Accepts medical documents (PDFs, scans, images)
• Uses OCR to extract structured medical data
• Summarizes doctor’s notes and prescriptions using GPT
• Generates personalized:
  • Health summaries
    
  • Few health habit suggestions
    
  • Few food/nutrition suggestions
• Stores data per user profile and avoids duplicate document processing
• Recommends habits based on entire user history, not just one report

How we built it

• Backend: Python
• Frontend: HTML UI
• OCR: Tesseract for extracting text from images
• AI Engine: OpenAI GPT/Llama2 local model for summarization and suggestions
• Model Management: MLflow for model versioning and deployment
• Platform: Deployed and tested on HP AI Studio

The pipeline automatically connects uploaded files → OCR → GPT prompt creation → response processing → user-specific output generation.

Challenges we ran into

• Parsing inconsistent formats across medical reports
• Handling poor-quality scanned images for OCR
• Designing prompts that produce medically accurate summaries and suggestions
• Avoiding redundant processing of the same report multiple times
• Deployment issues with GitHub and HP AI Studio integration

Accomplishments that we're proud of

• Fully functional end-to-end system: upload → analyze → actionable insights
• Automatically detects duplicate uploads to save compute and time
• Personalized health & food suggestions based on real medical context
• Intuitive and clean interface with minimal setup
• Packaged model using MLflow for reproducibility

What we learned

• Effective ways to combine OCR + LLMs for real-world document processing
• How to build self-adaptive prompts for variable medical data
• Streamlining model lifecycle with MLflow
• Leveraging HP AI Studio’s deployment stack for app hosting and testing
• Importance of user-centric design for medical applications

What's next for Medical Report

• Add multi-language support for reports in regional languages
• Enable continuous habit tracking and notifications via email or mobile app
• Integrate basic health visualizations (charts/trends)
• Include more advanced medical models for deeper insights
• Add voice-based interaction for elderly or visually impaired users

Built With

• llama
• llm
• openai
• prompt
• python