MedGenie

MedGenie: An AI-powered medical assistant that helps doctors make faster, more accurate diagnoses by intelligently summarizing clinical records and providing interactive disease prediction support.

Comment

Inspiration

Our journey with MedGenie began with a simple observation: doctors spend an inordinate amount of time reviewing patient histories and clinical reports. In an era where AI is transforming industries, we saw an opportunity to give doctors their time back while enhancing their diagnostic capabilities. We were particularly inspired by conversations with healthcare professionals who described the challenges of quickly digesting extensive patient histories during time-constrained consultations.

What it does

MedGenie serves as an AI-powered medical assistant that:

  • Transforms lengthy clinical reports into concise, actionable summaries using Facebook's BART Large CNN model
  • Enables doctors to interact with these summaries, adding their professional insights
  • Predicts possible diseases based on the summarized history and doctor's input
  • Provides a chat-based interface for doctors to explore potential diagnoses
  • Features a lab report and prescription data extraction capability (ready for future integration)

How we built it

Our development journey involved several key components:

Core Summarization Engine:

  • Implemented Facebook's BART Large CNN model for accurate medical text summarization
  • Fine-tuned the model to handle medical terminology and clinical report formats
  • Optimized the model for mobile deployment

Disease Prediction System:

  • Developed a custom model that combines report summaries with doctor's annotations
  • Implemented a conversational interface for interactive diagnosis exploration
  • Integrated real-time processing capabilities

Android Application:

  • Built a native Android application for seamless user experience
  • Implemented efficient data handling and processing
  • Created an intuitive interface for doctors to review and interact with summaries

Challenges we ran into

  1. Model Optimization: Adapting the BART model for mobile deployment while maintaining accuracy was particularly challenging. We had to carefully balance model size with performance.

  2. Integration Complexity: Combining multiple AI models (summarization, disease prediction, data extraction) within a single Android application required careful architecture planning and optimization.

  3. Medical Data Handling: Ensuring accurate processing of various medical report formats while maintaining the integrity of critical medical information was a significant challenge.

  4. Performance Optimization: Achieving real-time performance for the summarization and prediction features on mobile devices required extensive optimization.

Accomplishments that we're proud of

  • Successfully integrated advanced AI models into a user-friendly mobile application
  • Developed a working system that significantly reduces report review time
  • Created an interactive system that combines AI capabilities with human medical expertise
  • Built an additional data extraction model ready for future integration

What we learned

  • Deep insights into deploying large language models on mobile devices
  • Techniques for optimizing AI model performance without sacrificing accuracy
  • The complexities of handling medical data and terminology
  • The importance of user experience in medical applications

What's next for MedGenie

  • Integration of our lab report and prescription data extraction model
  • Expansion of the disease prediction capabilities
  • Addition of multilingual support for broader accessibility

Built With

Share this project:

Updates