Glucose

Glucose in action
Real time data
Historical data and AI analysis
Landing Page

Members

Keith Khadar, Alexander Salinas, Eli Campos, Gabriella Conde

Inspiration

Last year, Keith had a freshman roommate named Cayson who had played high school football until a knee injury sidelined him. While his condition improved in college—allowing him to walk, he couldn’t run. Keith remembered how he often had to make a 30-minute walk to his physical therapist. It was through witnessing his struggle and through Keiths experience working on medical devices in Dream Team Engineering, a Club at the University of Florida dedicated to improving patient care, and our curiosity to work on real world problems with AI, that we began to think about this issue.

What it does

Our device tracks an injured athlete's movements and provides personalized advice, comparable to that of a world-class physical therapist, ensuring the patient recovers effectively and safely. Our device helps users perform physical therapy exercises at home safely while AI analyzes their movements to ensure they operate within their expected range of motion and effort values.

How we built it

Using web technologies (Angular, Python, Tune) and microcontrollers (Flex-sensor + ESP32) to track, give insights, and show improvement over time.

Challenges we ran into

Bluetooth Implementation: Establishing a reliable and efficient Bluetooth connection between the microcontroller and our web application proved more complex than anticipated. Sleeve Assembly: Designing and constructing a comfortable, functional sleeve that accurately houses our sensors while maintaining flexibility was a delicate balance. Data Interpretation: Translating raw sensor data into meaningful, actionable insights for users required extensive algorithm development and testing. Cross-platform Compatibility: Ensuring our web application functioned seamlessly across various devices and browsers presented unexpected complications. Specifically browser as well as

Accomplishments that we're proud of

Seamless Bluetooth Integration: We successfully implemented robust Bluetooth communication between our hardware and software components, enabling real-time data transfer. Real Time Digital Signal Processing: Our team developed sophisticated algorithms to analyze data from our sensors, providing a comprehensive view of the user's movements and progress. Intuitive User Interface: We created a user-friendly interface that clearly presents complex data and personalized recommendations in an easily digestible format. Rapid Prototyping: Despite time constraints, we produced a fully functional prototype that demonstrates the core capabilities of our concept. Tune AI Integration: We are proud of our connection to tune ai and using their llama ai model to provide insights into the patients movements.

What we learned

Full-Stack Development: We gained valuable experience in integrating frontend and backend technologies, particularly in using Python for backend operations and Angular for the frontend. Interdisciplinary Collaboration: We learned the importance of effective communication and teamwork when combining expertise from various fields (e.g., software development, hardware engineering, and physical therapy). Real-world Problem Solving: This experience reinforced the value of addressing genuine societal needs through innovative technological solutions.

What's next for Glucose

Enhanced Sensor Array: Integrate additional sensors (e.g., accelerometers, gyroscopes) for more comprehensive movement tracking and analysis. Machine Learning Integration: Implement more advanced ML algorithms to improve personalization and predictive capabilities of our advice engine. Clinical Trials: Conduct rigorous testing with physical therapists and patients to validate and refine our system's effectiveness. Mobile App Development: Create dedicated iOS and Android apps to increase accessibility and user engagement. Expanding Use Cases: Explore applications beyond athletic injuries, such as rehabilitation for stroke patients or elderly care. Another use case is to help correct diagnostic error. We will have a lot of data on how the patient moves and we can train machine learning models to then analyze that data and affirm the diagnostic that the doctor gave.