Inspiration
The inspiration for OstoGuard came from listening to the overlooked voices of people living with colostomies—individuals who live each day with a silent fear: the fear of leaking. We learned that even with premium bags on the market, many users experience unexpected leaks that lead to embarrassment, skin infections, and social withdrawal. Worse yet, high-quality colostomy solutions are often financially out of reach, especially in underserved areas. We asked ourselves: what if we could combine accessible AI and low-cost sensor technology to bring peace of mind, dignity, and confidence back to these individuals? OstoGuard was born from that question and the deep belief that no one should have to suffer in silence due to an invisible yet solvable problem.
What it does
OstoGuard is an AI-powered prototype system that simulates sensor data readings for pressure, moisture, and chemical levels inside colostomy and ileostomy bags. Using a lightweight machine learning model trained on these simulated readings, it predicts potential leaks in real time. This early warning system lays the groundwork for a future device that can discreetly alert users before leaks occur, helping them avoid discomfort and maintain quality of life.
How we built it
We developed a modular simulation environment that generates sensor data representing realistic pressure, moisture, and chemical changes. We trained a simple but effective machine learning model to detect patterns associated with leaks based on this data. The codebase includes scripts for data generation, model training, and leak prediction, enabling easy iteration and extension. This approach allows us to experiment and refine leak detection algorithms before integrating them with physical sensors.
Challenges we ran into
Developing a reliable simulation that realistically mimics sensor behavior for pressure, moisture, and chemical levels was a significant challenge due to the variability of real-world leak conditions. Balancing the machine learning model’s sensitivity to detect leaks early while minimizing false alarms required iterative tuning and careful feature selection. Additionally, designing a modular codebase that could easily integrate with actual hardware and scale for future development demanded thoughtful architectural planning.
Accomplishments that we're proud of
We successfully built an end-to-end prototype that simulates sensor input, trains a leak detection model, and predicts leaks with measurable confidence. The project demonstrates the feasibility of AI-powered leak detection using basic sensor data and establishes a strong foundation for transitioning toward hardware integration. Our modular design supports easy updates and improvement, making this a solid starting point for further research and development.
What we learned
This project deepened our understanding of integrating simulated sensor data with machine learning for predictive maintenance applications. We learned the importance of crafting robust feature representations to improve model accuracy in environments with noisy or incomplete data. Moreover, we gained valuable experience in designing modular, scalable code that anticipates future hardware integration. Beyond the technical skills, the project underscored the critical role of user-centered design principles in developing solutions that are not only effective but also practical and accessible.
What’s next for OstoGuard
Moving forward, we plan to incorporate real sensor data from prototype hardware units and improve the AI model’s accuracy and robustness. We aim to enhance the simulation with more realistic environmental factors and user scenarios. Further steps include developing an alert interface and exploring partnerships for clinical testing and deployment. Our vision is to evolve OstoGuard into a fully integrated, affordable leak detection system that empowers users worldwide.
Log in or sign up for Devpost to join the conversation.