ApneaAid

Logo With Name
Logo

Inspiration

ApneaAid was inspired by a team member’s loved one who struggled for years with undiagnosed sleep apnea, a condition affecting roughly 1 billion people worldwide, with 80% of cases going undiagnosed. Left untreated, it increases the risk of heart disease, stroke, and early mortality, yet traditional sleep studies remain costly and inaccessible. Determined to make diagnosis easier, we built ApneaAid using Flutter and Dart for a seamless front end and TensorFlow for AI-driven real-time apnea detection. Our CNN model, trained on thousands of hours of sleep data, achieves 94% accuracy, using Mel spectrograms for efficient on-device processing. With secure cloud storage and optimized audio handling, ApneaAid delivers clinical-level accuracy without expensive hardware, making early detection more accessible than ever.

What it does

ApneaAid uses a smartphone’s microphone to monitor breathing patterns while people sleep, analyzing them to detect signs of sleep apnea. Our AI model, trained on thousands of hours of data, ensures highly accurate results without the need for expensive equipment or clinical visits. The app also includes history from all previous nights for the user categorized into low, medium, or high risk of sleep irregularities, a resources tab to help users with their health, and a settings page for easy customization.

How we built it

We built ApneaAid using Flutter and Dart for the front end, ensuring a smooth and intuitive user experience, while the backend was developed with Android Studio and TensorFlow to handle real-time audio processing and machine learning inference. Our AI model, a convolutional neural network (CNN) trained on thousands of hours of sleep data, was developed in Python using TensorFlow, allowing us to detect subtle breathing irregularities with 94% accuracy. To optimize performance, we implemented on-the-fly audio feature extraction using Mel spectrograms, reducing storage requirements and improving classification speed. The backend processes incoming audio data, converts it into spectrograms, and runs it through our trained CNN model to identify apnea episodes in real-time. We leveraged GitHub to facilitate seamless version control and collaboration. By combining advanced AI techniques with efficient mobile development, we created an app that delivers clinical-level accuracy and guidance without the need for expensive hardware or sleep studies.

Challenges we ran into

One of the biggest challenges we faced was managing our busy schedules. Juggling school, work, and this project required strong time management and coordination. Balancing these commitments while ensuring the app was high-quality took a lot of effort, but we pushed through and made it happen.

Accomplishments that we're proud of

We shared ApneaAid with sleep medicine researchers in the UNC and Duke health care system and they were impressed by its accuracy and potential to improve lives. They emphasized how this app could help people live a healthier life and take control. We're also proud that ApneaAid achieves a 94% accuracy rate, comparable to traditional sleep studies.

What we learned

Through our preliminary research, we discovered how widespread undiagnosed sleep apnea is and how limited access to sleep studies leaves many people unaware of serious health risks. We also learned a myriad of technical skills such as integrating AI models with mobile applications, real-time audio processing, and UI/UX design to create an intuitive and engaging user experience.

What's next for ApneaAid

We’re focused on expanding our reach and conducting larger trials to gather data and refine the user experience. These trials will help us improve the tool and ensure it meets the needs of more people. We’re also developing new features, including ‘Sleep Trends,’ which will track sleep patterns over time, making it easier to identify long-term changes. Another exciting feature is the ‘AI Sleep Coach,’ which will provide personalized recommendations based on individual sleep data to improve sleep quality. These updates will give users more insights into their sleep and make it easier to take action for better health.