Inspiration
We were inspired by the global health inequity in respiratory disease diagnosis. Millions in low-resource and remote areas lack access to doctors and diagnostic tools, leading to late detection of pneumonia, tuberculosis, and asthma. We realized that while advanced medical equipment is scarce, nearly everyone has a microphone on their phone.
What it does
Cough2Care transforms any microphone into a preliminary respiratory screening tool. Users record their cough, and our AI model analyzes acoustic patterns to assess respiratory health risk. The system flags individuals who may need medical attention, helping healthcare workers prioritize cases and enabling early intervention before conditions worsen.
How we built it
We built a deep learning model trained on cough sound datasets, focusing on acoustic features that correlate with respiratory conditions. We developed a user-friendly interface for easy audio recording and implemented backend processing for real-time analysis. The system outputs a risk score with recommendations for medical follow-up when necessary.
Challenges we ran into
Obtaining diverse, quality-labeled cough audio data was difficult. We also faced challenges in balancing model sensitivity and specificity to minimize false positives while catching genuine cases. Ensuring the system works across different microphone qualities and background noise levels required extensive testing and audio preprocessing techniques.
Accomplishments that we're proud of
We created a functional prototype that accurately identifies respiratory risk patterns from cough sounds. We achieved a system that's truly accessible, requiring only a basic microphone and working offline after initial setup. Most importantly, we developed something that could genuinely save lives in underserved communities.
What we learned
We learned the complexities of audio signal processing and the importance of data quality in medical AI applications. We gained insights into responsible AI development for healthcare, understanding the ethical considerations of building screening tools. We also learned how critical user experience is when designing health technology for diverse populations.
What's next for Cough2Care
We plan to expand our training dataset with more diverse samples across different demographics and respiratory conditions. We'll pursue clinical validation studies and partnerships with healthcare organizations in underserved regions. We also aim to develop a mobile app with multilingual support and integrate telemedicine features to connect flagged users directly with healthcare providers.
Built With python, react, pytorch, numpy, pandas
Log in or sign up for Devpost to join the conversation.