neuroaid-cognitive-compass

"NeuroAid – Catch the signs before the silence."

Comment

📖 About the Project

🎯 Inspiration

We were inspired by the growing number of families affected by dementia and Alzheimer’s — often realizing it too late. With elderly populations rising globally and limited access to early diagnostics, we saw a major gap in cognitive health monitoring. What if there was a friendly AI companion that could screen for mental decline during everyday conversations and activities — long before symptoms became critical?

That’s how NeuroAid was born — an AI copilot for cognitive wellness.

🧠 What We Learned

  • How subtle linguistic, behavioral, and vocal patterns can reveal early signs of cognitive disorders.
  • How to integrate multiple AI modalities — speech, NLP, and behavioral modeling — into one seamless experience.
  • The importance of building accessible tools, especially for users with low digital literacy or disabilities.
  • How to balance privacy and data sensitivity in health-related applications.

🛠️ How We Built It

  • Frontend: A simple and accessible UI using Streamlit for fast iteration.
  • Cognitive Tests: We designed interactive tasks inspired by established neuropsychological screening tools (like MMSE and MoCA) and gamified them.
  • Speech Analysis: Integrated OpenAI Whisper and Perplexity Sonar API to transcribe and extract features like pause duration, hesitation, and fluency.
  • ML Models: Used Python and Scikit-learn to score cognitive performance and flag risk indicators.
  • LLM Integration: GPT-4 was used as a conversational agent to simulate natural interactions and prompt cognitive engagement.
  • Dashboard: Plotted user performance over time using Matplotlib and Streamlit’s native components to visualize risk trends.

🚧 Challenges We Faced

  • Speech data variability: Different accents, ambient noise, and senior users’ speech patterns made audio parsing tricky.
  • Balancing depth with usability: We had to simplify tests and dashboards so they remained senior-friendly while still capturing useful data.
  • LLM alignment: Ensuring that the GPT-powered assistant stayed on-topic and safe, without overstepping its role.
  • Data sensitivity: Designing the platform in a way that would later support HIPAA/GDPR compliance — even in a prototype phase.

🌟 What’s Next?

We aim to:

  • Expand our dataset with real cognitive assessments for clinical validation.
  • Develop a mobile-friendly version with voice-first interaction.
  • Partner with eldercare providers for pilot testing.
  • Add multilingual support for global reach.

With NeuroAid, we believe we can shift cognitive health from reactive to proactive — empowering individuals and families to act early, with clarity and confidence.

Built With

Share this project:

Updates