Cerebrix

Inspiration

Dementia and Alzheimer's is always devastating for patients and their families. This project was inspired by the need to bring gentle cognitive stimulation and meaningful engagement to dementia and Alzheimer’s patients in assisted living facilities. We wanted to create a tool that helps patients connect with familiar memories and track cognitive changes, while empowering caretakers with insights into their patients' well-being. By providing adaptable, comforting games, we aim to support both patients and caregivers on this challenging journey.

What it does

Cerebrix has a set of brain-stimulating minigames crafted for dementia and Alzheimer's patients, designed to encourage mental engagement through familiar and accessible activities. The suite includes games like card memory matching, a recognition game where users match acquaintances to their names, and a math challenge. Each minigame adapts its difficulty based on the player’s recent performance, creating a personalized experience that stays challenging yet manageable. This adaptive design aims to support memory retention, recognition skills, and cognitive stimulation in a gentle, supportive way. It additionally contains a voice-to-voice AI companion that the player can talk to for social activity and emotional support, and a minigame to guess the musical instrument playing in a song.

A caretaker sets up and manages the patient’s account, personalizing it to enhance relevance and usability. They can also monitor cognitive changes over time, as the application tracks game performance and provides feedback on any progress or cognitive shifts. This feature enables caretakers to observe patterns and receive insights. By combining engaging gameplay with helpful progress tracking, this project aims to support both patients and caretakers in their experience with dementia care.

How we built it

We used React.js/Next.js and Vercel to build the application, using a Cloudflare Worker running Meta Llama 3 for generation of math problems, Clerk and Supabase for authentication and users, Spotify API for the musical instrument guessing game, and Hume AI for the voice-to-voice AI.

Tech stack

The cognitive progress or decline is measured by accuracy in the games over time, which is stored in the Supabase database.

Challenges we ran into

We were having trouble with getting the Hume AI WebSocket working, and also had lots of issues with Next.js debugging and other APIs. There was also a big issue with Vercel where they didn't let us update the deployment unless we're committing to the GitHub from the Vercel owner's account. We had to get around this by committing, and then recommitting from the owner's account.

Accomplishments that we're proud of

Getting the Hume AI's WebSocket working, getting the Spotify API working, and getting the Cloudflare Worker working were all big accomplishments. Getting our minigames working with the varying difficulty and reports of cognitive changes was also very important for the project, as it helps give more information to caretakers.

What we learned

We learned a lot about using React components for games, integrating real-time AI APIs with games, and using images with Supabase.

What's next for Cerebrix

We want to implement more games, have more reporting for caretakers, have more aspects to track to calculate the cognitive changes, and advertise the application to assisted living facilities.