Inspiration
More than 10 million people live with Parkinson's worldwide. It's the world's fastest-growing neurodegenerative disease and early detection significantly improves treatment options. Clinical tests are short term, but what if there was a solution that monitors behavior over a period of time? In our digital age, many are constantly typing away on their keyboard, and we wanted to use this data to predict early stages of Parkinson's.
What it does
Cadence monitors your keystroke patterns in the background and uses machine learning to detect early stages of Parkinson's. This is particularly useful for the elderly who are at a higher risk of Parkinson's.
How we built it
We used Figma make and converted it to Typescript for the frontend and FastAPI for the backend. We also leveraged Electron as the desktop wrapper.
Challenges we ran into
Our dataset had more non-Parkison's user than Parkinson's users and the model initially always predicted that the user didn't have Parkinson's. To overcome this, we increased the penalty when the model predicted that the user didn't have Parkinson's when they actually do.
Accomplishments that we're proud of
We created a model with a 74% accuracy and implemented this in a user-friendly application to target a pressing medical problem.
What we learned
Through this experience, we learned how to quickly ship a desktop application.
What's next for Cadence
We plan to open-source this technology to allow the general public to use our application.
Built With
- electron
- fastapi
- figma
- python
- react
- typescript
Log in or sign up for Devpost to join the conversation.