ParkInsight

Could you have Parkinson’s 5,10 or 20 years down the line? Do you need to see your doctor today? Get your answers from ParkInsight, a simple phone based AI powered tool. Backed by medical research

Comment

Inspiration

Parkinson’s affects around 10 million people worldwide. Symptoms are often attributed to ageing. People don’t go to the clinic until it’s too late. Till then 60-70% of the neurons are dead.

Earlier identification and intervention can prove extremely helpful in prolonging a good quality of life. Though Parkinson’s cannot be reversed, speech therapies, exercise, brain focused diet etc can help extend a better quality of life when the disease worsens.

Hence, we questioned, what if there’s a tool that could help people screen whether they need to go to the doctor or not, based on the way they move, the way they talk, and the way they tap? What if, there’s a research and ML backed tool, that could give you explainable data analysis powered by ML, on whether your symptoms are to be worried about. Hence, we built ParkInsight.

What it does

Need to have smartphone and laptop. Smartphone uses a web UI to conduct three research backed tests :

  1. GAIT test: you walk for 30 seconds with your phone in the pocket. The data gets sent to the laptop for analysis and ML model inference.
  2. Tap test: you tap on your phone with your left and right thumb alternatively, as quickly as possible.
  3. Voice test: you say “Aaaaaaaah” to your phones’s microphone for 5 seconds. Data gets sent to the laptop again for ML model inference.

After the three tests are complete, the dashboard on the laptop gives out SHAP based explainable analysis on which features lead to the predictions of the model. There are ML models for the GAIT and the voice test, and there’s a threshold based analysis for the tap test.

How we built it

We used python and flask for backend, react and

Challenges we ran into

We struggled with accessing the device motion API (which is provided free by every browser) since that API can only be accessed on HTTPS, and we were trying on HTTP.

What we learned

We learned a lot about the science behind Parkinson’s. We also learned how early intervention can possibly prolong a better quality of life for people. We did our part to give people a science backed tool that could give them insights on their symptoms, and potentially motivate them to go to the clinic if any action is required.

Built With

Share this project:

Updates