Call Me - Parkinsons

The Problem

Estimated 10 million people suffering with Parkinson’s There are currently no treatments that can stop or slow Parkinson’s disease Catching it early and quickly can improve outcome and quality of life Currently the process of being diagnosed is very slow

Our Solution

Patients request a call through our online form krupa-17.github.io Patients reply to a short automated phone call - powered by Nexmo The sound file is processed using parselmouth.praat - a python library returning phonetic features We train a model using KNN from data supplied by University of California, Irvine - producing an accuracy of 70% We classify the patients speech against this model, sending the diagnosis and likelihood to their doctor

The Future

This technology could be used for early diagnosis of parkinson's disease, giving the potential for treatments to be started sooner. This can be rolled out on other devices to a mobile phone, such as an Amazon Alexa skill or Web Application - further increasing accessibility and user base.

Share this project:

Updates