We wanted to make something that can really make a difference and potentially change the lives of people. We wanted to focus on the prospects of the applications of machine learning to the medical industry, and provide a convenient and accurate model that can detect signs of Parkinson's disease in a user.
What it does
The application tests the user's motor skills through a simple drawing application, as well as analyzing patterns in the user's speech to compare against the results of a data set of those with and without Parkinson's disease.
How I built it
The application works from two different data sets that were use to train and test 3 different machine learning models to identify the user's likelihood of having symptoms of Parkinson's disease. The data sets contained information on how individuals with Parkinson's fared in a task to test their motor skills, as well as data outlining the patterns of their speech. The majority of the program was built using React and Node.js, as well as Python in order to train the TenserFlow machine learning models.
Challenges I ran into
A challenge the team ran into when developing the application was interfacing between the trained models and the user input, as each had specific requirements for how data was to be modeled, input, and formatted. Another issue was interfacing with the Google Cloud in order to host the application online, as the limitations on processing power and storage required some workarounds.
Accomplishments that I'm proud of
We are proud of our application of machine learning technology into an idea with realistic benefit and meaningful experience when applied to real world situations. The use of computer programming and machine learning in medical applications is continuing to advance, and we are proud to create something as a part of it.
What I learned
We learned that when tackling projects such as this, it is important to carefully prioritize tasks and ideas as well as work and activity, so as to not rush or fall behind on a realistic schedule and maintain composure throughout.
What's next for ParkinSun'sLife
The design of the project allows for endless iteration and improvement, as the machine learning models can be trained further to yield better and more accurate results. We hope to continue iterating and improving the project in the future.