Inspiration
A friend of mine went hiking and was bitten by an insect. He thought nothing of it, nor the rash that appeared in a completely different location on his body from where he was bitten.
Later that week, he started complaining about headaches and muscle aches. He assumed he was tired. He was not. He had contracted Lyme disease.
The rash that appeared on his skin didn't have the characteristic bulls-eye shape that Lyme disease is known for. But what if we could identify different dangerous rashes, based of an image? Enter, machine learning.
What it does
We ask a user to take a photo or select a photo from the gallery on their smartphone, or somewhere on their computer. Rashional will then provide a confidence level for our four trained concepts. Regardless of outcome, we will strongly suggest that the user visit a real health practioner - and we provide the contact information and addresses of nearby free medical clinics.
How does it solve Access to Care
Access to Care is a challenge at this years Hackathon with an aim to nulify social and regional determinants from impacting ones health. Our mobile/desktop app, Rashional, tackles this problem by using a Machine Learning Algorithim to identify different skin rashes, the severity, and utilizes a Free Clinic API to provide the user with a list of the nearest free healthcare clinics available to them. The app helps to provide initial insight and direction to a medical problem, in a field in which there is not a readily available solution that is accessible through the internet and works via a single picture upload.
The app is also available in both English and Spanish so it can be accommodate a larger user base and provide Access to Care to a larger growing population of native English and Spanish speakers.
How we built it
We scraped the internet for hundreds of photos of various skin conditions, and focused on training our model for several dangerous skin conditions that are easily overlooked: Lyme Disease, Ringworm, Shingles, and Rocky Mountain Spotted Fever.
We used react.js to create a smooth and dynamic front end application that interfaces with our backend MongoDB Database and Machine Learning Software to provide the user with a seamless easy-to-use application.
Challenges we ran into
- React has a steep learning curve.
- Domain name service changes take more 24h to come into effect.
- Scraping the internet for valid pictures to train the model for and against
Accomplishments that we're proud of
We are able to accurately identify over 90% of confirmed cases of the 4 concepts we have trained in. We built our frontend with the React framework, despite nobody on our team having experience in the framework.
What we learned
Cheese pizza looks frighteningly similar to shingles and ringworm to our trained model. Artificial Intelligence isn't as frightening to use as many would believe. React.js has a very steep learning curve with poor documentation, however once installed and setup correctly React.js provides a very flexible framework in allowing you integrate numerous components of an application.
What's next for Rashional
We're hoping to implement a Spanish plugin, to eliminate the language barrier and provide easier Access to Care.
Log in or sign up for Devpost to join the conversation.