In many areas, there is a dire shortage of COVID-19 nasal swab test kits. Many people are left untested, putting them at a huge risk. Even in areas where people can get access to test kits, very few provide immediate results. With a testing backlog in many parts of the US, patients get results in anywhere between 24 hours or to 2 weeks. We were inspired to create a COVID-19 test alternative to help patients gain access to an expedited diagnosis.

What it does

Previous SARS and COVID research has shown that by using a chest x-ray, you can identify through the lungs if a patient has contracted the virus or not. Because x-ray machines are commonly available at every hospital or clinic, we built RADIAN, a machine learning software that helps radiologists make a highly accurate prediction on whether or not a patient has COVID-19. RADIAN uses a simple web interface where healthcare providers (or patients alike) can upload a chest x-ray and get immediate results.

How we built it

We used python and tensorflow to build the convolutional neural net and trained the model with chest X-rays of different diseases from a database from the NIH. We then used HTML5 and JavaScript to build and design a user interface to upload x-ray images.

Challenges we ran into

We had trouble finding a server to properly run our model, since it was fairly large. We decided to host the server locally and eventually scale later.

To test it out, please download from our GitHub!

Accomplishments that we're proud of

We are proud of creating a model with 90% accuracy. Although real-life practical use is questionable for now, we are glad to have made steps in creating a model that hospitals can utilize to cheaply diagnosis COVID-19.

What's next for RADIAN

We would love to see RADIAN deployed at local hospitals or clinics, especially those with long wait lines for nasal test kits. While we understand that many healthcare providers would still want to perform a nasal test to confirm our results, RADIAN will be useful in identifying high risk patients.

In the long run, RADIAN can be expanded to assist radiologists and other medical professionals in identifying all different kinds of diseases, with the ability to point out anomalies that are easily missed.

Built With

Share this project: