RxPlain

Prescriptions, Plainly

Comment

RxPlain

Prescriptions, Plainly Submission for MedHacks 2020; Track 2 - Personalized medicine using data driven healthcare

Inspiration

Health literacy is an issue for many patients for using their medications correctly/at all. The complicated details and instructions printed on the prescriptions and drug labels are hard to interpret and understand. Therefore, if a system exists that can translate this complex in a simple format, then the patients can be empowered into making educated decisions regarding their healthcare.

What it does

The app captures images of the prescriptions and drug labels and refers a database to display the information on the prescriptions in a simple terms. The RxPlain app displays information such as the purpose of the medication, the potential side effects/risks, and the intended dosage. It will also display a graphic that explains the drug working and metabolism in a simplified way.

How I built it

We built this using a Python backend and a Firebase web app that runs on html and Node.js. The Firebase web app allows the user to submit a prescription label of their choice for processing. This is processed by our backend, starting with pyteressact for the initial optical character recognition of the pictures. The output was fed into the med7 model, an information extraction model for clinical NLP trained on MIMIC-III free-text electronic health records, to extract the drug name and other information. This was then fed into the Drugs@FDA API to get a list of properties for the drug. The data is sent to the real time database for storage and retrieval.

We used Figma to create a demo "prototype" of the app, designing its initial layout.

Challenges I ran into

For the programming aspect, implementing the frontend was hard as none of us were familiar with html or Node.js, and even less with Firebase. We wanted to get the Firebase retreival working with a smooth user interface using CSS but ran out of time. For the other aspects, this was a first time hackathon for a lot of us, so brainstorming and coming up with a workable issue that we value was a challenge.

Accomplishments that I'm proud of

Irrespective of our limitations in the programming aspect, we were able to build the web app in a shot span of time that serves as a strong proof-of- concept of our idea. We are proud to be able to create an interface that is user friendly and generates a robust data visualization.

What I learned

We learnt about using Firebase to create a web app that is user friendly and is based on a robust proof-of-concept idea. We also learned about how to collaborate with a fully interdisciplinary team to create a workable product that benefits all sides of the problem.

What's next for RxPlain

Maybe learn a little more to optimize the frontend, then dig around in the backend for bug fixes and implement spell check for the NLP, and more streamlined updating of databases.

Built With

Share this project:

Updates