Inspiration

HealthDex wants to move all paper records to digital ones in the easiest way possible.

In total, the team has spent over 500 hours volunteering at various hospitals across Canada over the past few years. We know that paper records are a big problem across hospitals because we are the ones who were responsible for organizing, preserving, and maintaining them. The process is extremely cumbersome for everyone involved and expensive.

What it does

HealthDex solves this problem by not only creating a health records platform to store data, but enabling a simple way to translate files into digital records using machine learning.

How we built it

In order to do this, we leveraged the incredible power of Google Cloud to create Healthdex. PDF files are uploaded to Firebase Store, and are accessed by Google’s Document AI. We then use text recognition and intent detection to parse form values, handwriting, and marks into mutable JSON objects. This is then sent to Firebase FireStore, where a healthcare professional can then search through the data (sorted by each patient) easily through a gorgeous frontend made using React.

Challenges we ran into

Form inputs, especially handwriting, are often parsed into non-ASCII characters, which cause many problems downstream in the stack. This was solved using regex filters. Thanks to our mentors from Google for their help!

Accomplishments that we're proud of

A clean, simple to use UI that does its job well. We’re proud to have leveraged document parsing ML to bring value and convenience to medical professionals.

What we learned

It’s really important to build from the user’s perspective. We learned that there are huge inefficiencies in the medical industry, and that it is hard to display verbose data in a clean interface.

What's next for HealthDex

The team is really excited to continue working on the project. Feature-wise, the goal is to implement an ML model that will categorize documents for each patient, which can help reduce clutter. Exporting digital profiles to PDFs, and enabling an option to create private records for each patient are also important. We plan on integrating with existing professional physician associations, and streamline the information sharing within the departmental, national, and international medical community.

Built With

Share this project:

Updates

posted an update

If you want to see how medical records are displayed, look for the ID 234567! "Yash Dani" also works. We didn't have time to really make our search robust, but that's an easy fix with more time.

Log in or sign up for Devpost to join the conversation.