Inspiration

The inspiration for our project was to improve the state of healthcare within emergency rooms. By helping to streamline the process of triage within these emergency rooms, we hope to eliminate human error that could result from the high-stress environment of a hospital.

What it does

The application itself takes historical patient data that is aggregated by a hospital emergency center over time and uses the information collected from patients to predict the maximum amount of time that can be allowed before a new patient's condition exacerbates beyond control. Furthermore, it updates an ongoing queue of patients based on this value in order to help guide emergency response teams accordingly in the most efficient way possible.

How we built it

The core machine learning regression algorithm was actually built from scratch. The algorithm centers around a k-Nearest Neighbors process that enables us to liken new patients to old patients and proceed accordingly. Each of the devices that run the application post queries to the Flask server with new patient information. The application interface was built using Swift and other iOS development kits in order to enable triage nurses to interact with the application efficiently. Finally, the Flask server sends information back to each of the connected devices and updates the queues for all of the devices.

Challenges we ran into

The main problem that we ran into was reading JSON files from each of the devices and parsing through the JSON file for information that we wanted to use. This required effective communication between the team members working on the backend of the application and the frontend of the application to ensure that our methodology was as efficient and effective as possible.

Accomplishments that we're proud of

The accomplishment that we were most proud of us was the development of the k-Nearest Neighbors algorithm entirely from scratch. We did not use any machine learning libraries or pre-written algorithms in developing the process which we viewed as a central accomplishment in our application.

What we learned

In order to develop the application, we had to learn Flask and integrate our knowledge of Python with our newly acquired understanding of the Flask interface. Furthermore, we had to learn about machine learning principles that guided the kNN regression model of the application.

What's next for TriageAI

The next step for TriageAI will be to better integrate connected devices by enabling nurses on individual devices to update the status of patients manually which enables overriding of the machine-produced queue in the event of unexpected changes in a patient's condition.

Built With

  • python-swift-flask-xcode
Share this project:

Updates