Heart AI: Distributed Health Services

Inspiration

In 2019, the first alleged instance of a death due to a cyberattack occurred in Springhill Medical Center. We aim to create tools for healthcare professionals that are resilient against cyberattacks, so that preventable deaths stay prevented.

What it does

Our aim is to create reliable, scalable, redundant applications and infrastructure for healthcare facilities. We began by taking a look at the number one cause of death in the United States, heart disease, and creating an easily deployable, easy to use AI-driven application that can predict the chances of an individual experiencing a heart attack with a few pieces of key information.

How we built it

Model was built using Tensorflow.js, with the training code and demo code written in Node.js.

Challenges we ran into

Data processing, JavaScript funny business, scheduling conflicts and a tight turnaround.

Accomplishments that we're proud of

Model achieves 70% accuracy with test data.

What we learned

When performing normalization on a matrix, make sure the data type is set to float and not int.

What's next for Heart AI: Distributed Health Services

• Tools for other diagnostic tasks (blood tests, mental illnesses, etc.)
• Web application UI, Mobile application UI
• Fine-tuning of prediction model to produce more accurate results

Built With

• javascript
• node.js
• tensorflow.js