ACE | Attention Cognitive Evaluation

Blink time graph
Head orientation graph

Inspiration

Our inspiration came from one of our teammates who has ADHD, and our collective passion for technology and biology to explore how data can help understand human behaviour. This connection inspired us to create a tool that makes behavioural assessment more engaging and accessible.

What it does

The app provides children with mini-game activities that include following a star while capturing eye tracking data, blinks per second, and head movements. These analytics can later be useful to help doctors with diagnosis of neurocognitive conditions including ADHD.

How we built it

We built the app by researching facial biological markers of ADHD and other neurocognitive disorders. We then used OpenCV to capture such markers. We used a CSS, HTML and JavaScript front end for the electron app and python for the backend which includes counting times the head was moved and blinks made. Then, a custom algorithm is applied to denoise the data created by the backend for doctors to view.

Challenges we ran into

A major challenge in our process was finding a proper way to use eye tracking without an external device. After long searches through the internet we found a python library that was made using OpenCV called Eyetrax. Although the eye tracking from this library was better than most, it was still inaccurate. One of the mentors suggested the idea of a finding the closest element to the gaze estimation, which we implemented to receive a more accurate understanding of where the patient is looking. However this eye tracking technology would be significantly improved with external components.

Accomplishments that we're proud of

We are proud of combining the various markers from OpenCV through our custom algorithm to organize the data to a frontend. It was harder then expected and we had to overcome several errors.

What we learned

After building this project we learned a more about OpenCV and how it works to calculate the three markers we focused on. One thing we learned was plotting points on a face mesh and using threshold equations to calculate the Eye Aspect Ratio (EAR).