It is important to detect concussions promptly. Much too often those that are affected go on doing what they were doing, blissfully unaware of being impaired and in serious danger for other injuries and oncoming symptoms of the brain trauma. Now a team at University of Washington has created a smartphone app that tracks the movement of the pupils to identify whether someone may or may not be concussed.
The PupilScreen app measures the eye’s pupillary light reflex, or how the pupil changes its shape in response to changes in the light striking the eye. The person being assessed simply looks at the phone, the phone produces a camera flash and records the person’s face, after which deep learning algorithms process the captured video, measuring the increase in the size of the pupil following the flash. Currently, a plastic box is used to place the smartphone into that blocks out surrounding light in order to improve the accuracy of the app, but the team believes it will be possible to eventually to alleviate this necessity and the investigators are already teaching the app to work in an environment with surrounding ambient lighting.
The algorithm that accurately differentiates between the pupil and surrounding areas of the eye relies on deep learning and a lot of previously expanded manual labor. Specifically, the team went through about 4,000 eye images, annotating them so that the app can learn to understand what the pupil is and how to outline it.
An initial review of the app demonstrated it being able to spot significant traumatic brain injury, but a larger study involving EMTs, coaches, and doctors will help to assess the real abilities of PupilScreen.
Here’s a University of Washington video about the new app:
Link to the white paper: PupilScreen: Using Smartphones to Assess Traumatic Brain Injury…