Using Computers to Count Eye Tics

Community Summary of: Automated Quantification of Eye Tics Using Computer Vision and Deep Learning Techniques 

Authors of the article in Movement Disorders: Official Journal of the International Parkinson and Movement Disorder Society: Christine Conelea PhD, Hengyue Liang MS, Megan DuBois BA, Brittany Raab BA, Mia Kellman BS, Brianna Wellen PhD, Suma Jacob MD, PhD, Sonya Wang MD, Ju Sun PhD, Kelvin Lim MD

Published online: 25 December 2023

Community summary posted on December 3rd, 2024

Word count: 510 (4 min)

Reading level: 8th grade

Full Article here

Why does this study matter?

Counting tics is important for diagnosing and treating Tourette Syndrome (TS). A diagnosis can help people with TS get support. Doctors usually ask patients to make their best guess about how many tics they are having. This way of doing things can cause mistakes because tics can be hard to count. Some people use videos to count tics, but this way of doing things takes a lot of time and people can still make mistakes. Multiple studies show that when people report their own tic symptoms they are different from observed tic symptoms. Computers might help find atypical movements (tics) and they might be more accurate. Our research team wanted to know if a computer could be trained to find eye tics, the most common type of tics.

What happened?

Our research team made a computer program that can find and organize human motion in videos. They did this using steps that have been used in other studies. 

Youth with TS were recruited and consented to participate. The youth sat down in front of a camera and a laptop. Researchers left the room. The youth followed the instructions to press record and "for the next three minutes let your tics happen naturally." Then they followed instructions "for the last three minutes, try to stop your tics.” Our team made a computer program to tell the difference between eye tics and other eye movements in these videos. The videos were coded by trained humans and analyzed using deep learning. 

This means that knowledgeable staff people watched the videos and created codes (names and numbers) for all the types of tics. Then, the video images were transformed into numbers. Next, a computer program was trained by these numbers to find and organize the human movements from the videos, such as “eye blinking tics” vs. “non-tic eye blinking”. Finally, the software program was tested for accuracy compared to what the humans saw in the first step.  

Our research team received information from 54 sessions and 11 participants aged 13-19 years old. 

What did our research team learn?

The computer program was successful in finding eye tics in the videos that were  unseen validation sets. This means that the program found tics in videos that it hadn’t seen in the training process. This suggests that computer programs can be used to tell tic from non-tic movements in people with TS. This might help more doctors and therapists screen, diagnose, and help people with TS.

What can our research team do next?

Our research team is continuing to develop these tools to better measure tics in people with TS. We have a grant from the National Institute of Neurological Disorders to do a larger study. Our hope is to develop an easy-to-use computer program that can help doctors and therapists measure tics more accurately and consistently, especially doctors and therapists who don’t see people with tics very often. This kind of tool could help with diagnosis and measuring how well a treatment for tics is working. In the future, this research might also help create brand new treatments that we haven’t thought of yet to help people manage their tics.

How can you find out more?

Please see the full article here