This week for Northern Exposure, we welcomed Dr. Ziteng Wang, an associate professor in Industrial and Systems Engineering, for his first in-person presentation at NIU covering AI-powered, video-based gait analysis to transform autism screening.

Dr. Wang began by introducing some of the statistics and current screening processes for atisum. We learned that 1 out of 31 children in the U.S., less than eight years old, has autism, and the current screening process is a subjective series of evaluations and tests that may lead to a diagnosis. Parent roles in early childhood screening can be plagued with delays, referrals, coverage, and evaluations, which can all be affected by demographics of the family, such as gender, race, and socioeconomic conditions. From these factors, proven to affect the screening process for children, Dr. Wang and the team of researchers working in the AI-Lab for Early Autism Diagnosis Screening (AI-LEADS) argue for objective, culturally responsive, timely, and equitable assessments for autism.

The goal of the research Dr. Wang presented is to establish gait, the specific patterns of limb movements during walking, as a validated, objective signal for autism screening that is translated into a scalable web-based tool that can be easily accessible. Their hypothesis is that using 2D video-based gait features can reliably distinguish autism, and that embedding these features in ML models within the web app will yield an objective, accessible, scalable screener tool for autism. Additionally, Dr. Wang explained the methods used in iGait, a web-based app for home recordings submitted for analysis using computer vision and AI to identify established posture and gait features. The use and assessment through iGait are private, the data is secure, and it is free to use.

For a portion of Dr. Wang’s talk, he engaged us in a fun “test” to see whether, based on the skeletal walking data from 10 subjects, we could deduce whether an individual has autism. After viewing the 10 videos and filling out the forms, we also answered which aspects we focused on when determining whether an individual was autistic, including arm, hip, and knee movements, as well as step distance and regularity. As a result, Jordan York identified 8 of 10 individuals, winning our little challenge provided by Dr. Wang.

All in all, Dr. Wang gave a very engaging presentation on the work he and many others are actively pursuing. For further development of their work, they hope to collect additional data from more participants and conduct upper-body data analysis to raise their 70% screening detection rate to 80% or more, making it an effective screening platform for autism. Dr. Wang wanted to emphasize the fact that iGait is not a diagnosis for autism, but rather a free screening tool to support further dialogue for parents about their children’s possible autism with physicians and pediatricians.