Artificial intelligence is widely used in many sectors, including healthcare. The main goal of health-related AI applications is to investigate correlations between preventative or treatment methods and patient outcomes. During the analysis in healthcare, AI algorithms mimic human cognition in the analysis, presentation, and comprehension of complex medical and health care data.
Researchers at the University of Geneva have designed a new machine learning algorithm based on automated video analysis to study children’s nonverbal communication in an anonymous and standardized manner.
Autism is defined by nonverbal communication that is different from that of a normally growing child. It differs in various ways, including the difficulties in making eye contact, smiling, pointing to items, and the way in which they are engaged in their surroundings.
Communication and interaction problems are common in children with Autism Spectrum Disorder. Despite its prevalence, this disease is difficult to detect until the age of five.
To overcome this setback, the researchers developed an AI-based algorithm that analyses children’s movements on video and determines whether they indicate autism spectrum condition. The proposed algorithm classifies the given case solely based on the child’s movements when interacting with another person.
To achieve this, the researchers employed OpenPose technology, which separates the skeletons of moving humans from video and enables the analysis of movements by removing any distinguishing traits.
The algorithm was then tweaked to detect Autism. They put it to the test on 68 normally developing children and 68 children with Autism, all of whom were under the age of five. Each group was further divided into two groups. The first 34 children helped train the algorithm, while the following 34 children helped them test its accuracy.
According to the findings, the AI correctly classifies Autism in more than 80% of cases. In ten minutes, the team was able to acquire a preliminary screening that is available to everybody, regardless of where they live. This would allow parents concerned about their young children to get an automated assessment of their child’s autism symptoms.
Installation of movement sensors takes time and is delicate; it can also annoy children and affect the results. The computer vision-based analysis used in this case is non-invasive, and there is no need for the child to be directly involved in the algorithm. Moreover, because it does not require any unique setup, the algorithm may be used to evaluate previously recorded videos, which significantly benefits future research scope.
The team now aims to make this AI technology accessible to everyone. They plan to create software that would enable such analysis using only a 10-minute video shot on a smartphone.