Abstract
Clinical research in autism has recently witnessed promising digital phenotyping results, mainly focused on single feature extraction, such as gaze, head turn on name-calling or visual tracking of the moving object. The main drawback of these studies is the focus on relatively isolated behaviors elicited by largely controlled prompts. We recognize that while the diagnosis process understands the indexing of the specific behaviors, ASD also comes with broad impairments that often transcend single behavioral acts. For instance, the atypical nonverbal behaviors manifest through global patterns of atypical postures and movements, fewer gestures used and often decoupled from visual contact, facial affect, speech. Here, we tested the hypothesis that a deep neural network trained on the non-verbal aspects of social interaction can effectively differentiate between children with ASD and their typically developing peers. Our model achieves an accuracy of 80.9% (F1 score: 0.818; precision: 0.784; recall: 0.854) with the prediction probability positively correlated to the overall level of symptoms of autism in social affect and repetitive and restricted behaviors domain. Provided the non-invasive and affordable nature of computer vision, our approach carries reasonable promises that a reliable machine-learning-based ASD screening may become a reality not too far in the future.
Highlights
Clinical research in autism has recently witnessed promising digital phenotyping results, mainly focused on single feature extraction, such as gaze, head turn on name-calling or visual tracking of the moving object
Previous studies have demonstrated a linear relationship between age at diagnosis and cognitive gain[6,7], whereby children diagnosed before the age of two years can gain up to 20 points in intellectual quotient (IQ) on average over the first year following diagnosis, while children diagnosed after the age of four will not show any substantial cognitive gain even with adequate intervention[7]
The retained model was trained over the Training Set videos (68 ADOS videos, split between Autism spectrum disorders (ASD) and typically developing (TD) groups; see Table S1) that contained solely skeletal information on the black background, without
Summary
Clinical research in autism has recently witnessed promising digital phenotyping results, mainly focused on single feature extraction, such as gaze, head turn on name-calling or visual tracking of the moving object. The combined use of motion capture and CV have provided insights on (1) the atypical midline postural control in autism[11,15], (2) highly variable gait patterns in ASD16 and (3) unique spatio-temporal dynamics of gestures in girls with A SD17 that has not been highlighted in standard clinical assessments These studies demonstrate how computer vision and machine learning technologies can advance the understanding of autism, as they have the potential to provide precise characterizations of complex behavioral phenotypes. The authors used motion tracking to measure the approach and avoidance behaviors and the directedness of children’s facial affect during the diagnostic assessment—the Autism Diagnosis Observation Schedule (ADOS,[19,20]) With these objective measures, the authors accounted for 30% of the standardized scores measuring the severity of autistic symptoms from only 5-min excerpts of the free play interaction with the examiner. There is a critical need to take a more holistic stance to tackle the complex task of measuring how the child with autism interacts socially in settings close to everyday situations to advance towards a fully ecological and scalable approach
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
