Is Smiling the Key? Machine Learning Analytics Detect Subtle Patterns in Micro-Expressions of Infants with ASD
Gianpaolo Alvari1,2, Cesare Furlanello3,4, Paola Venuti1
1Department of Psychology and Cognitive Sciences, University of Trento, 38068 Rovereto, Italy.
Journal of Clinical Medicine
|April 30, 2021
Summary
Early detection of Autism Spectrum Disorder (ASD) is vital. This study used AI to analyze infant facial expressions, finding reduced social smiles in autistic infants aged 6-12 months, aiding early screening.
Area of Science:
- Developmental Psychology
- Computational Neuroscience
- Pediatric Neurology
Background:
- Early detection of Autism Spectrum Disorder (ASD) is critical for successful intervention.
- Identifying reliable early markers for ASD manifestation remains challenging.
- Artificial intelligence (AI) offers promising avenues for behavioral screening.
Purpose of the Study:
- To investigate facial expressions in infants aged 6-12 months.
- To identify early behavioral markers for ASD using AI.
- To assess the potential of AI in analyzing social smile dynamics.
Main Methods:
- Analysis of facial micro-movements in home videos of 18 autistic and 15 typical infants (6-12 months).
- Utilized Openface, an AI-based software for systematic facial expression analysis.
- Employed Machine Learning models to map facial behavior and detect subtle differences.
Main Results:
- Autistic infants exhibited a reduced frequency and activation intensity of social smiles compared to typical infants.
- AI-powered analysis revealed subtle early differences in facial behavior.
- Machine learning models successfully mapped these early behavioral distinctions.
Conclusions:
- AI tools like Openface can detect early signs of ASD through facial expression analysis.
- Reduced social smile dynamics in infancy may serve as an early marker for ASD.
- AI holds significant potential as a supportive tool in clinical ASD diagnosis and screening.
