Autism Spectrum Disorder (ASD) is increasingly prevalent [1], but long waitlists hinder children’s access to expedient diagnosis and treatment. To begin addressing this problem, we developed an automated system to detect ASD using acoustic and text features drawn from short, unstructured conversations with naïve conversation partners (confederates). Seventy children (35 with ASD and 35 typically developing (TD)) discussed a range of generic topics (e.g., pets, family, hobbies, and sports) with confederates for approximately 5 minutes. A total of 624 features (352 acoustic + 272 text) were incorporated into a Gradient Boosting Model. To reduce dimensionality and avoid overfitting, we dropped insignificant features and applied feature reduction using Principal Component Analysis. Our final model was accurate substantially above chance levels. Predictive features were both acoustic-phonetic and lexical, from both participants and confederates. The goal of this project is to develop an automatic detection system for ASD that relies on very brief, generic, and natural conversations, which can eventually be used for ASD prescreening and triage in real-world settings such as doctor’s offices and schools.
Automatic Detection of Autism Spectrum Disorder in Children Using Acoustic and Text Features from Brief Natural Conversations
Background: The male bias in autism spectrum disorder (ASD) diagnoses is well documented. As a result, less is known about the female ASD phenotype. Recent research suggests that conclusions drawn from predominantly male samples may not accurately capture female behavior. In this study, we explore potential sex differences in attention to social stimuli, which is generally reported to be diminished in ASD. Population-based sex differences in attention to faces have been reported, such that typically developing (TD) females attend more to social stimuli (including faces) from infancy through adulthood than TD males. It is yet unknown whether population-based sex differences in the face domain are preserved in ASD.
Methods: A dynamic, naturalistic infrared eye-tracking paradigm measured attention to social stimuli (faces) in 74 school-aged males and females with ASD (male N = 23; female N = 19) and without ASD (male N = 16; female N = 16). Two kinds of video stimuli were presented that varied in social content: rich social scenes (dyadic play between two children) and lean social scenes (parallel play by two children).
Results: Results revealed a significant 3-way interaction between sex, diagnosis, and condition after controlling for chronological and mental age. ASD females attended more to faces than ASD males in the socially lean condition. This effect was not found in the typically developing (TD) group. ASD males attended less to faces regardless of social context; however, ASD females only attended significantly less to faces compared to TD females in the socially rich condition. TD males and ASD females did not differ in their attention to faces in either condition.
Conclusions: This study has implications for how the field understands core social deficits in children with ASD, which should ideally be benchmarked against same-sex peers (male and female). Social attention in ASD females fell on a continuum-greater than their ASD male peers, but not as great as TD females. Overall, their social attention mirrored that of TD males. Improved understanding of the female social phenotype in ASD will enhance early screening and diagnostic efforts and will guide the development of sex-sensitive experimental paradigms and social interventions.
Keywords: Autism spectrum disorder; Eye gaze; Sex differences; Social attention; Social cognition.