Presentation
Search Abstracts | Symposia | Slide Sessions | Poster Sessions | Lightning Talks
Differences in Neural Encoding of Speech Emerged in Preschool Years in Children Speaking a Tone Language
Poster E65 in Poster Session E, Thursday, October 26, 10:15 am - 12:00 pm CEST, Espace Vieux-Port
Hoyee W. Hirai1, Eric C.H. Poon2, Eric C.L. Lai3, Carol K.S. To4, Patrick C.M. Wong1,2; 1Brain and Mind Institute, The Chinese University of Hong Kong, Shatin, Hong Kong SAR, China, 2Department of Linguistics and Modern Languages, The Chinese University of Hong Kong, Shatin, Hong Kong SAR, China, 3Department of Child & Adolescent Psychiatry, Castle Peak Hospital, Hong Kong Hospital Authority, Hong Kong SAR, China, 4Division of Speech and Hearing Sciences, The University of Hong Kong, Hong Kong SAR, China
Introduction: Aspects of sound processing such as perception of lexical tones by tone-speaking children has shown to be a feature of Autism Spectrum Disorder (ASD). Neurophysiological research also found less robust speech encoding in both tone-language speaking school-age children and adults with ASD, suggesting that speech encoding deficits may emerge early in life (Lau, 2020). The current study examined neural encoding of speech in preschool children with and without ASD in order to better understand its developmental time course. Methods: We recruited fifty Cantonese-speaking children with ASD aged between 2 and 5 years; Autism Diagnostic Observation Schedule, Second Edition was conducted to further confirm their ASD condition. Forty-nine typically developing (TD) children with no reported developmental conditions were drawn from a larger pool of subjects as control. The children in the two groups were matched in age, gender, and gestation age using propensity score. Demographics data, including age, sex, household income, parents’ age and education level, were collected, which showed no significant differences between groups. The mean age was 36.29 months; 69.7% of the subjects were male. During EEG recordings, children were seated on their caregivers’ laps watching a silent movie of their choice or sleeping while listening to Cantonese speech stimuli in three lexical tones (/ga2/, /ga3/ and /ga4/). A Neuroscan system was used to examine their neural encoding of sound. Continuous EEG data were collected from Ag/AgCl electrodes at Cz, M1 and M2 at a 20 kHz sampling rate with CPz as a reference and Fpz as a ground. Cz data was re-referenced offline with the average of M1 and M2 for the subsequent analysis. A number of time- and frequency-domain measures were extracted for each tone including the frequency-following response (FFR) Signal-to-noise ratio (SNR), Noise root-mean-square, Fast Fourier transform power (low, middle and high), inter-trial phase coherence (low middle, high and maximal), pitch strength, peak amplitude, response consistency, pitch tracking, pitch error, response correlation, stimulus-response delay, peak latency and long-latency response SNR. Support Vector Machine (SVM) was applied, with EEG measures and demographic data as input to predict binary diagnostic outcome (ASD vs TD). A leave-one-out cross validation procedure was employed. The classification performance was assessed by sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and diagnostic accuracy. Results: Using solely demographics data, the diagnostic accuracy was only 60.6%; by adding FFR measures from different lexical tones into the SVM models, the classification accuracy improved. The model that included demographics data and FFR measures from all lexical tones outperformed the others with a sensitivity of 84.0%, specificity of 79.6%, PPV of 80.8%, NPV of 83.0%, and diagnostic accuracy of 81.8%. Conclusion: Differences in neural encoding of speech between ASD and TD children seem to emerge as early as preschool years in tone-learning children. Future research should examine whether these differences emerge from infancy and toddlerhood. EEG, coupled with machine learning, has the potential to be an automatic, objective tool to augment the autism diagnostic process to classify children at individual-subject level.
Topic Areas: Disorders: Developmental, Speech Perception