18742
Measures of Signal Complexity in Resting-State EEG Recordings from Young Children with ASD
Multiscale sample entropy (MSE), a measure of signal complexity, is a promising biomarker of both autism spectrum disorder (ASD) diagnosis (Ghanbari et al., 2013) and ASD risk (Bosl et al., 2011), yet few studies have characterized its trajectory in either typical development (TD) or in ASD. Furthermore, relationships between alternative measures of signal complexity and MSE remain uncharacterized in either population. We have introduced frequency variance (FV), a surrogate measure of phase resetting and metastable dynamics, as an alternative measure of signal complexity for comparison with MSE. Metastability is a mechanism underlying the synchronization of neuronal assemblies serving as substrates for cognitive states (Werner 2007) and thus may relate to the cognitive inflexibility observed in ASD.
Objectives:
In this cross-sectional study, we sought to correlate FV and MSE with age in cohorts of ASD and TD children ages 2–6. We also examined the relationship between FV and MSE within each cohort to judge the extent to which FV and MSE measure different aspects of signal complexity.
Methods:
Age matched children with ASD (n = 23, age = 55 ± 9.0 months) and TD controls (n = 34, age = 55 ± 12 months) were recruited through the UCLA Center for Autism Research and Treatment (CART). EEG was recorded while children watched a video of soap bubbles for 2 minutes. After artifact rejection, recordings were bandpass filtered in the beta-gamma range (12 – 48 Hz), as these frequencies carry phase resets, captured by FV, in spontaneous EEG (Freeman et al., 2003). A brain-related signal from an independent component analysis (ICA) was selected for further analysis. FV was computed as the variance of the time derivative of the analytic phase calculated by the Hilbert transform. MSE was computed for 20 time scales following Ahmed and Mandic (2011).
Results:
While FV was negatively correlated with age in TD (r = -0.38, p = 0.026), no correlation was found between FV and age in ASD, nor was any correlation found between MSE and age in either cohort. In the ASD cohort, a strong negative correlation exists between MSE at short time scales and FV (r = -0.72, p = 0.00013) whereas, in TD, MSE correlation with FV was instead observed at small time scales corresponding mostly to beta oscillations (r = -0.56, p = 0.00062).
Conclusions:
In TD children, FV exhibits a stronger correlation with age than MSE and, therefore, may serve as a more promising biomarker of typical development. This correlation was not observed in the ASD cohort, possibly because this biomarker does not appear in children with delayed language or social communication. Relationships between FV and MSE in each cohort suggest that (1) frequency components within the beta-gamma bands of EEG signals from TD children exhibit greater signal power at high frequencies than is the case in ASD children and (2) MSE and the reciprocal of FV both quantify similar aspects of signal complexity.