Phase Reset As a Biomarker of ASD

Background:   Rigid, inflexible behavior and interests represent a core deficit in autism spectrum disorder (ASD). Whether inflexible behavior is reflective or resultant of inflexible brain dynamics, as measured by electroencephalography (EEG), remains an open question in autism research. Prior work by Thatcher and colleagues (2009) has shown that children with ASD, aged 2.6 – 11 years, generally have longer durations of phase locking between EEG channels than those with typical development. We hypothesize that EEG dynamics in individuals with ASD exhibit greater metastability than those of typically developed (TD) individuals as measured by number of state transitions. Analytic phase serves as a useful state variable for detecting state transitions (specifically, “phase resets”). 

Objectives:  Our principal objective is to characterize the phase reset of resting state EEG in young children with ASD and to explore the heterogeneity in phase reset, with particular interest in the association between phase reset and severity of repetitive behaviors and inflexibility in this population. Specifically, we wish to correlate the number of phase resets (an inverse measure of metastability) in resting state EEG with repetitive behavioral symptoms. We also wish to correlate number phase resets with age. Doing so will not only allow effects of age to be regressed out, but will also allow changes in metastability to be cross-sectionally mapped through typical development.

Methods:  Children with ASD and TD controls, ages 2-6, were recruited through the UCLA Center for Autism Research and Treatment (CART). Resting state EEG data were cleaned for artifacts and analyzed using the Hilbert transform, a linear transform for computing the analytic phase and analytic amplitude of a time series. By dividing the unwrapped analytic phase by 2π and computing the second derivative with respect to time, i.e., the rate of linear frequency change, one can inspect the new signal for large “spikes” indicative of phase resets. Such phase resets are spatiotemporally correlated. Counting spatiotemporal clusters of phase resets in low gamma (30-50 Hz) band resting state EEG datasets quantifies their metastability as a consequence of the inverse relationship between metastability and number of phase resets.

Results:   After disqualifying datasets with excessive EMG artifacts and bootstrapping pilot data from 12 ASD (aged 27-72 months, mean = 49 months) and 9 TD subjects (aged 31-74 months, mean = 56 months), we found a trend of greater metastability in ASD subjects (not significant, p = 0.083). No significant difference was found between ages of TD and ASD subjects using Student’s t-test (p = 0.3). Examining only subjects at least as young as the median age (53 months) and bootstrapping gave a similar but nearly significant result (p= 0.051). 

Conclusions:  Metastability as measured by spatiotemporal clusters of phase resets in frequency filtered resting state EEG shows potential as a biomarker of ASD. Future work will focus on discriminating between different subpopulations in ASD, as well as cross-sectionally mapping changes in metastability in the TD population.