Modeling intra-individual inter-trial EEG response variability in autism.

Area of Science:

• Neuroscience
• Biostatistics
• Developmental Psychology

Background:

• Autism spectrum disorder (autism) is a neurodevelopmental condition marked by social and communication impairments.
• Electroencephalography (EEG) is a non-invasive method to study brain function in autism.
• Increased trial-to-trial variability in EEG responses is a potential biomarker for autism.

Purpose of the Study:

• To introduce nonlinear mixed effects (NLME) models for analyzing trial-level EEG data in autism.
• To quantify intra-individual inter-trial variability in EEG responses.
• To develop computationally feasible methods for analyzing large EEG datasets.

Main Methods:

• Application of multilevel nonlinear (shape-invariant) mixed effects (NLME) models.
• Utilizing a novel minorization-maximization (MM) algorithm for scalable estimation.
• Analysis of trial-level EEG data, focusing on features like latency and amplitude.

Main Results:

• Children with autism showed significantly higher intra-individual inter-trial variability in P1 latency during a visual evoked potential (VEP) task.
• The proposed NLME models effectively quantified response variability from noisy trial-level EEG data.
• Simulations confirmed the efficacy of the MM algorithm for large datasets.

Conclusions:

• NLME models offer a precise method to assess EEG response variability in autism.
• Greater inter-trial variability in specific EEG components may serve as an objective marker for autism.
• The developed computational methods enable scalable analysis of complex EEG data for autism research.