Automated identification for autism severity level: EEG analysis using empirical mode decomposition and second order...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Developmental Psychology

Background:

• Automated diagnosis of autism spectrum disorders (ASD) using nonlinear EEG analysis has historically struggled to differentiate severity levels.
• Previous methods could only distinguish ASD from typical development, not varying degrees of autistic features.

Purpose of the Study:

• To identify EEG differences between mild and severe ASD using Empirical Mode Decomposition (EMD) and Second-Order Difference Plot (SODP).
• To determine the accuracy of these EEG analysis models in distinguishing ASD severity.

Main Methods:

• Resting-state EEG data from 36 children (equally divided into mild and severe ASD groups) were analyzed.
• EMD was used to extract Intrinsic Mode Functions (IMFs) features, SODP patterns, elliptical area, and Central Tendency Measure (CTM) values.
• An Artificial Neural Network (ANN) was employed to assess the diagnostic accuracy of the derived EEG measures.

Main Results:

• Children with severe ASD exhibited less IMF oscillation, more stochastic SODP plotting, lower CTM values, and larger ellipse areas compared to those with mild ASD.
• These EEG patterns suggest greater variability and impaired behavioral control in severe ASD.
• The ANN achieved 100% sensitivity, 94.7% specificity, and 97.2% overall accuracy in differentiating between mild and severe ASD groups.

Conclusions:

• Distinct IMF features, SODP patterns, elliptical area, and CTM values differentiate between mild and severe ASD in children.
• These EEG-derived measures show promise as a sensitive automated tool for classifying ASD severity levels.