Sambavi Arulnandhy, a Master of Science candidate in the Department of Mathematics and Statistics, will defend the Statistics thesis titled Enhanced EEG Spectral Decomposition with Applications to Neurodevelopmental Changes in ADHD on Tuesday, July 15, 2025 at 9:30 AM.
The examination committee includes Chair Melanie Pilkington, Supervisor William Marshall, External Examiner (Brock University) Stephen M. Emrich, and Committee Members Jan Vrbik and S. Ejaz Ahmed.
Students (both graduate and undergraduate) as well as other members of the Brock Community are invited to attend. If you are interested in the presentation, please contact [email protected] for the room location.
Keywords: ADHD, Aperiodic Activity, Customized Loss Function, EEG Spectral Decomposition, Modified Specparam
Abstract:
Aperiodic activity in electroencephalogram (EEG) signals, typically modeled as a 1/f-like trend in the power spectrum, reflects the brain’s excitation-inhibition balance, and provides valuable insights into neurodevelopment and cognitive function. This thesis focuses on two main goals: (a) to propose algorithmic improvements to EEG spectral decomposition methods; and (b) to apply the most effective method to investigate age- and sex-related differences in aperiodic EEG activity among children with and without Attention-Deficit Hyperactivity Disorder (ADHD). A comparative simulation study evaluated four spectral decomposition algorithms — Better Oscillation Detection Method (BOSC), Irregular Resampling Auto-Spectral Analysis (IRASA), Specparam and modified Specparam — on simulated EEG-like signals. BOSC and IRASA consistently underestimated components in multi-peak contexts. Specparam and modified Specparam demonstrated the highest accuracy with the modified version introducing a customized loss function that penalizes localized spectral dips differently than peaks and uses subgradient descent for adaptive fitting. This led to more stable parameter estimates and reduced mean squared error across simulations. The second objective is addressed in a registered report, applying Specparam to resting-state EEG data from the Healthy Brain Network. This study includes children aged 5-18 years, grouped by sex and ADHD diagnosis. Notably, it aims to address the long-standing underrepresentation of girls in ADHD research by analyzing aperiodic EEG activity in both sexes. The aperiodic estimates will be analyzed as functions of age, sex, and ADHD status, controlling for socioeconomic status, IQ, and the use of psychiatric medications. Statistical hypothesis testing will include Analysis of Variance (ANOVA) or non-parametric equivalents, with adjustments for group imbalance. These contributions enhance EEG spectral analysis techniques and offer insights into sex- and age-related neurodevelopmental differences in ADHD.
