Schedule of Events | Search Abstracts | Invited Symposia | Symposia | Poster Sessions | Data Blitz Sessions
Novel approach for detection of topographic outliers for spatiotemporal EEG analyses
Poster Session F - Tuesday, April 1, 2025, 8:00 – 10:00 am EDT, Back Bay Ballroom/Republic Ballroom
Jiong Yan Yap1 (jiongyan@usc.edu), Delara Aryan1, Sahana Nagabhushan Kalburgi2; 1University of Southern California, 2Children's Hospital Los Angeles
Electroencephalography (EEG) microstates are increasingly being used to quantify millisecond-level spatiotemporal dynamics on the scalp. Outliers in individual microstates can affect group-level mean microstate maps and subsequently affect computed temporal dynamics of microstates. However, topographical outlier detection on individual-level microstate maps has yet to be explored, and only one microstate analysis toolbox provides outlier detection as a built-in feature. In this study, we propose a novel outlier detection method for microstates. Our method computes the topographic similarity between individual microstate maps and published normative maps to identify outlier topographies. We validated our novel algorithm against a publicly available dataset, in which outliers were identified by manual visual inspection, and then compared the results against an existing outlier detection method which utilized multidimensional scaling (MDS). Our method exhibited good sensitivity (mean = 67.7%) and specificity (mean = 92.5%) to outliers across all microstate classes. Our method also outperformed the traditional MDS method in sensitivity (mean = 2.4%) across all microstate classes except for class C. The findings suggest that our novel method is capable of identifying outlier microstates with good accuracy and is an improvement over existing methods. Further investigation into determining the optimal topographic similarity cutoff for each microstate class may further increase the specificity of our method. Overall, our study highlights the importance of outlier detection for improving reproducibility in EEG microstate studies.
Topic Area: METHODS: Electrophysiology