Dae Hyeon Kim RESEARCH 한국어

Multiscale entropy for short EEG

Used moving averages to retain samples when estimating multiscale entropy from short EEG segments.

Paper DOI GitHub: multiscale-distribution-entropy
Role
Method comparison and co-first authorship
Contribution
Analyzed short EEG segments, compared entropy methods, performed statistical tests and contributed to the manuscript.
Period & affiliation
2022.01 ~ 2022.02 NeuroAI Research internship, MBE 2024 Co-first author

01 / RESEARCH DETAILMDE-MA: multiscale analysis of short EEG

Challenge
Coarse-graining reduces sample counts at larger scales, limiting entropy estimation in short segments.
Approach
Used moving averages to retain N-s+1 samples and compared scale-wise ROC AUC with MSE and coarse-grained MDE.
Contribution
Analyzed short segments, compared methods and performed statistical tests. Co-first author, MBE 2024.
Moving-average construction with overlapping samples
Moving-average construction with overlapping samples
View full-size figure
Normal versus ictal EEG: five-second segment centered on the first quartile
Normal versus ictal EEG: five-second segment centered on the first quartile
View full-size figure
Focal versus non-focal EEG: discrimination across scales
Focal versus non-focal EEG: discrimination across scales
View full-size figure
Scale-wise sample counts and evaluation scope
Samples at scale s For a sequence of length N
Coarse-graining floor(N / s)
Moving average N - s + 1
Evaluation State separation in public EEG
Evaluation conditions and scopeThe Bonn A segment is a five-second interval centered on the first quartile. ROC AUC measures state separation in public EEG datasets. Bern-Barcelona evaluation distinguishes focal from non-focal EEG.