Three-Phase Seizure Segmentation in Stereotactic EEG Using Envelope-Based Multivariate Changepoint Analysis
Published in Annals of Biomedical Engineering, 2026
Published online 02 April 2026. DOI: 10.1007/s10439-026-04097-7
This paper presents a semi-supervised framework for automated three-phase seizure segmentation in SEEG. Seven envelope-based features (RMS, relative bandpower in θ/α/β/γ, line length, spectral entropy) are extracted and fed into the PELT changepoint algorithm with RBF kernel. Evaluated on 179 SOZ bipolar channels across 32 seizures from 10 Engel Class I patients. Achieves onset MAE of 4.19 ± 2.69 s (71.6% within 5 s) and termination MAE of 3.82 ± 4.24 s (75.0% within 5 s).
Recommended citation: Kumar, H., Seshadri, N.P.G., Martinez, D., Najm, I., Alexopoulos, A., Bulacio, J.C., Serletis, D., & Krishnan, B. (2026). "Three-Phase Seizure Segmentation in Stereotactic EEG Using Envelope-Based Multivariate Changepoint Analysis." Annals of Biomedical Engineering. DOI: 10.1007/s10439-026-04097-7
Download Paper