Developmental domain shifts in epileptiform discharges: multidomain EEG differences between neonates and adults
Purpose Interictal epileptiform discharges (IEDs) differ across brain development, yet the electrophysiological characteristics distinguishing neonatal from adult discharges have not been comprehensively quantified within a unified framework. This study aimed to characterize this developmental domain shift and identify the feature domains most relevant for adapting adult-trained EEG models to neonatal populations. Methods Scalp EEG recordings from 12 neonatal patients (45 annotated IEDs) and 39 adult patients (164 IEDs) were analyzed. An automated waveform-extraction algorithm combining smooth nonlinear energy (SNE) with morphological criteria isolated each discharge. Thirty-four features spanning morphological, time-domain, frequency-domain, and spatial characteristics were computed. Group differences were evaluated using false-discovery-rate-corrected Mann–Whitney tests, patient-cluster permutation testing, and bootstrap confidence intervals. Multivariate separation, feature relationships, and classification performance were further assessed using cluster-aware PERMANOVA, correlation analysis, and leave-one-patient-out cross-validation. Results Twenty of 34 features remained significant after patient-cluster adjustment. Neonatal discharges exhibited greater signal complexity, gamma-band activity, and distinct spectral organization, whereas adult discharges showed markedly greater SNE peak energy and source isolation. Multivariate separation between groups was significant ( F = 20.59, p < 0.001). Morphological and time-domain features alone distinguished neonatal from adult discharges with near-perfect performance (AUC = 0.999), while frequency-domain and spatial features provided little additional improvement. This high discriminability reflects substantial electrophysiological domain separability rather than clinically validated diagnostic performance. Conclusion The neonatal–adult domain shift in IEDs is driven primarily by discharge morphology and spectral/complexity characteristics rather than spatial organization. These findings provide a quantitative basis for developing transfer-learning and domain-adaptation strategies for neonatal EEG and underscore the importance of patient-cluster-aware validation when analyzing repeated discharges from individual patients.
Authors
- Malek Adjouadi (ORCID: https://orcid.org/0000-0001-5380-3155)
- Ahmed Hossam Mohammed (ORCID: https://orcid.org/0000-0002-6427-3003)
- Armando B. Barreto (ORCID: https://orcid.org/0000-0002-5572-284X)
- Anuj Jayakar
- Mercedes Cabrerizo (ORCID: https://orcid.org/0000-0002-0302-7300)
- Naphtali David Rishe (ORCID: https://orcid.org/0000-0002-1611-4067)
- Micheal Adeyosoye
- Mina Ezzati Asl
- Thony Yan Liang
- Xueting Cui
Institutions
- Florida International University (US)
- Thomas Jefferson National Accelerator Facility (US)
- Miami Children's Hospital (US)
Publication Details
- Journal
- Frontiers in Neuroscience
- Published
- 2026-11-19
- DOI
- https://doi.org/10.3389/fnins.2026.1977415
- Primary Topic
- EEG and Brain-Computer Interfaces
- Type
- article
- Field-Weighted Citation Impact
- 0.00