Abstracts

Rationale: Scalp EEG is an integral part of clinical care and diagnostics for children with genetic epilepsies. How quantitative electrophysiological biomarkers differ between specific genetic diagnoses remains largely unknown. Such biomarkers could inform the diagnostics, treatment, and care of individuals with rare genetic epilepsies such as STXBP1, SCN1A, and SYNGAP1.


Methods: We developed an automated pipeline to collect and analyze 1294 EEGs from 1059 individuals seen at the Children’s Hospital of Philadelphia (CHOP) Care Network from four primary cohorts: neurotypical healthy controls (1000 EEG, n=952) and individuals with pathogenic variants in STXBP1 (95 EEG, n=20), SCN1A (153 EEG, n=66), and SYNGAP1 (46 EEG, n=21). After removing artifacts and epochs of EEG with excess noise or altered state, we extracted spectral features and localized them to the primary lobes of the brain. We first compared the ratio of alpha-delta power across genetic disorders. We then used an ensemble of spectral features to train machine learning models to differentiate genetic disorders from age-matched controls, using leave-one-patient-out cross-validation. From these models, we identified the most important electrographic features.


Results: We found that individuals with STXBP1 have a significantly lower alpha-delta ratio compared to controls across all age groups (p< 0.001; Figure 1A). Furthermore, individuals with a protein-truncating variant in STXBP1 tend to have a higher alpha-delta ratio then individuals with a missense variant in the gene (Figure 1B). Random Forest models comparing EEGs between unseen controls and individuals with a genetic epilepsy predicted a diagnosis of STXBP1 (AUC=0.91), SYNGAP1 (AUC=0.85), and SCN1A (AUC=0.83; Figure 2). From these models, we isolated highly correlated biomarkers for these respective genetic disorders including beta-theta ratio in the frontal lobe with STXBP1 and alpha-theta ratio in the occipital lobe with SYNGAP1 (Figure 2).


Conclusions: These findings suggest that the genetic epilepsies have distinct quantitative electrographic signatures. In future analyses, we plan to use these features to approximate severity of comorbidities and outcomes such as gross motor functions and seizure control. In summary, quantitative EEG may be used to differentiate genetic disorders and quantify and track severity in individuals with genetic epilepsies.


Funding: This work was supported by the Center for Epilepsy and Neurodevelopmental Disorders (ENDD) at CHOP and the University of Pennsylvania and the STXBP1 Disorders Foundation. Erin Conrad received support from the NINDS (K23 NS121401-01A1) and the Burroughs Wellcome Fund.