Authors :
Presenting Author: Vikram Jakkamsetti, MD, PhD – Weill Cornell Medicine
Rawan Elhag, MD – New York Presbyterian Brooklyn Methodist Hospital
Oksana Nulman, MD – New York Presbyterian Brooklyn Methodist Hospital
Juan Pascual, MD, PhD – Weill Cornell Medicine
Asim Shahid, MD – Weill Cornell Medicine
Rationale:
Pathological seizure activity often involves basal high frequency activity that can be captured as increased high frequency oscillations (HFOs). Whereas the majority of early HFO studies were done with intracortical recordings of humans, recent work has shown that scalp EEG recordings too can capture pathological HFOs reliably. However, these recordings are expected to be at a sampling rate >300 Hz to ensure a sampling rate at least 3 times that of the upper frequency of interest and avoid aliasing. While a high sampling rate is ideal, it is beyond the current traditional clinical office/inpatient standard sampling rate of ~256 Hz. We theorized that evidence of biomarkers for these traditional clinical office/inpatient recordings could expand the applicability of HFO detection beyond a research setting. Specifically, we hypothesized that there would be evidence of more synchronized high frequency oscillation activity above 80 Hz in epilepsy in scalp EEGs recorded in a non-research setup.
Methods:
To test our hypothesis, we analyzed EEGs for HFOs from a publicly available database at Temple University Hospital that offered EEGs from patients with and without seizures. The files were downloaded in the available .edf (European Data Format) file format. The first EEG in a set of EEGs for a patient was taken for analysis with the reasoning that it would best reflect initial brain insult and be least confounded by cumulative brain damage consequent to seizures. Only one EEG per patient and only EEGs with similar recording parameters and electrode labels were included in the analysis. For data analysis, the .edf EEG file was imported into the MATLAB workspace and the EEG data extracted as time-series data. Coherence was determined for 1 to 120 Hz between all scalp electrodes for a total of 171 electrode pairing possibilities.
Results:
One hundred subjects with epilepsy versus ninety-four in the non-epilepsy group were examined for synchronous HFO activity. For average referenced (AR) EEGs, there was no significant difference in coherence between the epilepsy and non-epilepsy group (n=71 patient EEGs each) for all studied frequencies and electrode pairings. For the linked ear (LE) referenced EEGs, synchrony of temporal electrodes with most non-prefrontal electrodes was significantly increased for a range of frequencies above 30 Hz, including from 80 to 90 Hz (p< 0.01, Mann-Whitney U test for non-parametric data for each frequency for each electrode comparison). The extent of significant high frequencies was interrupted between 55 to 65 Hz, reflecting the inadequate electrode grounding and consequent 60 Hz electrical noise seen in field non-research settings, yet the electrical noise did not impact significant differences between the two groups in 60 Hz flanking frequencies, especially for intra-temporal electrode coherences.
Conclusions:
Notwithstanding a lower sampling rate (~256) and imperfect electrode grounding in scalp EEGs done in traditional clinical office/inpatient settings, EEGs with linked ear (LE) referencing can be analyzed to detect the greater presence of synchronous high-frequency oscillations ( >80 Hz) in subjects with epilepsy.
Funding: NIH grant RM1NS133593