A prospective controlled study for EEG records comparison between Scalp-EEG and Ear-EEG wearable device, for subsequent analysis by an artificial intelligence-based system: SERAS-EEG study

Abstract

ABSTRACTmjn-SERAS is an earpiece shaped as a hearing-aid device, which continuously records the electrical brain activity in two channels placed in the external auditory. It uses an artificial intelligence algorithm (AI), based system for early detection of preictal period of seizures. Sixteen patients with drug-resistant focal epilepsy and 14 control subjects were simultaneously studied with the mjn-SERAS device and a standard 24-channel EEG using the 10-20 system. Data from channels F8-T4 or F7-T3, according to the laterality of the epileptic focus was extracted from the standard EEG. We analyzed the average signal correlation (AC) between the two types of records, with and without artefact removal (filtered records [FR]), comparing inter-subject and subjects recordings (SR), as well between ictal and interictal periods in epilepsy patients.AC was 0.90 [0.88 - 0.91] and 0.88 [0.86 - 0.90] in the FR and the whole cohort, respectively. No differences in the correlation of signals were found between controls and patients in the FR (-0.01 [-0.04;0.01], p=0.261) or the SR (-0.03 [-0.06;0.01], p=0.09). In addition, in the subset of patients with epilepsy, no differences in AC were noted between interictal activity and seizures (-0.02 [- 0.06; 0.02], p=0.352). Only AC during sleep in controls was found to be smaller compared to repose (-0.04 [-0.08;-0.01], p=0.01). No adverse events were reported. Our study supports an adequate correlation between the information recorded with both methods, providing technical support for use of the mjn-SERAS to record EEG signals.