Stimulus-induced narrowband gamma oscillations are test-retest reliable in healthy elderly in human EEG

Abstract

Visual stimulus-induced narrowband gamma oscillations in electroencephalogram (EEG) recordings have been recently shown to be compromised in subjects with Mild Cognitive Impairment or Alzheimer′s Disease (AD), suggesting that gamma could be an inexpensive and easily accessible biomarker for early diagnosis of AD. However, to use gamma as a biomarker, its characteristics should remain consistent across multiple recordings, even when separated over long intervals. Previous magnetoencephalography studies in young subjects have reported that gamma power remains consistent over recordings separated by a few weeks to months. Here, we assessed the consistency of slow (20-35 Hz) and fast gamma (36-66 Hz) oscillations induced by static full-field gratings in male (N=20) and female (N=20) elderly subjects (>49 years) in EEG recordings separated by more than a year and tested the consistency in the magnitude of gamma power, its temporal evolution and spectral profile. Gamma oscillations had distinct spectral and temporal characteristics across subjects, which remained consistent across recordings (average intraclass correlation, ICC of ~0.7). Alpha oscillations (8-12 Hz) and steady-state-visually-evoked-potentials (SSVEPs) were also found to be reliable. We further tested how EEG features can be used to identify two recordings as belonging to the same versus different subjects and found high classifier performance (area under ROC curve of ~0.89), with the temporal evolution of slow gamma and spectral profile emerging as the most informative features. These results suggest that EEG gamma oscillations are reliable across recordings and can be used as a clinical biomarker as well as a potential tool for subject identification.

Stimulus-induced narrow-band gamma oscillations in humans can be recorded using open-hardware low-cost EEG amplifier

“…Computing the cross-correlation between the two traces led to a median correlation (± standard error; computed using bootstrapping) of 0.79 ± 0.05, with OpenBCI traces lagging by 8 ± 0.5 milliseconds (cross-correlation value and the lag for each subject are indicated in the figure), corresponding to a lag of 2 sample points because our sampling frequency was 250 Hz. There was substantial variation in ERPs across subjects, consistent with our previous study where we show substantial variation in alpha and gamma responses across subjects which were nevertheless consistent across two recordings of the same subject[32],…”

“…Slow gamma and fast gamma bands were both found to be reduced in Alzheimer’s disease and Mild Cognitive Impairment patients in a study done by Murty and colleagues [ 14 ]. Further, previous studies have shown that slow gamma band is more reliable with age than fast gamma [ 13 ] and shows more inter-subject variability and better test-retest reliability [ 32 ]. This raises the possibility of using OpenBCI to detect slow gamma as a biomarker or screening tool in low resource settings.…”

“…Similarity between OpenBCI and Brain Product recordings was quantified using the Spearman correlation of the data points between the two sessions. Although interpretation of correlation coefficients can be complicated [ 41 ], we used these analyses to be consistent with our previous paper in which we compared the consistency of gamma oscillations in two recordings taken a year apart from the same subjects [ 32 ]. For band powers, the data points were the change in band power between the stimulus and baseline averaged across electrodes (yielding one value per frequency band for each subject).…”

“…However, a high correlation between OpenBCI and Brain Products traces may be confounded by the possibility that the spectral profile in response to visual stimulus is common to all subjects. A previous study suggests that spectral and temporal profiles are unique to subjects and different in different subjects [ 32 ]. To assert whether these correlations indeed represent subject-specific trends and not a general similarity of all traces, we further compared correlations between the two recordings of the same subject and between different subjects.…”

“…The same electrode cap, stimulus presentation software, and downstream analyses were used in both cases so that any difference was attributable only to the amplifiers. Full-screen gratings induce two distinct bands in the gamma range [12], termed slow gamma (20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30)(31)(32)(33)(34) and fast gamma (35-66 Hz) bands, which have characteristic spectral and temporal profiles [32]. Therefore, in addition to comparing the amplitude of change in band powers between stimulus and baseline, we also compared the spectral pattern and temporal evolution of the gamma bands between the two recording systems.…”