MEG, myself, and I: individual identification from neurophysiological brain activity

Abstract

Large, openly available datasets and current analytic tools promise the emergence of population neuroscience. The considerable diversity in personality traits and behaviour between individuals is reflected in the statistical variability of neural data collected in such repositories. This variability challenges the sensitivity and specificity of analysis methods. Yet, recent studies with functional magnetic resonance imaging (fMRI) have concluded that patterns of resting-state functional connectivity can both successfully identify individuals within a cohort and predict their individual traits, yielding the notion of a neural fingerprint. Here, we aimed to clarify the neurophysiological foundations of individual differentiation from features of the rich and complex dynamics of magnetoencephalography (MEG) resting-state brain activity in 158 participants. The resulting neurophysiological functional connectomes enabled the identification of individuals with similar identifiability rates to fMRI. We also show that individual identification was equally successful from simpler measures of the spatial distribution of neurophysiological spectral signal power. Our data indicate that identifiability can be achieved from brain recordings as short as 30 seconds, and that it is robust over time: individuals remain identifiable from recordings performed weeks after their baseline reference data was collected. We can anticipate a vast range of diverse applications in personalized, clinical and basic neuroscience of individual differentiation from large-scale neural electrophysiology, in future longitudinal and cross-section studies.

Stability of spectral estimates in resting-state magnetoencephalography: recommendations for minimal data duration with neuroanatomical specificity

“…Overall, we find that spectral estimates of task-free MEG brain activity stabilize remarkably quickly when derived from durations as short as 30 seconds, and typically 120 seconds, of data. These results align well with a recent study reporting that participants can be differentiated from one another based on MEG spectral “fingerprints” obtained with as little as 30 seconds of resting-state data (da Silva Castanheira et al, 2021). We noted some exceptions in the high-gamma range over frontal and temporal cortices, where the data duration required to reach stability was longer than for the alpha and beta bands.…”

“…We emphasize, however, that spectral instability at higher frequencies should not be conflated to signals being affected by intractable levels of noise. Indeed, the variability of high-frequency gamma activity is well documented to be associated to various cognitive functions (Başar, 2013; Uhlhaas et al, 2011; Ward, 2003; Wiesman et al, 2020; Wiesman and Wilson, 2019), disease states (Herrmann and Demiralp, 2005; Mably and Colgin, 2018; Uhlhaas and Singer, 2010), and inter-individual differences (da Silva Castanheira et al, 2021; Hirschmann et al, 2020; Muthukumaraswamy et al, 2010; Perry et al, 2013; Shaw et al, 2017). Furthermore, we show that estimates of intra-session resting-state MEG stability in normative populations require samples of at least 50 participants.…”

“…MEG preprocessing procedures for both datasets were similar to those reported by da Silva Castanheira et al (2021), following good-practice guidelines (Gross et al, 2013). All processing steps were performed using Brainstorm (Tadel et al, 2011), unless otherwise specified.…”

“…Intra-session variability in resting-state activity has been of recent interest for “fingerprinting” (da Silva Castanheira et al, 2021; Finn et al, 2015), functional parcellation (Laumann et al, 2015), and enhanced prediction of behavior (Waschke et al, 2021) at the level of individual participants. Notwithstanding these recent efforts, most resting-state research derives and interprets summary statistics of neural fluctuations over data durations that vary considerably between studies, ranging at least from one (Ranasinghe et al, 2020) to fifteen (Marquetand et al, 2019) minutes.…”

Connectomics of Human Electrophysiology

“…Both simple features of the MEG/EEG power spectrum (Da Silva Castanheira et al, 2021;Demuru and Fraschini, 2020) as well as attributes of oscillation-based connectomes (Da Silva Castanheira et al, 2021;Demuru et al, 2017;Nentwich et al, 2020;Sareen et al, 2021) constitute highly subject-specific brain "fingerprints'' on the basis of which individuals can be identified. Inter-individual differences in e-phys attributes, including e-phys connectome organization, should thus be considered important to brain function rather than only reflective of noise.…”