Brief segments of neurophysiological activity enable individual differentiation

We all have the intuition that our brain makes us unique.

In our new article published today in Nature Communications with an Editor highlight, we show that seconds of brain activity captured with millisecond temporal resolution are sufficient to differentiate an individual in a large group of people, and that their neurophysiological fingerprint is stable and robust over weeks, months and years.

unsplash-image-D44kHt8Ex14.jpg

The project started when Hector Domingo Orozco Perez visited Bratislav Misic and Sylvain Baillet in 2018 for an internship sponsored by the NSERC CREATE program for Complex Dynamics in Brain & Behavior in the context of his Masters at McMaster University.

Hector was interested in mining large datasets of electrophysiological resting state activity to discover markers of complex behavior.

The idea came up quickly to emulate exciting developments in the fMRI world around the idea of brain fingerprinting, whereby brain connectivity patterns tend to be specific of a person, and associated with complex traits. The neurophysiological equivalent of these findings, based on fast brain electrical brain activity as captured with EEG or magnetoencephalography (MEG), remained a question mark.

Hector worked with the OMEGA open repository and derived several significant analytical steps, from pre-preprocessing, quality control, source imaging to connectome extraction. The early results were very encouraging but Hector had to move back to McMaster to complete his MSc.

Enter Jason Da Silva Castanheira as new PhD student in the Baillet lab. Jason followed in Hector’s footsteps and pushed the envelope further, testing that differentiation results were robust over time, immune to several confounds (head movements, environmental noise signatures, etc.), could be achieved with considerably shorter recordings (down to 30 seconds!) and essentially equivalent when derived from simple spectral measures of brain activity across a broad frequency range.

Here is the rest of the story…

MEG data sets were split in short segments (2 minutes, 30 seconds), and connectivity features (from amplitude-envelope correlations), then spectral features were derived from cortical source activity obtained with Brainstorm.

A differentiability score was computed for each individual from data extracted from the same session, or different MEG visits days, weeks or months apart (>200 days on average across participants.)

Screen Shot 2021-09-29 at 8.29.34 AM.png

The first “easy” challenge consisted in differentiating between participants from data segments collected during the same MEG visit. The differentiation accuracy of connectome and spectral fingerprinting based on broadband and narrowband brain signals was remarkably high, 95% and above—very much equivalent to that obtained in fMRI. In panel a below, horizontal gray bars indicate reference differentiation levels obtained from empty-room data recorded on the same days as participants. The differentiability scores were not related to typical confounds such as head motion, eye movements, and heartbeats (panel b).

We then took on the harder challenges to differentiate between participants from MEG connectome and spectral fingerprints taken days (>200 days on average) apart.

The fingerprinting performances were similar to those from the within-session challenge above (panel a; note the y-axis range starts at 75% accuracy). There was no association between differentiability and the delay between session recordings.

We then raised the bar using only 30-second data segments collected days apart (>200 days on average). Each data point represents one combination of datasets used for fingerprinting (panel c).

Panel d shows scatter plots of all fingerprinting challenges across frequency bands for source (brain) and sensor (scalp) level fingerprinting.

Overall, performances from beta-band activity and connectome measures were the highest and the most robust, although simpler power-spectrum measures fared decently as well.

Screen Shot 2021-09-29 at 8.39.10 AM.png

We also looked at which brain regions and canonical resting-state networks contributed to individual differentiation the most. The data show that the dorsal attention and visual networks were the most specific across individuals for connectome fingerprinting, in all frequency bands (figure below). Beta-band connectivity of the limbic network was particularly distinctive of individuals. For spectral fingerprinting, theta, alpha, beta, and gamma bands discriminated individuals along midline, parietal, lateral temporal, and visual areas.

Screen Shot 2021-09-29 at 8.48.44 AM.png

Finally, working with Brastislav Misic, we could not refrain from using partial-least-squares (PLS) to identify neurophysiological fingerprinting features associated with demographics.

We found three significant latent components, which were distinct for connectome and spectral fingerprinting. The first latent component in connectome fingerprinting was related to clinical population (some participants were patient volunteers) and handedness, via reduced beta-band functional connectivity over the frontal-parietal network. For spectral fingerprinting, the first salient latent component was related to a younger age, sex and clinical population. This demographic profile was associated with stronger expressions of broadband neurophysiological signal power in superior parietal regions and the pericalcarine gyrus bilaterally, and reduced neurophysiological signals in the isthmus cingulate.

We hope this type of analysis paves the way to using fingerprinting procedures in patients, using short brain data segments that are easy and practical to collect (e.g., possible transfer of methods to EEG). The rationale would be to measure how patient fingerprints evolve with treatment and hopefully, follow a trajectory that brings them closer to normal variants in the healthy population.

This is where we leave that story for now.

To be continued!


From the paper — 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.

Recent studies with functional magnetic resonance imaging (fMRI) have concluded that patterns of resting-state functional connectivity can both successfully distinguish individual participants within a cohort and predict some individual traits, yielding the notion of an individual’s neural fingerprint.

Here, we aim to clarify the neurophysiological foundations of individual differentiation from features of the rich and complex dynamics of resting-state brain activity using magnetoencephalography (MEG) in 158 participants.

We show that akin to fMRI approaches, neurophysiological functional connectomes enable the differentiation of individuals, with rates similar to those seen with fMRI. We also show that individual differentiation is equally successful from simpler measures of the spatial distribution of neurophysiological spectral signal power. Our data further indicate that differentiation can be achieved from brain recordings as short as 30 seconds, and that it is robust over time: the neural fingerprint is present in recordings performed weeks after their baseline reference data was collected.

This work, thus, extends the notion of a neural or brain fingerprint to fast and large-scale resting-state electrophysiological dynamics.

(Source mapping powered by Brainstorm.)

Previous
Previous

Multimodal pre-surgical protocol has positive impact on patient outcome in epilepsy

Next
Next

Lab graduate trainee Max Levinson receives multiple awards