Luc Wilson

Graduate Student
McGill Univ
Email author

Time-resolved parameterization of aperiodic and periodic brain activity

Luc E Wilson, Jason da Silva Castanheira, Sylvain Baillet

Hi there, thanks for stopping by!

My name is Luc, and I am a Master's student in the Integrated Program in Neuroscience at McGill University. I study in the NeuroSPEED (Baillet) lab, where I develop and deploy accessible, open-source tools to study dynamics in neurophysiological signals. 

Are you interested in understanding the different contributions of rhythmic and arrhythmic signal components underlying brain activity?

In this presentation, I focus on a tool we developed in the lab for decomposing neural activity into periodic (rhythmic) and aperiodic (arrhythmic) components, and tracking the features of these cardinal components as they evolve across time. This is important because brain activity and behaviour vary from moment to moment; our method provides a principled approach to align the timing of behavioural changes with the contents of the spectrum of brain activity.

This tool, which we call SPRiNT (for Spectral Parameterization Resolved in Time), is open-source and available for use in the Brainstorm toolbox if you are interested in trying it on your own datasets. You can also find a brief tutorial on how to use SPRiNT here. If you'd like to learn more about SPRiNT, the full preprint is available here.

I hope to see you at the poster session!

Time-resolved parameterization of aperiodic and periodic brain activity

Luc E Wilson, Jason da Silva Castanheira, Sylvain Baillet

Brain activity is characterized by complex dynamics across multiple, coexisting time scales. Recent work has shown that brain activity can be decomposed into periodic (rhythmic) and aperiodic (arrhythmic) components. Advances in parameterizing the neural power spectrum offer practical tools for evaluating the features of these signal components separately. Although neural signals express non-stationarity in relation to ongoing behaviour and perception, current methods for spectral parameterization are constrained to static spectral profiles. We present Spectral Parameterization Resolved in Time (SPRiNT) as a new, open source tool for decomposing neural dynamics into periodic and aperiodic spectral elements in a time-resolved manner. SPRiNT operates on time series and returns a parameterized spectrogram. It derives short-time fast Fourier transforms (FFTs) over sliding time windows averaged locally in time produce a temporally smoothed spectrogram. The spectrogram is subsequently parameterized into periodic and aperiodic components using the specparam algorithm. To evaluate algorithm performance, we recovered nonstationary spectral parameters from 10,000 naturalistic time series simulations, varying all dimensions of aperiodic and periodic signal components within realistic parameter bounds. SPRiNT, validated on simulated and empirical data, is an open-source plug-in library and interoperates with Brainstorm, facilitating its use in a variety of research contexts. SPRiNT addresses growing interests for parameterizing spectrograms derived from multiple signal modalities and advances the quantitation of complex neural dynamics at the natural timescales of behaviour.