Enhancing oscillations in intracranial electrophysiological recordings with data-driven spatial filters

Abstract

In invasive electrophysiological recordings, a variety of neural oscillations can be detected across the cortex, with overlap in space and time. This overlap complicates measurement of neural oscillations using standard referencing schemes, like common average or bipolar referencing. Here, we illustrate the effects of spatial mixing on measuring neural oscillations in invasive electrophysiological recordings and demonstrate the benefits of using data-driven referencing schemes in order to improve measurement of neural oscillations. We discuss referencing as the application of a spatial filter. Spatio-spectral decomposition is used to estimate data-driven spatial filters, a computationally fast method which specifically enhances signal-to-noise ratio for oscillations in a frequency band of interest. We show that application of these data-driven spatial filters has benefits for data exploration, investigation of temporal dynamics and assessment of peak frequencies of neural oscillations. We demonstrate multiple use cases, exploring between-participant variability in presence of oscillations, spatial spread and waveform shape of different rhythms as well as narrowband noise removal with the aid of spatial filters. We find high between-participant variability in the presence of neural oscillations, a large variation in spatial spread of individual rhythms and many non-sinusoidal rhythms across the cortex. Improved measurement of cortical rhythms will yield better conditions for establishing links between cortical activity and behavior, as well as bridging scales between the invasive intracranial measurements and noninvasive macroscale scalp measurements.

Overcoming harmonic hurdles: genuine beta-band rhythms vs. contributions of alpha-band waveform shape

“…The most important aspect we aim to emphasize is that in order to arrive at strong, methodologically valid interpretations of potential functional roles, that in analysis of higher frequency rhythms the relationship to lower frequency rhythms needs to be clarified. This is necessary, because there are numerous non-sinusoidal rhythms present in the human brain, as quantified using EEG or intracranial recordings (Schaworonkow and Nikulin, 2019;Schaworonkow and Voytek, 2021).…”

A supervised data-driven spatial filter denoising method for speech artifacts in intracranial electrophysiological recordings

“…(i) Performance declines for broadband artifactual sources (Extended Data Fig. 4), a limitation inherited from SSD, which is able to enhance the power only in narrow frequency bands 11,21 . This limitation could also partially explain the differences in performance between simulated and real data.…”

“…We assessed the robustness of the PCD pipeline by sweeping key parameters of the model within a meaningful range (see details in Methods). The algorithm's performance was quantified by different metrics that describe (i) the agreement in the temporal domain between the denoised neural signals and the ground-truth neural signals (𝑙𝑜𝑔(𝜒)) (see Methods equation ( 9)), (ii) the magnitude coherence estimate (MSCE, see Methods equation ( 10)), and (iii) the phase-locking value (PLV) between the estimated vibration artifact source and z (see Methods equation (11)). For pure sinusoidal artifacts, PCD perfectly removed the artifact regardless of the AGR, fundamental frequency and number of channels (Supplementary Figs.…”

“…They can be used for denoising, using linear transformations to estimate the data subspace related to the "noise" and discard it for the subsequent signal analyses [14][15][16] . This approach has been used primarily for noninvasive electrophysiology 14,15,17,18 , and more recently for iEEG signal processing 13,19,20 . A typical pipeline for artifact removal in non-invasive EEG consists of using principal component analysis (PCA) and independent component analysis (ICA) for identifying artifact components 17,21 .…”

Resting-state is not enough: alpha and mu rhythms change shape across development, but lack diagnostic sensitivity

“…Time series for each SSD component can be obtained from broadband activity across electrodes through matrix multiplication of the transposed spatial filters: Crucially, because the SSD component time series is a linear transformation on the broadband activity rather than filtered activity, nonsinusoidal features of the waveforms, such as harmonics, are retained in the SSD component time series, allowing for fine-grained analysis of waveform shape. Finally, we can determine the spatial patterns A for each participant by inverting the matrix of spatial filters: These spatial patterns can then be used to interpret the spatial origin of the extracted SSD components (Schaworonkow and Voytek, 2021a), and in our case, classify the SSD components as occipital alpha or sensorimotor mu (see below).…”