Manifold-regression to predict from MEG/EEG brain signals without source modeling

Magnetoencephalography and electroencephalography (M/EEG) can reveal neuronal dynamics non-invasively in real-time and are therefore appreciated methods in medicine and neuroscience. Recent advances in modeling brain-behavior relationships have highlighted the effectiveness of Riemannian geometry for summarizing the spatially correlated time-series from M/EEG in terms of their covariance. However, after artefact-suppression, M/EEG data is often rank deficient which limits the application of Riemannian concepts. In this article, we focus on the task of regression with rank-reduced covariance matrices. We study two Riemannian approaches that vectorize the M/EEG covariance between-sensors through projection into a tangent space. The Wasserstein distance readily applies to rank-reduced data but lacks affine-invariance. This can be overcome by finding a common subspace in which the covariance matrices are full rank, enabling the affine-invariant geometric distance. We investigated the implications of these two approaches in synthetic generative models, which allowed us to control estimation bias of a linear model for prediction. We show that Wasserstein and geometric distances allow perfect out-of-sample prediction on the generative models. We then evaluated the methods on real data with regard to their effectiveness in predicting age from M/EEG covariance matrices. The findings suggest that the data-driven Riemannian methods outperform different sensor-space estimators and that they get close to the performance of biophysics-driven source-localization model that requires MRI acquisitions and tedious data processing. Our study suggests that the proposed Riemannian methods can serve as fundamental building-blocks for automated large-scale analysis of M/EEG.

[1]  Pierre-Antoine Absil,et al.  Quotient Geometry with Simple Geodesics for the Manifold of Fixed-Rank Positive-Semidefinite Matrices , 2020, SIAM J. Matrix Anal. Appl..

[2]  György Buzsáki,et al.  Space and time in the brain , 2017, Science.

[3]  Mark W. Woolrich,et al.  Adding dynamics to the Human Connectome Project with MEG , 2013, NeuroImage.

[4]  Fernando Maestú,et al.  Choice of Magnetometers and Gradiometers after Signal Space Separation , 2017, Sensors.

[5]  Luc Berthouze,et al.  Human EEG shows long-range temporal correlations of oscillation amplitude in Theta, Alpha and Beta bands across a wide age range , 2010, Clinical Neurophysiology.

[6]  Bart Vandereycken,et al.  Embedded geometry of the set of symmetric positive semidefinite matrices of fixed rank , 2009, 2009 IEEE/SP 15th Workshop on Statistical Signal Processing.

[7]  Cam-CAN Group,et al.  The Cambridge Centre for Ageing and Neuroscience (Cam-CAN) data repository: Structural and functional MRI, MEG, and cognitive data from a cross-sectional adult lifespan sample , 2017, NeuroImage.

[8]  R. Ilmoniemi,et al.  Interpreting magnetic fields of the brain: minimum norm estimates , 2006, Medical and Biological Engineering and Computing.

[9]  Mehrtash Harandi,et al.  Dimensionality Reduction on SPD Manifolds: The Emergence of Geometry-Aware Methods , 2016, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[10]  William D. Marslen-Wilson,et al.  The Cambridge Centre for Ageing and Neuroscience (Cam-CAN) study protocol: a cross-sectional, lifespan, multidisciplinary examination of healthy cognitive ageing , 2014, BMC Neurology.

[11]  R. Ilmoniemi,et al.  Magnetoencephalography-theory, instrumentation, and applications to noninvasive studies of the working human brain , 1993 .

[12]  Alexandre Gramfort,et al.  Autoreject: Automated artifact rejection for MEG and EEG data , 2016, NeuroImage.

[13]  Thomas E. Nichols,et al.  Statistical Challenges in “Big Data” Human Neuroimaging , 2018, Neuron.

[14]  Sylvain Baillet,et al.  Magnetoencephalography for brain electrophysiology and imaging , 2017, Nature Neuroscience.

[15]  R. Bhatia,et al.  On the Bures–Wasserstein distance between positive definite matrices , 2017, Expositiones Mathematicae.

[16]  E. Gordon,et al.  Spontaneous alpha peak frequency predicts working memory performance across the age span. , 2004, International journal of psychophysiology : official journal of the International Organization of Psychophysiology.

[17]  F. Yger,et al.  Riemannian Approaches in Brain-Computer Interfaces: A Review , 2017, IEEE Transactions on Neural Systems and Rehabilitation Engineering.

[18]  Julia M. Huntenburg,et al.  A mind-brain-body dataset of MRI, EEG, cognition, emotion, and peripheral physiology in young and old adults , 2019, Scientific Data.

[19]  K.-R. Muller,et al.  Optimizing Spatial filters for Robust EEG Single-Trial Analysis , 2008, IEEE Signal Processing Magazine.

[20]  Francis R. Bach,et al.  Low-Rank Optimization on the Cone of Positive Semidefinite Matrices , 2008, SIAM J. Optim..

[21]  Silvere Bonnabel,et al.  Riemannian Metric and Geometric Mean for Positive Semidefinite Matrices of Fixed Rank , 2008, SIAM J. Matrix Anal. Appl..

[22]  Alexandre Gramfort,et al.  Automated model selection in covariance estimation and spatial whitening of MEG and EEG signals , 2015, NeuroImage.

[23]  Bamdev Mishra,et al.  Manopt, a matlab toolbox for optimization on manifolds , 2013, J. Mach. Learn. Res..

[24]  Martin Luessi,et al.  MNE software for processing MEG and EEG data , 2014, NeuroImage.

[25]  Alexandre Barachant,et al.  A New Generation of Brain-Computer Interface Based on Riemannian Geometry , 2013, ArXiv.

[26]  Stefan Haufe,et al.  On the interpretation of weight vectors of linear models in multivariate neuroimaging , 2014, NeuroImage.

[27]  Jian Kong,et al.  Maturation trajectories of cortical resting-state networks depend on the mediating frequency band , 2018, NeuroImage.

[28]  Stefan Haufe,et al.  SPoC: A novel framework for relating the amplitude of neuronal oscillations to behaviorally relevant parameters , 2014, NeuroImage.

[29]  R. Bhatia Positive Definite Matrices , 2007 .

[30]  Adam Gazzaley,et al.  Age-Related Changes in 1/f Neural Electrophysiological Noise , 2015, The Journal of Neuroscience.

[31]  Levent Tunçel,et al.  Optimization algorithms on matrix manifolds , 2009, Math. Comput..

[32]  G. Buzsáki,et al.  The log-dynamic brain: how skewed distributions affect network operations , 2014, Nature Reviews Neuroscience.

[33]  Masashi Sugiyama,et al.  Geometry-aware principal component analysis for symmetric positive definite matrices , 2017, Machine Learning.

[34]  W. Förstner,et al.  A Metric for Covariance Matrices , 2003 .

[35]  R. Ilmoniemi,et al.  Signal-space projection method for separating MEG or EEG into components , 1997, Medical and Biological Engineering and Computing.

[36]  Riitta Hari,et al.  MEG–EEG Primer , 2017 .

[37]  Gaël Varoquaux,et al.  Scikit-learn: Machine Learning in Python , 2011, J. Mach. Learn. Res..

[38]  Gene H. Golub,et al.  Generalized cross-validation as a method for choosing a good ridge parameter , 1979, Milestones in Matrix Computation.

[39]  Daniel S. Margulies,et al.  Predicting brain-age from multimodal imaging data captures cognitive impairment , 2016, NeuroImage.

[40]  Christian Jutten,et al.  Multivariate Time-Series Analysis Via Manifold Learning , 2018, 2018 IEEE Statistical Signal Processing Workshop (SSP).

[41]  L. Parra,et al.  Human Neuroscience Original Research Article Correlated Components of Ongoing Eeg Point to Emotionally Laden Attention – a Possible Marker of Engagement? , 2022 .

[42]  Alexandre Barachant,et al.  Riemannian geometry for EEG-based brain-computer interfaces; a primer and a review , 2017 .

[43]  S. Taulu,et al.  Presentation of electromagnetic multichannel data: The signal space separation method , 2005 .

[44]  Moritz Grosse-Wentrup,et al.  Multiclass Common Spatial Patterns and Information Theoretic Feature Extraction , 2008, IEEE Transactions on Biomedical Engineering.

[45]  Tzyy-Ping Jung,et al.  Independent Component Analysis of Electroencephalographic Data , 1995, NIPS.