Aapo Hyvärinen: Publications

Publications by topic

Publications front page

Home page

Brain imaging data analysis

[The papers here are concentrated on MEG/EEG, and unsupervised learning like ICA. However, a couple of papers consider decoding in MEG/EEG/EMG, and unsupervised learning in fMRI as well.]

Unsupervised deep learning methods for feature extraction

Yongjie Zhu, Tiina Parviainen, Erkka Heinilä, Lauri Parkkonen, Aapo Hyvärinen. Unsupervised representation learning of spontaneous MEG data with nonlinear ICA. NeuroImage, 274:120142, 2023
open access journal article
[Shows how nonlinear ICA is useful in EEG/MEG analysis by learning a representation from big resting-state data that can then be used to classification on smaller data sets.]

Hubert Banville, Omar Chehab, Aapo Hyäinen, Denis-Alexander Engemann, Alexandre Gramfort. Uncovering the Structure of Clinical EEG Signals with Self-supervised Learning. J. Neural Engineering.18:046020, 2021
preprint pdf
[Deep learning methods, similar to our nonlinear ICA work, learns a representation from EEG in an unsupervised (self-supervised) manner, which is useful for sleep analysis.]

Hiroshi Morioka, Vince Calhoun, Aapo Hyvärinen. Nonlinear ICA of fMRI reveals primitive temporal structures linked to rest, task, and behavioral traits. NeuroImage, 218:116989, 2020.
[A deep learning method based on time-contrastive learning (see "unsupervised deep learning" in the menu on the left) is developed for fMRI analysis.]

Neurofeedback / Mindfulness

A. Zhigalov, E. Heinilä, T. Parviainen, L. Parkkonen, and A.Hyvärinen. Decoding attentional states for neurofeedback: Mindfulness vs. wandering thoughts. NeuroImage, 185:565-574, 2019.
[Work towards building a neurofeedback system that alerts the user when wandering thoughts invade the mind during mindfulness meditation.]

Resting-state connectivity analysis (fMRI and EEG/MEG)

Hugo Richard, Pierre Ablin, Bertrand Thirion, Alexandre Gramfort, Aapo Hyvärinen. Shared Independent Component Analysis for Multi-Subject Neuroimaging. NeurIPS 2021.
[A new method for group analysis by ICA which can both separate Gaussian sources based on covariance structure as well as non-Gaussian sources based on sparsity. A principled method featuring a likelihood in closed form and an efficient algorithm.]

Hugo Richard, Luigi Gresele, Aapo Hyvärinen, Bertrand Thirion, Alexandre Gramfort, Pierre Ablin. Modeling Shared Responses in Neuroimaging Studies through MultiView ICA. NeurIPS 2020.
[A method for group analysis by ICA. A principled method featuring a likelihood in closed form and an efficient algorithm.]

Ricardo Pio Monti, Alex Gibberd, Sandipan Roy, Matthew Nunes, Romy Lorenz, Robert Leech, Takeshi Ogawa, Motoaki Kawanabe, Aapo Hyvärinen. Interpretable brain age prediction using linear latent variable models of functional connectivity. PLoS ONE, 2020.
[Uses our new method for finding components together with causal connection on fMRI age prediction. It is not only a principled method featuring a likelihood in closed form and an efficient algorithm, but also enables easier interpretation of the results than many other methods.]

J. Hirayama and T. Ogawa and A. Hyvärinen. Unifying Blind Separation and Clustering for Resting-State EEG/MEG Functional Connectivity Analysis. Neural Computation, 27:1373-1404, 2015.
pdf     Matlab code
[A new method for analysing nonstationarity of power-coherence in EEG and MEG. An extension of ICA where the variance patterns are clustered.]

P. Ramkumar, L. Parkkonen, and A. Hyvärinen. Group-level spatial independent component analysis of Fourier envelopes of resting-state MEG data. NeuroImage, 86:480-491, 2014.
[Analyses Fourier envelopes of MEG data by spatial ICA, in some sense emulating the analysis usually dones for fMRI resting-state data.]

P. Ramkumar, L. Parkkonen, R. Hari, and A. Hyvärinen. Characterization of neuromagnetic brain rhythms over time scales of minutes using spatial independent component analysis.
Human Brain Mapping, 33(7):1648-1662, 2012.
[Shows an alternative way of applying ICA on EEG/MEG signals, based on the idea of spatial ICA well-known in fMRI literature.]

A. Hyvärinen, P. Ramkumar, L. Parkkonen, and R. Hari. Independent component analysis of short-time Fourier transforms for spontaneous EEG/MEG analysis. NeuroImage, 49(1):257-271, 2010.
pdf   Matlab code
[Proposes a source separation method tailor-made to EEG and MEG signals. Basically, preprocess the data by short-time Fourier transforms and then do ICA. Shows that this takes temporal correlations into account, and combines them with non-Gaussianity.]

V. Kiviniemi, J. H. Kantola, J. Jauhiainen, A. Hyvärinen and O. Tervonen. Independent component analysis of nondeterministic fMRI signal sources. NeuroImage, 19(2):253-260, 2003.
[Probably the first paper to analyze resting state activity in the brain using ICA. ICA is especially suited for this kind of analysis where no stimulation schedule is known.]

Analysing changing (dynamic) connectivity

A. Hyvärinen, J. Hirayama , V. Kiviniemi and M. Kawanabe. Orthogonal Connectivity Factorization: Interpretable decomposition of Variability in Correlation Matrices. Neural Computation, 28:445-484, 2016.
pdf   Matlab code
[A new method for analysing nonstationarity of connectivity in timeseries, by finding linear components of the data such that their connectivity changes the most. Equally applicable to intersubject differences, or any kinds of differences, in connectivity.]

J. Hirayama, A. Hyvärinen, V. Kiviniemi, M. Kawanabe and O. Yamashita. Characterizing Variability of Modular Brain Connectivity with Constrained Principal Component Analysis. PLoS ONE, 2016.
Open Access Article
[A generalization of OCF above.]

Decoding EEG, MEG, EMG (using linear ICA representation)

J.-P. Kauppi, L. Parkkonen, R. Hari, and A. Hyvärinen. Decoding MEG rhythmic activity using spectrospatial information. NeuroImage, 83:921-936, 2013.
pdf   Matlab code
[A decoding method and toolbox for neuroscientific data analysis of EEG or MEG, with special emphasis on optimal use of spectral information.]

J.-P. Kauppi, J. Hahne, K.-R. Müller, and A. Hyvärinen. Three-Way Analysis of Spectrospatial Electromyography Data: Classification and Interpretation. PLoS ONE, 83:921-936, 2015.
Open access article   Matlab code (same as above)
[Develops the decoding methods in the preceding paper further, and applies them on EMG.]

H. Celikkanat, H. Moriya, T. Ogawa, J.-P. Kauppi, M. Kawanabe and A. Hyvärinen. Decoding Emotional Valence from Electroencephalographic Rhythmic Activity. 39th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (IEEE-EMBC'17), Jeju, Korea, 2017.
[An application of the method on detection of affects from EEG.]

Testing and validating ICA

A. Hyvärinen. Testing the ICA mixing matrix based on inter-subject or inter-session consistency. NeuroImage, 58:122-136, 2011.
pdf  Matlab code
[A method for assigning a statistical significance (p-value) to each independent component, based on whether an independent component with the same mixing coefficients was found in different data sets. The datasets can be from different subjects in brain imaging, or just different parts of the same larger data set. A probabilistic extension of the Esposito et al method below.]

A. Hyvärinen and P. Ramkumar. Testing independent component patterns by inter-subject or inter-session consistency. Frontiers in Human Neuroscience, 7:94, 2013.
Open Access article   Matlab code
[Extends the theory of the preceding paper to testing the values of the independent components themselves.]

F. Esposito, T. Scarabino, A. Hyvärinen, J. Himberg, E. Formisano, S. Comani, G. Tedeschi, R. Goebel, E. Seifritz and F. Di Salle. Independent component analysis of fMRI group studies by self-organizing clustering. NeuroImage, 25(1):193-205, 2005.
[Proposed an original method for ICA analysis of fMRI group studies (i.e. several subjects) which does ICA separately for the subject and analyzes the consistency of the results.]

J. Himberg, A. Hyvärinen and F. Esposito. Validating the independent components of neuroimaging time-series via clustering and visualization. NeuroImage 22(3):1214-1222, 2004.
pdf    Matlab code
[A method for analyzing the reliability and stability of estimated independent components by re-running the algorithm many times and visualizing the relationships of the obtained component. Features an easy-to-use software package for Matlab.]

Hyperscanning in EEG/MEG

C. Campi and L. Parkkonen and R. Hari and A. Hyvärinen. Non-linear canonical correlation for joint analysis of MEG signals from two subjects. Frontiers in Brain Imaging Methods 7:107, 2013.
Open Access article
[A method for finding maximally energy-correlated components in two EEG/MEG recordings, either in hyperscanning or when the subjects are given the same stimulation.]