CanICA: Model-based extraction of reproducible group-level ICA patterns from fMRI time series

Gaël Varoquaux (1, 2) , Sepideh Sadaghiani (3) , Jean Baptiste Poline (2) , Bertrand Thirion (1, 2)

1 PARIETAL - Modelling brain structure, function and variability based on high-field MRI data
2 LNAO - Laboratoire de Neuroimagerie Assistée par Ordinateur
3 LCogn - Neuroimagerie cognitive

Spatial Independent Component Analysis (ICA) is an increasingly used data-driven method to analyze functional Magnetic Resonance Imaging (fMRI) data. To date, it has been used to extract meaningful patterns without prior information. However, ICA is not robust to mild data variation and remains a parameter-sensitive algorithm. The validity of the extracted patterns is hard to establish, as well as the significance of differences between patterns extracted from different groups of subjects. We start from a generative model of the fMRI group data to introduce a probabilistic ICA pattern-extraction algorithm, called CanICA (Canonical ICA). Thanks to an explicit noise model and canonical correlation analysis, our method is auto-calibrated and identifies the group-reproducible data subspace before performing ICA. We compare our method to state-of-the-art multi-subject fMRI ICA methods and show that the features extracted are more reproducible.

Mots clés: fMRI; ICA; blind source separation; cross-validation; resting-state

Domaines: Neurosciences; Vision par ordinateur et reconnaissance de formes [cs.CV]; Applications [stat.AP]

Gaël Varoquaux, Sepideh Sadaghiani, Jean Baptiste Poline, Bertrand Thirion. CanICA: Model-based extraction of reproducible group-level ICA patterns from fMRI time series. Medical Image Computing and Computer Aided Intervention, Sep 2009, London, United Kingdom. pp.1. ⟨hal-00435262⟩