Spatially regularized wavelet leader scale-free analysis of fMRI data

Herwig Wendt (1, 2) , Patrice Abry (3) , Philippe Ciuciu (4)

1 IRIT-MINDS - CoMputational imagINg anD viSion
2 CNRS - Centre National de la Recherche Scientifique
3 Phys-ENS - Laboratoire de Physique de l'ENS Lyon
4 PARIETAL - Modelling brain structure, function and variability based on high-field MRI data

Slow brain dynamics has received considerable interest in the recent years, with the scale-free paradigm playing a crucial role for analysis of various neuroimaging modalities. However , assessing the role of slow arrhythmic fluctuations requires the use of a large continuum of time scales and thus of long time series, hence raising concerns regarding the use of scale-free tools on fMRI data. Further, scale-free analysis remained so far mostly univariate, that is, voxels are analyzed independently, hence neglecting their spatial organization. The present contribution aims to propose a spatially regularized estimation of the self-similarity parameter, based on a recently formalized formalism combining wavelet leaders and Bayesian models. The spatially regularized estimates permit to quantify the modulations of the scale-free dynamics from rest to a working memory task from fMRI data collected for 21 healthy volunteers. These modulations are significant in the default mode network and in some regions involved in task performance such as primary visual regions or the supplementary motor area.

Mots clés: multivariate; scale-free; fMRI; spatial regularization; Bayesian inference

Domaines: Imagerie; Imagerie médicale; Traitement du signal et de l'image [eess.SP]; Sciences cognitives

Herwig Wendt, Patrice Abry, Philippe Ciuciu. Spatially regularized wavelet leader scale-free analysis of fMRI data. IEEE International Symposium on Biomedical Imaging, Apr 2018, Washington, DC, United States. ⟨hal-01782332⟩