Skip to main content
Workshops & seminars, Conferences & lectures

Multifractal Analysis of Neural Activity (MEG) reveals Convergence to an Optimal Regime in Cortex that predicts Learning

Wednesday, December 20, 2017
4 p.m. – 5 p.m.

Philippe Ciuciu


This event is free




Wendy Kunin
514-848-2424 ext. 5295


7200 Sherbrooke St. W.
Room 2.115



Philippe Ciuciu

Mesoscopic brain activity recorded using magnetoencephalography (MEG) is usually analyzed through oscillatory regimes in different frequency bands (α,β,γ...). Brain activity also comprises arrhythmic or scale-free dynamics (1/f power spectrum) whose characterization is usually quantified by a single parameter (e.g., Hurst exponent) which reflects the long range temporal autocorrelation or self-similarity. Here, we propose to enrich this description with multifractality that measures transient oscillating temporal structures.
The functional relevance of scale-free dynamics (both self-similarity and multifractality) in learning was questioned by MEG recordings obtained while participants improved on a difficult visual motion discrimination task [Zilber et al, NeuroImage 2014]. We report several major findings. First, a fronto-occipital decrease of self-similarity typically reported during resting-state was here strengthened during task performance. Importantly, local decreases in self-similarity correlated with individuals’ performance, suggesting that the capacity of a brain region to decrease its temporal autocorrelation is indicative of perceptual improvements. Second, multifractality was found in several brain regions and neural activity converged to modeled local multifractal attractors (e.g. visual motion area MT, parietal cortices): after training, the distance between an individual’s multifractality and group-level multifractal attractors correlated with the individual’s performance. Thus, the distance of multifractality measured after training to its asymptotic value provides a predictive marker of learning. Altogether, we show that self-similarity and multifractality are functionally relevant measures of scale-free brain activity and may provide useful indexes of computational limitations in brain plasticity.

Speaker Bio:

Philippe Ciuciu has been Director of Research since 2014 in Biomedical Engineering at the French Atomic Energy Commission (CEA), a government-funded technological research organization. Since 2007, he has been with NeuroSpin, the largest neuroimaging center dedicated to ultra-high field MRI and its applications to neuroscience in France. Dr Ciuciu has also a joint appointment with INRIA (Parietal team), which is the research institution dedicated to computer science and automatic control in France.
In 2012-13, he held a visiting professor position at the University of Toulouse (Department of Applied Mathematics). He is currently principal investigator of interdisciplinary research projects at the interface between MR physics, signal processing, neuroimaging and neuroscience. His team has recently developed new compressed sensing solutions to speed up Magnetic Resonance Image (MRI) examinations. He also leads research projects for developing signal processing methods that aim to quantify and understand the functional role of arrhythmic brain activity from MEG and fMRI time series. Ultimately, his goal is to provide new insights on the neural basis of multiperceptual learning.
Dr Ciuciu is also scientific advisor to pharmaceutical groups in the context of pharmacological MRI studies for clinical trials.

Back to top Back to top

© Concordia University