Impact of perceptual learning on resting-state fMRI connectivity: A supervised classification study

International audiencePerceptual learning sculpts ongoing brain activity [1]. This finding has been observed by statistically comparing the functional connectivity (FC) patterns computed from resting-state functional MRI (rs-fMRI) data recorded before and after intensive training to a visual attention task. Hence, functional connectivity serves a dynamic role in brain function, supporting the consolidation of previous experience. Following this line of research, we trained three groups of individuals to a visual discrimination task during a magneto-encephalography (MEG) experiment [2]. The same individuals were then scanned in rs-fMRI. Here, in a supervised classification framework, we demonstrate that FC metrics computed on rs-fMRI data are able to predict the type of training the participants received. On top of that, we show that the prediction accuracies based on tangent embedding FC measure outperform those based on our recently developed multivariate wavelet-based Hurst exponent estimator [3], which captures low frequency fluctuations in ongoing brain activity too