Corrélats neuronaux de la mémoire de travail en magnétoencéphalographie à l’état de repos

Purpose: There are few studies demonstrating the link between neural oscillations in magnetoencephalography (MEG) at rest and cognitive performance. Working memory is one of the most studied cognitive processes and is the ability to manipulate information on items kept in short-term memory. Heister & al. (2013) showed correlation patterns between brain oscillations at rest in MEG and performance in a working memory task (n-back). These authors showed that delta/theta activity in fronto-parietal areas is related to working memory performance. In this study, we use resting state MEG oscillations to validate these correlations with both of verbal (VWM) and spatial (SWM) working memory, and test their specificity in comparison with other cognitive abilities. Methods: We recorded resting state MEG and used clinical neuropsychological tests to assess working memory performance in 18 volunteers (6 males and 12 females). The other neuropsychological tests of the WAIS-IV were used as control tests to assess the specificity of the correlation patterns with working memory. We calculated means of Power Spectrum Density for different frequency bands (delta, 1-4Hz; theta, 4-8Hz; alpha, 8-13Hz; beta, 13-30Hz; gamma1, 30-59Hz; gamma2, 61-90Hz; gamma3, 90-120Hz; large gamma, 30-120Hz) and correlated MEG power normalised for the maximum in each frequency band at the sensor level with working memory performance. We then grouped the sensors showing a significant correlation by using a cluster algorithm. Results: We found positive correlations between both types of working memory performance and clusters in the bilateral posterior and right fronto-temporal regions for the delta band (r2 =0.73), in the fronto-middle line and right temporal regions for the theta band (r2 =0.63) as well as in the parietal regions for the alpha band (r2 =0.78). Verbal working memory and spatial working memory share a common fronto-parietal cluster of sensors but also show specific clusters. These clusters are specific to working memory, as compared to those obtained for other cognitive abilities and right posterior parietal areas, specially in slow frequencies, appear to be specific to working memory process. Conclusions: Slow frequencies (1-13Hz) but more precisely in delta/theta bands (1-8Hz), recorded at rest with magnetoencephalography, predict working memory performance and support the role of a fronto-parietal network in working memory.