Neuronal avalanches support cognitive processes during speech and music listening

984 

Abstract Submission 

Matteo Neri¹, Claudio Runfola², Noemie te Rietmolen³, Pierpaolo Sorrentino², Daniele Schon², Benjamin Morillon², Giovanni Rabuffo²

¹Aix Marseille Université, CNRS, INT, Institut de Neurosciences de la Timone, Marseille, France, ²Aix Marseille Université, INSERM, INS, Institut de Neurosciences des Systèmes, Marseille, France, ³Language and Computation in Neural Systems Group, Max Planck Institute for Psycholinguistics, Nijmegen, Netherlands

Matteo Neri  
Aix Marseille Université, CNRS, INT, Institut de Neurosciences de la Timone
Marseille, France

Claudio Runfola  
Aix Marseille Université, INSERM, INS, Institut de Neurosciences des Systèmes
Marseille, France

Noemie te Rietmolen  
Language and Computation in Neural Systems Group, Max Planck Institute for Psycholinguistics
Nijmegen, Netherlands

Pierpaolo Sorrentino  
Aix Marseille Université, INSERM, INS, Institut de Neurosciences des Systèmes
Marseille, France

Daniele Schon  
Aix Marseille Université, INSERM, INS, Institut de Neurosciences des Systèmes
Marseille, France

Benjamin Morillon  
Aix Marseille Université, INSERM, INS, Institut de Neurosciences des Systèmes
Marseille, France

Giovanni Rabuffo  
Aix Marseille Université, INSERM, INS, Institut de Neurosciences des Systèmes
Marseille, France

Neuronal avalanches consist of collective network events propagating across the brain in short-lived and aperiodic instances [1]. These salient events have garnered a great interest in the study of cortical dynamics. They have been observed across different imaging modalities and scales, and have been used to successfully distinguish between healthy and pathological conditions or between resting wakefulness and sleep states [4]. While a growing body of literature investigated neuronal avalanches in task-free conditions, whether they index cognitive functions or purely reflect physiological states remains an open question. In this work we investigated neuronal avalanches to index cognition, analyzing an intracranial stereo electroencephalography (sEEG) dataset collected during speech and music listening (naturalistic stimuli of about 10 min. each) and resting state, in 19 epileptic patients.

Neuronal avalanches were estimated by binarizing the z-scored neural activity, estimated as high-frequency activity (80-120 Hz; obtained as in [6]; Fig. 1A left). We then defined the activity profile (AP) as the sum of the binarized data across channels (Fig. 1A right). To investigate whether avalanches occurred in a stimulus-driven manner, we assessed the extent to which neuronal avalanches were similarly distributed in time across participants listening to the same stimuli. To this end, we computed the average inter-subject correlation of the APs for each condition, assessed results significance with a null model based on random permutations of time steps and finally correlated the APs across participants in a time-resolved way, using a sliding window approach (Fig. 1B-C). To assess the engagement of different brain regions in speech and music listening, we compared the number of times a channel was above threshold during highly correlated time windows (red dots in Fig. 1C) with respect to rest (in Fig. 1D). To investigate how avalanches propagate in the different cognitive states, we computed the avalanche transition matrices (ATMs), which estimate the transition probability of an avalanche across any pair of channels (Fig 2A-B) [5]. We correlated the ATMs across conditions for each participant (Fig. 2C). To disentangle the contribution of local (intra-areal) and distributed (inter-areal) processes, we separated auditory (Heschl's gyrus, H) and non-auditory (noH) channels, and estimated ATMs within (H-H, noH-noH) and between (H-noH; noH-H) them (Fig. 2D).

Firstly, we observed that avalanches relate to cognitive processes insofar as they are similarly distributed in time across patients while listening to speech and music, but not during rest (Fig. 1B). Secondly, we found that there are time windows in which avalanches are particularly coordinated across participants, and this is the case for both music and speech in partially overlapping but distinct distributed networks involving auditory and non-auditory regions (Fig. 1D). Then, we observed that ATMs tend to be overall more similar during music and rest, than speech (Fig. 2C). Furthermore this analysis revealed that the directed functional connections that differ the most across conditions (speech, music and resting state) are the ones between auditory and non auditory regions (Fig. 2D). This underlines the importance of feedforward and feedback mechanisms in shaping brain dynamics during speech and music perception.

With the present work we contribute to extend an approach based on neuronal avalanches, adopted in the past mainly in the study of resting state, to the investigation of cognitive functions, here speech and music [2]. Moreover, given the broad range of different contexts in which neuronal avalanches have been studied [3], our work provides the cognitive and system neuroscience communities with a set of theoretical and computational tools to investigate local and global properties of neural activity, with a particular focus on cognitive functions and naturalistic stimuli.

Music ¹

Speech Perception

Methods Development ²

Language

Other - Neuronal Avalanches