Movie Content Influence Relation Between Connectome Similarities and Metastability in fMRI

1743 

Abstract Submission 

Tiago Pereira¹, André Cravo², Claudinei Biazoli²

¹Federal University of ABC, São Bernardo do Campo, São Paulo, ²Center of Mathematics, Computing and Cognition. Federal University of ABC, São Bernardo do Campo, São Paulo

Tiago Pereira  
Federal University of ABC
São Bernardo do Campo, São Paulo

André Cravo  
Center of Mathematics, Computing and Cognition. Federal University of ABC
São Bernardo do Campo, São Paulo

Claudinei Biazoli  
Center of Mathematics, Computing and Cognition. Federal University of ABC
São Bernardo do Campo, São Paulo

Better characterizing brain dynamics, particularly in fMRI, is a crucial step to advancing neuroscience [1]. A promising approach models the brain as a metastable system that alternates between integration and segregation of regions. [2] Recent work with MEG links non-linear features of the signal with brain FC identification (i.e. connectome brain fingerprinting). [3,4] Movie stimuli provide technical, methodological, cognitive and affective advantages. [5,6] However, each film has a specific narrative and different range of low and high features that must be considered. This work intends to investigate the correlation between metastability, identifiability at the global and large-scale functional network levels with naturalistic fMRI. Moreover, we sought to investigate how this relation is modulated by movie content.

The Naturalistic Neuroimaging Database, comprising 86 participants (42 females, 18–58 years, M = 26.81, SD = 10.09 years, TR  = 1s, 10 movies stimuli), was used. The spatial coherence was calculated as the distance between each ROI value at each timepoint and the standard deviation (metastability) was based on a specific time window. [7] The Similarity Between-Subjects and Stability Within-Subjects were respectively defined as the mean of the correlation's values of the FC matrix in 333 regions between (86 pairs) and within subjects (7310 pairs), on the same time window as the metastability for the whole-brain and for 8 functional networks (>23 ROIs). Both measures were calculated throughout 91 minutes (shortest film) merging the 10 movies, and then separately for two movies, a romantic-comedy (500 Days of Summer, 91 minutes, 5470 timepoints) and a documentary (Citizenfour, 113 minutes, 6780 timepoints).

For the whole brain, metastability and the similarities between subjects were moderately correlated (r = 0.634, p<0.001) at a 11-minute sliding-window. At the network level, 7 out of the 8 networks analyzed presented moderate to strong correlations between these measures, with the highest values observed for the networks related with attentional process: Fronto Parietal (r = 0.866, p<0.001), Dorsal Attentional (r = 0.710, p<0.001) and Cingulo Opercular (r = 0.658, p<0.001). On the other hand, evidence for coupling between metastability and Stability Within-Subjects was found only for Primary Motor (0.534, p<0.001) and Auditory (0.428, p<0.001) networks. In the 2 films separately, the romantic-comedy showed a much higher correlation (r=0.837, p<0.001) between metastability and Similarity Between-Subjects compared to the documentary (r=0.566, p<0.001). When correlating global metastability with the Similarity Between-Subjects of each network for each movie, all the 8 networks presented a correlation in the first movie while the documentary only correlated in 2 networks.

In this work and for this sample, a significant correlation was found between metastability levels and Similarity Between-Subjects during movie watching for the whole-brain and at network levels, more distinctly in attentional networks. Furthermore, metastability and Stability Within-Subjects correlated only for primary sensory-motor cortical networks. These results expand the finding of a relation between identifiability and non-linear dynamics observed in MEG. The results also match with developmental processes, with sensory-motor networks that stabilized earlier in brain development showing greater Stability Within-Subjects while attentional networks are more strongly shaped by environmental stimuli. Finally, the relation between identifiability and metastability varied with the content of the stimuli, suggesting that different features of the movies are related with fMRI nonlinear dynamics.

Connectivity (eg. functional, effective, structural) ²

fMRI Connectivity and Network Modeling ¹

Cognition

Computational Neuroscience

Data analysis

FUNCTIONAL MRI