Predicting Variance in Behaviour and Cognition from Task-Based fMRI-Derived Brain Networks

1363 

Abstract Submission 

Maiya Rasheed¹, Zhangyu Yang¹, Kyle Mackenzie¹, Abhijit Chinchani², Luca Besso³, Todd Woodward¹

¹University of British Columbia, Vancouver, British Columbia, ²The University of British Columbia (UBC), Vancouver, British Columbia (BC), ³The University of Manchester, Manchester, United Kingdom

Maiya Rasheed  
University of British Columbia
Vancouver, British Columbia

Zhangyu Yang  
University of British Columbia
Vancouver, British Columbia

Kyle Mackenzie  
University of British Columbia
Vancouver, British Columbia

Abhijit Chinchani  
The University of British Columbia (UBC)
Vancouver, British Columbia (BC)

Luca Besso  
The University of Manchester
Manchester, United Kingdom

Todd Woodward, Ph.D.  
University of British Columbia
Vancouver, British Columbia

The field of neuroscience has been moving towards understanding behaviour and cognition by examining underlying brain activity. Beyond deepening the understanding of individual differences in these constructs, examining what brain activity relates to cognitive or behaviour traits unlocks important translational applications such as personalized neuromodulation. One way to examine brain activity is using fMRI measurements during task paradigms, where networks underlying the cognitive operations required to complete a task may be extracted (Percival et al., 2020). Task-based fMRI can provide precise brain activity parameters that describe hemodynamic activity in different networks over different phases of the task. Importantly, a supervised multivariate approach called iterative constrained principal component analysis (CPCA) may be used to derive fine relationships between brain activity and behaviour. Iterative CPCA provides benefits over traditional analyses. Rather than reducing many variables into a single summary score and restricting the analysis to only variance in one variable, CPCA includes all individual items for both sets of variables (Chinchani et al., 2021).
In this study, we aimed to evaluate the relationship between brain activity in four task-based networks extracted from a social task from the human connectome project and measures of behaviour/cognition.

1000 participants were included in this analysis. During the experiment, participants were presented with a short video (20 s) of objects either moving randomly or interacting socially, and had to respond with one of three options: 1) objects had a social interaction, 2) objects had no social interaction, or 3) unsure (Van Essen et al., 2013).
fMRI-CPCA (Percival et al., 2020) was used to extract brain networks from fMRI variance related to task timing. Four components were found and classified as the auditory attention for response network, sustained attention network, re-evaluation network, and traditional default mode network. Features of each network's estimated hemodynamic response were extracted to describe brain activity (Figure 1).
This project collected behavioural data such as demographics, emotions, personality, cognition, and task performance. The predictor variables were made up of the brain network parameters, while criterion variables were made up of behavioural data. First, variance in the criterions was constrained to what is explained by the predictors.(Hunter & Takane, 2011). Next, components were extracted to summarize overlap between the two sets of variables. Finally, rigorous split-half reliability and permutation testing were performed to determine specific, reliable relationships between items.

Activity at the end of the trial in the sustained attention network, when participants watched videos denoting social interaction, was related to perceptions of social support (predictor loading=0.57, predictor loading reliability proportion=0.82, p=0.020). Specifically, perceptions of emotional (0.20) and instrumental support (0.15) were positively associated with this network activity, while perceptions of rejection (-0.19), loneliness (-0.13), and hostility (-0.18) were negatively associated with this network activity.

In this study, we aimed to delineate fine relationships between two sets of variables describing brain activity and behaviour, specifically examining relationships between network variables extracted in a social cognition task and emotions. The association with end-trial activity in the sustained attention network in the social condition suggests that people who exhibit more hostile traits stop attending to social stimuli faster than people who have higher perceived social support. These findings shed light on the detailed interplay between brain activity and behavior, in particular, emphasizing the relevance of Theory of Mind in deciphering how individuals with varying social traits engage with and process social stimuli.

Social Cognition

Activation (eg. BOLD task-fMRI) ¹

Multivariate Approaches ²

Attention: Visual

ADULTS

Cognition

Data analysis

Emotions

FUNCTIONAL MRI

Multivariate