FMRI-based brain signature of divergent thinking
Presented During:
Monday, June 24, 2024: 5:45 PM - 7:00 PM
COEX
Room:
Hall D 2
Poster No:
955
Submission Type:
Abstract Submission
Authors:
Cheng Liu1, KAIXIANG ZHUANG2, XUEYANG WANG1, Jiang Qiu1
Institutions:
1Southwest University, Chongqing, China, 2The Institute of Science and Technology for Brain-inspired Intelligence (ISTBI), Shanghai, China
First Author:
Co-Author(s):
KAIXIANG ZHUANG
The Institute of Science and Technology for Brain-inspired Intelligence (ISTBI)
Shanghai, China
The Institute of Science and Technology for Brain-inspired Intelligence (ISTBI)
Shanghai, China
Introduction:
Divergent thinking constitutes a vital component of creativity – a complex cognitive process that necessitates the collaborative engagement of multiple brain regions involved in distinct functions (1, 2). Prior studies employed connectivity measures at rest and have implicated the involvement of default, salience and executive systems (3). However, the neural signature of divergent thinking during task performance remains elusive, requiring further characterization of this higher cognitive process. Here, we employed fMRI data from two large samples in conjunction with machine learning techniques to identify and delineate a neural marker capable of predicting divergent thinking ability both at the group and individual levels. We then further described this marker in the context of cortical connectivity gradients and meta-analytic decoding to unravel its architectural principals within the hierarchical organization of the human brain.
Methods:
Across two study samples (n=55 and n=31), we acquired fMRI data while participants performed an AUT inside the scanner. In this task, participants were asked to generate either a "novel" or a "general" use example for an everyday object. Outside the scanner, participants were then asked to rate the originality of their responses on a 1-5 Likert scale. For data analysis, we used an MVPA-based neural decoding technique to identify a brain pattern that successfully classified the two conditions (subject-level beta map) with highest accuracy. To investigate the organizational principles of this neural signature, we then conducted a spatial correlation analysis between the brain pattern and cortical connectivity gradients, and used cognitive terms from the Neurosynth database for meta-analytic decoding. Finally, to assess the generalizability of our findings from group to individual-level, we tested the accuracy of the brain pattern in predicting originality ratings using relevance vector regression.
Results:
To identify the multivariate patterns of fMRI activation, we applied linear SVMs to discriminate novel and general use conditions. The classification models have high accuracy that is 80%±3.8% on sample1 and 85%±4.6% on sample2. Notably, the weighted average of two models was calculated as the group-level neural signature of divergent thinking due to high correlation (Figure1A,r=0.838,p<0.001). Novel versus general use prediction weights were positive in bilateral DLPFC, bilateral DMPFC, left VLPFC, bilateral ACC, bilateral OFC, left AG, left MTG, and bilateral thalamus, right cerebellum, and negative in the right SPL, right precuneus, right ILOC (Figure1B). In this brain pattern, feature weights were mainly distributed across the default network and frontal-parietal control networks (Figure1C), associated with higher cognitive processes, such as judgment, retrieval, memory and semantic (Figure1D). More importantly, there was a high correlation between brain patterns and the principal connectivity gradient (r=0.60,p<0.0001), suggesting that regions closer to a segment of the default network may play an important role in divergent thinking (Figure1E). Finally, to validate the effectiveness of the brain pattern as biomarker for predicting divergent thinking, we applied RVR with single-trial beta maps (only novel use condition for each subject) as features to predict originality ratings. The distribution of the correlation (r) between the predicted and true value in 10×10-fold cv ranged from 0.21-0.39 on sample1 (Figure2A), and 0.09-0.17 on sample2 (Figure2B).
·Figure 1. Brain pattern sensitive to predict divergent thinking. (A) Pearson correlation between thresholded feature weights of the two models across two samples. (B) The weight map shows the final p
·Figure 2. Divergent thinking brain pattern predict AUT self-rating score. (A) The distribution of the correlation(r) between the predicted value and the true value in 10×10-fold cross-validations on s
Conclusions:
Our analysis has identified a comprehensive neural representation and organizational principle of divergent thinking. We show that divergent thinking is inextricably linked to a variety of higher cognitive processes and that its neural patterns are organized in the default and frontoparietal control networks in a manner that is consistent with the principal gradient of functional connectivity.
Brain Stimulation:
Non-Invasive Stimulation Methods Other
Higher Cognitive Functions:
Executive Function, Cognitive Control and Decision Making
Higher Cognitive Functions Other 1
Modeling and Analysis Methods:
Classification and Predictive Modeling 2
Novel Imaging Acquisition Methods:
BOLD fMRI
Keywords:
Cognition
Cortex
Data analysis
Meta-Cognition
Multivariate
1|2Indicates the priority used for review
Provide references using author date format
2. Kenett YN. (2018), 'Flexibility of thought in high creative individuals represented by percolation analysis', Proceedings of the National Academy of Sciences, 115(5):867-72.
3. Beaty RE. (2018), 'Robust prediction of individual creative ability from brain functional connectivity', Proceedings of the National Academy of Sciences, 115(5):1087-92.