Predicting Real-Life Cognitive Scores From Functional Connectivity.

Area of Science:

• Neuroscience
• Cognitive Science
• Psychology

Background:

• Functional connectivity (FC) from fMRI is widely used to predict cognitive traits.
• Previous studies often used lab-based cognitive measures, limiting real-world applicability.
• Ecologically valid assessments are needed to understand cognitive performance in naturalistic settings.

Purpose of the Study:

• To investigate connectome-based predictions of cognitive performance in real-world settings.
• To assess the predictive power of functional connectivity for a standardized academic entrance exam.
• To identify specific connectivity features contributing to different cognitive domains.

Main Methods:

• Collected fMRI data from 194 participants.
• Utilized functional connectivity patterns to predict performance on the Psychometric Entrance Test (quantitative reasoning, verbal reasoning, foreign language proficiency).
• Employed four distinct prediction approaches and analyzed edge- and node-level connectivity contributions.
• Implemented a transfer learning approach for global score prediction.

Main Results:

• Functional connectivity significantly predicted overall test performance and its three cognitive domains.
• Predictions were robust across multiple analytical methods.
• Distinct cognitive abilities were associated with unique connectivity patterns, with feature similarity correlating with behavioral score correlations.
• Transfer learning improved global score prediction compared to direct FC prediction.

Conclusions:

• The functional connectome effectively captures real-world variability in global and domain-specific cognitive abilities.
• Brain connectivity serves as a potential objective marker for real-world cognitive performance.
• This approach advances the understanding of brain-behavior relationships in ecologically valid contexts.