Learning Neural Representations of Human Cognition across Many fMRI Studies

TL;DR

This paper introduces a machine learning framework that integrates diverse fMRI datasets to learn universal neural representations of cognition, improving prediction accuracy and interpretability across multiple studies.

Contribution

It presents a scalable multi-task learning approach with multi-scale dimension reduction to unify heterogeneous brain imaging data into cognitive-relevant low-dimensional representations.

Findings

Achieves superior prediction performance on large reference datasets.

Enhances analysis accuracy for small datasets.

Enables identification of universal cognitive concepts.

Abstract

Cognitive neuroscience is enjoying rapid increase in extensive public brain-imaging datasets. It opens the door to large-scale statistical models. Finding a unified perspective for all available data calls for scalable and automated solutions to an old challenge: how to aggregate heterogeneous information on brain function into a universal cognitive system that relates mental operations/cognitive processes/psychological tasks to brain networks? We cast this challenge in a machine-learning approach to predict conditions from statistical brain maps across different studies. For this, we leverage multi-task learning and multi-scale dimension reduction to learn low-dimensional representations of brain images that carry cognitive information and can be robustly associated with psychological stimuli. Our multi-dataset classification model achieves the best prediction performance on several large reference datasets, compared to models without cognitive-aware low-dimension representations, it brings a substantial performance boost to the analysis of small datasets, and can be introspected to identify universal template cognitive concepts.