Meta-RegGNN: Predicting Verbal and Full-Scale Intelligence Scores using Graph Neural Networks and Meta-Learning

TL;DR

Meta-RegGNN leverages graph neural networks and meta-learning to accurately predict intelligence scores from brain connectomes, addressing heterogeneity and topological properties for improved generalization across neurotypical and autistic populations.

Contribution

The paper introduces Meta-RegGNN, a novel meta-learning based GNN approach that enhances behavioral score prediction from brain connectomes, especially for heterogeneous and autistic subjects.

Findings

Outperforms existing methods in IQ prediction.

Ensures better generalization for autistic subjects.

Effective with limited training data.

Abstract

Decrypting intelligence from the human brain construct is vital in the detection of particular neurological disorders. Recently, functional brain connectomes have been used successfully to predict behavioral scores. However, state-of-the-art methods, on one hand, neglect the topological properties of the connectomes and, on the other hand, fail to solve the high inter-subject brain heterogeneity. To address these limitations, we propose a novel regression graph neural network through meta-learning namely Meta-RegGNN for predicting behavioral scores from brain connectomes. The parameters of our proposed regression GNN are explicitly trained so that a small number of gradient steps combined with a small training data amount produces a good generalization to unseen brain connectomes. Our results on verbal and full-scale intelligence quotient (IQ) prediction outperform existing methods in both neurotypical and autism spectrum disorder cohorts. Furthermore, we show that our proposed approach ensures generalizability, particularly for autistic subjects. Our Meta-RegGNN source code is available at https://github.com/basiralab/Meta-RegGNN.