Predicting cognitive scores with graph neural networks through sample selection learning

Abstract

Analyzing the relation between intelligence and neural activity is of the utmost importance in understanding the working principles of the human brain in health and disease. In existing literature, functional brain connectomes have been used successfully to predict cognitive measures such as intelligence quotient (IQ) scores in both healthy and disordered cohorts using machine learning models. However, existing methods resort to flattening the brain connectome (i.e., graph) through vectorization which overlooks its topological properties. To address this limitation and inspired from the emerging graph neural networks (GNNs), we design a novel regression GNN model (namely RegGNN) for predicting IQ scores from brain connectivity. On top of that, we introduce a novel, fully modular sample selection method to select the best samples to learn from for our target prediction task. However, since such deep learning architectures are computationally expensive to train, we further propose a learning-based sample selection method that learns how to choose the training samples with the highest expected predictive power on unseen samples. For this, we capitalize on the fact that connectomes (i.e., their adjacency matrices) lie in the symmetric positive definite (SPD) matrix cone. Our results on full-scale and verbal IQ prediction outperforms comparison methods in autism spectrum disorder cohorts and achieves a competitive performance for neurotypical subjects using 3-fold cross-validation. Furthermore, we show that our sample selection approach generalizes to other learning-based methods, which shows its usefulness beyond our GNN architecture.

Early prediction of dementia using fMRI data with a graph convolutional network approach

“…Graphs [39] are a kind of non-Euclidean data structure composed by a set of nodes and edges, where nodes represent objects and edges represent the relationship between objects. Brain FC can be modeled as graphs, where nodes represent ROIs and edges correspond to correlations in activity between these ROIs [40][41][42]. GNNs [21,43] have recently become a widely used machine learning tool in graph analysis due to their persuasive performance.…”

Hand function after neonatal stroke: A graph model based on basal ganglia and thalami structure

“…GNN-based strategies are becoming increasingly popular in the medical field. Thus, recently, GNNs were utilized for graph regression to predict IQ scores from graphs that represent brain connectivity ( Hanik et al, 2022 ). In previous work, we introduced a GNN-based classification model for predicting cerebral palsy in children with NAIS ( Coupeau et al, 2023 ).…”

StawGAN: Structural-Aware Generative Adversarial Networks for Infrared Image Translation

“…As part of preprocessing, techniques for identifying important and explanatory training examples have been developed for natural images [15], [16] aiming at discarding superfluous data samples and improving the generalization ability of neural networks. Similarly, several works try to establish the most representative patches (or whole sample) to be involved in the training of medical deep learning approaches showing the ability to improve models performance [17], [18]. Simultaneously, a further approach consists of fusing different image modalities in a single and more informative sample [19], [20].…”