28 September 2020 in Nature Neuroscience
6 April 2023 in Nature Medicine
NLNicola T. Lautenschlager
1 July 2019 in Biological Psychiatry
22 September 2020 in Molecular Psychiatry
1 February 2021 in NeuroImage
While the function of most biological systems is tightly constrained by their structure, current evidence suggests that coupling between the structure and function of brain networks is relatively modest. We aimed to investigate whether the modest coupling between connectome structure and function is a fundamental property of nervous systems or a limitation of current brain network models. We developed a new deep learning framework to predict an individual's brain function from their structural connectome, achieving prediction accuracies that substantially exceeded state-of-the-art biophysical models (group: R=0.9±0.1, individual: R=0.55±0.1). Crucially, brain function predicted from an individual's structural connectome explained significant inter-individual variation in cognitive performance. Our results suggest that structure-function coupling in human brain networks is substantially tighter than previously suggested. We establish the margin by which current brain network models can be improved and demonstrate how deep learning can facilitate investigation of relations between brain function and behavior.
1 December 2021 in NeuroImage
Cognitive performance can be predicted from an individual's functional brain connectivity with modest accuracy using machine learning approaches. As yet, however, predictive models have arguably yielded limited insight into the neurobiological processes supporting cognition. To do so, feature selection and feature weight estimation need to be reliable to ensure that important connections and circuits with high predictive utility can be reliably identified. We comprehensively investigate feature weight test-retest reliability for various predictive models of cognitive performance built from resting-state functional connectivity networks in healthy young adults (n=400). Despite achieving modest prediction accuracies (r=0.2-0.4), we find that feature weight reliability is generally poor for all predictive models (ICC< 0.3), and significantly poorer than predictive models for overt biological attributes such as sex (ICC≈0.5). Larger sample sizes (n=800), the Haufe transformation, non-sparse feature selection/regularization and smaller feature spaces marginally improve reliability (ICC< 0.4). We elucidate a tradeoff between feature weight reliability and prediction accuracy and find that univariate statistics are marginally more reliable than feature weights from predictive models. Finally, we show that measuring agreement in feature weights between cross-validation folds provides inflated estimates of feature weight reliability. We thus recommend for reliability to be estimated out-of-sample, if possible. We argue that rebalancing focus from prediction accuracy to model reliability may facilitate mechanistic understanding of cognition with machine learning approaches.