Connectome-based machine learning models are vulnerable to subtle data manipulations

Power and reproducibility in the external validation of brain-phenotype predictions

Brain-phenotype predictive models seek to identify reproducible and generalizable brain-phenotype associations. External validation, or the evaluation of a model in external datasets, is the gold standard in evaluating the generalizability of models in neuroimaging. Unlike typical studies, external validation involves two sample sizes: the training and the external sample sizes. Thus, traditional power calculations may not be appropriate. Here we ran over 900 million resampling-based simulations in functional and structural connectivity data to investigate the relationship between training sample size, external sample size, phenotype effect size, theoretical power and simulated power. Our analysis included a wide range of datasets: the Healthy Brain Network, the Adolescent Brain Cognitive Development Study, the Human Connectome Project (Development and Young Adult), the Philadelphia Neurodevelopmental Cohort, the Queensland Twin Adolescent Brain Project, and the Chinese Human Connectome Project; and phenotypes: age, body mass index, matrix reasoning, working memory, attention problems, anxiety/depression symptoms and relational processing. High effect size predictions achieved adequate power with training and external sample sizes of a few hundred individuals, whereas low and medium effect size predictions required hundreds to thousands of training and external samples. In addition, most previous external validation studies used sample sizes prone to low power, and theoretical power curves should be adjusted for the training sample size. Furthermore, model performance in internal validation often informed subsequent external validation performance (Pearson's r difference <0.2), particularly for well-harmonized datasets. These results could help decide how to power future external validation studies.

Using a deep generation network reveals neuroanatomical specificity in hemispheres

Asymmetry is an important property of brain organization, but its nature is still poorly understood. Capturing the neuroanatomical components specific to each hemisphere facilitates the understanding of the establishment of brain asymmetry. Since deep generative networks (DGNs) have powerful inference and recovery capabilities, we use one hemisphere to predict the opposite hemisphere by training the DGNs, which automatically fit the built-in dependencies between the left and right hemispheres. After training, the reconstructed images approximate the homologous components in the hemisphere. We use the difference between the actual and reconstructed hemispheres to measure hemisphere-specific components due to asymmetric expression of environmental and genetic factors. The results show that our model is biologically plausible and that our proposed metric of hemispheric specialization is reliable, representing a wide range of individual variation. Together, this work provides promising tools for exploring brain asymmetry and new insights into self-supervised DGNs for representing the brain.

Power and reproducibility in the external validation of brain-phenotype predictions

Identifying reproducible and generalizable brain-phenotype associations is a central goal of neuroimaging. Consistent with this goal, prediction frameworks evaluate brain-phenotype models in unseen data. Most prediction studies train and evaluate a model in the same dataset. However, external validation, or the evaluation of a model in an external dataset, provides a better assessment of robustness and generalizability. Despite the promise of external validation and calls for its usage, the statistical power of such studies has yet to be investigated. In this work, we ran over 60 million simulations across several datasets, phenotypes, and sample sizes to better understand how the sizes of the training and external datasets affect statistical power. We found that prior external validation studies used sample sizes prone to low power, which may lead to false negatives and effect size inflation. Furthermore, increases in the external sample size led to increased simulated power directly following theoretical power curves, whereas changes in the training dataset size offset the simulated power curves. Finally, we compared the performance of a model within a dataset to the external performance. The within-dataset performance was typically within r=0.2 of the cross-dataset performance, which could help decide how to power future external validation studies. Overall, our results illustrate the importance of considering the sample sizes of both the training and external datasets when performing external validation.

Identifiability in Functional Connectivity May Unintentionally Inflate Prediction Results

Functional magnetic resonance (fMRI) is an invaluable tool in studying cognitive processes in vivo. Many recent studies use functional connectivity (FC), partial correlation connectivity (PC), or fMRI-derived brain networks to predict phenotypes with results that sometimes cannot be replicated. At the same time, FC can be used to identify the same subject from different scans with great accuracy. In this paper, we show a method by which one can unknowingly inflate classification results from 61% accuracy to 86% accuracy by treating longitudinal or contemporaneous scans of the same subject as independent data points. Using the UK Biobank dataset, we find one can achieve the same level of variance explained with 50 training subjects by exploiting identifiability as with 10,000 training subjects without double-dipping. We replicate this effect in four different datasets: the UK Biobank (UKB), the Philadelphia Neurodevelopmental Cohort (PNC), the Bipolar and Schizophrenia Network for Intermediate Phenotypes (BSNIP), and an OpenNeuro Fibromyalgia dataset (Fibro). The unintentional improvement ranges between 7% and 25% in the four datasets. Additionally, we find that by using dynamic functional connectivity (dFC), one can apply this method even when one is limited to a single scan per subject. One major problem is that features such as ROIs or connectivities that are reported alongside inflated results may confuse future work. This article hopes to shed light on how even minor pipeline anomalies may lead to unexpectedly superb results.