Variability in fMRI data analysis from different research teams on the same data set makes the scientific findings of similar studies using fMRI less persuasive. We want to comment that this reproductivity issue is associated with the hypothesis testing paradigm, which can be potentially mitigated by a machine learning based predictive modeling approach. Furthermore, such an approach also opens new possibilities in detecting subtle spatial patterns across the brain and in estimating more useful individual effects in neurology and psychiatry.
Keywords: Machine learning; Psychiatry; Reproducibility; fMRI.
Copyright © 2020 Elsevier B.V. All rights reserved.