Current best practices and future opportunities for reproducible findings using large-scale neuroimaging in psychiatry

Abstract

Research into the brain basis of psychopathology is challenging due to the heterogeneity of psychiatric disorders, extensive comorbidities, underdiagnosis or overdiagnosis, multifaceted interactions with genetics and life experiences, and the highly multivariate nature of neural correlates. Therefore, increasingly larger datasets that measure more variables in larger cohorts are needed to gain insights. In this review, we present current "best practice" approaches for using existing databases, collecting and sharing new repositories for big data analyses, and future directions for big data in neuroimaging and psychiatry with an emphasis on contributing to collaborative efforts and the challenges of multi-study data analysis.

Fig. 1. Overview of differences between Federated and Centralized Data Resources.

Acronyms: BIDS (Brain Imaging Data Structure; https://bids.neuroimaging.io/), COINSTAC (Collaborative Informatics and Neuroimaging Suite Toolkit for Anonymized Computation) [132], ETL (extract, transform, and load operations), MTA (material transfer agreement), DUA (data use agreement).

Fig. 2. Continuous and virtuous research circles.

From inside (lighter gray) to outside (darker gray), we show how adding more data can continuously add to research findings and lead to new research directions and opportunities, by allowing more and more questions to be asked. In this diagram, “Question/Hypothesis 1” is asked of “Data 1” to get “Findings 1.” As more data are added (Data 2, Data 3), this question can be asked continuously to update the findings. Findings 1 may then lead to novel questions asked in “Question/Hypothesis 2” that can be addressed with more data “Data 1+ Data 2,” while “Data 2” can also be used to address the first question, “Question/Hypothesis 1” alongside “Data 1.” “Finding 2” may then lead to a new “Question/Hypothesis 3” that can be answered only with the addition of a third dataset “Data 3” to the original two. Not all datasets need to be used to answer every question, for example “Data 3” in this case is not used for “Question/Hypothesis 2” to obtain “Findings 2”.