Carnegie Mellon University
VanGulick_Borghi_OHBM18.pdf (205.47 kB)

Surveying data management practices and perceptions among neuroimaging researchers

Download (205.47 kB)
posted on 2018-06-19, 00:00 authored by Ana Van GulickAna Van Gulick, John BorghiJohn Borghi

Poster presented at the annual meeting of the Organization for Human Brain Mapping in Singapore in June 2018.

Introduction: The processes involved in using MRI data to answer questions about human neuroscience are complex - involving large datasets, multiple decision points, and a variety of software and analysis tools. In recent years, growing interest in open science and reproducibility has led the neuroimaging community to develop a range of new tools, platforms, and standards to support managing, sharing, and reusing valuable data (Nichols, 2017; Poldrack, 2017). While careful research data management (RDM) is clearly important given the complexities of MRI research, this study is the first to systematically examine the data-related practices and perceptions of active MRI researchers from a data curation perspective.

Methods: This assessment was done using an online survey comprised of 74 multiple choice questions, designed in consultation with active researchers in the field to elucidate the RDM practices at various stages of a typical MRI research project including planning, data collection, analysis, and publication. Questions focused on a range of RDM topics, including the type of data collected, analytical and research management tools, and the degree to which practices are standardized within a research group. Several questions also inquired about factors that motivate and limit their ability to manage data well, while the final section covered emerging scholarly communication practices including publishing pre-prints, datasets, and other research products. The survey was distributed to active MRI researchers in summer 2017 through email and social media. 144 subjects, a mix of graduate student, postdoc, and faculty researchers, mostly in the U.S. with some in Canada and Europe, completed a significant portion of the survey and are included in our results.

Results: In general, we found that the research practices employed in MRI research are highly heterogeneous and complex. Our results suggest that there is a large variety of data types and formats, software platforms and versions, and analysis and documentation tools being used in MRI research and that there is a lack of consistency in workflows, even within research groups and across projects. For example, in regards to documenting data analysis choices and versioning data or code, researchers frequently responded that their RDM practices were not the same as other lab members or collaborators. The respondents in our survey consistently rated the maturity of their practices as higher than the field as a whole (see Figure), which may be due to sampling a population more aware of open science practices. Faculty also rated their practices as more standardized than trainees, which may reflect different knowledge of true practices or variations in RDM maturity. Perceptions of new scholarly communications practices including data sharing, data reuse, and Open Access publishing, were largely positive but demonstrated little adoption of these activities into practice. Participants consistently reported that their RDM and open science practices are limited by the amount of time these activities take, and by the lack of training, community standards, and professional incentives. The most common motivations are avoiding data loss and providing access to everyone in a lab group.

Conclusions: Overall, we found a need for more training in RDM among MRI researchers, support for software and data tools that link together across the span of a research project, and more robust community standards for data documentation, processing, and sharing, and greater incentives for taking on this work. We hope that the survey instrument and results can be used to extend the understanding of current research practices and perceptions in neuroimaging and to build interoperable tools to better support open science in the field.

Related Materials: The survey and dataset as well as a preprint of the manuscript describing this work can be found in the references below.