Bridging Huntington’s disease research with big data science: Harmonized neuroimaging datasets from multiple studies

Abstract

Multiple neuroimaging datasets from Huntington’s disease (HD) studies are publicly available, but these datasets are in various formats, omit imaging metadata, and sometimes contain corrupt scans. We have created a platform to curate, harmonize, and distribute neuroimaging datasets from eight different studies: TRACK-HD, TRACKOn-HD, PREDICT-HD, IMAGE-HD, HD-YAS, SHIELD-HD, PEARL-HD, and LONGPDE10. The platform is organized into three conceptual levels to serve the research community with both raw and processed data. Raw data are converted into Brain Imaging Data Structure (BIDS) format, while processed data are obtained from pipelines such as Freesurfer and fmriprep. Studies that had followed the same participants were combined. After combining studies, the final six BIDS datasets include a total of 2,216 participants and 7,073 sessions. We outline tools, principles, and recommendations for future data management in HD research.