The information content of any functional imaging dataset far exceeds what any single investigative group can extract. This insight spurred creation of the fMRI Data Center, which was established to share fMRI data but not embraced by the neuroimaging community. Amid calls for data sharing from research funders, there is also increasing awareness of the problems inherent in underpowered samples: low likelihood of replication and inability to appreciate clinical heterogeneity. While applicable to psychiatric research in general, these challenges are particularly germane to Autism Spectrum Disorders (ASD) – which are marked by complex clinical manifestations and impressive heterogeneity. The recent emergence of resting state fMRI (R-fMRI) as a mainstream imaging modality has revived momentum towards open data sharing. Over the past four years, the 1000 Functional Connectomes Project (FCP) and the International Data-sharing Initiative (INDI) have pioneered an open neuroscience solution to the challenge of amassing large-scale clinical neuroimaging datasets efficiently. The initial FCP release of 1300+ R-fMRI datasets demonstrated the feasibility and utility of aggregating and openly sharing previously collected neuroimaging datasets for the purposes of comparing findings across sites and carrying out large-scale discovery analyses. More recently, the Attention Deficit Hyperactivity Disorder (ADHD)-200 initiative demonstrated this model could be applied to clinical research to demonstrate unique neural signatures underlying clinical subtypes. Based on the promise and success of this investigator initiated efforts, the Autism Brain Imaging Data Exchange (ABIDE) was founded.
Objectives:
To generate and provide a platform for a large-scale R-fMRI and morphometry data aggregation for ASD: ABIDE.
Methods:
Investigators willing to openly share awake R-fMRI data from at least 15 individuals with ASD and 15 matched typical controls were invited to participate in ABIDE. Institutional IRB approval or waiver was required prior to data contribution. Basic phenotypic measures that are standard in the ASD field (e.g., age at scan, sex, IQ, DSM-IV TR diagnosis, ADOS and ADI-R scores) were also aggregated for analyses and sample characterization. Scripts for full anonymization and a pipeline for data integrity checking and aggregation were developed. ABIDE data aggregation was based on the INDI platform.
Results:
Twenty previously collected autism brain imaging samples were received from 16 independent imaging sites, yielding a total of 1112 datasets (from ages 7 to 64). Following data upload, aggregation, organization, and verification, we publically released the ABIDE dataset on Aug 30, 2012 via www.nitrc.org at (http://fcon_1000.projects.nitrc.org/indi/abide/).
Conclusions:
The FCP/INDI model was successfully extended to the autism field. Similar to the initial FCP dataset which has yielded more than 40 publications since 2010, and the ADHD-200 release which has yielded 12 publications in the last year, we expect to witness a rapid proliferation of manuscripts examining the neural correlates of ASD using the large-scale ABIDE dataset. The ABIDE dataset will not only allow the testing and generation of hypotheses using an adequately powered imaging dataset, but facilitate the exploration of neural and clinical subtypes and the potentially confounding effects of variables such as those related to IQ.