A National Human Neuroimaging Collaboratory Enabled by the Biomedical Informatics Research Network (BIRN)

The aggregation of imaging, clinical, and behavioral data from multiple independent institutions and researchers presents both a great opportunity for biomedical research as well as a formidable challenge. Many research groups have well-established data collection and analysis procedures, as well as data and metadata format requirements that are particular to that group. Moreover, the types of data and metadata collected are quite diverse, including image, physiological, and behavioral data, as well as descriptions of experimental design, and preprocessing and analysis methods. Each of these types of data utilizes a variety of software tools for collection, storage, and processing. Furthermore sites are reluctant to release control over the distribution and access to the data and the tools. To address these needs, the biomedical informatics research network (BIRN) has developed a federated and distributed infrastructure for the storage, retrieval, analysis, and documentation of biomedical imaging data. The infrastructure consists of distributed data collections hosted on dedicated storage and computational resources located at each participating site, a federated data management system and data integration environment, an extensible markup language (XML) schema for data exchange, and analysis pipelines, designed to leverage both the distributed data management environment and the available grid computing resources.

I. Introduction

The Field of biomedical imaging has undergone explosive growth in the last 50 years, driven in part by emergent imaging technologies and the information superhighway. The ability to collect vast amounts of complimentary data has empowered scientists to ask multifaceted questions of the data that was difficult at best in earlier years. In recent years, this data explosion has grown, as more publicly accessible data have become available. Due to the size and specialized nature of certain individual datasets, multidisciplinary teams of domain scientists are required to work together toward a common goal of understanding the illness and the disease. Complicating this endeavor is the reality that these scientists are geographically dispersed, necessitating effective means for domain scientists to collaborate on such multidisciplinary research problems. Improvements in critical information technologies and internet speeds have made the construction of a national collaboratory of federated and distributed data resources connected through a common semantic fabric viable.