Portal de Periódicos da CAPES

P1‐232: THE HIVE DB IMAGE DATA MANAGEMENT AND ANALYSIS FRAMEWORK

2014; Wiley; Volume: 10; Issue: 4S_Part_10 Linguagem: Inglês

10.1016/j.jalz.2014.05.471

1552-5279

Andrew Simmons, J‐Sebastian Muehlboeck, Eric Westman,

Distributed and Parallel Computing Systems

The hive database system (theHiveDB) is a web-based brain imaging management framework for cross-sectional and longitudinal multi-center studies. It has been conceived with a focus on data aggregation across modalities (i.e. brain imaging, clinical and genetic data). TheHiveDB has been designed to manage imaging projects, individuals (study participants), scalar data and associated file assets.Data files pertaining to different projects may be stored on distinct networked computers. The system's activity and resource management component is capable of distributing processing across local computing resources and both private and public clouds. Data transfers are automated using the SSH-2 protocol. The system provides a framework for effective collaborations and resource sharing. It facilitates access to data at all levels, by providing pertinent meta information about image acquisitions, allowing to extract individual series in various formats from DICOM studies and offering direct file download and transfer to workstations for source data and processed output. The system supports a rich set of common tasks from image archival and format conversion to established processing pipelines, such as Freesurfer for extraction of scalar measures from magnetic resonance images (MRI). All file assets managed by the system are identified by means of unique identifiers. Checksum information about all files is stored in the database to warrant both file integrity and authenticity.Scalar data (variables) are associated with individuals, individuals at timepoints, or may be derived from assets or (e.g. activity output). Query functionality is delivered without the need to modify the database schema, as the data query module is based on entity-attribute-value (EAV) modeling. Queryable parameters are described as variables (with exportable data dictionary) and grouped into variable collections (e.g. questionnaire). An interactive query composition interface lets the user combine data across all modalities and download tabular data. N/A N/A