You are not logged in. (Sign In)

Northwestern University Neuroimaging Data Archive (NUNDA)

PI: Wang

Summary:
To establish a data archive for the neuroimaging community at Northwestern University, which has the capacity to efficiently and securely store collected MR scans, and to later retrieve them in time-sensitive datasets.

External Link: Northwestern University Neuroimaging Data Archive (NUNDA)

Papers:

  • The Northwestern University Neuroimaging Data Archive (NUNDA).
    Alpert K, Kogan A, Parrish T, Marcus D, Wang L.
    NeuroImage 2016; 124: 1131-1136
    View Article

Presentations:

  • Northwestern University Neuroimaging Data Archive "Great Leap Forward"
    Lei Wang, Kathryn Alpert CBMG (Chicago -- March 2012)
    Download Presentation

Funding:

  • Northwestern Memorial Hospital: Northwestern University Neuroimaging Data Archive (NUNDA) (PI: Csernansky/Wang) - To establish a data archive for the neuroimaging community at Northwestern University. (10/01/2008 - 10/01/2011).

Details:
NUNDA was developed jointly by the Departments of Psychiatry and Radiology, modeled after the Washington University Central Neuroimaging Data Archive. NUNDA has been created to integrate neuroimaging research programs, and to create the capacity to efficiently and securely store collected images of target brain structures, and to later retrieve them in time-sensitive datasets. Such datasets would also serve as an invaluable source of images for pilot studies, and support the residency and fellowship training programs in neuroradiology, neurology, neurosurgery, and psychiatry.

NUNDA contains the following image workflow: Acquisition → Archive → Automated Processing & Preliminary Structural Analysis → Integration → Exploration & Discovery. This process begins with data acquisition from subjects in IRB-approved protocols at the scanner. Image files are then transferred to the data archive after de-identification. Annotations are added as needed to identify the subject number, study, scanner, scanning parameters, and any other data needed for later dataset retrieval. The image data are then made available for automated image processing routines, and derived images and quantitative measures are stored within archive datasets so that they can be integrated with demographic, clinical and neurobiological measures. Finally, discovery and productivity tools are used to interact with the integrated database.
Currently, acquisition and archiving is available to CAMRI users. Details are below. Work on other components of the image flow is under way. To register with NUNDA. Once logon privileges are granted, create new project, new subject and new NMR sessions.

Acquisition. The migration from closed in-house data formats to DICOM by scanner manufacturers has created a bounty of opportunity for open-source development of ancillary applications. DICOM is an open industry standard that is massive and comprehensive, and includes specifications for image file formats, data transport, printing, worklists, and querying. We have adopted existing open-source tools which focus on the essential capabilities: receiving images, visualizing images, and accessing/editing image header content. DicomServer supports communication between the scanner and data archive. DicomBrowser is a general-purpose tool for working with DICOM data, with specific emphasis on contributing DICOM images to a research data archive. These tools are built on the open-source dcm4che library. We have customized FTP- and HTTP-based receivers to capture and organize images into a pre-archive and to generate XML in the same manner as DicomServer. We will also create a web-based user interface for uploading images.

Archive. For image archiving, we use the Extensible Neuroimaging Archive Toolkit (XNAT). XNAT is an open source Java-based application. It follows a three-tiered architecture that includes a data archive, user interface, and middleware engine, designed to facilitate management and exploration of neuroimaging and related data. It includes a secure database backend and a rich web-based user interface. XNAT uses an XML data model from which a relational database is generated. XNAT implements a workflow to support the quality, integrity, and security of data from acquisition, and storage of analysis and public sharing. Non-imaging data are entered via web-based forms, spreadsheet uploads, or XML. Newly entered data are placed in a “virtual quarantine” until an authorized user validates that the integrity of the data is intact. Once the data have been validated, they are moved into a secure archive. The archive unifies the data acquired from various sources associated with a study into a single-integrated resource. Archived data are made available to additional authorized users and project-specific automated processing and analysis pipelines. XNAT’s web-based user interface provides tools for monitoring the XNAT workflow and for exploring the resulting archive. As XNAT-managed studies progress, the data can be made available to successively broader groups of users, from collaborators to reviewers to the general scientific community.

Processing & Analysis. We will use an open-source pipeline tool, PipelineRunner, to direct and monitor automated image processing routines to reduce scanner artifacts, correct for head motion, register in atlas space, compensate for systematic, slice-dependent time shifts, and eliminate systematic odd/even slice intensity differences due to interleaved acquisition, brain region segmentation, atrophy measures, and white matter tract labeling. Thus, we will ensure that all image data undergo identical processing and analysis steps in an organized sequence to yield optimal stored images and derived measures. PipelineRunner executes pipelines defined in project-specific XML specification documents that describe the sequence of tasks that constitute a pipeline. It is also capable of pausing and entering pipelines at any step, which is extremely useful if a manual procedure is required or if a task needs to be rerun with different parameters.

Integration. The integration of images, measures derived from them, and non-imaging data will allow users to generate test their hypothesis but also mine datasets for unexpected patterns. XNAT’s search interface is designed to facilitate queries on the integrated database, allowing users to enter search criteria tailored to each data type and to request a result set that joins across data types. The search results would be presented to the user with the option of downloading the associated image data.

Exploration & Discovery. NUNDA will provide a valuable resource for exploring research study data. Discovery tools include data mining applications, image viewers and manipulators, plotting packages, and statistics applications. XNAT includes a simple, open-source web-based image viewer that can be extended to support custom image types. From a workflow perspective, the key to enabling discovery is to simplify the exchange of information and data between the archive and discovery tools. XNAT uses an XML data model, which has also been adopted by the BIRN.

Investigators are encouraged to put the following into their consent forms:
Please note that:
• By signing this consent, you agree that your unidentified research data (information with no names or other personal identifiers) will be stored in the Northwestern University Neuroimaging Data Archive (NUNDA) for research purposes only.