The open-source neuroimaging research enterprise

Abstract

While brain imaging in the clinical setting is largely a practice of looking at images, research neuroimaging is a quantitative and integrative enterprise. Images are run through complex batteries of processing and analysis routines to generate numeric measures of brain characteristics. Other measures potentially related to brain function - demographics, genetics, behavioral tests, neuropsychological tests - are key components of most research studies. The canonical scanner - PACS - viewing station axis used in clinical practice is therefore inadequate for supporting neuroimaging research. Here, we model the neuroimaging research enterprise as a workflow. The principal components of the workflow include data acquisition, data archiving, data processing and analysis, and data utilization. We also describe a set of open-source applications to support each step of the workflow and the transitions between these steps. These applications include DIGITAL IMAGING AND COMMUNICATIONS IN MEDICINE viewing and storage tools, the EXTENSIBLE NEUROIMAGING ARCHIVE TOOLKIT data archiving and exploration platform, and an engine for running processing/analysis pipelines. The overall picture presented is aimed to motivate open-source developers to identify key integration and communication points for interoperating with complimentary applications.

Figures

Fig 1.

The neuroimaging enterprise workflow.

Fig 2.

The data capture tools place incoming images into “prearchives” that can be accessed by data archive applications. The archive application securely stores the images and distributes them to various users and applications.

Fig 3.

DicomBrowser allows users to view and deidentify DICOM image files and to send them to DICOM storage providers. Deidentification can be done manually by entering values into the appropriate header fields or automatically using script files.

Fig 4.

XNAT is a three-tiered application for securely archiving, exploring, and distributing neuroimaging and related data. Its extensible XML data model allows the database to capture study-specific data types.

Fig 5.

The image viewer built into the XNAT web application uses Java applet technology to distribute images over the web. Its plug-in design enables developers to create new display modes for neuroimaging images. Here, a sagittal view of a T1 image is shown next to a transverse view of a FreeSurfer segmentation of the T1 image.