Introduction to using XNAT as an Institutional Repository
Imaging-based research is becoming increasingly important to research institutions. New imaging technologies are yielding insights into the realms of cancer and disease pathology, and research funding is growing rapidly for imaging projects.
The challenge for research institutions is to be able to manage this rapid growth of research data. Imaging data is exponentially larger than clinical data, and its analysis requires multiple rounds of computationally demanding processing. Moreover, the systems designed to store imaging data that are directly attached to scanners have not kept up with complex demands of imaging research. They can store files, but do little else, and offer no support for integrating imaging data with any associated subject metadata.
This is why XNAT was created.XNAT is a web-based software platform designed to facilitate common management and productivity tasks for in vivo imaging and associated data. It consists of an image repository to store raw and post-processed images, a database to store metadata and non-imaging measures, and user interface tools for accessing, querying, visualizing, and exploring data.
XNAT supports all common imaging methods (e.g. MRI, CT, PET), and its data model can be extended to capture virtually any related metadata (e.g. demographics, genetics). XNAT includes a DICOM workflow to enable exams to be sent directly from scanners, PACS, and other DICOM devices. XNAT's web application provides a number of quality control and productivity features, including data entry forms, searching, reports of experimental data, upload/download tools, access to standard laboratory processing pipelines, and an online image viewer.
The largest current installation of XNAT is the CNDA at Washington University, which currently supports more than 800 individual research projects, consisting of more than 16,000 subjects and 20,000 imaging sessions. The CNDA has more than 200 active users, and imports data from more than a dozen research centers around the world.
Why did you install XNAT?
(As a preface, it should be noted that the decision to try XNAT was made before Adam joined the team; however, that process was slow-going and not very well invested before Adam took it over.)
This instance of XNAT at the University of Iowa was set to be a "research PACS" for the many researchers and investigators at Iowa's Institute of Clinical Translation Services (ICTS). Unlike the XNAT that supports the PREDICT-HD project, this XNAT is not centrally managed to be a single canonical source of data, or to support a single data workflow, but to provide a set of widely available research tools & services. Users immediately see value in having an easy and portable way of establishing a secure means of accessing data, using the XNAT web application.
Installation was a several month process, playing with the configuration and testing with users, and waiting for security fixes before going live. Currently, this XNAT contains a mix of active data with archival data that is seen to have future use. "Everyone sees it as an archival system.... Long term storage." At this point, the goal of this XNAT is to "get people's data in, make sure we can accommodate it, and make sure they can access it."
Down the road, other use cases will emerge. For example, there are plans to create and customize an XNAT for ophthalmology research, as a stand-alone instance (but linked to same storage).
Who are our primary users?
There are very disparate levels of ability & knowledge & code savvy among users. A lot of python-savvy people are excited about using XNAT as a platform to fuel their research. However, most users are researchers attempting to use the web interface to browse through their data. (Mark Scully helped researchers upload all their old data into XNAT, having just done the same thing for the PREDICT-HD project.)
Uploaders: MRI techs send five sessions per day every day. However, this data is still stored in DICOM PACS for "query retrieve" purposes. There is also interest in some of the DICOM functionality provided by XNAT Gateway.
The user-management model is 180 degrees from the PREDICT-HD XNAT. That project is managed as a "monarchy" - can dictate usage policy centrally. This XNAT is a "service provider" -- when features we offer are attractive, researchers use them. As such, our strategy for driving user adoption is to focus on power users within the community who naturally drive wider adoption when their needs are met.
For some users, such as our research MR imaging group, XNAT is immediately useful for data accessibility out-of-the-box. Others, such as our ophthalmology group, engage with us to make XNAT meet their needs.
What inherent features of XNAT do your users find most valuable?
Being able to store research data securely is crucial -- imperative for researchers. (Security in this case refers to protection against data loss. A hardware setup combined with software policy that supports data integrity is key.) ICTS does a lot of IT work, flexibly managing hardware for a number of different groups.
Researchers also see value in having a way to search through & display metadata that they're interested in, helping them build pools of subjects, sessions, etc. to support research.
How have you customized XNAT to meet your needs?
Added an ophthalmic CT data model. Very little other customization to the XNAT production instance, but open to user requests.
How could XNAT improve for the future?
- Would love to be able to provide more help for uploaders.
- From Adam's perspective as a system admin: "under the hood" changes to maintenance, eliminating the deployment process, would be invaluable. Currently the process requires a lot of refactoring. This wouldn't improve the product for end users, but would make it much easier, much less costly to install. Despite "free" price tag, there is a large cost of development and deployment attached. The XNAT software project seems ready to grow from an open-sourced product by WUSTL into a community project heavily contributed to by WUSTL.
- Web UI improvements will drive user adoption.
- So will scriptability, rapid application development, to support web developers.