Print This Page

Purpose:

To solicit expert opinion regarding the need, feasibility and approach to a common archival system for storing and accessing imaging databases by the broad scientific community.

Goals:

1. Verify the need for an archive
2. Define the technical requirements for a virtual archive of images to be used in oncology
3. Define the policy issues for using these archives
4. Recommend a process and phases for implementation of a robust imaging archival system
5. To review how this effort could be expanded and coupled with other ongoing NIH wide efforts.
6. To determine if this could be coupled with other professional societies efforts in establishing imaging archives.

Background:

The idea of a common archive for images and related data is being considered because of a number of NCI CIP planned initiatives such as the ACRIN Digital Mammography Project and a recent RFA (NCI CIP: RFA: CA-01-001) entitled Lung Imaging Database Resource for Imaging Research: http://www.nci.nih.gov/bip/NCI-DIPinisumm.htm#all.

In addition the issue of inter-operability of and investigator access to image data bases funded by the different IC's at NIH has been recently discussed at the NIH wide inter-institute meetings on medical imaging. Other organizations are currently addressing similar archiving issues such as RSNA, ACR and DICOM Working Group 18.

The intent was to invite representative's from each organization to outline the issues that need to be considered. It was envisioned that these could be addressed in future NCI initiatives in order to create image databases as research and training resources for the broader community. We solicited input from the attendees on the process that should be used to establish these archives.

Workshop Recommendations:

1. Provide support to establish and maintain a distributed internet-accessible image archive to serve the investigator community at the most appropriate level of service for their specific research and training needs.
2. Populate the archive with data generated by NCI supported clinical trials and voluntarily contributed by individual investigators. Incorporate images and relevant data from NCI cooperative groups by stipulating the generation of archives as a prior requirement for support.
3. Consider the levels of data provided including (a) raw detector measurements, instrument characteristics and calibration information, (b) 2D and 3D static images, (c) Image sequences alone or before and after intervention, episodic and longitudinal, and (d) full clinical trial data sets with image and no-image data.
4. Consider the data types including (a) 2D, 3D, color, B/W, video and audio.
5. Consider who will use these archived datasets. Possibilities include (a) clinical research investigators, (b) educators, (c) clinicians, (d) agencies and (e) public.
6. Consider the use of common data elements and structured reporting data.
7. Provide incentives and necessary support to investigators that cover their real costs for developing and contributing to these archives.
8. Establish database formats that are self-describing and meet strict quality criteria that need to be continually reviewed.
9. Foster the development of a technical framework that should provide mechanisms for different levels of control and access to data consistent with an agreed upon distribute access policy.
10. Establish an access mechanism that respects investigator rights but is as "open" as possible to the wider imaging community. If possible there should be mechanisms to allow timely access to subsets of completed data. Consider the (a) application procedure, (b) review and approval process, (c) appeal process, (d) license or limitation, (e) physical transfer of data, (f) reporting requirements, (g) cost allocation in the access document.
11. Manage the resource centrally using the Internet.
12. Assure its continued viability with appropriate infrastructure support.
13. Evaluate the success of this effort based on its effect on the level of research synergy generated by investigators using the same image archive resource, the improved quality of research produced such as
    1. Number and quality of publications
    2. Number of investigators served
    3. New investigators using the resource
    4. Frequency of access
    5. Citation analysis
    6. Other
14. Undertake the implementation in stages, using selected pilot models supported by NC, providing the most elementary data types and simplest mechanisms first: with the intent to use the experience gained with a simple environment to guide further development and enhancements that may lead to wider acceptance by the imaging community, NIH and related imaging societies.
15. Adhere to current standards and utilize existing COTS technology wherever possible.
16. Consider the use of "open source" software tools to allow wider user access tot the image archive and data mining/ processing.

Action items for NCI program Staff:

NCI Image Archive Advisory Panel: Establish an expert advisory panel to review the recommendations from the workshop and advise the Cancer Imaging Program on a short term plan to explore the feasibility of developing NCI sponsored image archives. This should be small enough (7-9 individuals) to reach a consensus in a reasonable time period. This panel could also provide advice to ongoing projects namely: (a) collaborative groups (Uo1's) that are developing application specific image archives such as the ACRIN and Lung Image database Consortium (LIDC) (b) oncology clinical trial groups including NCI programs such as SPORES and CTEP that may involve imaging, and (c) any planned NIH wide effort for development of image achieves.

NCI/CIP: CA-01-001: Lung Imaging database Resource for Imaging Research: This is a consortium of institutions that will develop the necessary consensus and standards for a retrospective image database of spiral computed Tomography (CT) lung images. Consensus will be necessary on many issues including database design, access, metrics and statistical methods for evaluating image-processing algorithms. It was recommended that this project serve as a pilot NCI supported initiative to explore the feasibility of developing a more widely available and useable image archive. The NCI Image Archive Advisory Panel should interact with the steering committee of the Uo1 to provide advice on the development of the image archive and to ensure that the archive will be linked with future image archives developed through support from NCI. It is anticipated that some of the sites will be involved in ACRIN supported research and a linkage to the ACRIN archival systems will emerge. The use of such database resources for facilitating FDA approval for image processing algorithms should be explored.

Existing Image Data Sets and Archive Feasibility: Determine the number of NCI currently funded investigators that are developing image databases that might be entered into a more widely used archive. Provide this to the expert panel to consider which ones might be approached as a pilot study for the voluntary buildup of the image archive The panel should explore the image formats, the value of the data sets and barriers to their use. Working with the investigators it should establish the criteria by which the individual investigators would be willing to share their databases. It should be determined what supplemental funding would be necessary for this to occur.

Development of Technical Framework: Consider an NCI initiative to develop a technical framework and promote its use for image database archives. This could be either a distributed or central based archive. It should be designed to permit varying levels of data collection, access, data mining and analysis. It should consider the incorporation of existing archives considered of importance by the research community.