Abstract Archives of the RSNA, 2008
SSJ22-02
The Lung Image Database Consortium (LIDC): Lessons Learned from the Development of a Consensus-based Public Resource
Scientific Papers
Presented on December 2, 2008
Presented as part of SSJ22: Physics (CAD: Methods/Observer Studies)
Samuel George Armato PhD, Presenter: Nothing to Disclose
Poonam Batra MD, Abstract Co-Author: Nothing to Disclose
Philip Caligiuri MD, Abstract Co-Author: Nothing to Disclose
Cecilia Matilda Jude MD, Abstract Co-Author: Nothing to Disclose
Leslie Eisenbud Quint MD, Abstract Co-Author: Nothing to Disclose
Baskaran Sundaram MBBS, Abstract Co-Author: Research grant, Bayer AG
Michael F. McNitt-Gray PhD, Abstract Co-Author: Member, Physics Committee, American College of Radiology Imaging Network (ACRIN)
Consultant, ACRIN ACR Image Metrix
Consultant, Rapiscan Systems
UCLA Radiological Sciences has a research agreement with Siemens Medical Solutions
Research grants from National Cancer Institute (NCI) and National Institute of Biomedical Imaging and Bioengineering (NIBIB).
Charles R. Meyer PhD, Abstract Co-Author: Nothing to Disclose
Geoffrey McLennan MD, PhD, Abstract Co-Author: Shareholder, VIDA Diagnostics
Anthony P. Reeves PhD, Abstract Co-Author: Research grant, GlaxoSmithKline plc
Research grant, AstraZeneca PLC
Stockholder, VisionGate, Inc
Consultant, VisionGate, Inc
Patent agreement, General Electric Company
Laurence P. Clarke PhD, Abstract Co-Author: Nothing to Disclose
Barbara Yoder Croft PhD, Abstract Co-Author: Nothing to Disclose
Denise Ru Aberle MD, Abstract Co-Author: Nothing to Disclose
Ella A. Kazerooni MD, Abstract Co-Author: Consultant, General Electric Company
Consultant, Vital Images, Inc
Research funded, General Electric Company
Heber M. MacMahon MD, Abstract Co-Author: Consultant, Riverain Medical
Research support, MEDIAN Technologies
Stockholder, Hologic, Inc
Edwin J.R. Van Beek MD, PhD, Abstract Co-Author: Consultant, Schering-Plough Corporation
Medical Advisory Board, EDDA Technology, Inc
Medical Advisory Board, Vital Images, Inc
Speakers Bureau, Koninklijke Philips Electronics NV
Research support, General Electric Company
Research support, Siemens AG
Founder and owner, Quantitative Imaging of Iowa, Inc
David F. Yankelevitz MD, Abstract Co-Author: Researcher, OSI Pharmaceuticals, Inc
Researcher, Carestream Health, Inc
Researcher, GlaxoSmithKline plc
Researcher, AstraZeneca PLC
Medical Advisor, PneumRx, Inc
Shareholder, PneumRx, Inc
Patent agreement, General Electric Company
et al, Abstract Co-Author: Nothing to Disclose
The Lung Image Database Consortium (LIDC) has developed a database of thoracic computed tomography (CT) scans as a research resource. The scans have been marked by experienced thoracic radiologists to indicate the presence of lung nodules. Potential uses of the database include the training and testing of computer-aided diagnosis (CAD) systems, the development of volumetric nodule analyses, and the teaching of radiology residents. Users of the database should be aware of assumptions that went into the creation of the database, practical design issues, and best-use practices that have evolved through the LIDC experience.
To date, 85 publicly available CT scans have been reviewed through a two-phase process. During the initial "blinded read," each of four thoracic radiologists independently marked lesions they identified as "nodule > 3mm," "nodule 3mm." During the subsequent "unblinded read," the anonymous blinded read results of all radiologists were revealed, and each radiologist independently reviewed these marks to render a final opinion. Radiologists also subjectively evaluated nine characteristics for "nodules > 3mm."
The unblinded read has an intentionally different purpose from the blinded read; the distinction between these two reads and their impact on the overall annotation process will be described. From among the first 30 LIDC cases, 59 lesions were marked as “nodule > 3mm” by at least one radiologist; however, only 27 of these lesions (45.8%) were marked as such by all four radiologists. The reasons for this variability will be explained, and the implications of this finding for CAD development will be explored. Details of the radiologist nodule outlining procedure will be examined to understand the impact on nodule segmentation studies. The subjective rating of nodule characteristics used a fixed five-point scale for each characteristic; this approach is appreciably different from how radiologists function clinically.
Proper interpretation of results obtained using the LIDC database requires an understanding of how the database was created.
The LIDC database will become a research resource for the medical imaging research community. An understanding of how the database was created is required for proper use.
Armato, S,
Batra, P,
Caligiuri, P,
Jude, C,
Quint, L,
Sundaram, B,
McNitt-Gray, M,
Meyer, C,
McLennan, G,
Reeves, A,
Clarke, L,
Croft, B,
Aberle, D,
Kazerooni, E,
MacMahon, H,
Van Beek, E,
Yankelevitz, D,
et al, ,
The Lung Image Database Consortium (LIDC): Lessons Learned from the Development of a Consensus-based Public Resource. Radiological Society of North America 2008 Scientific Assembly and Annual Meeting, February 18 - February 20, 2008 ,Chicago IL.
http://archive.rsna.org/2008/6022510.html