Skip to Content

Programs & Resources

SPIE Medical Imaging Conference NCI Lung Image Database Consortium (LIDC): Open meeting, February 16, 2003

Sunday, February 16, 2003, 5:45 - 7:45, San Diego Room


Workshop Agenda

PresentationDuration

Presentation: LIDC Overview

10 minutes

Presentation: Nodule Inclusion Criteria
     Questions

10 minutes
5 minutes

Presentation: Spatial "Truth" Determination
     Questions

10 minutes
5 minutes

Presentation: Pathologic "Truth"
     Questions

10 minutes
5 minutes

Presentation: Process Model
     Questions

10 minutes
5 minutes

Presentation: Assessment Methodologies
     Questions

10 minutes
5 minutes

Presentation: The End Product
     Questions

10 minutes
5 minutes

Open Forum

5 minutes


Workshop Syllabus

LIDC Overview

  • motivation: the need for computerized lung nodule detection and evaluation methods in thoracic CT, the current research climate, and the need for research standards and an image resource
  • LIDC mission
  • LIDC members
  • overview of subsequent presentations

Nodule Inclusion Criteria

  • spectrum of focal abnormalities
  • nodule definition: visual review
  • size and number limitations
  • presence of other pathology

Spatial "Truth" Determination

  • location of all abnormalities to be indicated (by approximate center of mass)
  • spatial extent of "qualifying" nodules required for CAD research
  • manual vs. (semi-)automated margin outlines (vs. bounding box)
  • pitfalls in referring to such delineations as "truth"
  • consensus vs. probabilistic margins

Pathologic "Truth"

  • need for pathologic ground truth
  • World Health Organization (WHO) classification of lung tumors
  • confirmation procedure for histopathology and cytologic diagnoses
  • whole lung specimens for CT-pathology correlation

Process Model

  • clinical source of CT scans
  • CT scan inclusion criteria
  • CT scan task classification
  • data collection process model

Assessment Methodologies

  • the sample size issue and its practical impact on the efforts of the LIDC
  • how the many potential uses of the database complicate the ability to statistically characterize the database
  • the need for and limitations of a proposed pilot study
  • the role of score-keeping methods, truth-determination paradigms, and performance metrics in CAD research

The End Product

  • description of the anticipated results of the LIDC effort
  • expected manner in which the database will be used by researchers
  • timeline for database population and public release
  • mechanism for receiving feedback from the research community

Open Forum

  • initiate with discussion points from an LIDC panel directed at the audience, such as:
    • How will the existence of this database affect your own research?
    • What other information should be supplied with the database?
    • How can the database be used to positively direct CAD research within the community?
  • open floor