SummaryThe Lung Image Database Consortium research project (LIDC-IDRI) involves the generation of marked-up annotated lesions on the diagnostic and lung cancer screening thoracic CT scans found in the LIDC-IDRI image collection to create a web-accessible international resource for development, training, and evaluation of computer-assisted diagnostic (CAD) methods for lung cancer detection and diagnosis.
The following paper published in Medical Physics is effectively the authoritative user's manual for the database and should be cited in all manuscripts that make use of the database:
Armato SG III, McLennan G, Bidaut L, McNitt-Gray MF, Meyer CR, Reeves AP, Zhao B, Aberle DR, Henschke CI, Hoffman EA, Kazerooni EA, MacMahon H, van Beek EJR, Yankelevitz D, et al.: The Lung Image Database Consortium (LIDC) and Image Database Resource Initiative (IDRI): A completed reference database of lung nodules on CT scans. Medical Physics, 38: 915--931, 2011.
Additionally, please include the following attribution:
The authors acknowledge the National Cancer Institute and the Foundation for the National Institutes of Health and their critical role in the creation of the free publicly available LIDC/IDRI Database used in this study.
Reader Annotation and Markup
These links help describe how to use the .XML annotation files which are packaged along with the images in The Cancer Imaging Archive. The option to include annotation files in the download is enabled by default, so the XML described here will be included when downloading the LIDC-IDRI images unless you specifically uncheck this option. If you are only interested in the XML files or you have already downloaded the images you can obtain them here:
The following documentation explains the format and other relevant information about the XML annotation and markup files:
- XML File Documentation
- XML Base Schema - This file is called "voi array.xsd", and is central in defining tumors greater than or equal 3 mm in the datasets as well as defining the loci of non-nodules.
- Annotated XML File
- LIDC Radiologist Instructions for Spatial Location and Extent Estimates
Annotation and Markup Issues/Comments
- Please note that it is not currently possible to visualize this annotation and markup on top of the images themselves. The Cancer Imaging Program is exploring the possibility of converting this XML into a standardized format so that it could be visualized in commonly available workstations, but no definite estimate of completion is available at this time.
- For a subset of approximately 100 cases from among the initial 399 cases released, inconsistent rating systems were used among the 5 sites with regard to the spiculation and lobulation characteristics of lesions identified as nodules > 3 mm. The XML nodule characteristics data as it exists for some cases will be impacted by this error. We apologize for any inconvenience.
- Also note that the XML files do not store radiologist annotations in a manner that allows for a comparison of individual radiologist reads across cases (i.e., the first reader recorded in the XML file of one CT scan will not necessarily be the same radiologist as the first reader recorded in the XML file of another CT scan).
- March 2010: Contrary to previous documentation, the correct ordering for the subjective nodule lobulation and nodule spiculation rating scales stored in the XML files is 1=none to 5=marked. The issue of consistency noted in issue 1 still remains to be corrected.
- Note: On 2012-03-21 the XML associated with patient LIDC-IDRI-0101 was updated with a corrected version of the file.
- Nodule size list for the LIDC public cases - This link provides a list of available cases and the associated size of each identified nodule.
- lidc-idri nodule counts (6-23-2015).xlsx - This link provides an accounting of the total number of nodules for each LIDC-IDRI patient.
For a limited set of cases, LIDC sites were able to identify diagnostic data associated with the case.
- Note: This project has concluded and we are not able to obtain any additional diagnosis data beyond what is available in the above link.
Data was collected for as many cases as possible and is associated at two levels:
- Diagnosis at the patient level (diagnosis is associated with the patient)
- Diagnosis at the nodule level (where possible)
At each level, data was provided as to whether the nodule was:
- Unknown (no data is available)
- Benign or non-malignant disease
- A malignancy that is a primary lung cancer
- A metastatic lesion that is associated with an extra-thoracic primary malignancy
For each lesion, there is also information provided as to how the diagnosis was established including options such as:
- unknown - not clear how diagnosis was established
- review of radiological images to show 2 years of stable nodule
- surgical resection
- progression or response
MAX ("multi-purpose application for XML") performs nodule matching and pmap generation based on the XML files provided with the LIDC/IDRI Database. It also performs certain QA and QC tasks and other XML-related tasks.
MAX is written in Perl and was developed under RedHat Linux. It has been run under Windows.
Downloading MAX and its associated files implies acceptance of the following notice (also available here and in the distro as a text file):
DISCLAIMER: MAX is not guaranteed to process all input correctly. Possible errors include (but are not limited to) the inability to process correctly some types of nodule ambiguity (where nodule ambiguity refers to overlap between nodule markings having complicated shapes or to overlap between a nodule marking and a non-nodule mark).
LIDC 2 Image Toolbox (Matlab)
This tool is a community contribution developed by Thomas Lampert. It is designed for extracting individual annotations from the XML files and converting them, and the DICOM images, into TIF format for easier processing in Matlab (LIDC-IDRI dataset). It is available for download from: https://sites.google.com/site/tomalampert/code.