LIDC-IDRI
Lung Image Database Consortium and Image Database Resource Initiative
1,018 thoracic CT scans of 1,010 patients from seven US academic centers, each read by four radiologists who marked and outlined lung nodules and rated their characteristics. Includes a smaller set of chest radiographs.
Overview
LIDC-IDRI is a reference collection of thoracic CT scans with lung nodule annotations by several radiologists. It was assembled by the Lung Image Database Consortium, started by the US National Cancer Institute, and expanded through the Image Database Resource Initiative of the Foundation for the NIH with industry partners. It is hosted by The Cancer Imaging Archive and remains a standard benchmark for nodule detection, segmentation and malignancy rating.
Composition
The database holds 1,018 CT scans from 1,010 patients: eight patients were included twice by mistake, with scans from two time points. Both diagnostic and low-dose screening scans were accepted, with at most 3 mm collimation and reconstruction interval. Cases were chosen to contain roughly up to six nodules between 3 and 30 mm. TCIA also distributes chest radiographs for a subset of patients (1,308 studies in total) as well as patient- and nodule-level diagnosis data for a limited number of cases.
Acquisition
Seven US academic centers contributed scans from routine clinical care, so scanner models and parameters vary:
| Vendor | CT scans |
|---|---|
| GE (LightSpeed) | 670 |
| Siemens | 205 |
| Philips (Brilliance) | 74 |
| Toshiba (Aquilion) | 69 |
Slice thickness ranges from 0.6 to 5 mm, most often 1.25 or 2.5 mm.
Annotations
Four radiologists from different institutions read each scan in two phases. First they marked lesions independently as nodule of at least 3 mm, nodule under 3 mm, or non-nodule of at least 3 mm. Then each reviewed the anonymised marks of the others and gave a final opinion, without forced consensus. Nodules of 3 mm or more were outlined slice by slice and rated for features such as subtlety, spiculation and likelihood of malignancy. In total, 7,371 lesions were marked as a nodule by at least one reader, and 2,669 as a nodule of at least 3 mm.
Known limitations
- No consensus ground truth: readers often disagree, and users must choose how to combine the four reads.
- Reader order in the XML files is not consistent across scans.
- Spiculation and lobulation ratings are inconsistent for about 100 early cases, and a few XML files have known errors listed on the TCIA page.
- Diagnosis data exist only for a small subset.
Cohort
Aggregate numbers from the sources below. Bars are relative to the 1,010 subjects.
Modality
scans
- CT 1,018
License and access
Our reading of the license, not legal advice. Before you use the data, read the original license and confirm that your use is allowed. We take no responsibility for how you use a dataset. Full disclaimer
Download without an account
Download with the TCIA Data Retriever; annotations are XML files.
Creative Commons Attribution 3.0 Unported
Older version of CC BY. Use, share and adapt the data for any purpose, including commercial use, as long as you credit the creators.
What you can do
- Yes
- Yes
- Yes
- Yes
What you can share
- Yes
- Yes
- Yes
What you must do
- Yes
- Share alike No
- No
- No
- Manuscript review No
- Release code No
- Return results No
- Delete after use No
Limits
- No
- Location limits No
Citation
Armato SG 3rd, McLennan G, Bidaut L, 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). https://doi.org/10.1118/1.3528204. Data: Armato SG III, et al. (2015). Data From LIDC-IDRI [Data set]. The Cancer Imaging Archive. https://doi.org/10.7937/K9/TCIA.2015.LO9QL9SX
All numbers
Every number on this page, as stored in stats.csv, with its source.
| Measure | Breakdown | Value | Source |
|---|---|---|---|
| Subjects | total | 1,010 | tcia-lidc-idri |
| Studies | total CT and chest radiograph studies | 1,308 | tcia-lidc-idri |
| Scans | modality=CT 8 patients have two CT scans | 1,018 | armato2011 Section Materials and Methods, Patient image data |
| Scans | modality=CT;vendor=ge Seven GE LightSpeed models | 670 | armato2011 Section Materials and Methods, Patient image data |
| Scans | modality=CT;vendor=siemens Siemens Definition, Emotion and Sensation models | 205 | armato2011 Section Materials and Methods, Patient image data |
| Scans | modality=CT;vendor=philips Four Philips Brilliance models | 74 | armato2011 Section Materials and Methods, Patient image data |
| Scans | modality=CT;vendor=canon Toshiba Aquilion | 69 | armato2011 Section Materials and Methods, Patient image data |
Sources
The keys used in the table above.
- armato2011 Armato et al. 2011, Medical Physics paper
- tcia-lidc-idri TCIA LIDC-IDRI collection page (version 4) website