fidoptNSCLC-Radiomics-Interobserver1
The pretreatment CT of a patient with non-small cell lung cancer treated by radiotherapy at MAASTRO, with the primary tumour where at least three of five radiation oncologists, outlining it blind and by hand, placed it (NSCLC-Radiomics-Interobserver1, TCIA version 3).
10 test caseslungimage CTresearch only
Read before comparing (6)
- Truth. Five radiation oncologists (two then in training) outlined the tumour blind; the truth is where three or more of them put it. Their outlines agree at a median pairwise Dice of 0.82 on the test cases (0.47 for the worst pair; interobs14 has the lowest case median, 0.72), and one reader against the outline three of the other four agree on at a median of 0.83 (0.56 to 0.93): a method near 0.83 is as close as a sixth oncologist. The five outlines the same oncologists edited from an autosegmentation are not used.
- Cases. The test tumours hold 4.4 to 344 mL (median 30 mL), interobs33's a 4.4 mL nodule.
- Images. CERR exports of a 5 mm whole-body CT, head to thighs, 178 or 154 slices of 512 by 512 pixels of 0.98 mm, in Hounsfield units clipped at -1000; not Lung1's 3 mm planning CTs.
- Overlap. The collection comes from the MAASTRO study of Lung1 (NSCLCRadiomicsGTV, PleThoraCT) but carries new IDs and its own 5 mm CTs exported through CERR, so nothing in the DICOM links a patient to Lung1, and no image check with a positive control was made: identity rests on the clinical tables. Every patient with the sex and the T, N and M stage of a Lung1 patient in the corpus, the ages within 1.5 years, is held out: interobs19, 31 and 34 (PleThoraCT), 32 (NSCLCRadiomicsGTV's LUNG1-303, whose age Lung1 does not record), 05, 10, 15 and 21. interobs06, 27, 28 and 33 match no Lung1 patient within two years, so the two tables may not describe the same patients alike. interobs09 has no outline.
- Start. A mask of the whole image scores a Dice of 0.000 at the median of the test CTs (0.003 at most), an Otsu mask 0.001 (0.0001 to 0.011).
- Outside numbers. None: the collection has no challenge and no split.
No method has a result on every case of this dataset yet.
Case by case
Click a score to look at that result: the image, the truth in green and the method's map in red. Underlined scores have their image; for the others you see the case without the result.
About the data
- Test cases
- 10 test CTs drawn with seed 2026 among the 13 patients left, and the other 3 for development. The pool is the 21 patients with outlines less eight who may be patients of NSCLCRadiomicsGTV or PleThoraCT (Overlap). The collection has no test set.
- Truth
- contours by experts the majority of the five manual outlines of the primary gross tumour volume (segments GTV-1vis-1 to GTV-1vis-5 of the patient's SEG): a voxel is tumour where at least three radiation oncologists, working blind and independently, put it
- Source
- www.cancerimagingarchive.net/collection/nsclc-radiomics-interobserver1
doi.org/10.7937/tcia.2019.cwvlpd26
doi.org/10.1038/ncomms5006
doi.org/10.1002/mp.14322 - Licence
- CC BY-NC 3.0
3 cases for tuning, never tested
interobs12, interobs13, interobs28