fidoptCTPelvic1K, CLINIC pelvic bones
One preoperative pelvic CT from the CLINIC sub-dataset of CTPelvic1K, with the sacrum and the two hip bones annotated through an iterative network-and-expert correction workflow.
Read before comparing (5)
- Source. Only sub-dataset 6, CLINIC's 103 scans, is used; images and masks keep their released values and geometry, without resampling. The images include pelvic fractures.
- Labels. On the masks checked, label 2 lies on the patient's left and label 3 on the right. Label 4, the lumbar spine, stays in the truth file but is not scored.
- Overlap. 31 of the 103 CLINIC scans are PENGWIN images under other names: 27 identical voxel for voxel, 4 cropped and flipped (correlation 0.998 to 1.000). A case is kept when its root-mean-square log difference to every PENGWIN case, over the volume and three principal lengths of each bone, is at least 0.04; the shared images lie at 0.002 to 0.016, other images from 0.026. 43 cases pass.
- Overlap check. The twenty scans drawn, aligned rigidly to the corpus's twenty PENGWIN scans as one bone mask, reach a Dice of at most 0.71, where known shared images reach 0.97 to 0.98: none is a pelvis the corpus already holds. No provider gives a patient crosswalk, and PENGWIN's 50 scans that are not public could not be compared.
- Cohort. This dataset and PENGWIN share a source cohort; they are grouped apart rather than treated as independent populations.
| Accuracy | Cost | |||||||
|---|---|---|---|---|---|---|---|---|
| # | Method | Dice | Surface Dice | HD95 mm | Time s | RAM GB | GPU GB | Images |
| 11–3 | KonfAI segmentation app · TotalSegmentator (total), the authors' packageGPUapp | 0.928 ±0.007 | 0.899 ±0.019 | 1.78 ±0.23 | 19.0 | 4.5 | ||
| 21–3 | KonfAI segmentation app · TotalSegmentator (total), KonfAI appGPUapp | 0.927 ±0.008 | 0.898 ±0.020 | 1.75 ±0.24 | 28 | 3.7 | 7.1 | |
| 31–3 | CT-FM whole-body segmentationGPUapp | 0.930 ±0.019 | 0.885 ±0.041 | 14.60 ±10.67 | 159 | 15.7 | 6.9 | |
| 4 | KonfAI segmentation app · MRSegmentator, KonfAI appGPUapp | 0.902 ±0.007 | 0.875 ±0.016 | 4.48 ±0.49 | 24.3 | 7.8 | ||
| 5 | MONAI wholeBody CTGPUapp9 failed | 0.089 ±0.201 | 0.082 ±0.186 | 6.22 | 22 | 5.0 | 9.0 | |
The order averages each method's ranks on Dice, Surface Dice, HD95 mm, so the first one is not always the best in every column. ± is the 95% interval of the mean over the 10 cases. Two neighbours are tied when a paired Wilcoxon test on the shared cases does not find the upper one better on Dice (5%): the places a method could hold are then shown under its own. The best value of each column is in colour. Click a row to see its images, a column title to sort by it. Time and memory measured on Workstation.
1 result not ranked: an older run, cases still missing, or a failure on every case
| Accuracy | Cost | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| # | Method | Dice | Surface Dice | HD95 mm | Time s | RAM GB | GPU GB | Machine | Images |
| MONAI wholeBody CTGPUappolder run | 0.908 ±0.011 | 0.850 ±0.027 | 6.91 ±7.73 | 15 | 3.0 | 12.5 | Laptop | ||
Dice: how much the organ contours overlap, 1 is perfect. Surface Dice: the share of the contours' surfaces within a tolerance of each other, 1 is perfect. HD95: distance between the contours' surfaces.
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 and 10 development scans from the 43 CLINIC cases whose pelvic bone shapes differ from all 100 public PENGWIN cases. Test is the first ten in the seed-2026 order, development the last ten.
- Truth
- contours by a program, verified by people The released sacrum and hip-bone masks, initially threshold-assisted manual contours or network predictions, checked and corrected by human experts in the AID workflow. The CLINIC mask also retains label 4 for lumbar spine, which this pelvic-bone task does not score.
- Licence
- CC BY 4.0
10 cases for tuning, never tested
CLINIC_0027, CLINIC_0053, CLINIC_0059, CLINIC_0042, CLINIC_0068, CLINIC_0009, CLINIC_0023, CLINIC_0039, CLINIC_0091, CLINIC_0046