fidoptSynthRAD2025 CBCT/CT, head and neck
CT (fixed) and CBCT (moving) of the head and neck of one patient (SynthRAD2025, Task 2).
Read before comparing (3)
- Labels. They come from a segmentation network: a method that registers on that network's features is scored against its own view of the anatomy.
- Specks. The network also marks a few specks of the abdomen and the legs, 2 to 520 voxels, kept as segmented. A speck on one image only is never scored, since a structure is scored only where both images hold it; two test pairs hold one on both images, scored like the rest of the case: a femur in HN_2HNA014, colon and liver in HN_2HNE003.
- Mask. The challenge's body mask, the same for both images.
No method has a result on every pair of this dataset yet.
| Accuracy | Smoothness | Cost | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| # | Method | Dice | Change | HD95 mm | SDlogJ | Folds % | Time s | RAM GB | GPU GB | Images |
| No registration | 0.683 ±0.057 | 6.75 ±1.94 | 0.000 | 0.00 | ||||||
Dice: how much the organ contours overlap, 1 is perfect. HD95: distance between the contours' surfaces. SDlogJ: how irregular the deformation is, 0 is perfectly smooth. Folds: share of the image where the deformation folds over itself.
Pair by pair
Click a score to look at that result: the fixed image, and over it the moving image after registration, with the annotations. Underlined scores have their image; for the others you see the pair without the result. A green cell gained against no registration on that pair, a red one lost.
About the data
- Test pairs
- 10 test pairs picked by a fixed rule before any result, one a patient, centres alternated, and 10 development pairs.
- Truth
- contours by a program TotalSegmentator (task total) run by us on each image
- Licence
- CC BY-NC 4.0 (centres A, B, C, E)
10 pairs for tuning, never tested
HN_2HNA002, HN_2HNA005, HN_2HNA010, HN_2HNB007, HN_2HNB105, HN_2HNB086, HN_2HNC049, HN_2HNC051, HN_2HNE004, HN_2HNE066