fidoptBITE
Intra-operative 3D ultrasound before resection (fixed) and pre-operative T1-weighted MR with gadolinium (moving) of one patient with a brain tumour (MNI BITE database, group 2).
Read before comparing (5)
- Alignment. The provider moved both images into the MNI Talairach-like brain frame by rotation and translation, the ultrasound through the neuronavigation system's registration of the patient to the MR. That is the alignment a surgeon starts from: 1.5 to 9.4 mm at the landmarks as a patient's mean, 3.4 mm over the test pairs.
- Images. The ultrasound covers a part of the brain around the tumour; the field is asked on its 0.3 mm grid, and the mask is where the sweep recorded something. The MR comes resampled by the provider at 0.5 mm.
- Landmarks. Two experts chose their own points in every patient, a third in the first six; the corpus pools the sets (the dataset's NN_all.tag).
- Left out. The tumour masks, on the MR only, and the dataset's other groups: ultrasound before and after resection (group 1) and MR with ultrasound after resection (group 4).
- Overlap. RESECT poses the same problem on other patients, with a FLAIR MR.
No method has a result on every pair of this dataset yet.
| Accuracy | Smoothness | Cost | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| # | Method | TRE mm | Change | rTRE | SDlogJ | Folds % | Time s | RAM GB | GPU GB | Images |
| No registration | 3.41 ±0.83 | 0.0201 ±0.0061 | 0.000 | 0.00 | ||||||
TRE: distance between matching landmarks. rTRE: the TRE over the length of the image's diagonal, to compare datasets of other sizes and units. 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
- The 10 patients outside development are all test pairs, and 4 patients are development pairs.
- Truth
- landmarks by experts the dataset's homologous landmark pairs, chosen by two experts in every patient and by a third in the first six, the sets pooled: 19 to 40 a patient
- Source
- nist.mni.mcgill.ca/bite-brain-images-of-tumors-for-evaluation-database
www.bic.mni.mcgill.ca/uploads/Services/00_readme_group2.txt
doi.org/10.1118/1.4709600 - Licence
- none stated (the page asks to cite Mercier et al. 2012)
4 pairs for tuning, never tested
case01, case02, case03, case08