fidopt Interactive view

Mandibular CT Dataset Collection

A head and neck CT of a patient treated in the oral and maxillofacial surgery department in Graz, with the mandible outlined slice by slice by a clinical expert (Mandibular CT Dataset Collection, Wallner and Egger 2018).

10 test caseshead and neckimage CT
Read before comparing (7)
  • Pool. Ten CTs from the clinical routine of the Department of Oral and Maxillofacial Surgery, Medical University of Graz, 2013 to 2016 (6 men, 4 women), drawn at random among 20 with a healthy mandible, clear bone contours and no osteosynthesis material. Fewer than 20 patients: all are test cases.
  • Images. The NRRD files' voxels in Hounsfield units, 0.38 to 0.52 mm in plane and slices 0.75 to 2 mm apart; Pat_2 and Pat_4 are tilted. A box in front of the eyes and nose is set to 1024 HU on every case to hide the face; it touches no mandible.
  • Truth. Expert A's mandible (a radiologist), drawn slice by slice in MeVisLab in 35 to 48 minutes a patient. The deposit gives each expert's drawing as a surface of voxel cubes, which prepare.py turns back into exactly the voxels it encloses. Expert B (a maxillofacial surgeon) drew every case apart: A and B agree at a Dice of 0.918 to 0.956 (median 0.943). Graded expert.
  • Cases. The mandibles hold 17 to 47 mL (median 31 mL), at a median of 515 to 1,118 HU a case.
  • Start. A whole-image mask scores a Dice of 0.005 at the median, an Otsu mask 0.013.
  • Overlap. No other dataset of the corpus names Graz or this deposit; PDDCA's and HaNSeg's mandibles are other patients.
  • Outside numbers. GrowCut in 3D Slicer reached a mean Dice over 0.85 against both experts on these ten CTs (Wallner et al. 2018).

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
All 10 patients are test cases; there are no development cases. One CT a patient; the provider publishes no split.
Truth
contours by experts the mandible drawn by hand slice by slice in MeVisLab by clinical expert A, a radiologist, turned into voxels from the surface the deposit gives; expert B, a maxillofacial surgeon, drew it apart (Wallner et al. 2018)
Licence
CC BY 4.0