fidopt Interactive view
Lesions and tumours, segmentation

Vestibular-Schwannoma-MC-RC

T2-weighted head MR of a patient with a vestibular schwannoma, from routine follow-up at ten UK sites, the tumour outlined by a specialised company and reviewed by radiologists (Vestibular-Schwannoma-MC-RC, TCIA).

10 test casesbrainimage MR
Look at the images
Read before comparing (9)
  • Cohort. 160 adults with a single sporadic vestibular schwannoma, followed for at least five years, imaged at ten UK sites on many scanners at 1.0 to 3 T. The collection names no hospital, age or treatment, so whether a case comes before or after surgery or radiosurgery is unknown.
  • Cases. For 16 of the 71 patients the case is not their first examination in the collection: their first has a contrast-enhanced T1 only (test cases VS-MC-RC-012 and VS-MC-RC-098) or a mask not released (VS-MC-RC-048 and VS-MC-RC-080). Masks exist for 301 examinations of 124 patients; the challenge's validation and test examinations come without one.
  • Images. The T2 as delivered, 16-bit without unit, defaced by the provider. On the test cases nine are 3D sequences, 0.31 to 0.53 mm in plane, and one, VS-MC-RC-013, a 2D sequence of 15 slices 2.2 mm apart; most are axial slabs around the internal auditory canals, two coronal. No mask is given.
  • Truth. A Neuromorphometrics technician outlined each tumour in ITK-SNAP to the authors' guidelines, capping cysts included; one of three radiologists reviewed each mask until it was accepted, and an expert team annotated the ambiguous ones. The parts inside and outside the canal are one label.
  • Agreement. On 41 T2s four annotators agreed at a mean Dice of 0.845 ± 0.078, and the technician, drawing again five months later, at 0.844 ± 0.115. The T2 masks are jagged across the slices.
  • Checks. On the 20 built each mask, its axes put back, holds 0 and 1 on its image's grid and follows a mass in the cerebellopontine angle or the canal. Stray voxels in VS-MC-RC-012, VS-MC-RC-084 and VS-MC-RC-028 are scored as delivered.
  • Room. The tumours are small, 0.051 to 16.4 ml on the test cases (median 0.58 ml): a one-voxel shift keeps a Dice of 0.68 to 0.97 (mean 0.88), while the surface Dice at 1 mm, the provider setting no tolerance, does not see one in plane.
  • Outside numbers. The provider's nnU-Net reached a mean Dice of 0.850 on its own held-out T2s, on another split: its released models and any network trained on this collection may have seen the test cases. TotalSegmentator makes no vestibular schwannoma.
  • Overlap. 58 patients are also in Vestibular-Schwannoma-SEG, hence in the corpus's VestibularSchwannoma: all are left out, so no patient is in both datasets. The 20 series built name their VS-MC-RC IDs and match no Vestibular-Schwannoma-SEG series by scanner model and slice count.

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

The first test case of Vestibular-Schwannoma-MC-RC
The first test caseClick to look at every case.
Test cases
10 test T2s drawn with seed 2026 by scanner maker, 5 Siemens, 3 Philips and 2 GE, then 10 development cases, the last of the same seeded order, each maker in proportion (4 Siemens, 3 Philips, 3 GE). The pool is the 71 patients with a VS-MC-RC ID and a T2 mask in the archive, one examination a patient, the first with a T2 mask.
Truth
contours by experts the dataset's mask of the tumour on the T2, its parts inside and outside the internal auditory canal as one label, capping cysts included: outlined in ITK-SNAP by a technician of a company specialised in brain measurement, to the authors' guidelines, reviewed by one of three trained radiologists and amended until accepted, the doubtful ones annotated by an expert team of two consultant neuroradiologists and a consultant neurosurgeon
Licence
CC BY 4.0
10 cases for tuning, never tested

VS-MC-RC-062, VS-MC-RC-028, VS-MC-RC-005, VS-MC-RC-063, VS-MC-RC-007, VS-MC-RC-011, VS-MC-RC-052, VS-MC-RC-024, VS-MC-RC-086, VS-MC-RC-074