Skip to main content
[Preprint]. 2024 Apr 29:arXiv:2312.17670v3. Originally published 2023 Dec 29. [Version 3]

Table 2.

Inter-rater agreement results (mean ± standard deviation) on a subset of the testset (n = 5) for MRA and CTA in terms of binary and class average Dice scores, centerline Dice (clDice) scores, binary and class average errors of the zero-th Betti number β0. The binary Dice, binary β0 errors, and clDice scores were computed on the merged binary class. The class-average Dice scores and β0 errors were computed for each class separately and the average was taken per case. The arrow indicates the favorable direction.

Inter-rater Agreement for Challenge Metrics

Modality Binary Dice (%) ↑ Per case class-average Dice (%) ↑ clDice (%) ↑ Binary β0 error ↓ Per case class-average β0 error ↓

CTA 94.86 ± 2.24 90.91 ± 3.35 99.72 ± 0.24 0 ± 0 0 ± 0
MRA 96.49 ± 2.01 90.21 ± 4.02 99.49 ± 0.80 0.4 ± 0.55 0.07 ± 0.07