From: Practical utility of liver segmentation methods in clinical surgeries and interventions
Challenge | Reference | Dataset | Method | Performance (best results) |
---|---|---|---|---|
CHAOS | Conze et al. [62] | 80 patients (40 CT, 40 MRI scans) | Conditional generative adversarial network with a partially pre-trained generator | Dice: 97.95 ± 0.27 ASSD: 0.76 ± 0.16  (performance of cGv16pUNet1-1) |
FLARE | Zhang et al. [14] | 511 CT scans and annotations for 4 abdominal organs | Context-aware efficient encoder-decoder model with anisotropic pyramid pooling | Dice: 96.5 ± 6.1 NSD: 87.8 ± 11.2 (performance of efficientSegNet) |
KiTS | Chen et al. [63] | 300 CT scans | nnU-Net-based coarse-to-fine segmentation framework | Dice: 90.99 NSD: 83.48 |