Brain MRI Segmentation on Brain MRI scans 100 (test)
84.9Dice ScoreSupervised (upper bound)
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Supervised (upper bound)Description=Fully-supervised network using ground truth labels for all 101 training examples2019.02 | 84.9 | 0.089 | |
| Ours-indep + rand-augSampling Strategy=Independent sampling alternating with hand-tuned random data augmentation2019.02 | 81.5 | 0.056 | |
| Ours-indepSampling Strategy=Independent sampling of spatial and appearance target images2019.02 | 80.4 | 0.045 | |
| Ours-coupledSampling Strategy=Coupled sampling of spatial and appearance transformations from the same target image2019.02 | 79.5 | 0.036 | |
| SAS-augDescription=Data augmentation using SAS results as labels for unannotated brains2019.02 | 77.5 | 0.016 | |
| Rand-augDescription=Hand-tuned random data augmentation using smooth deformation fields and intensity variations2019.02 | 76.5 | 0.006 | |
| SASDescription=Single-atlas segmentation baseline using a state-of-the-art registration model2019.02 | 75.9 | — |