Table 4 Post-processing settings for various structures
From: Generalizable deep learning framework for 3D medical image segmentation using limited training data
Name | Probability threshold | n-largest object | Touch other objects |
|---|---|---|---|
Skeletal | 50% | - | No |
Orbita | 12.5% | 1 | No |
Eyes | 12.5% | 2 | No |
Mandible | 50% | 1 | No |
Cranium | 50% | 1 | No |
Congenital | 25% | 1 | No |
Fetal | 50% | 1 | No |
Parenchyma | 50% | 2 | No |
Trachea | 50% | 1 | Yes, Parenchyma |