Research Article
A General System for Automatic Biomedical Image Segmentation Using Intensity Neighborhoods
Table 4
Quantitative comparison of nuclei segmentation.
| Algorithm | RI | JI | Hausdorff | NSD () | Split | Merged | Added | Missing |
| AS manual | 95% | 2.4 | 9.7 | 0.5 | 1.6 | 1.0 | 0.8 | 2.2 | RC threshold | 92% | 2.2 | 34.8 | 1.2 | 1.1 | 2.4 | 0.3 | 5.5 | Otsu threshold | 92% | 2.2 | 34.9 | 1.2 | 1.1 | 2.4 | 0.3 | 5.6 | Mean threshold | 96% | 2.2 | 26.5 | 1.0 | 1.3 | 3.4 | 0.9 | 3.6 | Watershed (direct) | 91% | 1.9 | 34.9 | 3.6 | 13.8 | 1.2 | 2.0 | 3.0 | Watershed (gradient) | 90% | 1.8 | 34.6 | 3.0 | 7.7 | 2.0 | 2.0 | 2.9 | Active masks | 87% | 2.1 | 148.3 | 5.5 | 10.5 | 2.1 | 0.4 | 10.8 | Merging algorithm | 96% | 2.2 | 12.9 | 0.7 | 1.8 | 2.1 | 1.0 | 3.3 | Our result | 97% | 2.5 | 119.3 | 0.8 | 0.8 | 2.8 | 3.5 | 0.3 |
|
|