Table 3 Performances of three open-source software for nuclei detection (CellProfiler, QuPath, Fiji)
From: Multi-tissue and multi-scale approach for nuclei segmentation in H&E stained images
Method | Computational Time (s) | Object-level performance | Pixel-level performance | ||||
|---|---|---|---|---|---|---|---|
Recall | Precision | F1-score | CS (%) | US (%) | SE (%) | ||
CellProfiler (automatic) | 21.13 ± 3.78 | 0.6866 ± 0.2421 | 0.8274 ± 0.0820 | 0.7154 ± 0.2030 | 65.47 ± 23.23 | 4.79 ± 3.21 | 29.73 ± 23.44 |
QuPath (automatic) | 11.37 ± 1.96 | 0.9248 ± 0.0552 | 0.7120 ± 0.0916 | 0.8004 ± 0.0638 | 74.24 ± 10.41 | 9.29 ± 7.70 | 6.47 ± 3.23 |
Fiji (semi-automatic) | 252.73 ± 76.11 | 0.9462 ± 0.0386 | 0.8658 ± 0.0424 | 0.9030 ± 0.0248 | 82.59 ± 7.05 | 12.69 ± 5.88 | 4.65 ± 3.30 |
MANA (proposed) | 16.89 ± 5.72 | 0.9372 ± 0.0288 | 0.9253 ± 0.0293 | 0.9305 ± 0.0161 | 81.97 ± 7.05 | 12.43 ± 7.32 | 5.53 ± 2.66 |