Table 3.
Evaluation of proposed deep learning network using precision, recall, and f-measure for nuclei marker locations and per-pixel measurements, calculated over the six time points in the test time-series. For the sake of comparison, the per-pixel F-measure values written within brackets represent the best result achieved for this time stamp using machine learning.
| Volumetric Time Stamp | Nuclei in Ground Truth | Nuclei in Predicted Output | True Positives | False Positives | False Negatives | Precision | Recall | F-Measure | Per-Pixel Measurements | ||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Precision | Recall | F-Measure | |||||||||
| 1 | 5 | 8 | 5 | 3 | 0 | 0.63 | 1.0 | 0.77 | 0.65 | 0.26 | 0.37 (0.25) |
| 2 | 11 | 9 | 8 | 1 | 2 | 0.88 | 0.8 | 0.84 | 0.84 | 0.21 | 0.34 (0.33) |
| 3 | 9 | 9 | 8 | 1 | 0 | 0.88 | 1.0 | 0.94 | 0.84 | 0.25 | 0.39 (0.43) |
| 4 | 6 | 6 | 5 | 1 | 0 | 0.83 | 1.0 | 0.90 | 0.77 | 0.35 | 0.49 (0.43) |
| 5 | 3 | 6 | 3 | 3 | 0 | 0.50 | 1.0 | 0.66 | 0.61 | 0.45 | 0.52 (0.18) |
| 6 | 2 | 5 | 2 | 3 | 0 | 0.40 | 1.0 | 0.57 | 0.46 | 0.42 | 0.44 (0.11) |