Table 2 Test results for each type of forceps, and corresponding recall and precision.
From: Automated recognition of objects and types of forceps in surgical images using deep learning
Forceps | Number of forceps in images | Number of forceps identified correctly | Number of forceps not identified | False positive | Recall (95% CI) | Precision (95% CI) |
|---|---|---|---|---|---|---|
Grasping | 648 | 636 | 12 | 13 | 98.1% (97.1–99.2) | 98.0% (96.9–99.1) |
Ultrasonic scalpel | 170 | 169 | 1 | 11 | 99.4% (98.3–100.6) | 93.9% (90.4–97.4) |
Clip | 53 | 51 | 2 | 4 | 96.2% (91.1–101.4) | 92.7% (85.9–99.6) |
Angled | 138 | 131 | 7 | 0 | 94.9% (91.3–98.6) | 100% (100–100) |
Spatula | 53 | 52 | 1 | 3 | 98.1% (94.5–101.8) | 94.5% (88.5–100.5) |
Total | 1062 | 1039 | 23 | 31 | 97.8% (97.0–98.7) | 97.1% (96.1–98.1) |
