Table 3.
Diagnostic performances of the AI-based evaluation method using contrastive learning and MOSE for obtaining a histological diagnosis.
| Histology | Histology | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Diagnosable | Undiagnosable | Total | Diagnosable | Undiagnosable | Total | ||||
| AI | Diagnosable | 131 | 13 | 144 | MOSE | Diagnosable | 129 | 13 | 142 |
| Undiagnosable | 14 | 15 | 29 | Undiagnosable | 16 | 15 | 31 | ||
| Total | 145 | 28 | 173 | Total | 145 | 28 | 173 | ||
| Sensitivity | 90.3% | Sensitivity | 88.9% | ||||||
| Specificity | 53.5% | Specificity | 53.5% | ||||||
| Accuracy | 84.4% | Accuracy | 83.2% | ||||||
| PPV | 90.9% | PPV | 90.8% | ||||||
| NPV | 51.7% | NPV | 48.4% | ||||||
AI: artificial intelligence, MOSE: macroscopic on-site evaluation, PPV: positive predictive value, NPV: negative predictive value.