Table 3 Accuracy, sensitivity, specificity, positive predictive value, and negative predictive value for the six classifier models in the validation set.
From: Development of machine learning-based clinical decision support system for hepatocellular carcinoma
Accuracy (%) | Sensitivity (%) | Specificity (%) | PPV (%) | NPV (%) | |
|---|---|---|---|---|---|
Classifier 1 (RFA/PEIT or resection vs. not RFA/PEIT or resection) | 81.0 ± 2.6 | 77.4 ± 4.1 | 83.7 ± 3.3 | 77.8 ± 3.6 | 83.5 ± 2.5 |
Classifier 2 (RFA/PEIT vs. resection) | 88.4 ± 3.1 | 56.2 ± 11.6 | 95.8 ± 2.7 | 76.8 ± 12.1 | 90.6 ± 2.3 |
Classifier 3 (TACE vs. not TACE) | 76.8 ± 2.9 | 82.3 ± 4.1 | 69.3 ± 5.5 | 78.3 ± 4.0 | 74.6 ± 4.9 |
Classifier 4 (TACE + EBRT vs. not TACE + EBRT) | 76.6 ± 4.7 | 43.9 ± 12.6 | 89.4 ± 3.9 | 61.6 ± 10.8 | 80.4 ± 4.3 |
Classifier 5 (Sorafenib vs. Not sorafenib) | 80.0 ± 4.2 | 12.3 ± 13.3 | 95.0 ± 4.0 | 44.0 ± 37.7 | 83.1 ± 3.0 |
Classifier 6 (Supportive care vs. Other therapies) | 80.1 ± 6.3 | 53.0 ± 17.6 | 90.4 ± 5.2 | 67.7 ± 15.8 | 83.7 ± 5.6 |
