Table 2 The performance evaluation for CNN, ANN, MLR, SVM and RF-based classification models.
From: Deep learning enables automated scoring of liver fibrosis stages
Sensitivity | Specificity | PPV | NPV | ||
|---|---|---|---|---|---|
Morphological features | |||||
F0 vs F1–4 | CNN | 85.0% | 100.0% | 100.0% | 96.4% |
ANN | 95.0% | 90.5% | 70.4% | 98.7% | |
MLR | 90.0% | 98.8% | 94.7% | 97.5% | |
SVM | 95.0% | 100.0% | 100.0% | 98.8% | |
RF | 95.0% | 100.0% | 100.0% | 98.8% | |
F0–1 vs F2–4 | CNN | 80.0% | 90.0% | 84.2% | 87.1% |
ANN | 82.5% | 91.7% | 86.8% | 88.7% | |
MLR | 87.5% | 90.0% | 85.4% | 91.5% | |
SVM | 97.5% | 100.0% | 100.0% | 98.4% | |
RF | 92.5% | 91.7% | 88.1% | 94.8% | |
F0–2 vs F3–4 | CNN | 91.7% | 97.5% | 98.2% | 88.6% |
ANN | 100.0% | 100.0% | 100.0% | 100.0% | |
MLR | 96.7% | 100.0% | 100.0% | 95.2% | |
SVM | 100.0% | 95.0% | 96.8% | 100.0% | |
RF | 100.0% | 97.5% | 98.4% | 100.0% | |
F0–3 vs F4 | CNN | 96.3% | 80.0% | 95.1% | 84.2% |
ANN | 98.8% | 100.0% | 100.0% | 95.2% | |
MLR | 100.0% | 100.0% | 100.0% | 100.0% | |
SVM | 100.0% | 90.0% | 97.6% | 100.0% | |
RF | 100.0% | 95.0% | 98.8% | 100.0% | |
Textural features | |||||
F0 vs F1–4 | CNN | 85.0% | 100.0% | 100.0% | 96.4% |
ANN | 80.0% | 98.8% | 94.1% | 95.2% | |
MLR | 0.0% | 98.8% | 0.0% | 80.0% | |
SVM | 100.0% | 2.5% | 20.4% | 100.0% | |
RF | 95.0% | 98.8% | 95.0% | 98.8% | |
F0–1 vs F2–4 | CNN | 80.0% | 90.0% | 84.2% | 87.1% |
ANN | 60.0% | 96.0% | 92.3% | 78.4% | |
MLR | 77.5% | 48.3% | 90.0% | 45.0% | |
SVM | 100.0% | 16.7% | 44.4% | 100.0% | |
RF | 77.5% | 88.3% | 81.6% | 85.5% | |
F0–2 vs F3–4 | CNN | 91.7% | 97.5% | 98.2% | 88.6% |
ANN | 1.7% | 97.5% | 50.0% | 40.0% | |
MLR | 90.0% | 45.0% | 71.1% | 75.0% | |
SVM | 100.0% | 20.0% | 65.2% | 100.0% | |
RF | 95.0% | 87.5% | 91.9% | 92.1% | |
F0–3 vs F4 | CNN | 96.3% | 80.0% | 95.1% | 84.2% |
ANN | 100.0% | 60.0% | 91.0% | 100.0% | |
MLR | 95.0% | 50.0% | 88.4% | 71.4% | |
SVM | 100.0% | 55.0% | 89.9% | 100.0% | |
RF | 100.0% | 95.0% | 96.4% | 100.0% | |
All features | |||||
F0 vs F1–4 | CNN | 85.0% | 100.0% | 100.0% | 96.4% |
ANN | 100.0% | 100.0% | 100.0% | 100.0% | |
MLR | 0.0% | 98.8% | 0.0% | 80.0% | |
SVM | 100.0% | 37.5% | 28.6% | 100.0% | |
RF | 90.0% | 98.8% | 94.7% | 97.5% | |
F0–1 vs F2–4 | CNN | 80.0% | 90.0% | 84.2% | 87.1% |
ANN | 97.5% | 98.3% | 97.5% | 98.3% | |
MLR | 77.5% | 48.3% | 50.0% | 76.3% | |
SVM | 100.0% | 16.7% | 44.4% | 100.0% | |
RF | 82.5% | 86.7% | 80.5% | 88.1% | |
F0–2 vs F3–4 | CNN | 91.7% | 97.5% | 98.2% | 88.6% |
ANN | 100.0% | 100.0% | 100.0% | 100.0% | |
MLR | 90.0% | 45.0% | 71.1% | 75.0% | |
SVM | 100.0% | 5.0% | 61.2% | 100.0% | |
RF | 100.0% | 95.0% | 98.8% | 100.0% | |
F0–3 vs F4 | CNN | 96.3% | 80.0% | 95.1% | 84.2% |
ANN | 98.7% | 100.0% | 100.0% | 95.2% | |
MLR | 95.0% | 50.0% | 88.4% | 71.4% | |
SVM | 100.0% | 17.6% | 82.5% | 100.0% | |
RF | 100.0% | 95.0% | 98.8% | 100.0% | |
