Table 2.
Performance of machine learning classifiers for liver fibrosis staging.
| Machine Learning Classifier Type | Training Features | Accuracy (%; 95% CI) |
Microaveraged AUC (95% CI) |
AUC (95% CI) | Accuracy (%; 95% CI) | ||||
|---|---|---|---|---|---|---|---|---|---|
| Significant Fibrosis | Advanced Fibrosis | Cirrhosis | Significant Fibrosis | Advanced Fibrosis | Cirrhosis | ||||
| Logistic regression classifier with L1 penalty | Selected liver features | 76 (70,83) |
0.94 (0.93,0.97) |
0.95 (0.93,0.98) |
0.92 (0.88,0.95) |
0.91 (0.86,0.95) |
86 (78,90) |
80 (75,88) |
82 (78,100) |
| Selected liver & spleen features | 84 (80,90) |
0.95 * (0.93,0.98) |
0.93 (0.90,0.98) |
0.88 (0.84,0.94) |
0.94 (0.90,0.99) |
92 (83,95) |
86 (83,95) |
90 (88,100) |
|
| Logistic regression classifier with L2 penalty | Selected liver features | 78 (73,85) |
0.95 (0.93,0.97) |
0.96 (0.95,0.99) |
0.91 (0.88,0.95) |
0.93 (0.90,0.97) |
88 (78,93) |
82 (78,88) |
84 (80,100) |
| Selected liver & spleen features | 80 (75,85) |
0.95 ** (0.93,0.97) |
0.95 (0.93,0.98) |
0.90 (0.86,0.95) |
0.94 (0.91,0.97) |
88 (78,93) |
81 (78,90) |
86 (83,100) |
|
| AdaBoosting | Selected liver features | 74 (68,80) |
0.82 (0.79,0.86) |
0.74 (0.67,0.80) |
0.72 (0.63,0.79) |
0.82 (0.73,0.88) |
76 (70,84) |
80 (75,100) |
74 (67,80) |
| Selected liver & spleen features | 76 (70,83) |
0.84 *** (0.82,0.88) |
0.57 (0.48,0.66) |
0.61 (0.52,0.69) |
0.84 (0.75,0.90) |
78 (73,85) |
82 (78,100) |
57 (48,66) |
|
| Gradient Boosting | Selected liver features | 76 (70,83) |
0.88 (0.85,0.93) |
0.85 (0.81,0.92) |
0.83 (0.78,0.89) |
0.88 (0.78,0.93) |
80 (75,85) |
80 (75,100) |
85 (81,92) |
| Selected liver & spleen features | 78 (73,85) |
0.91 **** (0.88,0.95) |
0.84 (0.80,0.90) |
0.88 (0.85,0.94) |
0.86 (0.78,0.90) |
80 (75,88) |
84 (80,100) |
84 (80,90) |
|
| XGBoosting | Selected liver features | 80 (75,88) |
0.86 (0.82,0.91) |
0.78 (0.70,0.87) |
0.79 (0.71,0.89) |
0.90 (0.83,0.95) |
84 (80,90) |
86 (83,100) |
78 (70,87) |
| Selected liver & spleen features | 82 (78,88) |
0.93 ***** (0.90,0.96) |
0.86 (0.82,0.92) |
0.90 (0.86,0.94) |
0.90 (0.80,0.95) |
84 (80,93) |
88 (85,100) |
86 (82,92) |
|
* p = 0.243 when comparing liver features with liver and spleen features; ** p = 0.278 when comparing liver features with liver and spleen features; *** p = 0.025 when comparing liver features with liver and spleen features; **** p = 0.024 when comparing liver features with liver and spleen features; ***** p < 0.001 when comparing liver features with liver and spleen features. Abbreviations: AUC = area under the receiver operating characteristic curve; CI = confidence interval.