Table 3.
Comparison of the performance metrics of the three classifiers.
| Feature extraction methods | Classifiers | AUC | 95% CI | p | Sensitivity | Specificity | Accuracy | PPV | NPV |
|---|---|---|---|---|---|---|---|---|---|
| Fisher | PCA | 0.471 | 0.336,0.605 | 0.672 | 50% | 40.6% | 47.2% | 51.3% | 39.4% |
| LDA | 0.669 | 0.542,0.796 | 0.014 | 67.4% | 65.5% | 66.7% | 74.4% | 57.6% | |
| NDA | 0.709 | 0.585,0.833 | 0.002 | 82.8% | 65.1% | 72.2% | 61.5% | 84.8% | |
| MI | PCA | 0.649 | 0.520,0.778 | 0.030 | 68.4% | 61.7% | 65.3% | 66.7% | 63.6% |
| LDA | 0.512 | 0.377,0.646 | 0.865 | 55% | 46.9% | 51.4% | 56.4% | 45.5% | |
| NDA | 0.744 | 0.626,0.862 | <0.001 | 80% | 70.3% | 75% | 71.8% | 78.8% | |
| POE + ACC | PCA | 0.520 | 0.385,0.655 | 0.773 | 57.1% | 48.6% | 52.8% | 51.3% | 54.5% |
| LDA | 0.645 | 0.515,0.774 | 0.036 | 70.6% | 61.5% | 65.3% | 61.5% | 69.7% | |
| NDA | 0.812 | 0.706,0.919 | <0.001 | 88.2% | 76.3% | 81.9% | 76.9% | 87.9% |
MI, mutual information; POE + ACC, classification error probability combined average correlation coefficients; PCA, principal component analysis; LDA, linear discriminant analysis; NDA, non-linear discriminant analysis; AUC, area under curve; CI, confidence interval; PPV, positive predictive value; NPV, negative predictive value.