Research Article
Classification of Thyroid Nodules with Stacked Denoising Sparse Autoencoder
Table 5
Performance of the classifiers based on the two common algorithms of machine learning.
| | Parameter | LR | SVM | | Value | CI of 95% | Value | CI of 95% |
| | Proportion of malignancy | 85.7% | — | 85.7% | — | | Area under the curve | 0.576 | 0.465–0.687 | 0.576 | 0.465–0.687 | | Accuracy | 86.9% | 86.6–87.2% | 86.9% | 86.6–87.2% | | Sensitivity | 98.6% | 95.9–101.3% | 98.6% | 95.9–101.3% | | Specificity | 16.7% | −4.4–37.8% | 16.7% | −4.4–37.8% | | Positive likelihood ratio | 1.183 | 0.917–1.526 | 1.183 | 0.917–1.526 | | Negative likelihood ratio | 0.083 | 0.008–0.849 | 0.083 | 0.008–0.849 | | Positive predictive value | 87.7% | 80.5–94.8% | 87.7% | 80.5–94.8% | | Negative predictive value | 66.7% | 13.3–120.0% | 66.7% | 13.3–120% |
|
|
LR: logistic regression; SVM: support vector machine; CI: confidence interval.
|
