Table 2.
Summary and comparison of the computed areas under ROC curves (AUC) and overall classification accuracy (ACC) along with the standard deviations (STD) after applying an operation threshold (T = 0.5) to the classification scores generated by six models tested in this study.
| Model (Output Score) | Feature Description | AUC ± STD | ACC (%) ± STD |
|---|---|---|---|
| Model-I (S1) | PCA-generated feature vector | 0.77 ± 0.02 | 71.23 ± 2.44 |
| Model-II (S2) | Transfer learning classification of ResNet50 | 0.85 ± 0.02 | 77.31 ± 2.65 |
| Model-III.1 (S3.1) | SVM (S1, S2) | 0.85 ± 0.01 | 77.42 ± 2.47 |
| Model-III.2 (S3.2) | W1 × S1 + W2 × S2 | 0.85 ± 0.01 | 77.31 ± 2.83 |
| Model-III.3 (S3.3) | Min (S1, S2) | 0.83 ± 0.02 | 73.35 ± 2.17 |
| Model-III.4 (S3.4) | Max (S1, S2) | 0.85 ± 0.02 | 74.07 ± 2.24 |