Table 3 Classification results according to the different window sizes of 3 × 3, 5 × 5, 7 × 7, 9 × 9, and 11 × 11. The condition of 3 × 3 mask size had the worst results, while the condition of 5 × 5 provided the best results. This result indicates that the mask size can largely influence the classification performance. Statistically significant differences (p < 0.05) compared with ‘3 × 3’, ‘7 × 7’, ‘9 × 9’, and ‘11 × 11’ are indicated by ‘*’, ‘γ’, ‘†’, and ‘ψ’, respectively, as determined from the student’s t-test
Mask Size | PPV (%) | NPV (%) | Sensitivity (%) | Specificity (%) | Accuracy (%) | AUC |
|---|---|---|---|---|---|---|
3 × 3 | 86.7 ± 0.1 | 82.5 ± 0.1 | 81.6 ± 0.1 | 87.3 ± 0.1 | 84.5 ± 0.1 | 0.846 ± 0.002 |
7 × 7 | 85.4 ± 0.1 | 91.8 ± 0.1 | 92.5 ± 0.1 | 84.1 ± 0.1 | 88.3 ± 0.1 | 0.886 ± 0.001 |
9 × 9 | 89.1 ± 0.1 | 84.0 ± 0.1 | 82.9 ± 0.1 | 89.8 ± 0.1 | 86.3 ± 0.1 | 0.865 ± 0.001 |
11 × 11 | 87.1 ± 0.1 | 87.4 ± 0.1 | 87.6 ± 0.1 | 87.0 ± 0.1 | 87.3 ± 0.1 | 0.873 ± 0.001 |
5 × 5 | 86.0 ± 0.1†| 91.2 ± 0.1*†ψ | 92.8 ± 0.1*†ψ | 85.1 ± 0.1 | 88.4 ± 0.1*†| 0.886 ± 0.001*†|