Table 6.
Area under the ROC performance measures for the multi-class scenario following the validation scheme in Bron et al. (2015).
| AUC (CI) | AUCHC (CI) | AUCMCI (CI) | AUCAD (CI) | stdAUC (%) | |
|---|---|---|---|---|---|
| ML + SVM | 74.7 (72.1–77.0) | 77.6 (74.2–80.7) | 61.3 (57.6–64.9) | 82.3 (78.7–84.9) | 11.0 |
| Pretrained 1-layer NN | 78.5 (76.3–80.3) | 82.4 (79.8–84.9) | 64.5 (61.0–67.7) | 85.2 (82.5–87.5) | 11.2 |
| Pretrained 2-layer NN | 77.9 (75.8–79.8) | 81.5 (78.5–83.9) | 62.7 (59.1–65.9) | 85.5 (82.9–87.9) | 12.2 |
| Soerensen-equal | 78.8 (75.6–82.0) | 86.3 (81.8–89.3) | 63.1 (56.6–68.3) | 87.5 (83.4–91.1) | 13.8 |
| Abdulkadir | 77.7 (74.2–81.0) | 85.6 (81.4–89.0) | 59.9 (54.1–66.4) | 86.7 (82.3–90.3) | 15.2 |
| Ledig-ALL | 76.7 (73.6–79.8) | 86.6 (82.7–89.8) | 59.7 (53.3–65.1) | 84.9 (79.7–88.7) | 15.1 |
| Wachinger-step1 | 74.6 (70.8–78.1) | 79.1 (73.5–83.1) | 55.0 (48.5–61.4) | 89.2 (85.3–92.3) | 17.6 |
The gray shade denotes the best result for each performance measure.