Table 4.
Average accuracy and runtime of networks and physicians for six-class classification.
| Approach | Average accuracy (%) | Runtime (S, mean ± SD) |
|---|---|---|
| 2D ResNet-18 (25) | 63.6 | 2.6 ± 0.4 |
| 2D Faster R-CNN (28) | 62.2 | 1.8 ± 0.3 |
| 3D VGG-16 (35) | 72.0 | 2.4 ± 0.3 |
| Physician 1 | 72.7 | 21.4 ± 9.2 |
| Physician 2 | 83.2 | 27.2 ± 9.3 |
| Our 3D ResNet-18 | 73.4 | 2.0 ± 0.5 |
“Physician 1” and “Physician 2” represent the junior and senior physicians, respectively. SD, standard deviation.
The bold values represent the results obtained from our proposed model.