Table 5.
Performance Comparison of the Lymphocytes Detection Task between Saltz et al and Our Models
Method | F1 score | Accuracy | AUC |
---|---|---|---|
Saltz et al31 | 0.770 | 74.9% | 0.808 |
L-VGG16 | 0.891 | 88.4% | 0.943 |
L-ResNet34 | 0.893 | 89.0% | 0.950 |
L-Incepv4 | 0.879 | 87.5% | 0.938 |
Bold numbers indicate the best results. L indicates the lymphocyte detection models. AUC, area under the receiver-operating characteristic curve; ResNet34, 34-layer ResNet; VGG16, 16-layer VGG.