Table 8 ACA (%) of Different Frameworks for CCG Testing Set
From: Reverse active learning based atrous DenseNet for pathological image classification
Patch-level | Slice-level | ||||||
|---|---|---|---|---|---|---|---|
Normal | Level I | Level II | Level III | Ave. ACA | Ave. ACA | F-mea. | |
AlexNet [1] | 91.75 | 42.24 | 69.88 | 70.91 | 68.70 | 50 | 41.67 |
VGG-16 [10] | 97.80 | 63.65 | 71.25 | 78.39 | 77.77 | 75 | 66.67 |
ResNet-50 [12] | 97.82 | 46.86 | 75.05 | 68.57 | 72.08 | 50 | 50 |
ResNet-101 [12] | 96.64 | 67.34 | 75.57 | 58.66 | 74.55 | 50 | 41.67 |
DenseNet [13] | 98.81 | 56.62 | 72.20 | 71.04 | 74.67 | 75 | 66.67 |
ADN (ours) | 99.29 | 71.51 | 76.51 | 73.81 | 80.28 | 75 | 66.67 |
ADN+DRAL (ours) | 99.95 | 80.35 | 85.31 | 82.60 | 87.05 | 100 | 100 |