Table 7 ACA (%) of Different Frameworks for BACH Testing Set
From: Reverse active learning based atrous DenseNet for pathological image classification
Patch-level | Slice-level | ||||||||
|---|---|---|---|---|---|---|---|---|---|
Nor. | Ben. | C. in situ | I. car. | Ave. ACA | Ave. ACA | Pre. | Rec. | F-mea. | |
CNN [24] | 61.70 | 56.70 | 83.30 | 88.30 | 72.50 | 80 | 79.52 | 80.00 | 79.76 |
CNN+SVM [24] | 65.00 | 61.70 | 76.70 | 88.30 | 72.93 | 85 | 86.61 | 85.00 | 85.80 |
AlexNet [1] | 60.00 | 58.33 | 85.00 | 95.00 | 74.58 | 80 | 82.86 | 80.00 | 81.40 |
VGG-16 [10] | 75.00 | 61.67 | 75.00 | 90.00 | 75.42 | 85 | 86.61 | 85.00 | 85.80 |
ResNet-50 [12] | 63.33 | 65.00 | 80.00 | 95.00 | 75.83 | 85 | 86.67 | 85.00 | 85.83 |
ResNet-101 [1] | 65.00 | 70.00 | 75.00 | 90.00 | 75.00 | 85 | 87.86 | 85.00 | 86.41 |
DenseNet [13] | 66.67 | 76.67 | 73.33 | 88.33 | 76.25 | 85 | 90.00 | 80.33 | 84.89 |
ADN (ours) | 60.00 | 66.67 | 88.33 | 93.33 | 77.08 | 85 | 86.67 | 85.00 | 85.83 |
ADN+DRAL (ours) | 71.67 | 73.33 | 88.33 | 96.67 | 82.50 | 90 | 92.86 | 90.00 | 91.41 |