Table 1 The performance of the five types of CNNS in the tile identification module for the internal and external testing datasets

From: AI assistance enhances histopathologic distinction between sebaceous and squamous cell carcinoma of the eyelid

Model

Class

Accuracy

Internal testing

External testing

ResNet101

SC

29/33, 29/33

2/9, 2/9

SGC

148/201, 153/201

10/27, 10/27

ResNeXt101_32X8d

SC

30/33, 30/33

2/9, 2/9

SGC

146/201, 154/201

13/27, 13/27

Wide ResNet101

SC

28/33, 28/33

2/9, 2/9

SGC

151/201, 158/201

21/27, 22/27

DenseNet121

SC

30/33, 30/33

10/10, 10/10

SGC

90/201, 87/201

0/27, 0/27

EfficientNetb6

SC

29/33, 28/33

2/9, 2/9

SGC

151/201, 156/201

14/27, 16/27

  1. Bold highlights the results of the best-performing model.
Back to article page