Table 4 The macro-average performance metrics of DeepOpacityNet and other networks on the external validation datasets.

From: A deep network DeepOpacityNet for detection of cataracts from color fundus photographs

Dataset

Size

Network

Accuracy

Precision

Recall

F1 score

κ

AUC

AP

SiMES Dataset

5752 CFPs

VGG16

0.78

0.71

0.75

0.72

0.44

0.84

0.76

ResNet50

0.74

0.70

0.78

0.70

0.43

0.86

0.79

ResNet152V2

0.77

0.70

0.75

0.71

0.43

0.83

0.77

InceptionResNetV2

0.74

0.67

0.73

0.68

0.38

0.80

0.74

InceptionV3

0.77

0.70

0.74

0.71

0.42

0.82

0.76

Dense201

0.80

0.70

0.70

0.70

0.40

0.80

0.75

Xception

0.80

0.71

0.68

0.69

0.39

0.81

0.76

DeepOpacityNet

0.83

0.75

0.71

0.73

0.46

0.86

0.80

SCES Dataset

5745 CFPs

VGG16

0.74

0.65

0.77

0.66

0.35

0.85

0.77

ResNet50

0.75

0.67

0.80

0.68

0.39

0.88

0.80

ResNet152V2

0.83

0.69

0.71

0.70

0.40

0.81

0.73

InceptionResNetV2

0.78

0.66

0.74

0.68

0.37

0.80

0.72

InceptionV3

0.82

0.68

0.73

0.70

0.40

0.82

0.74

Dense201

0.67

0.62

0.71

0.59

0.25

0.80

0.72

Xception

0.80

0.67

0.73

0.69

0.38

0.81

0.74

DeepOpacityNet

0.88

0.82

0.69

0.73

0.46

0.89

0.81

SINDI Dataset

5591 CFPs

VGG16

0.82

0.69

0.77

0.71

0.43

0.85

0.76

ResNet50

0.80

0.69

0.81

0.71

0.44

0.88

0.78

ResNet152V2

0.87

0.73

0.66

0.69

0.38

0.83

0.73

InceptionResNetV2

0.86

0.71

0.68

0.70

0.39

0.82

0.73

InceptionV3

0.87

0.73

0.69

0.71

0.42

0.83

0.74

Dense201

0.80

0.66

0.73

0.68

0.36

0.82

0.73

Xception

0.86

0.72

0.69

0.70

0.41

0.81

0.73

DeepOpacityNet

0.88

0.79

0.67

0.71

0.42

0.88

0.78

  1. Bold font denotes the highest scores before rounding.