Table 3 Performance comparison of the proposed CNN and VGG19 and VGG16 as feature extraction models. These values were obtained using a combination of Circlet and 2D-DWT bases as the input to the models. Italic values represent the best-achieved results for each dataset.

From: A new convolutional neural network based on combination of circlets and wavelets for macular OCT classification

Feature extraction model

Dataset

ACC (%)

SE (%)

SP (%)

PR (%)

F1-score (%)

ROAUC (%)

Proposed CNN

A

94.5

96

89.5

90

90

98

B

90

92

84.5

86

85

96

VGG19

A

93

94.5

88

88

88

96.5

B

88.5

91

83

84

84.5

95

VGG16

A

91.5

93

87.5

87

87

96

B

87

90

82

83

82.5

94

  1. CNN Convolutional Neural Network; ACC Accuracy; SP Specificity; SE Sensitivity; PR Precision; ROAUC Area-under-the-Receiver-Operating-Characteristic curve.
Back to article page