Table 8 Ablation studies – performance analysis.

From: Innovative deep learning classifiers for breast cancer detection through hybrid feature extraction techniques

Configuration

Accuracy (%)

Sensitivity (%)

Specificity (%)

F1 score (%)

No preprocessing

87.62

85.34

86.4

83.21

Only contrast stretching

89.15

86.75

87.92

85.73

Contrast + median filtering

91.32

89.82

90.1

88.91

Contrast + median + otsu

93.01

92.18

91.88

91.2

Full preprocessing (Shearlet + Otsu + Canny)

95.91

96.82

94.55

93.34

Back to article page