Table 8 The results applied using the DenseNet and Adam optimizer per class.

From: Adapting the pre-trained convolutional neural networks to improve the anomaly detection and classification in mammographic images

CNN

Class

Performance of the Classifier

Accuracy (%)

Sensitivity

Specificity

Precision

AUC

Before pre-processing

Benign

59.76

0.35

0.57

0.33

0.42

Malignant

54.2

0.33

0.54

0.35

0.46

Normal

52.72

0.31

0.69

0.39

0.44

Average

55.56

0.33

0.61

0.356

0.44

After pre-processing

Benign

95.8

0.93

0.93

0.91

0.99

(SM)

Malignant

94.1

0.92

0.94

0.90

0.98

Normal

96.1

0.95

0.96

0.90

0.99

Average

95.3

0.93

0.94

0.903

0.986

After pre-processing

Benign

97.3

0.99

0.96

0.90

0.999

(MSVM)

Malignant

96.9

0.981

0.96

0.91

0.997

Normal

97.8

0.972

0.95

0.93

0.998

Average

97.33

0.98

0.956

0.913

0.998

After pre-processing

Benign

98.2

0.98

0.95

0.91

0.999

(RF)

Malignant

96.3

0.97

0.96

0.92

0.998

Normal

96.9

0.96

0.96

0.939

0.998

Average

97.13

0.97

0.956

0.923

0.998

Back to article page