Table 3 Our model achieves the state-of-the-art accuracy (%) in the binary classification task.

From: Breast Cancer Multi-classification from Histopathological Images with Structured Deep Learning Model

Accuracy at

Methods

Magnification factors

40X

100X

200X

400X

Image level

AlexNet17

85.6 ± 4.8

83.5 ± 3.9

83.1 ± 1.9

80.8 ± 3.0

CSDCNN

95.8 ± 3.1

96.9 ± 1.9

96.7 ± 2.0

94.9 ± 2.8

Patient level

PFTAS + QDA12

83.8 ± 4.1

82.1 ± 4.9

84.2 ± 4.1

82.0 ± 5.9

PFTAS + SVM12

81.6 ± 3.0

79.9 ± 5.4

85.1 ± 3.1

82.3 ± 3.8

GLCM + 1-NN12

74.7 ± 1.0

76.8 ± 2.1

83.4 ± 3.3

81.7 ± 3.3

PFTAS + RF12

81.8 ± 2.0

81.3 ± 2.8

83.5 ± 2.3

81.0 ± 3.8

AlexNet17

90.0 ± 6.7

88.4 ± 4.8

84.6 ± 4.2

86.1 ± 6.2

CSDCNN

97.1 ± 1.5

95.7 ± 2.8

96.5 ± 2.1

95.7 ± 2.2

  1. Comparison with mean recognition rates of the classifiers trained with different descriptors: parameter-free threshold adjacency statistics (PFTAS)22 and gray-level co-occurrence matrix (GLCM)10 are traditional feature descriptors. Quadratic discriminant analysis (QDA)38, support vector machine (SVM)14, 1-nearest neighbor (1-NN)39 and random forests (RF)40 are traditional classifiers.
Back to article page