Skip to main content

Table 1 APR, ARR and F1 scores of different methods under three settings

From: An experimental study on breast lesion detection and classification from ultrasound images using deep learning architectures

Method

Benign

Malignant

Benign+ Malignant

 

APR

ARR

F1

APR

ARR

F1

APR

ARR

F1

Auto ROI [21]

66.95

14.16

23.38

78.22

19.23

30.87

71.86

16.36

26.65

Fast+ZFNet

87.25

65.47

74.81

89.02

53.54

66.86

91.11

62.60

74.21

Fast+VGG16

90.17

66.39

76.47

71.00

40.83

51.84

88.70

61.97

72.96

Faster+ZFNet

93.14

66.25

77.43

86.37

46.83

60.73

92.42

62.23

74.38

Faster+VGG16

93.01

67.08

77.95

90.36

52.05

66.05

92.37

62.54

74.58

YOLO

95.59

68.85

80.05

96.46

57.73

72.23

96.81

65.83

78.37

YOLOv3

96.89

68.81

80.47

94.56

54.21

68.91

96.58

65.85

78.31

SSD300+ZFNet

97.20

70.56

81.76

96.44

54.91

69.97

96.89

67.23

79.38

SSD300+VGG16

96.03

69.76

80.82

97.56

58.96

73.50

96.42

66.70

78.85

SSD500+ZFNet

95.98

70.04

80.98

94.22

54.90

69.38

95.09

65.06

77.26

SSD500+VGG16

94.58

69.57

80.17

94.67

55.82

70.23

96.42

66.70

78.85

  1. Note–Boldface data indicate the best results