Skip to main content

Table 2 Confusion matrix of Model 1 results, Crowe classification

From: Application of deep learning for automated diagnosis and classification of hip dysplasia on plain radiographs

 

True

 

Normal

Borderline

Crowe 1

Crowe 2

Crowe 3

Crowe 4

Class precision

Predicted

       

Normal

48

13

5

0

0

0

72.7%

Borderline

3

1

2

0

0

0

16.7%

Crowe 1

0

1

13

3

1

0

72.2%

Crowe 2

0

0

0

4

3

0

57.1%

Crowe 3

0

0

0

1

3

1

60.0%

Crowe 4

0

0

0

0

0

1

100.0%

Class recall

94.1%

6.7%

65.0%

50.0%

42.9%

50.0%

Â