Skip to main content

Table 6 Accuracy and precision of deep learning models for 10-class classification of MRI images with fine-tuning and training from scratch

From: Deep Learning Approaches for Automatic Quality Assurance of Magnetic Resonance Images Using ACR Phantom

 

Fine-tuned layers

From scratch (batch size = 3700 images)

 

Number of fine-tuned layers

Accuracy

Precision

Accuracy

Precision

VGG16

5

88

76

73

70

VGG19

6

91

84

71

72

Resenet50

50

10

8

35

35

InceptionV3

50

57

37

21

19

EfficinetNetB0

50

49

56

19

18

EfficinetNetB5

50

100

100

90

88