Table 6 Calculated metrics of all the models in classifying stroke vascular territory.

From: Automatic detection and vascular territory classification of hyperacute staged ischemic stroke on diffusion weighted image using convolutional neural networks

 

Inception-v3

Efficientnet-b0

Modified lenet

ACI

Normal

PCI

ACI

Normal

PCI

ACI

Normal

PCI

Accuracy

89.6%

87.1%

95.9%

87.8%

84.2%

95.1%

89.4%

85.3%

95.5%

Precision

84.0%

87.1%

88.0%

79.3%

85.7%

79.1%

82.4%

86.3%

86.9%

Sensitivity

79.4%

93.5%

48.8%

78.9%

90.3%

42.2%

81.0%

91.6%

44.4%

Specificity

93.7%

75.6%

99.4%

91.5%

73.4%

99.1%

92.9%

74.3%

99.4%

F1 score

81.9%

90.9%

63.0%

79.1%

88.2%

55.2%

81.9%

89.2%

58.8%

Overall accuracy

 

86.3%

  

83.6%

  

85.2%

 

Weighted F1 score

 

86.2%

  

83.0%

  

84.7%

 

Κ score

 

0.715

  

0.662

  

0.693

 

Total time elapsed (min)

 

335

  

410

  

35

 
Back to article page