Skip to main content

Table 2 The diagnostic performance of eNPM-DM and/or oncologists in nasopharyngeal malignancy

From: Development and validation of an endoscopic images-based deep learning model for detection with nasopharyngeal malignancies

Evaluation indicators

Test seta

Prospective test seta

eNPM-DM

Oncologist level

eNPM-DM plus experts

 

Expertsb

Residentsb

Internsb

Accuracy

88.7 (87.8, 89.5)

88.0 (86.1, 89.6)

80.5 ± 0.8 (77.0, 84.0)

72.8 ± 2.5 (66.9, 78.6)

66.5 ± 4.3 (48.0, 84.9)

89.0 (87.2, 90.5)

Sensitivity

91.3 (90.3, 92.2)

90.2 (87.8, 92.2)

89.5 ± 0.5 (87.4, 91.7)

88.8 ± 2.4 (83.1, 94.5)

92.2 ± 2.3 (82.1, 100.0)

87.9 (85.3, 90.2)

Specificity

83.1 (81.1, 84.8)

85.5 (82.7, 88.0)

70.8 ± 1.8 (63.0, 78.6)

55.5 ± 7.2 (38.6, 72.5)

38.9 ± 11.0 (8.5, 86.3)

90.0 (87.5, 92.1)

PPV

92.2 (91.2, 93.0)

86.9 (84.3, 89.2)

76.6 ± 1.1 (71.9, 81.3)

69.5 ± 3.1 (62.2, 76.8)

62.3 ± 3.9 (45.4, 79.2)

90.4 (87.9, 92.4)

NPV

81.3 (79.3, 83.1)

89.2 (86.5, 91.4)

86.4 ± 0.5 (84.0, 88.7)

83.2 ± 1.6 (79.4, 87.0)

82.2 ± 2.4 (71.9, 92.4)

87.5 (84.8, 90.0)

Time(min)

 

0.67 (~ 40 s)

110.0 ± 5.8 (85.2, 134.8)

99.3 ± 6.3 (84.3, 114.2)

106.7 ± 8.8 (68.7, 144.6)

 
  1. eNPM-DM endoscopic images-based nasopharyngeal malignancies detection model, PPV positive predictive value, NPV negative predictive value
  2. aThe numbers in parenthesis are the corresponding 95% confidence interval
  3. bThe performance of the oncologists is presented as mean ± standard error