Skip to main content
Insights into Imaging

Table 2 Diagnostic accuracy of reader study

From: A deep learning model using chest X-ray for identifying TB and NTM-LD patients: a cross-sectional study

 

Internal test

All doctors

Senior

Junior

DNN

p*

Overall

49.2 ± 3.4%

50.8 ± 3.0%

47.5 ± 2.8%

66.5 ± 2.5%

 < 0.001

AFS

     

 AFS ( +)

46.8 ± 9.0%

47.1 ± 11.8%

46.6 ± 4.9%

73.0 ± 3.8%

 < 0.001

 AFS (-)

50.1 ± 5.4%

52.3 ± 5.4%

47.9 ± 4.4%

64.0 ± 2.9%

 < 0.001

Diagnosis

     

 Imitator

59.2 ± 19.7%

61.2 ± 19.5%

57.1 ± 19.7%

60.6 ± 5.1%

0.816

 TB

55.6 ± 13.5%

56.7 ± 8.0%

54.6 ± 17.3%

74.0 ± 6.7%

 < 0.001

 NTM-LD

32.7 ± 12.5%

34.6 ± 15.5%

30.8 ± 8.0%

65.0 ± 5.5%

 < 0.001

  1. DNN, deep neural network; AFS, acid-fast smear; TB, tuberculosis; NTM-LD, nontuberculous mycobacterial lung disease
  2. p* compared between all doctors and DNN on diagnosis accuracy
Back to article page