Skip to main content

Table 2 Summary of the changes in the decision-making of radiologists before and after AI assistance

From: Deep learning radiomics based on contrast-enhanced ultrasound images for assisted diagnosis of pancreatic ductal adenocarcinoma and chronic pancreatitis

 

Reader number

True negative

True positive

CP

PDACa

PDAC

CPb

Sensitivity (%)

Specificity (%)

Internal validation cohort

1

33→28

52→70↑

23

0

71.2→95.9↑*

91.7→77.8

2

19→27↑

65→73↑

9

9

89.0→100.0↑*

52.8→75.0↑*

3

22→27↑

70→73↑

3

5

95.9→100.0↑

61.1→75.0↑

4

32→33↑

63→69↑

6

1

86.3→94.5↑*

88.9→91.7↑

5

22→33↑

58→64↑

5

13

79.5→84.2↑

61.1→91.7↑*

External validation cohort 1

1

5→8↑

34→36↑

4

5

87.2→92.3↑

45.5→72.7↑

2

8→8

31→34↑

4

1

79.5→87.2↑

72.7→72.7

3

10→10

33→35↑

2

0

84.6→89.7↑

90.9→90.9

4

10→10

33→36↑

3

0

84.6→92.3↑

90.9→90.9

5

8→9↑

28→33↑

9

5

71.8→84.6↑

72.7→81.8↑

External validation cohort 2

1

7→7

31→35↑

6

2

81.6→92.1↑

70.0→70.0

2

6→6

32→35↑

3

0

84.2→92.1↑

60.0→60.0

3

9→8

31→35↑

7

2

81.6→92.1↑

90.0→80.0

4

6→8↑

34→36↑

2

2

89.5→94.7↑

60.0→80.0↑

5

7→4

29→36↑

13

3

76.3→94.7↑*

70.0→40.0

  1. The upward arrow (↑) represents indicators that improved owing to AI assistance
  2. PDAC pancreatic ductal adenocarcinoma, CP chronic pancreatitis, AI artificial intelligence
  3. *P<.05
  4. aNumber of patients for whom the radiologists altered their decision from CP to PDAC
  5. bNumber of patients for whom the radiologists altered their decision from PDAC to CP