Table 2 Diagnostic performances of each subgroup after conducting AI-unassisted or AI-assisted evaluation for differentiating CCM and AIH.

From: Improved differentiation of cavernous malformation and acute intraparenchymal hemorrhage on CT using an AI algorithm

Reviewers

AI-unassisted

AI-assisted

P value

Total reviewers (%) (95% confidence interval)

 Accuracy

79.86 (75.03, 84.05)

86.92 (82.77, 90.28)

 < 0.001

 Sensitivity

71.39 (68.56, 74.05)

81.21 (78.72, 83.47)

 Specificity

92.61 (90.41, 94.33)

95.51 (93.69, 96.81)

Emergency department physicians

 Accuracy

72.57 (67.29, 77.38)

80.73 (75.86, 84.96)

0.006

 Sensitivity

59.53 (54.29, 64.58)

70.23 (65.21, 74.81)

 Specificity

92.17 (87.97, 94.99)

96.52 (93.29, 98.22)

Radiology residents

 Accuracy

75.35 (69.95, 80.21)

84.21 (79.63, 88.04)

 < 0.001

 Sensitivity

63.01 (57.80, 67.92)

77.17 (72.46, 81.28)

 Specificity

93.91 (90.04, 96.33)

94.78 (91.10, 96.99)

Neuroradiologists

 Accuracy

91.67 (87.88, 94.57)

95.83 (92.83, 97.83)

0.56

 Sensitivity

91.62 (88.22, 94.10)

96.24 (93.68, 97.79)

 Specificity

91.74 (87.46, 94.65)

95.22 (91.64, 97.31)

  1. AI, artificial intelligence; CCM, cerebral cavernous malformation; AIH, acute intraparenchymal hemorrhage.