Skip to main content
Insights into Imaging

Table 2 Performance of the raters on external validation set

From: Improving the diagnosis of acute ischemic stroke on non-contrast CT using deep learning: a multicenter study

 

Baseline

Aided

p value

Expert 1

   

Sensitivity

0.163 (0.105, 0.236)

0.319 (0.241, 0.404)

 < .001

Specificity

0.882 (0.865, 0.898)

0.913 (0.898, 0.927)

0.002

Expert 2

   

Sensitivity

0.185 (0.124, 0.261)

0.304 (0.228, 0.389)

0.006

Specificity

0.932 (0.919, 0.944)

0.950 (0.937, 0.960)

0.022

Expert 3

   

Sensitivity

0.378 (0.296, 0.465)

0.370 (0.289, 0.458)

1

Specificity

0.855 (0.837, 0.872)

0.880 (0.863, 0.896)

0.019

Expert 4

   

Sensitivity

0.289 (0.214, 0.373)

0.341 (0.261, 0.427)

0.281

Specificity

0.913 (0.898, 0.927)

0.918 (0.903, 0.931)

0.63

Average

   

Sensitivity

0.254 (0.167, 0.341)

0.333 (0.248, 0.419)

0.196

Specificity

0.896 (0.884, 0.907)

0.915 (0.904, 0.926)

0.014

Region-based ICC

0.741

0.98

–

Score-based ICC

0.614

0.809

–

  1. Sensitivity and specificity were expressed as point estimate with lower and upper bound of 95% confidence interval (CI) in parentheses. CI for average sensitivity and specificity was calculated by the Jackknife method
Back to article page