Skip to main content
Insights into Imaging

Table 1 Balancing of training and test sets compared to real world epidemiology

From: Reproducibility of artificial intelligence models in computed tomography of the head: a quantitative analysis

Data set

Mean ± SD

Training sets

0.50 + 0.31

Test sets

0.47 + 0.30

Real word epidemiology

0.22 + 0.28

t test

Training sets/test sets

0.45 (.66)

Training sets/real world epidemiology

3.78 (.0004***)

Test sets/real world epidemiology

3.32 (.002**)

  1. We statistically compared the prevalences of diseases of the training and test sets in the article selection with the respective real world epidemiology. We used a Welch Two sample t-test
  2. *p < .05. **p < .01. ***p < .001; n = 30
Back to article page