Table 1.
Age, sex, study duration, experience with radiography, traumatology, radiology.
Fig 1.
Inverse exponential curves fitted within the regression analyses of the main parameters of interest.
For the control, a linear fitting was applied as baseline. R2 values are given for the three groups: teams (circles), individual raters (crosses), and control (boxes).
Table 2.
Summary of fundamental parameters across all ten days and in total, split into the seven raters.
Sums are shown for false negatives and false positives, mean values for sensitivity, specificity, and F1 score, and mean values plus SD for IoU.
Fig 2.
Image labeling precision during study period (a) and F1 scores and IoU metrics in relation to patient age (b). On image b quadratic curve fittings with 95% CI are displayed.
Fig 3.
Regression analyses of annotation (a) and correction times (b). Logarithmic curve fittings are given for all three rater groups.
Fig 4.
The stars mark the areas of bone injury. a, c & d) Dorsal compression fractures of the distal radius. b & f) Overlooked scaphoid fractures. e) Missed epiphysiolysis, Salter-Harris type 2.
Fig 5.
Different cases of erroneously annotated fractures in pediatric wrist radiographs.
a) The cast was mistaken for a fracture. The + sign indicates the second, correctly labeled bone injury. b, c & f) Students marked the ulnar and radial growth plates as a fracture. d) A Madelung’s deformity was mimicking a fracture. e) The carpal bones were mistaken for an injury. g) A so-called Harris line thought to be a fracture. * The stars indicate missed injuries.