Pediatric hand fractures detection on radiographs: do localization cues improve diagnostic performance?

To compare the diagnostic accuracy and interpretation time for detection of pediatric fractures on hand radiographs with and without localization cues.

Consecutive children, who underwent radiographic examinations after injury, over 2 years (2019-2021) and with > 2 weeks of follow-up to confirm the presence or absence of a fracture, were included. Four readers, blinded to history and diagnosis, retrospectively reviewed all images twice, without and with cue, at least 1 week apart and after randomization, to determine the presence or absence of a fracture, and if present, anatomic location and diagnostic confidence were recorded. Interpretation time for each study was also recorded and averaged across readers. Inter-reader agreement was calculated using Fleiss' kappa. Diagnostic accuracy and interpretation time were compared between examinations using sensitivity, specificity, and Mann-Whitney U correlation.

Study group included 92 children (61 boys, 31 girls; 10.8 ± 3.4 years) with and 40 (31 boys, 9 girls; 10.9 ± 3.7 years) without fractures. Cue improved inter-reader agreement (κ = 0.47 to 0.62). While the specificity decreased (63 to 62%), sensitivity (75 to 78%), diagnostic accuracy (71 to 73%), and confidence improved (78 to 87%, p < 0.01), and interpretation time (median: 40 to 22 s, p < 0.001) reduced with examinations with localization cue. Specifically, examinations with fracture and cue had the shortest interpretation time (median: 16 s), whereas examinations without fracture and without cue had the longest interpretation time (median: 48 s).

Localization cues increased inter-reader agreement and diagnostic confidence, reduced interpretation time in the detection of fractures on pediatric hand radiographs, while maintaining diagnostic accuracy.