Comparative efficacy of conventional and handheld ultrasound devices in detecting bone fractures: an in vitro study

Abstract

Aim: Ultrasound (US) is increasingly used for fracture diagnosis, yet the comparative performance of handheld and conventional US, particularly by novice users, remains unclear. This study evaluated their diagnostic accuracy using in vitro bone models.

Material and methods: Five blinded junior sonographers examined seven bone samples (comprising four fractured, comminuted, transverse, obliquely displaced, and linear, as well as three intact samples) using both types of devices. Diagnostic sensitivity, specificity, and accuracy were calculated. McNemar’s test evaluated differences in performance, Cohen’s Kappa measured inter-device agreement, and McNemar’s test along with mixed-effects logistic regression analysed the impact of fracture types. Inter-sonographer reliability was assessed using the intraclass correlation coefficient (ICC).

Results: Both devices showed 85% sensitivity (95% CI: 61.1–96%). Specificity was low (conventional: 20.0%; handheld: 33.3%). Accuracy was moderate (conventional: 57.1%; handheld: 62.9%). No significant difference was found (p = .21), and inter-device agreement was poor (Kappa = 0.05±0.20). Comminuted fractures were significantly more detectable than intact bones (OR=2.25, p=0.03). Inter-operator reliability was low (ICC: 0.28–0.35).

Conclusion: Handheld US showed slightly higher specificity and accuracy, but both systems demonstrated limited agreement and variable novice performance. These findings highlight the need for improved training and optimisation when using US for fracture detection.