Accuracy evaluation of American College of Radiology Bone Reporting and Data System (ACR Bone-RADS) for extremity...

The aim of this study was to evaluate the accuracy and inter-reader reliability of the Bone Reporting and Data System supported by the American College of Radiology (ACR) in classifying benign and malignant bone tumours of the extremities, across observers with varying levels of experience.

This retrospective study analysed radiographs from 405 patients diagnosed with extremity bone tumours at our institution between January 2017 and January 2023, following the World Health Organization (WHO) classification based on pathology. Sixteen observers initially conducted a visual assessment and then received training in ACR Bone-RADS to assess relevant radiological features. Interobserver agreement was measured using Conger's kappa, and diagnostic accuracy was compared using generalised estimating equations (GEEs).

Of the 405 tumours, 239 were benign and 166 were malignant. Observers using ACR Bone-RADS demonstrated substantial agreement in distinguishing between benign and malignant cases (κ = 0.63). Compared with visual assessment, ACR Bone-RADS improved diagnostic accuracy for both junior and attending physicians but decreased accuracy for senior physicians. Significant differences in diagnostic accuracy were observed among junior, attending, and senior physicians during the initial visual assessment. However, these differences were not significant after implementing ACR Bone-RADS. In a separate analysis of indeterminate cases, junior physicians achieved significantly higher accuracy than both attending and senior physicians.

Implementing ACR Bone-RADS improved diagnostic accuracy for junior and attending physicians compared to visual assessment but reduced accuracy for senior physicians.