Bone Reporting and Data System on MRI (Bone-RADS-MRI): a validation study by four readers on 275 cases from three...

To evaluate the reproducibility and effectiveness of the Bone Reporting and Data System on MRI (Bone-RADS-MRI) for incidental solitary bone lesions in adults.

We retrospectively included 275 MRI cases from three local and two public databases, respectively. All the cases were histopathologically or clinically confirmed bone lesions, or "do not touch" lesions with typical appearance and remained stable for at least two years. Each lesion with gender, age, and clinical history was categorized according to the Bone-RADS algorithm by two musculoskeletal radiologists and two non-musculoskeletal radiologists. The Bone-RADS categories were as follows: Bone-RADS-1, likely benign, leave alone; Bone-RADS-2, incompletely assessed on imaging, perform different imaging modality; Bone-RADS-3, intermediate, perform follow-up imaging; Bone-RADS-4, suspicious for malignancy or need for treatment, biopsy and/or oncologic referral. Inter-reader agreement was evaluated. The diagnostic performance of the Bone-RADS-MRI was measured for distinguishing intermediate or malignant lesions or osteomyelitis from benign lesions. The histopathology results, clinical diagnosis, or follow-up were used as a standard reference.

There were 165 intermediate or malignant lesions or osteomyelitis, and 110 benign lesions, respectively. The inter-reader agreements between two musculoskeletal and between two non-musculoskeletal radiologists were both moderate (weighted kappa 0.572 and 0.520). The diagnostic performance for identifying intermediate or malignant lesions or osteomyelitis ranged according to radiologists with sensitivities of 88.5% to 94.5%, specificities of 55.5% to 74.5%, and accuracies of 76.4% to 82.9%.

Bone-RADS-MRI is effective for identifying bone lesions that need further treatment, but it has only moderate reliability for readers with different specialties and experience.

With local and public databases, Bone-RADS-MRI has been demonstrated to be a reliable algorithm for musculoskeletal and non-musculoskeletal radiologists with varying experience and an effective tool for identifying incidental solitary bone lesions that "need treatment" in adults.

Bone-RADS-MRI needs clinical validation for inter-reader agreement and diagnostic performance. Bone-RADS-MRI achieved moderate agreements between musculoskeletal and non-musculoskeletal radiologists, respectively. Bone-RADS-MRI presented high sensitivities but low specificities for identifying "need-for-treatment" bone lesions.