This retrospective study was conducted using cases from three local institutional databases and two public databases. The study has been approved by the institutional review board, and the written informed consents from participants were waived. The two public databases, Eurorad (https://www.eurorad.org) and Radiopaedia (https://radiopaedia.org), were under the Creative Commons License CC BY-NC-SA 4.0 and CC BY-NC-SA 3.0, respectively.

The sample selection was conducted by two musculoskeletal radiologists with 6 and 7 years of experience. All the cases were histopathologically or clinically confirmed bone lesions, “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 Bone-RADS algorithm by four readers, including two musculoskeletal radiologists with 6 and 8 years of experience, and two non-musculoskeletal radiologists both with 9 years of experience. The readers were told to rate the Bone-RADS categories for each lesion according to following criteria: Bone-RADS-1, likely benign, leave alone; Bone-RADS-2, incomplete 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. During formal evaluation, the PowerPoint document of images and Excel document of patient information were provided to the readers but the source of the case, final diagnosis, and databases were blinded. The cases were randomly presented to the readers, and there were no time limits for the rating process of each case. Inter-reader agreement was evaluated. The diagnostic performance of the Bone-RADS-MRI for distinguishing intermediate or malignant lesions or osteomyelitis, from benign lesions, were measured, using histopathology results, clinical diagnosis, or follow-up as a standard reference.