Performance and Reproducibility of Bone-RADS MRI Algorithm for Vertebral Lesions
Emil J Barkovich1, William A Mehan2, Karen Buch2
1From the Department of Radiology and Biomedical Imaging (E.J.B.), University of California, San Francisco, 505 Parnassus Ave, San Francisco, CA 94143 and Department of Radiology (W.A.M., K.B.), Massachusetts General Hospital, Harvard Medical School, 55 Fruit St, Boston, MA 02114. embarko@comcast.net.
AJNR. American Journal of Neuroradiology
|August 5, 2026
Summary
The Bone-RADS MRI algorithm shows excellent negative predictive value for vertebral lesions, indicating reliability in identifying benign cases. However, moderate inter-reader reliability suggests caution is needed when applying this tool to the spine.
Area of Science:
- Radiology
- Oncology
- Medical Imaging
Background:
- Vertebral lesions present diagnostic challenges.
- The Bone-RADS MRI algorithm aims to standardize assessment.
- Inter-rater reliability and accuracy are crucial for clinical application.
Purpose of the Study:
- To evaluate the inter-rater reliability of the SSR Bone-RADS MRI algorithm.
- To assess the accuracy of the Bone-RADS MRI algorithm for vertebral lesions.
- To determine the predictive values of the Bone-RADS MRI algorithm in a retrospective cohort.
Main Methods:
- Retrospective review of 538 MRI exams with "indeterminate" vertebral lesions.
- Exclusion of studies lacking T1WI/T2WI or with known vertebral metastases.
- Independent review by two neuroradiologists using Bone-RADS, excluding clinical criteria initially.
- Calculation of predictive values and inter-rater reliability using Cohen's kappa.
Main Results:
- 282 vertebral lesions met inclusion criteria; 60% had follow-up ≥2 months.
- Bone-RADS 1 demonstrated 100% negative predictive value (NPV).
- Positive predictive value for indeterminate lesions was 3.4%, decreasing to 0% with clinical criteria; inter-reader reliability (Cohen's kappa) was 0.60.
Conclusions:
- Bone-RADS 1 score offers excellent NPV for vertebral lesions.
- Lesions scored 2/3 had low malignancy rates, further reduced to 0% with clinical criteria.
- Moderate inter-reader reliability necessitates caution; a vertebral-specific algorithm may be beneficial.
