What is BTI RADS 2.0 ?
This a machine-learning-based approach to distinguish benign from malignant focal bone lesions. The algorithm was developed on a 1113 patient cohort and yielded high performance for the benign/malignant differentiation with multimodal imaging, CT/radiographs only and an MRI only approaches (F1 score=0.81). This algorithm led to proposing the Bone Tumor Imaging Reporting And Data System (BTI-RADS) v.2.0 with increasing malignancy rates among classes from 0%, to 8%, to 45% to 92%.Prerequisites
This classification was tested on patient with solitary bone tumor with imaging evaluation centered on a focal bone lesion using MRI with contrast injection and with CT or conventionnal radiographs (i.e. radiographs were not accepted for spinal or pelvic lesions). The classification was NOT tested on soft tissue tumors, known multifocal bone lesion (i.e. multifocal metastatic disease or systemic conditions), and lesions previously treated surgically or by adjuvant therapy.Give it a try with the Calculator.
Publication
For detailed methodology and results, see our multicenter study published in Radiology:“Enhanced CT and MRI Focal Bone Tumor Classification with Machine Learning–based Stratification”
https://doi.org/10.1148/radiol.232834
Access the code here: https://github.com/IADI-Nancy/BTIRADS2
If you use BTI-RADS 2.0 in your research, please cite the original article to acknowledge the work of the authors and contributors.
Disclaimers
All information is provided for educational purposes only. This information should not be used for the diagnosis or treatment of any health problem or disease. This information is not intended to replace clinical judgment or guide individual patient care in any manner. The user is hereby notified that the information contained herein may not meet the user's needs. The user is advised that, although the information is derived from medical research and literature we cannot guarantee either its correctness, comprehensiveness or currency. The User assumes sole responsibility for any decisions made or actions taken. The algorithm’s performance may vary according to the user’s radiological expertise.
