Objective To compare the malignancy risk stratification performance and inter-reader reliability of four Reporting and Data System (RADS) algorithms for solitary bone lesions: CT Bone-RADS, MRI Bone-RADS, Osseous Tumor (OT)-RADS, and Bone Tumor Imaging (BTI)-RADS. Materials and methods This retrospective analysis included patients with solitary bone lesions who underwent both CT and MRI between March 2005 and September 2021. Three radiologists independently categorized each lesion using CT Bone-RADS (1–4), MRI Bone-RADS (1–4), OT-RADS (2–5), and BTI-RADS (1–4). Categories were dichotomized into high- versus low-risk for malignancy. Diagnostic performance metrics and area under the receiver operating characteristic curve (AUC) were calculated for each reader as well as for a consensus interpretation generated using a majority-vote method. The reference standard was either histopathologic confirmation or imaging surveillance. Inter-reader reliability was assessed using Gwet’s AC1 statistic. Results A total of 207 patients (mean age, 49 ± 18 years; 111 men and 96 women) were included. Consensus malignancy risk stratification performance (AUC; sensitivity/specificity/positive predictive value/negative predictive value/accuracy, %) was as follows: CT Bone-RADS (0.52; 95/9/43/73/44), MRI Bone-RADS (0.60; 98/12/44/88/47), OT-RADS (0.91; 93/71/69/94/80), and BTI-RADS (0.89; 98/39/53/96/63). Inter-reader reliability (AC1) was excellent for CT Bone-RADS (0.978), MRI Bone-RADS (0.931), and BTI-RADS (0.822), and moderate for OT-RADS (0.585). Conclusion Among the evaluated bone tumor-RADS, OT-RADS demonstrated the most balanced diagnostic performance with moderate inter-reader reliability. CT Bone-RADS, MRI Bone-RADS, and BTI-RADS showed excellent inter-reader reliability. Key Points <Emphasis Type="BoldItalic">Question</Emphasis> Evaluation of solitary bone lesions is important but often challenging. This study compared four bone tumor-RADS algorithms to determine which provides the best malignancy risk stratification. <Emphasis Type="BoldItalic">Findings</Emphasis> Among the four RADS algorithms, OT-RADS demonstrated the most balanced overall diagnostic performance in consensus analysis, while CT Bone-RADS, MRI Bone-RADS, and BTI-RADS showed excellent inter-reader reliability. <Emphasis Type="BoldItalic">Clinical relevance</Emphasis> Knowledge of each RADS system’s performance characteristics helps clinicians apply these algorithms appropriately to optimize the assessment of solitary bone lesions. Graphical Abstract

Comparative analysis of RADS classification systems for solitary bone lesions: malignancy risk stratification performance and clinical utility

期刊论文
发表日期：

2026年3月6日

查看全文

Arash Azhideh,
Sara Haseli,
Chankue Park,
Hyeyun Lee,
Hyun Su Kim,
Peyman Mirghaderi,
Majid Chalian

摘要

Objective

To compare the malignancy risk stratification performance and inter-reader reliability of four Reporting and Data System (RADS) algorithms for solitary bone lesions: CT Bone-RADS, MRI Bone-RADS, Osseous Tumor (OT)-RADS, and Bone Tumor Imaging (BTI)-RADS.

Materials and methods

This retrospective analysis included patients with solitary bone lesions who underwent both CT and MRI between March 2005 and September 2021. Three radiologists independently categorized each lesion using CT Bone-RADS (1–4), MRI Bone-RADS (1–4), OT-RADS (2–5), and BTI-RADS (1–4). Categories were dichotomized into high- versus low-risk for malignancy. Diagnostic performance metrics and area under the receiver operating characteristic curve (AUC) were calculated for each reader as well as for a consensus interpretation generated using a majority-vote method. The reference standard was either histopathologic confirmation or imaging surveillance. Inter-reader reliability was assessed using Gwet’s AC1 statistic.

Results

A total of 207 patients (mean age, 49 ± 18 years; 111 men and 96 women) were included. Consensus malignancy risk stratification performance (AUC; sensitivity/specificity/positive predictive value/negative predictive value/accuracy, %) was as follows: CT Bone-RADS (0.52; 95/9/43/73/44), MRI Bone-RADS (0.60; 98/12/44/88/47), OT-RADS (0.91; 93/71/69/94/80), and BTI-RADS (0.89; 98/39/53/96/63). Inter-reader reliability (AC1) was excellent for CT Bone-RADS (0.978), MRI Bone-RADS (0.931), and BTI-RADS (0.822), and moderate for OT-RADS (0.585).

Conclusion

Among the evaluated bone tumor-RADS, OT-RADS demonstrated the most balanced diagnostic performance with moderate inter-reader reliability. CT Bone-RADS, MRI Bone-RADS, and BTI-RADS showed excellent inter-reader reliability.

Key Points

Question Evaluation of solitary bone lesions is important but often challenging. This study compared four bone tumor-RADS algorithms to determine which provides the best malignancy risk stratification.

Findings Among the four RADS algorithms, OT-RADS demonstrated the most balanced overall diagnostic performance in consensus analysis, while CT Bone-RADS, MRI Bone-RADS, and BTI-RADS showed excellent inter-reader reliability.

Clinical relevance Knowledge of each RADS system’s performance characteristics helps clinicians apply these algorithms appropriately to optimize the assessment of solitary bone lesions.

Graphical Abstract