Elverici Eda, Zengin Betul, Nurdan Barca Ayse, Didem Yilmaz Pinar, Alimli Aysegul, Araz Levent
Department of Radiology, Ankara Numune Education and Research Hospital, Ankara, Turkey.
Department of Radiology, Yozgat State Hospital, Yozgat, Turkey.
Iran J Radiol. 2013 Sep;10(3):122-7. doi: 10.5812/iranjradiol.10708. Epub 2013 Aug 30.
BI-RADS was first developed in 1993 for mammography and in 2003 it was redesigned for ultrasonography (US). If the observer agreement is high, the method used in the classification of lesion would be reproducible.
The aim of this study is to evaluate the inter- and intraobserver agreement of sonographic BI-RADS lexicon in the categorization and feature characterization of nonpalpable breast lesions.
We included 223 patients with 245 nonpalpable breast lesions who underwent ultrasound-guided wire needle localization. Two radiologists retrospectively described each lesion using sonographic BI-RADS descriptors and final assessment. The observers were blinded to mammographic images, medical history and pathologic results. Inter- and intraobserver agreement was assessed using Kappa (κ) agreement coefficient.
The interobserver agreement for sonographic descriptors changed between fair and substantial. The highest agreement was detected for mass orientation (κ=0.66). The lowest agreement was found in the margin (κ=0.33). The interobserver agreement for BI-RADS final category was found as fair (κ=0.35). The intraobserver agreement for sonographic descriptors changed between substantial and almost perfect. The intraobserver agreement of BI-RADS result category was found as substantial for observer 1 (κ=0.64) and excellent for observer 2 (κ=0.83).
Our results demonstrated that each observer was self-consistent in interpreting US BI-RADS classification, while interobserver agreement was relatively poor. Although it has been ten years since the description of sonographic BI-RADS lexicon, further training and periodic performance evaluations would probably help to achieve better agreement among radiologists.
乳腺影像报告和数据系统（BI-RADS）于1993年首次用于乳房X线摄影，并于2003年重新设计用于超声检查（US）。如果观察者之间的一致性较高，那么用于病变分类的方法将具有可重复性。
本研究旨在评估超声BI-RADS词典在不可触及乳腺病变的分类和特征描述方面观察者间和观察者内的一致性。
我们纳入了223例患有245个不可触及乳腺病变的患者，这些患者均接受了超声引导下钢丝针定位。两名放射科医生回顾性地使用超声BI-RADS描述符和最终评估来描述每个病变。观察者对乳房X线摄影图像、病史和病理结果不知情。使用Kappa（κ）一致性系数评估观察者间和观察者内的一致性。
超声描述符的观察者间一致性在中等和实质性之间变化。在肿块方向上检测到最高的一致性（κ=0.66）。在边缘方面一致性最低（κ=0.33）。BI-RADS最终分类的观察者间一致性为中等（κ=0.35）。超声描述符的观察者内一致性在实质性和几乎完美之间变化。BI-RADS结果分类的观察者内一致性对于观察者1为实质性（κ=0.64），对于观察者2为优秀（κ=0.83）。
我们的结果表明，每个观察者在解释超声BI-RADS分类时都是自我一致的，而观察者间的一致性相对较差。尽管超声BI-RADS词典的描述已经有十年了，但进一步的培训和定期的性能评估可能有助于放射科医生之间达成更好的一致性。
