人工智能辅助BI-RADS分类指导乳腺肿物活检的初步研究  

A Preliminary Study of Artificial Intelligence (S-detect) Assisted BI-RADS Categorization in Guiding Breast Lesion Biopsy

作　　者：冯杰 吴宏 王心怡[1] 朱璐[1] 王淑敏[1] 陈文[1] 崔立刚[1] Feng Jie;Wu Hong;Wang Xinyi;Zhu Lu;Wang Shumin;Chen Wen;Cui Ligang(Department of Ultrasound,Peking University Third Hospital,Beijing 100191,China)

机构地区：[1]北京大学第三医院超声科,北京市100191

出　　处：《中国超声医学杂志》2020年第4期325-328,共4页Chinese Journal of Ultrasound in Medicine

摘　　要：目的探讨人工智能(S-Detect技术)辅助BI-RADS分类指导乳腺肿物活检的价值。方法常规超声BI-RADS分类均为4类的51例患者,共52个病灶均进行穿刺活检。穿刺活检前行S-Detect辅助结节的BI-RADS再次分类,以病理结果作为金标准比较两种分类方法的差异。结果BI-RADS 4类结节遵照指南均应进行穿刺活检,活检率100%(52/52)、阳性预测值38.5%(20/52)、假阳性率61.5%(32/52)、漏诊率0%(0/20)。S-Detect辅助分类后,52个病灶中24个降类为3类,则活检率53.8%(28/52)、阳性预测值67.9%(19/28)、假阳性率32.1%(9/28)、漏诊率5%(1/20)均发生明显变化,其中活检率及假阳性率显著减低(P<0.01),漏诊率的变化无统计学意义(P>0.05)。结论S-Detect辅助乳腺结节再次分类有助于减少不必要的穿刺活检,提高活检效能。Objective To determine the value of using S-Detect as a reference criteria for BI-RADS categorization and breast lesion biopsy.Methods A total of 51patients with 52lesions classified as BI-RADS category 4on conventional ultrasound(US)were evaluated before US-guided biopsy.The differences between conventional US results and S-Detect assisted US results were compared.The pathological results were regarded as gold standard.Results As all lesions classified as BI-RADS category 4by conventional US were proposed for biopsy,the biopsy rate,positive predictive value,false positive rate and misdiagnosis rate were 100%(52/52),38.5%(20/52),61.5%(32/52)and 0%(0/20),respectively.After reclassification with S-Detect,24lesions were downgraded to BI-RADS category 3.The biopsy rate,positive predictive value,false positive rate and misdiagnosis rate turned to 53.8%(28/52),67.9%(19/28),32.1%(9/28)and 5%(1/20),respectively.With the assist of S-Detect,the biopsy rate and false positive rate reduced significantly(P<0.01),while the misdiagnosis rate increased with no statistical difference(P>0.05).Conclusions Using S-Detect as a reference criteria for BI-RADS category reduced unnecessary breast lesion biopsies and increased the efficiency of biopsy.

关 键 词：BI-RADS分类 S-Detect 活检 

分 类 号：R445.1[医药卫生—影像医学与核医学] R737.9[医药卫生—诊断学] TP18[医药卫生—临床医学]