基于深度学习的乳腺X线智能检测系统对钙化的检出能力评估.pdfVIP

贵州医药2023年1月第47卷第1期 ・13・ 基于深度学习的乳腺X线智能检测系统对钙化的检出能力评估 邹雪瑾王齐艳郑勇蒲杨梅康敏△ （四川省妇幼保健院放射科，四川成都610045） 摘 要 目的评价基于深度学习（DL）的乳腺X线智能检测系统对乳腺钙化的检出效能 。方法回顾分析140 例经病理组织学证实的患者 ，并进行乳腺X线检查患者的头尾位（CC）和内外斛位（ML0） 图像资料，以2名经验丰 富高年资放射科医师对图像钙化病变的一致意见作为标准组 ，由低年资、高年资住院医师各1老和DL系统盲法独 立阅片 。比较DL系统与低年资和高年资住院医师检出敏感度差异，结合双向表#检验，评价不同ACR构成、BI- RADS分类、钙化形态和分布影响 。结果140例病例检出BI-RADS 3类以上钙化共275个，低年资住院医师 、高年 资住院医师和AI系统的检出敏感度分别是73. 45%（202/275）、96. 00%（264/275）和96. 36%（265/275）,低年资住 院医师检出敏感度低于高年资医师、DL系统（PV0.05） 。DL系统及高年资住院医师检出敏感度不受ACR腺体 构成、钙化形态、分布.BI-RADS分类等因素的影响（P0.05） ,而低年资住院医师检出敏感度受其影响（PV 0. 05） o结论 基于深度学习的乳腺智能钙化检测系统对BI-RADS 3类以上钙化灶的检出敏感度较高且稳定，可辅 助低年资住院医师减少钙化漏诊，尤其对于致密型乳腺、点状和无定形钙化、簇集和区域分布钙化的漏检。 关键词 可疑钙化； 乳腺X线摄影； 深度学习； 人工智能 Evaluation of mammography artificial intelligencedetection system based on deep learning for breast calcifications Zou Xuejin , Wang Qiyan , Zheng Yong , Pu Yangmei , Kang Min. Department of Radiology , Sichuan Maternal and Child Health Hospital, Chengdu 610045 , Sichuan, China. Abstract Objective To evaluate the detection efficiency of mammography calcification detection system based on deep learning for breast calcification. Methods To retrospectively analyze the breast images of cranio-caudal (CC) and medio-lateral oblique (ML0) view of 140 women patients who had been confirmed by histopathology. The unanimous opinion of two experienced senior radiologists about calcified lesions of breast images was used as the standard group. These images were respectively interpreted by one junior resident, one senior resident and the DL system. The sensitivities of the two residents and DL model were compared. Then two-way tests were used to analyze differences un