人工智能辅助超声对中国女性乳腺病灶识别的有效性研究

doi:10.19401/j.cnki.1007-3639.2023.11.005

中国癌症杂志 ›› 2023, Vol. 33 ›› Issue (11): 1002-1008.doi: 10.19401/j.cnki.1007-3639.2023.11.005

人工智能辅助超声对中国女性乳腺病灶识别的有效性研究

沈洁¹(), 刘雅静², 莫淼¹, 周瑾², 王泽洲¹, 周昌明¹, 周世崇², 常才², 郑莹¹^,³()

1.复旦大学附属肿瘤医院肿瘤预防部，复旦大学上海医学院肿瘤学系，上海 200032
2.复旦大学附属肿瘤医院超声科，复旦大学上海医学院肿瘤学系，上海 200032
3.上海肿瘤疾病人工智能工程技术研究中心，上海 200032

收稿日期:2023-07-13 修回日期:2023-09-05 出版日期:2023-11-30 发布日期:2023-12-14
通信作者: 郑莹（ORCID: 0000-0002-6408-8510），主任医师，复旦大学附属肿瘤医院肿瘤预防部主任。
作者简介:沈洁（ORCID: 0000-0003-2504-4491），主管医师。
基金资助:
上海市老龄化和妇儿健康研究专项(2020YJZX0206);上海市申康医院发展中心市级医院诊疗技术推广及优化管理项目(SHDC22022308);上海申康医院发展中心管理研究项目(2022SKMR-24)

Effectiveness of artificial intelligence-assisted ultrasound for breast cancer screening in Chinese women

SHEN Jie¹(), LIU Yajing², MO Miao¹, ZHOU Jin², WANG Zezhou¹, ZHOU Changming¹, ZHOU Shichong², CHANG Cai², ZHENG Ying¹^,³()

1. Department of Cancer Prevention, Fudan University Shanghai Cancer Center, Department of Oncology, Shanghai Medical College, Fudan University, Shanghai 200032, China
2. Department of Ultrasound, Fudan University Shanghai Cancer Center, Department of Oncology, Shanghai Medical College, Fudan University, Shanghai 200032, China
3. Shanghai Engineering Research Center of Artificial Intelligence Technology for Tumor Diseases, Shanghai 200032, China

Received:2023-07-13 Revised:2023-09-05 Published:2023-11-30 Online:2023-12-14

摘要/Abstract

摘要：

背景与目的：人工智能（artificial intelligence，AI）技术可辅助影像学诊断。本研究探讨AI辅助超声对中国女性乳腺病灶的识别能力及其应用于乳腺癌筛查的可能性。方法： 采用平行对照诊断性试验和前瞻性随访的研究设计，纳入至肿瘤专科医院就诊、并行乳腺超声检查的非乳腺癌女性。所有女性首先接受AI辅助超声检查，然后接受常规超声检查，比较AI辅助超声和常规超声识别乳腺病灶的差异；随访1年内乳腺癌发生情况，比较两种超声方式诊断乳腺癌的灵敏度和特异度。结果： 研究纳入360人，共发现2 504个乳腺病灶，其中AI辅助超声报告2 217个病灶，病灶报告率为88.5%；常规超声报告1 090个病灶，病灶报告率为43.5%。以常规超声为标准，AI辅助超声识别乳腺影像报告和数据系统（Breast Imaging Reporting and Data System，BI-RADS）4级以上乳腺病灶的灵敏度为93.3%（95% CI：80.7%~98.3%），特异度为100.0%（95% CI：99.5%~100.0%）；随访发现10例乳腺癌，AI辅助超声和常规超声均判定为阳性的有8例，灵敏度均为80.0%（95% CI：44.2%~96.4%），特异度均为88.6%（95% CI：84.6%~91.6%）。结论： AI辅助超声对于BI-RADS 4A以上的高危乳腺病灶及早期乳腺癌的识别能力与常规超声相当，是一种有效的乳腺癌辅助诊断手段，并具有应用于人群乳腺癌筛查的潜力。

关键词: 乳腺癌筛查, 人工智能, 乳腺超声

Abstract:

Background and purpose: Artificial intelligence (AI) technology is increasingly being used in the medical field. This study aimed to assess the effectiveness of artificial intelligence ultrasound system for identifying breast lesions in Chinese women and its role in breast cancer early detection. Methods: A prospective study was conducted on healthy women aged 35-74 years who came to Fudan University Shanghai Cancer Center from August 2020 to December 2020 for breast ultrasonography. All the women were examined by AI-assisted ultrasound first, and then by conventional ultrasonography. We compared the differences between AI-assisted ultrasound and conventional ultrasonography in identifying breast lesions in Chinese women. One year later, we looked up the hospital medical history and Shanghai Cancer Registration Management System for the final diagnosis of breast cancer. Results: A total of 360 women were included in the study and received breast examinations using both AI-assisted ultrasound and conventional ultrasound. A total of 2 504 breast lesions were detected, of which, 2 217 were detected by AI-assisted ultrasound, with a lesion recognition rate of 88.5%. Conventional ultrasound identified 1 090 lesions, with a lesion recognition rate of 43.5%. Using conventional ultrasound as the standard, the sensitivity and specificity of AI-assisted ultrasound for Breast Imaging Reporting and Data System (BI-RADS) level 4 and above lesions were 93.3% (95% CI: 80.7-98.3) and 100.0% (95% CI: 99.5-100.0), respectively. During one-year follow-up, 10 patients were diagnosed with breast cancer, and 8 of whom were identified by both AI-assisted ultrasound and conventional B ultrasound. The sensitivity of AI-assisted ultrasound and conventional ultrasound for breast cancer was 80.0% (95% CI: 44.2-96.4), and the specificity was 88.6% (95% CI: 84.6-91.6). Conclusion: AI-assisted ultrasound has good identification ability for breast lesions in Chinese women. The recognition ability for high-risk breast lesions (BI-RADS 4A and above) and early breast cancer is equivalent to that of conventional ultrasound, which is suitable for breast cancer screening in large-scale community of women with general risk.

Key words: Breast cancer screening, Artificial intelligence, Breast ultrasonography

中图分类号:

R737.9

沈洁, 刘雅静, 莫淼, 周瑾, 王泽洲, 周昌明, 周世崇, 常才, 郑莹. 人工智能辅助超声对中国女性乳腺病灶识别的有效性研究[J]. 中国癌症杂志, 2023, 33(11): 1002-1008.

SHEN Jie, LIU Yajing, MO Miao, ZHOU Jin, WANG Zezhou, ZHOU Changming, ZHOU Shichong, CHANG Cai, ZHENG Ying. Effectiveness of artificial intelligence-assisted ultrasound for breast cancer screening in Chinese women[J]. China Oncology, 2023, 33(11): 1002-1008.

图/表 8

图1

图2

表1

表2

表3

表4

表5

表6

参考文献 21

[1]	SUNG H, FERLAY J, SIEGEL R L, et al. Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries[J]. CA Cancer J Clin, 2021, 71(3): 209-249. doi: 10.3322/caac.v71.3
[2]	ZHENG R S, ZHANG S W, ZENG H M, et al. Cancer incidence and mortality in China, 2016[J]. J Natl Cancer Cent, 2022, 2(1): 1-9.
[3]	HABBEMA J D, VAN OORTMARSSEN G J, VAN PUTTEN D J, et al. Age-specific reduction in breast cancer mortality by screening: an analysis of the results of the Health Insurance Plan of Greater New York study[J]. J Natl Cancer Inst, 1986, 77(2): 317-320. pmid: 3461193
[4]	NYSTRÖM L, ANDERSSON I, BJURSTAM N, et al. Long-term effects of mammography screening: updated overview of the Swedish randomised trials[J]. Lancet, 2002, 359(9310): 909-919. doi: 10.1016/S0140-6736(02)08020-0 pmid: 11918907
[5]	SIEGEL R L, MILLER K D, JEMAL A. Cancer statistics, 2020[J]. CA A Cancer J Clin, 2020, 70(1): 7-30. doi: 10.3322/caac.v70.1
[6]	ZHANG X, LIN X, TAN Y J, et al. A multicenter hospital-based diagnosis study of automated breast ultrasound system in detecting breast cancer among Chinese women[J]. Chung Kuo Yen Cheng Yen Chiu, 2018, 30(2): 231-239.
[7]	WU G H, CHEN L S, CHANG K J, et al. Evolution of breast cancer screening in countries with intermediate and increasing incidence of breast cancer[J]. J Med Screen, 2006, 13(Suppl 1): S23-S27.
[8]	BAN K, TSUNODA H, WATANABE T, et al. Characteristics of ultrasonographic images of ductal carcinoma in situ with abnormalities of the ducts[J]. J Med Ultrason (2001), 2020, 47(1): 107-115.
[9]	杨萌, 晋思琦, 朱庆莉, 等. 乳腺癌超声诊断技术的发展历程及协和经验[J]. 中国科学(生命科学), 2021, 51(8): 1092-1100.
	YANG M, JIN S Q, ZHU Q L, et al. Development and experience of ultrasonic diagnosis technology for breast cancer[J]. Sci Sin Vitae, 2021, 51(8): 1092-1100. doi: 10.1360/SSV-2021-0204
[10]	FEI X Y, SHEN L, YING S H, et al. Parameter transfer deep neural network for single-modal B-mode ultrasound-based computer-aided diagnosis[J]. Cogn Comput, 2020, 12(6): 1252-1264. doi: 10.1007/s12559-020-09761-1
[11]	QIAO M Y, FANG Z, GUO Y, et al. Breast calcification detection based on multichannel radiofrequency signals via a unified deep learning framework[J]. Expert Syst Appl, 2021, 168: 114218. doi: 10.1016/j.eswa.2020.114218
[12]	GUO Y, HU Y, QIAO M, et al. Radiomics analysis on ultrasound for prediction of biologic behavior in breast invasive ductal carcinoma[J]. Clin Breast Cancer, 2018, 18(3): e335-e344. doi: 10.1016/j.clbc.2017.08.002
[13]	卢伟, 郑莹. 肿瘤命名与编码[M]. 上海: 第二军医大学出版社, 2011.
	LU W, ZHENG Y. Tumor Nomenclature and Coding[M]. Shanghai: Second Military Medical University Press, 2011.
[14]	中国抗癌协会乳腺癌专业委员会. 中国抗癌协会乳腺癌诊治指南与规范(2021年版)[J]. 中国癌症杂志, 2021, 31(10): 954-1040.
	The Society of Breast Cancer, China Anti-Cancer Association. Guidelines for breast cancer diagnosis and treatment by China Anti-Cancer Association (2021 edition)[J]. China Oncol, 2021, 31(10): 954-1040.
[15]	ZHANG X, YANG L, LIU S, et al. Evaluation of different breast cancer screening strategies for high-risk women in Beijing, China: a real-world population-based study[J]. Front Oncol, 2021, 11: 776848. doi: 10.3389/fonc.2021.776848
[16]	马兰, 连臻强, 赵艳霞, 等. 基于1 501 753名中国农村妇女乳腺癌筛查的乳腺超声优化流程分析[J]. 中华肿瘤杂志, 2021, 43(4): 497-503.
	MA L, LIAN Z Q, ZHAO Y X, et al. Breast ultrasound optimization process analysis based on breast cancer screening for 1 501 753 rural women in China[J]. Chin J Oncol, 2021, 43(4): 497-503
[17]	WANG P, BERZIN T M, GLISSEN BROWN J R, et al. Real-time automatic detection system increases colonoscopic polyp and adenoma detection rates: a prospective randomised controlled study[J]. Gut, 2019, 68(10): 1813-1819. doi: 10.1136/gutjnl-2018-317500 pmid: 30814121
[18]	ARDILA D, KIRALY A P, BHARADWAJ S, et al. End-to-end lung cancer screening with three-dimensional deep learning on low-dose chest computed tomography[J]. Nat Med, 2019, 25(6): 954-961. doi: 10.1038/s41591-019-0447-x pmid: 31110349
[19]	MCKINNEY S M, SIENIEK M, GODBOLE V, et al. International evaluation of an AI system for breast cancer screening[J]. Nature, 2020, 577(7788): 89-94. doi: 10.1038/s41586-019-1799-6
[20]	LEHMAN C D, WELLMAN R D, BUIST D S, et al. Diagnostic accuracy of digital screening mammography with and without computer-aided detection[J]. JAMA Intern Med, 2015, 175(11): 1828-1837. doi: 10.1001/jamainternmed.2015.5231 pmid: 26414882
[21]	中国抗癌协会乳腺癌专业委员会. 中国乳腺癌筛查与早期诊断指南[J]. 中国癌症杂志, 2022, 32(4): 363-372. doi: 10.19401/j.cnki.1007-3639.2022.04.010
	The Society of Breast Cancer, China Anti-Cancer Association. Screening and early diagnosis of breast cancer in China: a practice guideline[J]. China Oncol, 2022, 32(4): 363-372.

Item		Conventional ultrasound
Item		BI-RADS 4	BI-RADS 2-3
AI-assisted ultrasound	BI-RADS 4	TP	FP
AI-assisted ultrasound	BI-RADS 2-3	FN	TN

BI-RADS	0	2-3	4A	4B	4C	Total
AI-assisted ultrasound	6 (100.0)	2 169 (88.4)	31 (93.9)	8 (100.0)	3 (75.0)	2 217 (88.5)
Conventional ultrasound	6 (100.0)	1 039 (42.4)	33 (100.0)	8 (100.0)	4 (100.0)	1 090 (43.5)
Total	6	2 453	33	8	4	2 504

Item	Conventional ultrasound^#		Total
Item	BI-RADS 4	BI-RADS 2-3	Total
AI-assisted ultrasound
BI-RADS 4	42	0	42
BI-RADS 2-3	3	1 039	1 042
Total	45	1 039	1 084^*

Item	Breast cancer^*		Total
Item	Positive	Negative	Total
AI-assisted ultrasound
Positive	8	40	48
Negative	2	310	312
Total	10	350	360

Item	Breast cancer^*		Total
Item	Positive	Negative	Total
Conventional ultrasound
Positive	8	40	48
Negative	2	310	312
Total	10	350	360

人工智能辅助超声对中国女性乳腺病灶识别的有效性研究

Effectiveness of artificial intelligence-assisted ultrasound for breast cancer screening in Chinese women

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 8

参考文献 21

相关文章 4

编辑推荐

Metrics

本文评价

Accuracy	Conventional ultrasound	AI assisted ultrasound
Se	80.0 (44.2-96.4)	80.0 (44.2-96.4)
Sp	88.6 (84.6-91.6)	88.6 (84.6-91.6)
PPV	16.7 (8.0-30.8)	16.7 (8.0-30.8)
NPV	99.4 (97.4-99.9)	99.4 (97.4-99.9)

[1]	刘洋, 胡奕炀, 刘月平, 牛淑瑶, 丁平安, 田园, 郭洪海, 杨沛刚, 张泽, 郑涛, 檀碧波, 范立侨, 李勇, 赵群. 人工智能辅助技术在胃癌新辅助化疗患者HER2表达评估中的价值[J]. 中国癌症杂志, 2023, 33(4): 377-387.
[2]	徐文浩, 田熙, 艾合太木江·安外尔, 瞿元元, 施国海, 张海梁, 叶定伟. 人工智能在泌尿系统肿瘤中的应用研究进展[J]. 中国癌症杂志, 2022, 32(1): 68-74.
[3]	杨　鑫　 , 章　真 . 基于深度学习的人工智能在数字病理学中的进展[J]. 中国癌症杂志, 2021, 31(2): 151-155.
[4]	李毅,吕艳丽,赵越等. 不同组织模式下乳腺癌筛查结果分析[J]. 中国癌症杂志, 2014, 24(12): 944-950.