现代仪器与医疗

深度学习模型辅助基层超声医师鉴别乳腺结节良恶性的诊断效能研究

PDF下载

容炜 ¹ , 谷丽萍 ²^,^*

现代仪器与医疗 | 医学工程 2026,32(3): 53-60

收起

现代仪器与医疗 |医学工程 2026 , 32 (3) : 53 -60

深度学习模型辅助基层超声医师鉴别乳腺结节良恶性的诊断效能研究

全屏

[Author(id=1320381134832251469, tenantId=1146029695717560320, journalId=1319373289681944729, articleId=1320381132105953858, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1320381134916137551, tenantId=1146029695717560320, journalId=1319373289681944729, articleId=1320381132105953858, authorId=1320381134832251469, language=EN, stringName=Wei Rong, firstName=Wei, middleName=null, lastName=Rong, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=¹, address=^1.Medical Technology Department, Shanghai Pudong New Area Wanxiang Community Health Service Center, Shanghai 201313, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1320381135004217936, tenantId=1146029695717560320, journalId=1319373289681944729, articleId=1320381132105953858, authorId=1320381134832251469, language=CN, stringName=容炜, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=¹, address=^1.上海市浦东新区万祥社区卫生服务中心医技科，上海　201313, bio={"content":"

容炜，男，本科，主治医师，研究方向：超声，社区常见疾病风险分层及预警

"}, bioImg=null, bioContent=

容炜，男，本科，主治医师，研究方向：超声，社区常见疾病风险分层及预警

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1320381134605759046, tenantId=1146029695717560320, journalId=1319373289681944729, articleId=1320381132105953858, xref=1., ext=[AuthorCompanyExt(id=1320381134618341959, tenantId=1146029695717560320, journalId=1319373289681944729, articleId=1320381132105953858, companyId=1320381134605759046, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=^1.Medical Technology Department, Shanghai Pudong New Area Wanxiang Community Health Service Center, Shanghai 201313, China), AuthorCompanyExt(id=1320381134630924872, tenantId=1146029695717560320, journalId=1319373289681944729, articleId=1320381132105953858, companyId=1320381134605759046, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=^1.上海市浦东新区万祥社区卫生服务中心医技科，上海　201313)])]), Author(id=1320381135092298322, tenantId=1146029695717560320, journalId=1319373289681944729, articleId=1320381132105953858, orderNo=1, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=1, authorType=1, ext={EN=AuthorExt(id=1320381135188767316, tenantId=1146029695717560320, journalId=1319373289681944729, articleId=1320381132105953858, authorId=1320381135092298322, language=EN, stringName=Liping Gu, firstName=Liping, middleName=null, lastName=Gu, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=²^,^*, address=^2.Department of Ultrasound in Medicine, Sixth People's Hospital Affiliated to Medical College of Shanghai Jiaotong University, Shanghai 200233, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1320381135247487573, tenantId=1146029695717560320, journalId=1319373289681944729, articleId=1320381132105953858, authorId=1320381135092298322, language=CN, stringName=谷丽萍, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=²^,^*, address=^2.上海交通大学医学院附属第六人民医院超声医学科，上海　200233, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1320381134698033737, tenantId=1146029695717560320, journalId=1319373289681944729, articleId=1320381132105953858, xref=2., ext=[AuthorCompanyExt(id=1320381134706422346, tenantId=1146029695717560320, journalId=1319373289681944729, articleId=1320381132105953858, companyId=1320381134698033737, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=^2.Department of Ultrasound in Medicine, Sixth People's Hospital Affiliated to Medical College of Shanghai Jiaotong University, Shanghai 200233, China), AuthorCompanyExt(id=1320381134714810955, tenantId=1146029695717560320, journalId=1319373289681944729, articleId=1320381132105953858, companyId=1320381134698033737, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=^2.上海交通大学医学院附属第六人民医院超声医学科，上海　200233)])])]

容炜¹, 谷丽萍²^,^*

作者信息

^1.上海市浦东新区万祥社区卫生服务中心医技科，上海　201313

^2.上海交通大学医学院附属第六人民医院超声医学科，上海　200233

通讯作者:

谷丽萍，女，硕士研究生，副主任医师，研究方向：乳腺肿瘤，甲状腺肿瘤

作者简介:

容炜，男，本科，主治医师，研究方向：超声，社区常见疾病风险分层及预警

A study on the diagnostic efficacy of deep learning models in assisting primary ultrasound physicians in differentiating benign and malignant breast nodules

Wei Rong¹, Liping Gu²^,^*

Affiliations

^1.Medical Technology Department, Shanghai Pudong New Area Wanxiang Community Health Service Center, Shanghai 201313, China

^2.Department of Ultrasound in Medicine, Sixth People's Hospital Affiliated to Medical College of Shanghai Jiaotong University, Shanghai 200233, China

出版时间: 2026-06-10 doi: 10.11876/mimt202603007

文章导航

摘要

收起

目的

评估深度学习模型辅助基层超声医师鉴别乳腺结节良恶性的诊断效能。

方法

回顾性收集2022年9月至2024年9月在上海市第六人民医院进行常规超声检查时发现乳腺结节并诊断为BI-RADS 3类和4类的女性患者，收集患者临床资料及超声图像。构建GoogLeNet、VGG11、AlexNet、ResNet-34、ViT 5种深度学习模型，并通过对比曲线下面积（area under curve，AUC）等指标选取最佳模型。采用最佳模型对超声图像生成可视化热图，由社区卫生服务中心医师在不知病理情况下对患者的超声图像进行独立阅片，两周后再次借助最佳模型生成的可视化热图进行阅片，通过准确性、灵敏度、特异度和精确度等指标，评估和比较深度学习模型阅片、超声医师独立阅片、深度学习模型辅助超声医师阅片的诊断效能。

结果

共纳入430例乳腺结节患者，分为训练集（n=301）和验证集（n=129）。基于乳腺超声图像构建的不同深度学习模型中，ResNet-34模型在鉴别乳腺结节良恶性方面表现最佳，AUC、准确性和灵敏度在训练集和验证集中都优于其他模型，因此选择ResNet-34模型作为最佳模型。在3种阅片方案中，基层医师独立阅片的诊断效能最低，深度学习模型在准确性、灵敏度、精确度等指标上略优于基层医师，而深度学习模型辅助超声医师阅片的诊断效能最优，所有评估指标较医师独立阅片都得到了大幅度提升。从提升幅度来看，灵敏度和F1值提升最明显，分别提高34.3%和29.9%。

结论

基于超声图像的深度学习模型能够较好地区分乳腺结节的良恶性，结合其生成的可视化热图能够大幅提升基层超声医师鉴别乳腺结节良恶性的诊断效能，可为乳腺结节患者治疗方案的选择提供有利指导信息。

关键词

乳腺结节 / 超声 / 深度学习 / 可视化 / 社区

Abstract

收起

Objective

Evaluate the diagnostic performance of a deep learning model in assisting primary care ultrasound doctors in differentiating benign and malignant breast nodules.

Methods

This study retrospectively collected female patients who were diagnosed with BI-RADS category 3 and 4 breast nodules during routine ultrasound examinations at Shanghai Sixth People's Hospital from September 2022 to September 2024. The clinical data and ultrasound images of these patients were collected.. Five deep learning models (GoogLeNet, VGG11, AlexNet, ResNet-34, ViT) were constructed, and the optimal model was selected based on metrics such as area under curve (AUC). Visual heatmaps of the ultrasound images were generated using the optimal model. Community ultrasound physicians independently reviewed the images without knowledge of the pathological results. Two weeks later, the physicians re-reviewed the images with the assistance of the generated heatmaps. Diagnostic performance was compared among three approaches: deep learning model alone, physician independent review, and deep learning-assisted physician review, using accuracy, sensitivity, specificity, and precision as metrics.

Results

A total of 430 patients with breast nodules were included and divided into a training set (n=301) and a validation set (n=129). Among the different learning models constructed based on breast ultrasound images, the ResNet-34 model performed best in predicting benign and malignant breast nodules, with AUC, accuracy, and sensitivity superior to other models in both the training and validation sets. Therefore, the ResNet-34 model was selected as the optimal model. Among the three approaches, the diagnostic efficacy of independent review by primary care physicians was the lowest. The deep learning model slightly outperformed primary care physicians in terms of accuracy, sensitivity, and precision, while the diagnostic efficacy of the deep learning model assisting ultrasound physicians in reading was the strongest, with all evaluation indicators significantly improved. In terms of the improvement rate, the sensitivity and F1 value showed the most significant increase, rising by 34.3% and 29.9% respectively.

Conclusion

Deep learning models based on ultrasound images can effectively distinguish the benign and malignant nature of breast nodules. Combined with the generated visual heatmaps, they can significantly enhance the diagnostic efficacy of assisting primary ultrasound physicians in distinguishing between benign and malignant breast nodules, providing valuable guidance for selecting treatment plans for patients with breast nodules.

Key words

Breast nodules / Ultrasound / Deep learning / Visualization / Community

引用本文

容炜, 谷丽萍. 深度学习模型辅助基层超声医师鉴别乳腺结节良恶性的诊断效能研究. 现代仪器与医疗, 2026 , 32 (3) : 53 -60 . DOI: 10.11876/mimt202603007

Wei Rong, Liping Gu. A study on the diagnostic efficacy of deep learning models in assisting primary ultrasound physicians in differentiating benign and malignant breast nodules[J]. Modern Instruments & Medical Treatment, 2026 , 32 (3) : 53 -60 . DOI: 10.11876/mimt202603007

基金

收起

2024年度“浦东新区卫生系统优秀社区适宜人才培养计划”(PWRs2024-14)
“市六-临港”紧密型健康联合体临床类科研项目(ynlglht202401)
上海市浦东新区万祥社区卫生服务中心院内科研项目(WX2024-01)

参考文献

收起

文献

收起

参考文献引证文献

排序方式：

[1]

师琳, 钟李长, 马方, 等. 瘤周超声影像组学对乳腺结节良恶性的鉴别诊断价值[J]. 肿瘤影像学, 2023, 32(6): 485-491.

[2]

Li L, Niu Y, Tian F, et al. An efficient deep learning strategy for accurate and automated detection of breast tumors in ultrasound image datasets[J]. Front Oncol, 2025, 14: 1461542.

[3]

Stastny I, Zubor P, Kajo K, et al. Aberrantly methylated cfDNA in body fluids as a promising diagnostic tool for early detection of breast cancer[J]. Clin Breast Cancer, 2020, 20(6): e711-e722.

[4]

张莉敏. 基于CT平扫影像组学及机器学习预测急性缺血性脑卒中常规治疗患者结局研究[D]. 重庆: 重庆医科大学, 2024.

[5]

张烁, 张荣, 张岩波. 基于深度学习的医学图像识别研究综述[J]. 中国卫生统计, 2020, 37(1): 150-156.

[6]

尚建军, 邬旭日. 超声造影对于不同大小BI-RADS 4类乳腺结节的良恶性定性诊断分析[J]. 影像研究与医学应用, 2022, 6(14): 152-154.

[7]

米拉·也尔兰, 张海见, 胡峙珩, 等. 超声影像组学对BIRADS 4a类不规则乳腺结节良恶性的鉴别价值[J]. 分子影像学杂志, 2023, 46(1): 12-20.

[8]

Merjane V, Perin D M P, Bacha P M G E, et al. Breast Imaging Reporting and Data System (BI-RADS®): a success history and particularities of its use in Brazil[J]. Rev Bras Ginecol Obstet, 2024, 46: e-rbgo6.

[9]

师琳, 钟李长, 谷丽萍. 瘤内瘤周超声影像组学预测乳腺浸润导管癌Ki-67状态[J]. 中国医学物理学杂志, 2024, 41(9): 1139-1144.

[10]

林雅丹, 阮思敏, 陈立达, 等. 面向LR-M肝癌精准诊断的超声造影时空感知深度学习系统开发[J]. 现代仪器与医疗, 2025, 31(3): 1-9.

[11]

洪森, 漆万银, 周颖俊, 等. 基于TabPFN模型的CT影像组学联合临床影像特征对亚实性肺腺癌浸润性的术前预测研究[J]. 现代仪器与医疗, 2025, 31(6): 26-35, 81.

[12]

余美慧, 袁泉, 曾书娥, 等. 基于超声图像的迁移学习模型在乳腺肿块良恶性鉴别诊断中的价值[J]. 临床超声医学杂志, 2022, 24(9): 652-656.

[13]

Maani N, Westergard S, Yang J, et al. NF1 patients receiving breast cancer screening: insights from the ontario high risk breast screening program[J]. Cancers, 2019, 11(5): 707.

[14]

崔兆国, 吴昊, 汤敏, 等. 人工智能对基层影像医生筛查肺结节的辅助价值[J]. 中国现代医生, 2021, 59(1): 99-101, 105, 封3.

[15]

Shen J, Chen G, Lun H, et al. Dual-modal ultrasound-based deep learning radiomics for differentiation of benign and malignant breast lesions[J]. Gland Surgery, 2025, 14(10): 2035-2050.

[16]

Jin Z Y, Li J K, Niu R L, et al. Application of an artificial intelligence-assisted diagnostic system for breast ultrasound: a prospective study[J]. Gland Surgery, 2024, 13(12): 2221-2231.

[17]

Lu X, Lu Y, Zhao W, et al. Ultrasound-based deep learning radiomics for multi-stage assisted diagnosis in reducing unnecessary biopsies of BI-RADS 4A lesions[J]. Quantitative Imaging in Medicine and Surgery, 2025, 15(3): 2512-2528.

[18]

Liang Y, Wei Z, Dai Y, et al. An interpretable AI system reduces false-positive MRI diagnoses by stratifying high-risk breast lesions[J]. Nature Communications, 2026, 17: 2263.

[19]

Kim J, Han K, Kim K W, et al. Feasibility of using an AI system for breast ultrasonography interpretation according to clinical expertise: results of a pilot study[J]. Journal of the Korean Society of Radiology, 2026, 87(2): 314-327.

[20]

Ma Q, Wang J, Dong B, et al. YOLO AI model based on an automated breast volume scanner for the detection of benign and malignant breast lesions[J]. Quantitative Imaging in Medicine and Surgery, 2025, 15(10): 10156-10167.

[21]

Han C, Chen J, Hong M, et al. MRI radiomics for diagnosing small BI-RADS 4 breast lesions: an interpretable model[J]. Quantitative Imaging in Medicine and Surgery, 2025, 15(6): 5060-5072.

[22]

Lee J, Kim W H, Kim J, et al. Effect of a real-time artificial intelligence-assisted ultrasound system on BI-RADS C4 breast lesions based on breast density[J]. Cancers, 2026, 18(3): 536.

[23]

Krishna NK R, R.S. A, K S. Artificial intelligence in radiology: augmentation, not replacement[J]. Cureus, 2025, 17(6): e86247.

[24]

Fujioka T, Tsukada J, Hayashida T, et al. The future of breast cancer diagnosis in Japan with AI and ultrasonography[J]. JMA J, 2024, 8(1): 91-101.

2026年第32卷第3期

PDF下载

引用本文

BibTeX

文章信息

doi: 10.11876/mimt202603007

接收时间：2025-12-29
首发时间：2026-10-23
出版时间：2026-06-10

补充材料

相关文章

文章信息

作者

出版历史

收稿日期：2025-12-29

基金

2024年度“浦东新区卫生系统优秀社区适宜人才培养计划”(PWRs2024-14)

“市六-临港”紧密型健康联合体临床类科研项目(ynlglht202401)

上海市浦东新区万祥社区卫生服务中心院内科研项目(WX2024-01)

作者信息

^1.上海市浦东新区万祥社区卫生服务中心医技科，上海　201313

^2.上海交通大学医学院附属第六人民医院超声医学科，上海　200233

通讯作者:

谷丽萍，女，硕士研究生，副主任医师，研究方向：乳腺肿瘤，甲状腺肿瘤

参考文献

分享链接

https://castjournals.cast.org.cn/joweb/xdyqyyl/CN/10.11876/mimt202603007

分享至

全文二维码

扫描看全文

引用本文

BibTeX

本文的引用情况

2种不同金属材料的力学参数

科 Family	属数 Number of genus	种数 Number of species	占总种数比例 Percentage of total species (%)	属 Genus	种数 Number of species	占总种数比例 Percentage of total species (%)
鹅膏菌科Amanitaceae	2	11	5.26	鹅膏菌属 Amanita	10	4.78
小菇科 Mycenaceae	2	12	5.74	丝盖伞属 Inocybe	5	2.39
多孔菌科 Polyporaceae	8	14	6.70	蜡蘑属 Laccaria	5	2.39
红菇科 Russulaceae	3	23	11.00	小皮伞属 Marasmius	6	2.87
				小菇属 Mycena	11	5.26
				光柄菇属 Pluteus	5	2.39
				红菇属 Russula	17	8.13
				栓菌属 Trametes	5	2.39

关闭全屏

BibTeX
EndNote
RefWorks
TxT