Article(id=1149776902291943468, tenantId=1146029695717560320, journalId=1146123166801305609, issueId=1149776900194791454, articleNumber=null, orderNo=null, doi=10.12404/j.issn.1671-1815.2406118, pmid=null, cstr=null, oa=null, hot=null, price=null, onlineType=0, articleFormat=0, articleType=null, articleTypeStr=research-article, receivedDate=1723651200000, receivedDateStr=2024-08-15, revisedDate=1732377600000, revisedDateStr=2024-11-24, acceptedDate=null, acceptedDateStr=null, onlineDate=1752057775327, onlineDateStr=2025-07-09, pubDate=1744905600000, pubDateStr=2025-04-18, doiRegisterDate=null, doiRegisterDateStr=null, onlineIssueDate=1752057775327, onlineIssueDateStr=2025-07-09, onlineJustAcceptDate=null, onlineJustAcceptDateStr=null, onlineFirstDate=null, onlineFirstDateStr=null, sourceXml=null, magXml=null, createTime=1752057775327, creator=13701087609, updateTime=1752057775327, updator=13701087609, issue=Issue{id=1149776900194791454, tenantId=1146029695717560320, journalId=1146123166801305609, year='2025', volume='25', issue='11', pageStart='4397', pageEnd='4826', issueExtLink='null', onlineDate='null', pubDate='null', beforeIssueId=null, nextIssueId=null, price=null, status=1, issueComplete=1, articleOrder=1, issueType=-1, specialIssue=0, createTime=1752057774827, creator=13701087609, updateTime=1768456666677, updator=13701087609, preIssue=null, nextIssue=null, ext={EN=IssueExt(id=1218558837930512931, tenantId=1146029695717560320, journalId=1146123166801305609, issueId=1149776900194791454, language=EN, specialIssueTitle=, coverIllustrator=, specialIssueEditor=, specialIssueAbout=), CN=IssueExt(id=1218558837930512932, tenantId=1146029695717560320, journalId=1146123166801305609, issueId=1149776900194791454, language=CN, specialIssueTitle=, coverIllustrator=, specialIssueEditor=, specialIssueAbout=)}, issueFiles=null}, startPage=4476, endPage=4482, ext={EN=ArticleExt(id=1149776902619099181, articleId=1149776902291943468, tenantId=1146029695717560320, journalId=1146123166801305609, language=EN, title=Diagnosis of Interstitial Lung Disease Based on Multi-feature Fusion Contrast Learning Retrieval, columnId=1156262732384031076, journalTitle=Science Technology and Engineering, columnName=Papers·Medicine, runingTitle=null, highlight=null, articleAbstract=

At present, complex interstitial lung diseases have the problems of low classification accuracy and lack of auxiliary diagnostic information. To address these problems, an image retrieval framework based on multi-feature fusion and supervised contrastive learning methods was proposed. Interstitial lung disease features were extracted using Res-Net50 and radiomics feature extraction modules. In order to fuse two features of different modalities and scales, a feature fusion module was designed that can jointly represent the spatial calculation feature correlation of two features. The feature discrimination between interstitial lung disease categories was improved through supervised contrastive learning methods, and a typical interstitial lung disease database was retrieved. The highest precision, recall rate and F1 score were obtained in the retrieval task of interstitial lung disease data, and a silhouette coefficient of 0.482 was obtained in the feature vector discrimination index for image retrieval. The experimental results show that compared with the traditional deep learning single feature modality method, the proposed method can effectively improve the classification retrieval accuracy of interstitial lung disease images and improve the interpretability of interstitial lung disease diagnosis.

, correspAuthors=Hou-rong CAI, authorNote=null, correspAuthorsNote=null, copyrightStatement=null, copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=null, magXml=null, pdfUrl=null, pdf=null, pdfFileSize=null, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=null, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=null, mapNumber=null, authorCompany=null, fund=null, authors=null, authorsList=Ze-xiong CHEN, Ping WANG, Song JIANG, Yan-zhen CHEN, Xiao-feng XIE, Hou-rong CAI), CN=ArticleExt(id=1149776914547699818, articleId=1149776902291943468, tenantId=1146029695717560320, journalId=1146123166801305609, language=CN, title=基于多特征融合对比学习检索诊断间质性肺疾病, columnId=1156262732526637414, journalTitle=科学技术与工程, columnName=论文·医药、卫生, runingTitle=null, highlight=null, articleAbstract=

目前复杂间质性肺疾病存在分类精度不高,且缺少辅助诊断信息的问题,针对这些问题提出了基于多特征融合和有监督对比学习方法的图像检索框架。使用Res-Net50和影像组学特征提取模块提取间质性肺疾病特征。为了使不同模态不同尺度的两个特征进行融合,设计了一个可以通过两个特征联合表征空间计算特征相关性的特征融合模块。通过有监督对比学习方法,提升间质性肺疾病类别之间的特征区分度,并对典型间质性肺疾病数据库进行检索。在本次间质性肺疾病数据的检索任务中获得了最高的精确率、召回率和F1分数,在用于图像检索的特征向量区分度指标中,获得了0.482的轮廓系数。实验结果表明:与传统深度学习单一特征模态方法相比,所提方法能有效提高间质性肺疾病图像分类检索精度,并提高间质性肺疾病诊断的可解释性。

, correspAuthors=蔡后荣, authorNote=null, correspAuthorsNote=
* 蔡后荣(1961—),男,汉族,江西万安人,博士,主任医师。研究方向:间质性肺炎。E-mail:
, copyrightStatement=null, copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=zggkuY+bi4tEW1r9kV4xPA==, magXml=yHr98ASfQOYgkKwvzwy05Q==, pdfUrl=null, pdf=SsGe1qELz1eyCzR7xpSgmg==, pdfFileSize=5898532, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=TgOVMOd7kwoSNzGADiDorg==, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=vIJBl2u3mGpgqCg2S5abHg==, mapNumber=null, authorCompany=null, fund=null, authors=

陈泽雄(1999—),男,汉族,海南海口人,硕士研究生。研究方向:模式识别。E-mail:

, authorsList=陈泽雄, 王平, 江嵩, 陈颜真, 谢小峰, 蔡后荣)}, authors=[Author(id=1218843911389364411, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=q1979761632@126.com, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1218843911536165071, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, authorId=1218843911389364411, language=EN, stringName=Ze-xiong CHEN, firstName=Ze-xiong, middleName=null, lastName=CHEN, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1 Electrical and Mechanical College, Hainan University, Haikou 570228, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1218843911674577115, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, authorId=1218843911389364411, language=CN, stringName=陈泽雄, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1 海南大学机电工程学院,海口570228, bio={"content":"

陈泽雄(1999—),男,汉族,海南海口人,硕士研究生。研究方向:模式识别。E-mail:

"}, bioImg=null, bioContent=

陈泽雄(1999—),男,汉族,海南海口人,硕士研究生。研究方向:模式识别。E-mail:

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1218843910911213698, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, xref=1, ext=[AuthorCompanyExt(id=1218843910923796614, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, companyId=1218843910911213698, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1 Electrical and Mechanical College, Hainan University, Haikou 570228, China), AuthorCompanyExt(id=1218843910932185222, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, companyId=1218843910911213698, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1 海南大学机电工程学院,海口570228)])]), Author(id=1218843911821377771, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, 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=0, authorType=1, ext={EN=AuthorExt(id=1218843911972372731, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, authorId=1218843911821377771, language=EN, stringName=Ping WANG, firstName=Ping, middleName=null, lastName=WANG, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, address=2 Yi Zhi Yuan Health Technology Co., Ltd., Hangzhou 310000, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1218843912110784776, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, authorId=1218843911821377771, language=CN, stringName=王平, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, address=2 医智源健康科技有限公司, 杭州 310000, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1218843911062208661, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, xref=2, ext=[AuthorCompanyExt(id=1218843911087374488, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, companyId=1218843911062208661, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2 Yi Zhi Yuan Health Technology Co., Ltd., Hangzhou 310000, China), AuthorCompanyExt(id=1218843911095763098, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, companyId=1218843911062208661, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2 医智源健康科技有限公司, 杭州 310000)])]), Author(id=1218843912261779738, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, orderNo=2, 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=1218843912416969005, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, authorId=1218843912261779738, language=EN, stringName=Song JIANG, firstName=Song, middleName=null, lastName=JIANG, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1 Electrical and Mechanical College, Hainan University, Haikou 570228, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1218843912509243704, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, authorId=1218843912261779738, language=CN, stringName=江嵩, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1 海南大学机电工程学院,海口570228, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1218843910911213698, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, xref=1, ext=[AuthorCompanyExt(id=1218843910923796614, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, companyId=1218843910911213698, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1 Electrical and Mechanical College, Hainan University, Haikou 570228, China), AuthorCompanyExt(id=1218843910932185222, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, companyId=1218843910911213698, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1 海南大学机电工程学院,海口570228)])]), Author(id=1218843912614101315, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, orderNo=3, 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=1218843912731541842, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, authorId=1218843912614101315, language=EN, stringName=Yan-zhen CHEN, firstName=Yan-zhen, middleName=null, lastName=CHEN, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, address=2 Yi Zhi Yuan Health Technology Co., Ltd., Hangzhou 310000, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1218843912853176669, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, authorId=1218843912614101315, language=CN, stringName=陈颜真, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, address=2 医智源健康科技有限公司, 杭州 310000, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1218843911062208661, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, xref=2, ext=[AuthorCompanyExt(id=1218843911087374488, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, companyId=1218843911062208661, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2 Yi Zhi Yuan Health Technology Co., Ltd., Hangzhou 310000, China), AuthorCompanyExt(id=1218843911095763098, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, companyId=1218843911062208661, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2 医智源健康科技有限公司, 杭州 310000)])]), Author(id=1218843912962228583, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, orderNo=4, 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=1218843913109029231, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, authorId=1218843912962228583, language=EN, stringName=Xiao-feng XIE, firstName=Xiao-feng, middleName=null, lastName=XIE, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1 Electrical and Mechanical College, Hainan University, Haikou 570228, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1218843913222275447, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, authorId=1218843912962228583, language=CN, stringName=谢小峰, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1 海南大学机电工程学院,海口570228, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1218843910911213698, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, xref=1, ext=[AuthorCompanyExt(id=1218843910923796614, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, companyId=1218843910911213698, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1 Electrical and Mechanical College, Hainan University, Haikou 570228, China), AuthorCompanyExt(id=1218843910932185222, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, companyId=1218843910911213698, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1 海南大学机电工程学院,海口570228)])]), Author(id=1218843913490710916, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, orderNo=5, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=caihourong2013@163.com, emailSecond=null, emailThird=null, correspondingAuthor=1, authorType=1, ext={EN=AuthorExt(id=1218843913612345743, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, authorId=1218843913490710916, language=EN, stringName=Hou-rong CAI, firstName=Hou-rong, middleName=null, lastName=CAI, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=3, *, address=3 Department of Respiratory and Critical Care Medicine, Nanjing Drum Tower Hospital, Nanjing 210000, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1218843913750757783, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, authorId=1218843913490710916, language=CN, stringName=蔡后荣, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=3, *, address=3 南京鼓楼医院呼吸与危重症医学科, 南京 210000, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1218843911238369449, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, xref=3, ext=[AuthorCompanyExt(id=1218843911246758058, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, companyId=1218843911238369449, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3 Department of Respiratory and Critical Care Medicine, Nanjing Drum Tower Hospital, Nanjing 210000, China), AuthorCompanyExt(id=1218843911255146668, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, companyId=1218843911238369449, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3 南京鼓楼医院呼吸与危重症医学科, 南京 210000)])])], keywords=[Keyword(id=1218843914052747696, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, language=EN, orderNo=1, keyword=interstitial lung disease), Keyword(id=1218843914199548350, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, language=EN, orderNo=2, keyword=multi-feature fusion), Keyword(id=1218843914300211655, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, language=EN, orderNo=3, keyword=supervised contrastive learning), Keyword(id=1218843914396680656, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, language=EN, orderNo=4, keyword=image retrieval), Keyword(id=1218843914509926873, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, language=EN, orderNo=5, keyword=auxiliary diagnosis), Keyword(id=1218843914656727522, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, language=CN, orderNo=1, keyword=间质性肺疾病), Keyword(id=1218843914778362348, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, language=CN, orderNo=2, keyword=多特征融合), Keyword(id=1218843914891608564, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, language=CN, orderNo=3, keyword=有监督对比学习), Keyword(id=1218843915017437693, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, language=CN, orderNo=4, keyword=图像检索), Keyword(id=1218843915143266823, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, language=CN, orderNo=5, keyword=辅助诊断)], refs=[Reference(id=1218843917689209549, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, doi=null, pmid=null, pmcid=null, year=2022, volume=400, issue=10354, pageStart=769, pageEnd=786, url=null, language=null, rfNumber=[1], rfOrder=0, authorNames=Wijsenbeek M, Suzuki A, Maher T M, journalName=The Lancet, refType=null, unstructuredReference=Wijsenbeek M, Suzuki A, Maher T M. Interstitial lung diseases[J]. The Lancet, 2022, 400(10354): 769-786., articleTitle=Interstitial lung diseases, refAbstract=null), Reference(id=1218843917806650070, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, doi=null, pmid=null, pmcid=null, year=2018, volume=198, issue=5, pageStart=e44, pageEnd=e68, url=null, language=null, rfNumber=[2], rfOrder=1, authorNames=Raghu G, Remy-Jardin M, Myers J L, journalName=American Journal of Respiratory and Critical Care Medicine, refType=null, unstructuredReference=Raghu G, Remy-Jardin M, Myers J L, et al. Diagnosis of idiopathic pulmonary fibrosis. An official ATS/ERS/JRS/ALAT clinical practice guideline[J]. American Journal of Respiratory and Critical Care Medicine, 2018, 198(5): e44-e68., articleTitle=Diagnosis of idiopathic pulmonary fibrosis. An official ATS/ERS/JRS/ALAT clinical practice guideline, refAbstract=null), Reference(id=1218843917999588063, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, doi=null, pmid=null, pmcid=null, year=2005, volume=25, issue=12, pageStart=770, pageEnd=772, url=null, language=null, rfNumber=[3], rfOrder=2, authorNames=魏龙晓, 崔光彬, 魏经国, journalName=科学技术与工程, refType=null, unstructuredReference=魏龙晓, 崔光彬, 魏经国, 等. 实验性肺纤维化HRCT表现及其病理学基础相关性研究[J]. 科学技术与工程, 2005, 25(12): 770-772., articleTitle=实验性肺纤维化HRCT表现及其病理学基础相关性研究, refAbstract=null), Reference(id=1218843918104445669, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, doi=null, pmid=null, pmcid=null, year=2005, volume=25, issue=12, pageStart=770, pageEnd=772, url=null, language=null, rfNumber=[3], rfOrder=3, authorNames=Wei Longxiao, Cui Guangbin, Wei Jingguo, journalName=Science Technology and Engineering, refType=null, unstructuredReference=Wei Longxiao, Cui Guangbin, Wei Jingguo, et al. Experiment study on the characteristics of pulmonary fibrosis by HRCT and its relationship with pathological changes[J]. Science Technology and Engineering, 2005, 25(12): 770-772., articleTitle=Experiment study on the characteristics of pulmonary fibrosis by HRCT and its relationship with pathological changes, refAbstract=null), Reference(id=1218843918238663407, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, doi=null, pmid=null, pmcid=null, year=2013, volume=266, issue=3, pageStart=936, pageEnd=944, url=null, language=null, rfNumber=[4], rfOrder=4, authorNames=Watadani T, Sakai F, Johkoh T, journalName=Radiology, refType=null, unstructuredReference=Watadani T, Sakai F, Johkoh T, et al. Interobserver variability in the CT as-sessment of honeycombing in the lungs[J]. Radiology 2013, 266(3): 936-944., articleTitle=Interobserver variability in the CT as-sessment of honeycombing in the lungs, refAbstract=null), Reference(id=1218843918372881147, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, doi=null, pmid=null, pmcid=null, year=2013, volume=28, issue=5, pageStart=284, pageEnd=290, url=null, language=null, rfNumber=[5], rfOrder=5, authorNames=Lynch D A, Al-Qaisi M A, journalName=Journal of Thoracic Imaging, refType=null, unstructuredReference=Lynch D A, Al-Qaisi M A. Quantitative computed tomography in chronic obstructive pulmonary disease[J]. Journal of Thoracic Imaging, 2013, 28(5): 284-290., articleTitle=Quantitative computed tomography in chronic obstructive pulmonary disease, refAbstract=null), Reference(id=1218843918628733705, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, doi=null, pmid=null, pmcid=null, year=2024, volume=46, issue=8, pageStart=878, pageEnd=885, url=null, language=null, rfNumber=[6], rfOrder=6, authorNames=柳洪亚, 祝洁, 刘晨, journalName=陆军军医大学学报, refType=null, unstructuredReference=柳洪亚, 祝洁, 刘晨, 等. 高分辨率CT影像组学分类诊断类风湿关节炎相关间质性肺疾病的效能[J]. 陆军军医大学学报, 2024, 46(8): 878-885., articleTitle=高分辨率CT影像组学分类诊断类风湿关节炎相关间质性肺疾病的效能, refAbstract=null), Reference(id=1218843918737785610, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, doi=null, pmid=null, pmcid=null, year=2024, volume=46, issue=8, pageStart=878, pageEnd=885, url=null, language=null, rfNumber=[6], rfOrder=7, authorNames=Liu Hongya, Zhu Jie, Liu Chen, journalName=Journal of Army Medical University, refType=null, unstructuredReference=Liu Hongya, Zhu Jie, Liu Chen, et al. Efficacy of high-resolution CT imaging radiomics classification for diagnosis of rheumatoid arthritis-associated interstitial lung disease[J]. Journal of Army Medical University, 2024, 46(8): 878-885., articleTitle=Efficacy of high-resolution CT imaging radiomics classification for diagnosis of rheumatoid arthritis-associated interstitial lung disease, refAbstract=null), Reference(id=1218843918892974871, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, doi=null, pmid=null, pmcid=null, year=2021, volume=null, issue=null, pageStart=A1848, pageEnd=null, url=null, language=null, rfNumber=[7], rfOrder=8, authorNames=Poynton C, Ash S, Putman R K, journalName=Radiomics detects intersti-tial lung abnormalities (ILA) progression in high-risk patients, refType=null, unstructuredReference=Poynton C, Ash S, Putman R K, et al. Radiomics detects intersti-tial lung abnormalities (ILA) progression in high-risk patients[M]//TP026 Diagnosis, Assessment, and Prognosis of Fibrotic ILD. New York: American Thoracic Society, 2021: A1848., articleTitle=null, refAbstract=null), Reference(id=1218843918997832472, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, doi=null, pmid=null, pmcid=null, year=2020, volume=58, issue=null, pageStart=725, pageEnd=737, url=null, language=null, rfNumber=[8], rfOrder=9, authorNames=Huang S, Lee F, Miao R, journalName=Medical & Biological Engineering & Computing, refType=null, unstructuredReference=Huang S, Lee F, Miao R, et al. A deep convolutional neural network architecture for interstitial lung disease pattern classification[J]. Medical & Biological Engineering & Computing, 2020, 58: 725-737., articleTitle=A deep convolutional neural network architecture for interstitial lung disease pattern classification, refAbstract=null), Reference(id=1218843919153021730, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, doi=null, pmid=null, pmcid=null, year=2023, volume=14, issue=1, pageStart=2272, pageEnd=null, url=null, language=null, rfNumber=[9], rfOrder=10, authorNames=Mei X, Liu Z, Singh A, journalName=Nature Communications, refType=null, unstructuredReference=Mei X, Liu Z, Singh A, et al. Interstitial lung disease diagnosis and prognosis using an AI system integrating longitudinal data[J]. Nature Communications, 2023, 14(1): 2272., articleTitle=Interstitial lung disease diagnosis and prognosis using an AI system integrating longitudinal data, refAbstract=null), Reference(id=1218843919232713507, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, doi=null, pmid=null, pmcid=null, year=2023, volume=23, issue=16, pageStart=6834, pageEnd=6844, url=null, language=null, rfNumber=[10], rfOrder=11, authorNames=谷宇, 刘佳琪, 杨立东, journalName=科学技术与工程, refType=null, unstructuredReference=谷宇, 刘佳琪, 杨立东, 等. 融合注意力特征及动态卷积的肺结节辅助诊断[J]. 科学技术与工程, 2023, 23(16): 6834-6844., articleTitle=融合注意力特征及动态卷积的肺结节辅助诊断, refAbstract=null), Reference(id=1218843919379514156, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, doi=null, pmid=null, pmcid=null, year=2023, volume=23, issue=16, pageStart=6834, pageEnd=6844, url=null, language=null, rfNumber=[10], rfOrder=12, authorNames=Gu Yu, Liu Jiaqi, Yang Lidong, journalName=Science Technology and Engineering, refType=null, unstructuredReference=Gu Yu, Liu Jiaqi, Yang Lidong, et al. Fusion of attention and dynamic convolution for computer aided diagnosis of pulmonary nodules[J]. Science Technology and Engineering, 2023, 23(16): 6834-6844., articleTitle=Fusion of attention and dynamic convolution for computer aided diagnosis of pulmonary nodules, refAbstract=null), Reference(id=1218843919484371761, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, doi=null, pmid=null, pmcid=null, year=2010, volume=14, issue=12, pageStart=828, pageEnd=831, url=null, language=null, rfNumber=[11], rfOrder=13, authorNames=中华医学会风湿病学会, journalName=中华风湿病学杂志, refType=null, unstructuredReference=中华医学会风湿病学会. 多发性肌炎和皮肌炎诊断及治疗指南[J]. 中华风湿病学杂志, 2010, 14(12): 828-831., articleTitle=多发性肌炎和皮肌炎诊断及治疗指南, refAbstract=null), Reference(id=1218843919618589494, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, doi=null, pmid=null, pmcid=null, year=2010, volume=14, issue=12, pageStart=828, pageEnd=831, url=null, language=null, rfNumber=[11], rfOrder=14, authorNames=Chinese Medical Association Society of Rheumatology, journalName=Chinese Journal of Rheumatology, refType=null, unstructuredReference=Chinese Medical Association Society of Rheumatology. Guidelines for the diagnosis and treatment of polymyositis and dermatomyositis[J]. Chinese Journal of Rheumatology, 2010, 14(12): 828-831., articleTitle=Guidelines for the diagnosis and treatment of polymyositis and dermatomyositis, refAbstract=null), Reference(id=1218843919727641405, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, doi=null, pmid=null, pmcid=null, year=2018, volume=6, issue=2, pageStart=138, pageEnd=153, url=null, language=null, rfNumber=[12], rfOrder=15, authorNames=Lynch D A, Sverzellati N, Travis W D, journalName=The Lancet Respiratory Medicine, refType=null, unstructuredReference=Lynch D A, Sverzellati N, Travis W D, et al. Diagnostic criteria for idiopathic pulmonary fibrosis: a fleischner society white paper[J]. The Lancet Respiratory Medicine, 2018, 6(2): 138-153., articleTitle=Diagnostic criteria for idiopathic pulmonary fibrosis: a fleischner society white paper, refAbstract=null), Reference(id=1218843919857664836, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, doi=null, pmid=null, pmcid=null, year=1997, volume=9, issue=6, pageStart=527, pageEnd=535, url=null, language=null, rfNumber=[13], rfOrder=16, authorNames=Targoff I N, Miller F W, Medsger Jr T A, journalName=Current Opinion in Rheumatology, refType=null, unstructuredReference=Targoff I N, Miller F W, Medsger Jr T A, et al. Classification criteria for the idiopathic inflammatory myopathies[J]. Current Opinion in Rheumatology, 1997, 9(6): 527-535., articleTitle=Classification criteria for the idiopathic inflammatory myopathies, refAbstract=null), Reference(id=1218843919970911049, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, doi=null, pmid=null, pmcid=null, year=1993, volume=null, issue=null, pageStart=87, pageEnd=102, url=null, language=null, rfNumber=[14], rfOrder=17, authorNames=Horii S C, Bidgood W D, journalName=Enabling Technologies for High-Bandwidth Applications, refType=null, unstructuredReference=Horii S C, Bidgood W D. DICOM: A standard for medical imaging[C]// Enabling Technologies for High-Bandwidth Applications. Richardson: SPIE, 1993: 87-102., articleTitle=DICOM: A standard for medical imaging, refAbstract=null), Reference(id=1218843920084157264, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, doi=null, pmid=null, pmcid=null, year=2016, volume=null, issue=null, pageStart=770, pageEnd=778, url=null, language=null, rfNumber=[15], rfOrder=18, authorNames=He K, Zhang X, Ren S, journalName=Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, refType=null, unstructuredReference=He K, Zhang X, Ren S, et al. Deep residual learning for image recognition[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2016: 770-778., articleTitle=Deep residual learning for image recognition, refAbstract=null), Reference(id=1218843920201597780, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, doi=null, pmid=null, pmcid=null, year=2017, volume=77, issue=21, pageStart=e104, pageEnd=e107, url=null, language=null, rfNumber=[16], rfOrder=19, authorNames=Van Griethuysen J J M, Fedorov A, Parmar C, journalName=Cancer Research, refType=null, unstructuredReference=Van Griethuysen J J M, Fedorov A, Parmar C, et al. Computational radiomics system to decode the radiographic phenotype[J]. Cancer Research, 2017, 77(21): e104-e107., articleTitle=Computational radiomics system to decode the radiographic phenotype, refAbstract=null), Reference(id=1218843920302261079, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, doi=null, pmid=null, pmcid=null, year=2018, volume=41, issue=2, pageStart=423, pageEnd=443, url=null, language=null, rfNumber=[17], rfOrder=20, authorNames=Baltrušaitis T, Ahuja C, Morency L P, journalName=IEEE Transactions on Pattern Analysis and Machine Intelligence, refType=null, unstructuredReference=Baltrušaitis T, Ahuja C, Morency L P. Multimodal machine learning: a survey and taxonomy[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2018, 41(2): 423-443., articleTitle=Multimodal machine learning: a survey and taxonomy, refAbstract=null), Reference(id=1218843920423895902, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, doi=null, pmid=null, pmcid=null, year=2020, volume=33, issue=null, pageStart=18661, pageEnd=18673, url=null, language=null, rfNumber=[18], rfOrder=21, authorNames=Khosla P, Teterwak P, Wang C, journalName=Advances in Neural Information Processing Systems, refType=null, unstructuredReference=Khosla P, Teterwak P, Wang C, et al. Supervised contrastive learning[J]. Advances in Neural Information Processing Systems, 2020, 33: 18661-18673., articleTitle=Supervised contrastive learning, refAbstract=null), Reference(id=1218843920570696552, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, doi=null, pmid=null, pmcid=null, year=2008, volume=51, issue=1, pageStart=117, pageEnd=122, url=null, language=null, rfNumber=[19], rfOrder=22, authorNames=Andoni A, Indyk P, journalName=Communications of the ACM, refType=null, unstructuredReference=Andoni A, Indyk P. Near-optimal hashing algorithms for approximate nearest neighbor in high dimensions[J]. Communications of the ACM, 2008, 51(1): 117-122., articleTitle=Near-optimal hashing algorithms for approximate nearest neighbor in high dimensions, refAbstract=null), Reference(id=1218843920700719983, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, doi=null, pmid=null, pmcid=null, year=2022, volume=null, issue=null, pageStart=457, pageEnd=null, url=null, language=null, rfNumber=[20], rfOrder=23, authorNames=Chhabra M, Kumar R, journalName=Emergent Converging Technologies and Biomedical Systems: Select Proceedings of ETBS 2021, refType=null, unstructuredReference=Chhabra M, Kumar R. An advanced VGG16 architecture-based deep learning model to detect pneumonia from medical images[M]// Emergent Converging Technologies and Biomedical Systems: Select Proceedings of ETBS 2021. Singapore: Springer Singapore, 2022: 457-471., articleTitle=An advanced VGG16 architecture-based deep learning model to detect pneumonia from medical images, refAbstract=null), Reference(id=1218843920839132023, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, doi=null, pmid=null, pmcid=null, year=2012, volume=25, issue=null, pageStart=1106, pageEnd=1114, url=null, language=null, rfNumber=[21], rfOrder=24, authorNames=Krizhevsky A, Sutskever I, Hinton G E, journalName=Advances in Neural Information Processing Systems, refType=null, unstructuredReference=Krizhevsky A, Sutskever I, Hinton G E. Imagenet classification with deep convolutional neural networks[J]. Advances in Neural Information Processing Systems, 2012, 25: 1106-1114., articleTitle=Imagenet classification with deep convolutional neural networks, refAbstract=null)], funds=[Fund(id=1218843917315916471, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, awardId=2022YFC2010000, language=CN, fundingSource=国家重点研发计划(2022YFC2010000), fundOrder=null, country=null), Fund(id=1218843917429162685, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, awardId=2022YFC2010006, language=CN, fundingSource=国家重点研发计划(2022YFC2010006), fundOrder=null, country=null)], companyList=[AuthorCompany(id=1218843910911213698, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, xref=1, ext=[AuthorCompanyExt(id=1218843910923796614, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, companyId=1218843910911213698, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1 Electrical and Mechanical College, Hainan University, Haikou 570228, China), AuthorCompanyExt(id=1218843910932185222, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, companyId=1218843910911213698, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1 海南大学机电工程学院,海口570228)]), AuthorCompany(id=1218843911062208661, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, xref=2, ext=[AuthorCompanyExt(id=1218843911087374488, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, companyId=1218843911062208661, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2 Yi Zhi Yuan Health Technology Co., Ltd., Hangzhou 310000, China), AuthorCompanyExt(id=1218843911095763098, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, companyId=1218843911062208661, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2 医智源健康科技有限公司, 杭州 310000)]), AuthorCompany(id=1218843911238369449, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, xref=3, ext=[AuthorCompanyExt(id=1218843911246758058, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, companyId=1218843911238369449, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3 Department of Respiratory and Critical Care Medicine, Nanjing Drum Tower Hospital, Nanjing 210000, China), AuthorCompanyExt(id=1218843911255146668, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, companyId=1218843911238369449, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3 南京鼓楼医院呼吸与危重症医学科, 南京 210000)])], figs=[ArticleFig(id=1218843915365564952, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, language=EN, label=Fig.1, caption=Interstitial lung disease retrieval system block diagram, figureFileSmall=fA47AOOesrePc8OS3t2mPg==, figureFileBig=b6JHiWULk3rVk3FYeZSU2w==, tableContent=null), ArticleFig(id=1218843915474616864, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, language=CN, label=图1, caption=间质性肺疾病检索系统框图, figureFileSmall=fA47AOOesrePc8OS3t2mPg==, figureFileBig=b6JHiWULk3rVk3FYeZSU2w==, tableContent=null), ArticleFig(id=1218843915625611821, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, language=EN, label=Fig.2, caption=Interstitial lung disease feature fusion network, figureFileSmall=IllKxHqgdOkbtfuuwaynmA==, figureFileBig=gAsUOw2eaqhoIDENp/yaGQ==, tableContent=null), ArticleFig(id=1218843915759829561, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, language=CN, label=图2, caption=间质肺病特征融合网络, figureFileSmall=IllKxHqgdOkbtfuuwaynmA==, figureFileBig=gAsUOw2eaqhoIDENp/yaGQ==, tableContent=null), ArticleFig(id=1218843915889852995, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, language=EN, label=Fig.3, caption=Model confusion matrix comparison, figureFileSmall=+wJBZI0wNHs1HTsE+PUpdg==, figureFileBig=ht1J1PB7RRN7JDOYjs/l9A==, tableContent=null), ArticleFig(id=1218843916011487819, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, language=CN, label=图3, caption=模型混淆矩阵对比, figureFileSmall=+wJBZI0wNHs1HTsE+PUpdg==, figureFileBig=ht1J1PB7RRN7JDOYjs/l9A==, tableContent=null), ArticleFig(id=1218843916170871382, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, language=EN, label=Fig.4, caption=Activation map of retrieval features for each category, figureFileSmall=rN2/Tiin5l1COC0f2nfblQ==, figureFileBig=++log+VYxNcISVsf4/0sZA==, tableContent=null), ArticleFig(id=1218843916279923297, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, language=CN, label=图4, caption=各类别检索特征激活图, figureFileSmall=rN2/Tiin5l1COC0f2nfblQ==, figureFileBig=++log+VYxNcISVsf4/0sZA==, tableContent=null), ArticleFig(id=1218843916418335340, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, language=EN, label=Table 1, caption=

The interstitial lung disease database

, figureFileSmall=null, figureFileBig=null, tableContent=
间质肺病种类 数据集数量
ASS-ILD 2 594
IPF-ILD 2 779
DM/RP-ILD 2 696
Normal 4 411
), ArticleFig(id=1218843916531581560, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, language=CN, label=表1, caption=

间质性肺疾病数据库

, figureFileSmall=null, figureFileBig=null, tableContent=
间质肺病种类 数据集数量
ASS-ILD 2 594
IPF-ILD 2 779
DM/RP-ILD 2 696
Normal 4 411
), ArticleFig(id=1218843916657410690, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, language=EN, label=Table 2, caption=

Comparison with other unimodal methods

, figureFileSmall=null, figureFileBig=null, tableContent=
模型 精确率/% 召回率/% F1分数/%
VGG 57.7 58.9 57.8
AlexNet 61.9 60.7 60.1
ResNet50 69.7 69.6 69.1
Omic 72.1 70.3 69.9
ResNet50+ SCL 78.2 77.6 77.3
ILD-FFN + SCL 82.1 81.1 81.3
), ArticleFig(id=1218843916858737298, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, language=CN, label=表2, caption=

与其他单模态方法的对比

, figureFileSmall=null, figureFileBig=null, tableContent=
模型 精确率/% 召回率/% F1分数/%
VGG 57.7 58.9 57.8
AlexNet 61.9 60.7 60.1
ResNet50 69.7 69.6 69.1
Omic 72.1 70.3 69.9
ResNet50+ SCL 78.2 77.6 77.3
ILD-FFN + SCL 82.1 81.1 81.3
), ArticleFig(id=1218843916951011995, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, language=EN, label=Table 3, caption=

Differentiation and quantification of characteristics of different algorithms

, figureFileSmall=null, figureFileBig=null, tableContent=
模型 类内距离 类间距离 轮廓系数
VGG 43.209 69.673 0.379
AlexNet 29.847 50.678 0.411
ResNet50 22.201 37.409 0.406
Omic 10.565 14.333 0.262
ResNet50+ SCL 0.569 0.983 0.420
ILD-FFN + SCL 0.511 0.988 0.482
), ArticleFig(id=1218843917081035427, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149776902291943468, language=CN, label=表3, caption=

不同算法特征区分度量化

, figureFileSmall=null, figureFileBig=null, tableContent=
模型 类内距离 类间距离 轮廓系数
VGG 43.209 69.673 0.379
AlexNet 29.847 50.678 0.411
ResNet50 22.201 37.409 0.406
Omic 10.565 14.333 0.262
ResNet50+ SCL 0.569 0.983 0.420
ILD-FFN + SCL 0.511 0.988 0.482
)], attaches=null, journal=Journal(id=1146119176004939786, delFlag=0, nameCn=科学技术与工程, nameEn=Science Technology and Engineering, nameHistory1=null, nameHistory2=null, issn=1671-1815, eissn=, cn=11-4688/T, coden=null, periodic=4, language=CN, oaType=是, ccby=null, superviseOffice=null, ownerOffice=null, pubOffice=null, editorOffice=null, officeType=null, aims=null, clcCode=null, officeProv=null, officeCity=null, officeAddr=null, officeZip=null, officeEmail=null, officePhone=null, editDirector=null, officeDirector=null, officeDirectorPhone=null, officeStaffNum=null, officeEmpNum=null, coverPicUrl=UKU/O7GSka5polgCTkbIIw==, journalPrice=null, startedYear=null, abbrevIsoEn=Sci Technol Eng, journalRemark=null, publicationField=null, createdTime=null, updatedTime=1754445529766, createdBy=null, updatedBy=13701087609, firstLetterCn=S, firstLetterEn=S, subjectCode=Natural Sciences, subjectName=自然科学, subjectCodeEn=Natural Sciences, subjectNameEn=null, picCn=UKU/O7GSka5polgCTkbIIw==, picEn=5hwlULoNwcbj3xUmVi9MAQ==, jcr=null, cjcr=null, exts=[JournalExt(id=1159791870395564357, language=CN, name=科学技术与工程, nameHistory1=null, nameHistory2=null, managedBy=, sponsoredBy=, publishedBy=, editorOffice=, officeProv=null, officeCity=null, officeAddr=, officeZip=, editDirector=null, officeDirector=null, officePhone=null, coverPicUrl=null, journalRemark=, submitArticleUrl=null, websiteUrl=http://www.stae.com.cn/jsygc/home, createdTime=1754445529793, updatedTime=1754445529793, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=http://www.stae.com.cn/jsygc/site/menus/20090429150146001, submissionAuthorUrl=http://www.stae.com.cn/jsygc/author/login, submissionEditorUrl=http://www.stae.com.cn/jsygc/editor/login, submissionReviewUrl=http://www.stae.com.cn/jsygc/reviewer/login, submissionCeEditorUrl=, submissionAeEditorUrl=, option={"copyright":""}), JournalExt(id=1159791870441701702, language=EN, name=Science Technology and Engineering, nameHistory1=null, nameHistory2=null, managedBy=, sponsoredBy=, publishedBy=, editorOffice=, officeProv=null, officeCity=null, officeAddr=, officeZip=, editDirector=null, officeDirector=null, officePhone=null, coverPicUrl=null, journalRemark=, submitArticleUrl=null, websiteUrl=http://www.stae.com.cn/jsygc/home, createdTime=1754445529804, updatedTime=1754445529804, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=, submissionAuthorUrl=http://www.stae.com.cn/jsygc/author/login, submissionEditorUrl=http://www.stae.com.cn/jsygc/editor/login, submissionReviewUrl=http://www.stae.com.cn/jsygc/reviewer/login, submissionCeEditorUrl=, submissionAeEditorUrl=, option={"copyright":""})], databaseList=null, tenantJournalId=1146123166801305609, websiteList=[Website(id=1148243202391400884, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1146123166801305609, journalNameCn=null, journalNameEn=null, grayFlag=null, tenantId=1146029695717560320, platformId=null, journalGroupId=null, journalGroupNameCn=null, journalGroupNameEn=null, type=1, domain=https://castjournals.cast.org.cn/joweb/kxjsygc/CN, language=CN, createTime=1751692112777, createBy=18614031015, updateTime=1753520965431, updateBy=18614031015, name=科学技术与工程-中文站点, tplId=1146099689490845704, title=科学技术与工程, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1148622798802673703, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1148243202391400884, code=articleTextType, value=kx, createTime=1751782615614, updateTime=1751782615614, creator=18614031015, updator=18614031015), WebsiteProps(id=1148622798781702180, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1148243202391400884, code=banner, value=null, createTime=1751782615609, updateTime=1751782615609, creator=18614031015, updator=18614031015), WebsiteProps(id=1148622798769119267, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1148243202391400884, code=logo, value=https://castjournals.cast.org.cn/joweb/kjdb/CN/file/pic?fileId=j86gbwi+p0Idkyl5SzIlmQ==, createTime=1751782615606, updateTime=1751782615606, creator=18614031015, updator=18614031015), WebsiteProps(id=1148622798794285094, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1148243202391400884, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/kjdb/CN/file/pic, createTime=1751782615612, updateTime=1751782615612, creator=18614031015, updator=18614031015), WebsiteProps(id=1148622798790090789, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1148243202391400884, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_cn_619/, createTime=1751782615611, updateTime=1751782615611, creator=18614031015, updator=18614031015)]), Website(id=1155914124811976731, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1146123166801305609, journalNameCn=null, journalNameEn=null, grayFlag=null, tenantId=1146029695717560320, platformId=null, journalGroupId=null, journalGroupNameCn=null, journalGroupNameEn=null, type=1, domain=https://castjournals.cast.org.cn/joweb/kxjsygc/EN, language=EN, createTime=1753521003206, createBy=18614031015, updateTime=1753521003206, updateBy=18614031015, name=科学技术与工程-英文站点, tplId=1146101810881728533, title=Science Technology and Engineering, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1155914371227308235, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1155914124811976731, code=articleTextType, value=kx, createTime=1753521061952, updateTime=1753521061952, creator=18614031015, updator=18614031015), WebsiteProps(id=1155914371210531016, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1155914124811976731, code=banner, value=null, createTime=1753521061947, updateTime=1753521061947, creator=18614031015, updator=18614031015), WebsiteProps(id=1155914371202142407, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1155914124811976731, code=logo, value=https://castjournals.cast.org.cn/joweb/kjdb/CN/file/pic?fileId=j86gbwi+p0Idkyl5SzIlmQ==, createTime=1753521061945, updateTime=1753521061945, creator=18614031015, updator=18614031015), WebsiteProps(id=1155914371223113930, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1155914124811976731, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/kjdb/CN/file/pic, createTime=1753521061950, updateTime=1753521061950, creator=18614031015, updator=18614031015), WebsiteProps(id=1155914371218919625, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1155914124811976731, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_cn_619/, createTime=1753521061949, updateTime=1753521061949, creator=18614031015, updator=18614031015)])], journalTitle=科学技术与工程, weixinUrl=null, journalUrl=null, iacademicId=null, status=0, seqNo=null, journalTitleEn=Science Technology and Engineering, journalPhotoCn=UKU/O7GSka5polgCTkbIIw==, journalPhotoEn=5hwlULoNwcbj3xUmVi9MAQ==, journalFirstLetter=S, journalRecommend=null, journalNew=null, journalCollection=null, jcrJf=null, cjcrJf=null, jcrJfStr=null, cjcrJfStr=null, submissionFirstDecision=null, sciSubjectClassification=null, casSubjectClassification=null, citeScore=null, totalCitationFrequency=null, icpCode=null, psCode=null, advertisingLicenseCode=null, copyrightInformation=null, country=null, option=null, provinceCode=null, provinceName=null, collectFlag=false), detailUrlCn=https://castjournals.cast.org.cn/joweb/kxjsygc/CN/10.12404/j.issn.1671-1815.2406118, detailUrlEn=https://castjournals.cast.org.cn/joweb/kxjsygc/EN/10.12404/j.issn.1671-1815.2406118, pdfUrlCn=https://castjournals.cast.org.cn/joweb/kxjsygc/CN/PDF/10.12404/j.issn.1671-1815.2406118, pdfUrlEn=https://castjournals.cast.org.cn/joweb/kxjsygc/EN/PDF/10.12404/j.issn.1671-1815.2406118, aliStartDate=null, aliEndDate=null, collectionFlag=false, citedCount=null, citedUrl=null, reference=null)
收藏切换
基于多特征融合对比学习检索诊断间质性肺疾病
收藏切换
PDF下载
陈泽雄 1 , 王平 2 , 江嵩 1 , 陈颜真 2 , 谢小峰 1 , 蔡后荣 3, *
科学技术与工程 | 论文·医药、卫生 2025,25(11): 4476-4482
收起
收藏切换
科学技术与工程 | 论文·医药、卫生 2025, 25(11): 4476-4482
基于多特征融合对比学习检索诊断间质性肺疾病
全屏
陈泽雄1 , 王平2, 江嵩1, 陈颜真2, 谢小峰1, 蔡后荣3, *
作者信息
  • 1 海南大学机电工程学院,海口570228
  • 2 医智源健康科技有限公司, 杭州 310000
  • 3 南京鼓楼医院呼吸与危重症医学科, 南京 210000
  • 陈泽雄(1999—),男,汉族,海南海口人,硕士研究生。研究方向:模式识别。E-mail:

通讯作者:

* 蔡后荣(1961—),男,汉族,江西万安人,博士,主任医师。研究方向:间质性肺炎。E-mail:
Diagnosis of Interstitial Lung Disease Based on Multi-feature Fusion Contrast Learning Retrieval
Ze-xiong CHEN1 , Ping WANG2, Song JIANG1, Yan-zhen CHEN2, Xiao-feng XIE1, Hou-rong CAI3, *
Affiliations
  • 1 Electrical and Mechanical College, Hainan University, Haikou 570228, China
  • 2 Yi Zhi Yuan Health Technology Co., Ltd., Hangzhou 310000, China
  • 3 Department of Respiratory and Critical Care Medicine, Nanjing Drum Tower Hospital, Nanjing 210000, China
出版时间: 2025-04-18 doi: 10.12404/j.issn.1671-1815.2406118
文章导航
收藏切换

目前复杂间质性肺疾病存在分类精度不高,且缺少辅助诊断信息的问题,针对这些问题提出了基于多特征融合和有监督对比学习方法的图像检索框架。使用Res-Net50和影像组学特征提取模块提取间质性肺疾病特征。为了使不同模态不同尺度的两个特征进行融合,设计了一个可以通过两个特征联合表征空间计算特征相关性的特征融合模块。通过有监督对比学习方法,提升间质性肺疾病类别之间的特征区分度,并对典型间质性肺疾病数据库进行检索。在本次间质性肺疾病数据的检索任务中获得了最高的精确率、召回率和F1分数,在用于图像检索的特征向量区分度指标中,获得了0.482的轮廓系数。实验结果表明:与传统深度学习单一特征模态方法相比,所提方法能有效提高间质性肺疾病图像分类检索精度,并提高间质性肺疾病诊断的可解释性。

间质性肺疾病  /  多特征融合  /  有监督对比学习  /  图像检索  /  辅助诊断

At present, complex interstitial lung diseases have the problems of low classification accuracy and lack of auxiliary diagnostic information. To address these problems, an image retrieval framework based on multi-feature fusion and supervised contrastive learning methods was proposed. Interstitial lung disease features were extracted using Res-Net50 and radiomics feature extraction modules. In order to fuse two features of different modalities and scales, a feature fusion module was designed that can jointly represent the spatial calculation feature correlation of two features. The feature discrimination between interstitial lung disease categories was improved through supervised contrastive learning methods, and a typical interstitial lung disease database was retrieved. The highest precision, recall rate and F1 score were obtained in the retrieval task of interstitial lung disease data, and a silhouette coefficient of 0.482 was obtained in the feature vector discrimination index for image retrieval. The experimental results show that compared with the traditional deep learning single feature modality method, the proposed method can effectively improve the classification retrieval accuracy of interstitial lung disease images and improve the interpretability of interstitial lung disease diagnosis.

interstitial lung disease  /  multi-feature fusion  /  supervised contrastive learning  /  image retrieval  /  auxiliary diagnosis
陈泽雄, 王平, 江嵩, 陈颜真, 谢小峰, 蔡后荣. 基于多特征融合对比学习检索诊断间质性肺疾病. 科学技术与工程, 2025 , 25 (11) : 4476 -4482 . DOI: 10.12404/j.issn.1671-1815.2406118
Ze-xiong CHEN, Ping WANG, Song JIANG, Yan-zhen CHEN, Xiao-feng XIE, Hou-rong CAI. Diagnosis of Interstitial Lung Disease Based on Multi-feature Fusion Contrast Learning Retrieval[J]. Science Technology and Engineering, 2025 , 25 (11) : 4476 -4482 . DOI: 10.12404/j.issn.1671-1815.2406118
间质性肺疾病(interstitial lung disease,ILD)是以肺泡壁为主并包括肺泡周围组织及其相邻支撑结构病变的一组异质性肺疾病。在实际临床实践中,将 ILD 分为特发性ILD 及继发性ILD;引起继发性ILD主要包括结缔组织病,职业环境暴露,药物肺损伤的ILD 等[1]
高分辨率计算机断层扫描(computed tomography,CT)是间质性肺疾病诊断的核心[2],肺部CT是胸部各个水平位置采集纵隔窗的影像。患者的肺部CT完整的显示了疾病肺的各种形态以及病理特征。然而在CT图像上分辨间质性肺疾病并不容易。间质性肺疾病在影像学检查中的表现多种多样,包括斑点状、网状、磨玻璃样改变等[3],这些表现缺乏特异性,且相关合并症的研究较少。评估这些特征需要经验和专业知识[4],即使是经验丰富的放射科医生也很难快速的分辨出这些疾病模式的表征。因此,迫切需要一种方法,能将同样拥有ILD表征的几种相关疾病在CT影像上分辨出来,辅助医生诊断。
近年来,影像组学和深度学习技术在间质性肺疾病的辅助分类和分析中表现出了显著的优势。影像组学通过从影像数据中提取大量定量特征,为间质性肺疾病分析提供了组织分析的精准数值信息。例如,Lynch等[5]通过定量CT分析实现了对慢阻肺的精确评估和分类。柳洪亚等[6]使用胸部CT影像组学特征和机器学习算法构建的模型在类风湿关节炎相关间质性肺疾病的影像分型中具有良好的分类效能。Poynton等[7]对间质性肺异常(interstitial lung abnormalities,ILA)区域进行分割提取放射组学特征,对ILA进展类型进行分类,获得了0.6的一致性系数和0.94的曲线下面积系数。
与此同时,深度学习特征提取技术通过自动学习影像中的复杂模式,为间质性肺疾病分类提供了更为强大的工具。深度学习模型能够处理大量的高维数据,通过逐层提取特征进行分类。Huang等[8]为自动分类间质性肺疾病模式,设计了一种深度卷积神经网络(convolutional neural network,CNN)架构,并使用无监督两阶段迁移学习的方式对数据进行优化,改善了训练数据缺乏的问题。Mei等[9]在胸部高分辨率计算机断层扫描上使用P-Net(proposal network)网络多尺度特征提取的方法,评估像素级别上肺气肿、实变和纤维化病变的分数。谷宇等[10]为适应肺结节分类任务使用自适应调整卷积和迭代注意力融合改进了传统残差块,提高了肺结节检测的准确率。
目前的间质性肺疾病辅助系统往往采用单一特征进行分类,但由于不同疾病引起的ILD病变模式的难以描述和量化,基于单一特征的模型很难面对出现的形态差异很大的新间质性肺疾病影像,并且缺乏辅助诊断的信息。
基于上述问题,提出一种基于影像组学和深度学习多特征融合和有监督对比学习方法的图像检索框架。使用来自南京鼓楼医院权威专家提供的不同种类间质性肺疾病数据集作为典型基础数据库,以检索到的病例信息作为辅助诊断依据。提取间质性肺疾病多种特征丰富间质性肺疾病模式的表达,同时利用有监督对比学习优化融合特征中有效特征表达。
本节简要介绍多特征医学图像检索系统(multi-feature medical image retrieval system,MFMIRS),如图1所示。系统框架由两个阶段组成:①数据生成和训练阶段;②图像检索阶段。在数据生成和训练阶段,将间质性肺疾病影像进行预处理获取检索意义更大的肺实质部分。使用有监督对比学习的方法,将数据库用于训练特征提取和特征融合模块。在图像检索阶段,将预处理后的图片输入训练好的框架获取待测图片向量集{a1,a2,a3,…}与数据库图片向量集{y1,y2,y3,…},使用余弦距离来衡量检索内容之间的相似性,并通过局部敏感哈希加快检索的速度,最终从数据库中检索到与待测病例相似的间质肺病医学影像。
数据集来自南京鼓楼医院,入选标准为根据诊断指南[11-13]和可用的高分辨率CT使用高空间频率重建算法在1 mm或1.25 mm的层厚上以10 mm的间隔进行重建。研究选择抗合成酶综合征合并间质性肺疾病(anti-synthetase syndrome with interstitial lung disease,ASS-ILD)、特发性肺纤维化(idiopathic pulmonary fibrosis with interstitial lung disease,IPF-ILD)、抗MDA5抗体阳性皮肌炎合并快速进展性间质性肺疾病(antibody-positive dermatomyositis with rapidly progressive interstitial lung disease,DM/RP-ILD)患者以及正常人(Normal)做对比的四类数据,重建为标准医学数字化图像格式DICOM[14](digital imaging and communications in medicine),每类一共20个病例,含切面数据共28 942例,各个疾病的数量如表1所示。之后将医学DICOM数据转化成计算机图像,使用肺窗(L/W:-500 HU/1 500 HU),图像分辨率为(512×512)。设置聚类中心为2的K近邻聚类,计算出肺实质和肺部其他地方的灰度值质心,将两个质心的中值设置为区分肺实质和其他部分的阈值。通过二值化处理获取肺部实质掩膜,再通过图像形态学处理的腐蚀膨胀去除掉掩膜中的噪声。
特征提取模块包括两个部分:一是基于深度学习的影像抽象特征提取,二是影像组学的定量特征提取。原始的ResNet-50[15](residual network-50)网络结构由48个卷积层、1个平均池化层和1个线性层组成。在间质肺病图像检索框架的设计中,去除了网络中的最终全连接层和平均池化层,将网络的最后一个卷积层的输出视为一个维度为W×H×K的三维张量,其中K表示最后一层特征映射的数量。对通道维使用泛化均值池化,泛化均值池化公式为
GEM(x)= 1 H × W i = 1 H j = 1 W x i , j p 1 p
式(1)中:xi,j为特征图在空间维度上的数值;p为超参数,用于控制池化的数值范围。
影像组学是一种利用医学影像中的图像特征来提取疾病定量信息的技术。影像组学计算的特征可以分为3组:几何形状、强度、纹理。
提取特征的过程如下,选取肺实质部分作为影像计算的感兴趣区域(region of interest, ROI)。使用几何特征公式计算肺切片的形状以及肺部疾病表现的形状特征。强度特征公式计算原始影像数据中肺部的像素强度的一阶分布。纹理特征公式计算肺部疾病的图案,和强度的二阶分布和高阶空间分布。所有影像组学特征的计算都是使用专业的组学计算分析库 pyradiomics[16]完成的。
不同信息提取方式的数据之间具有不同的特征表示和信息结构。特征融合的关键挑战之一就是如何有效地整合这些异构数据[17],以提高综合信息的质量和可用性。ResNet-50网络和影像组学公式提取到对应的向量描述存在着特征形状尺度不统一的问题。针对这个问题,提出一种用于间质肺病检索的多特征融合网络(interstitial lung disease feature fusion network,ILD-FFN),用来融合这两种不同的向量表示,如图2所示。
首先对两个特征集合的位置方向进行外积,得到联合表征矩阵zq×p,计算方式为
zb,i,j=xb,iyb,j
式(2)中:xbybb批次的深度学习特征和影像组学特征。
对联合表征空间进行特征提取,使用3个连续的卷积块,将通道拓展为{16,32,64}同时将联合表征的大小减少为{q//8,p//8},再通过1×1大小的卷积核将通道融合,获得深度学习特征和影像组学特征的相关性特征R,再将相关性特征R和深度学习特征X与影像组学特征Y进行拼接,最后通过可学习线性层对特征进行映射和降维。所得到的特征可以表示为
F=(X+Y+R)w+b
式(3)中:w为线性层映射的权重;b为映射偏置。
有监督对比学习[18](supervised contrastive learning,SCL)可以充分利用被标注数据的标签信息,使得具有相同类别的样本,在特征空间上的距离尽可能接近,不同类别的样本距离尽可能远离。在间质性肺疾病特征对比中引入有监督学习,计算过程如下:首先随机抽样批量个待查询图像,计算其融合特征F,从相同类别中取得一个正例计算特征 F 1 +N-1个不同类别的负例中计算特征 F 2 -, F 3 -,…, F N - 1 -,其中N为类别数量。计算F F 1 +, F 2 -, F 3 -,…, F N - 1 -之间的欧式距离。公式为
D= ( F 1 - F n ) 2
构建损失函数为
$\begin{aligned} \text { Loss }= & \frac{1}{2} \sum_{i=1}^{n}\left(\operatorname{lbl}_{i} D_{i}^{2}\right)+\frac{1}{2} \sum_{i=1}^{n}\left\{\left(1-\operatorname{lbl}_{i}\right)\right. \\ & {\left.\left[\max \left(0,0.7-D_{i}\right)\right]^{2}\right\} } \end{aligned}$
当待查询特征与计算特征为相同类别时类别标签lbl=1,当两个特征属于不同类别时lbl=0。损失函数使得同一类别的距离尽可能缩小,而不同类的距离尽可能增大,当样本距离大于0.7时损失为0不进行反向传播。
使用局部敏感哈希[19](locality-sensitive Hashing,LSH)对待测向量集和数据库向量集进行相似检索。首先对数据库特征向量和待测向量进行归一化处理,以正态分布生成和待检测向量相同长度的随机法向量wi,将数据库特征向量xj乘以3个随机法向量的转置并输入符号函数中得到哈希桶的索引表示,即
hi(x)= 0 , w T i x j 0 1 , w T i x j > 0
$\text { Hash Signature: }\left\{h_{1}\left(\boldsymbol{x}_{i}\right), h_{2}\left(\boldsymbol{x}_{i}\right), h_{3}\left(\boldsymbol{x}_{i}\right)\right\}$
将待测向量乘以同样的法向量得到待测向量所在的哈希桶,将待测向量的哈希桶中存储的数据库向量与待测向量进行余弦相似度计算。余弦距离的计算公式为
dcos(xi,y)= x i Y T x i Y T
式(8)中:xi为待测向量Y为数据库向量的矩阵,因为特征提取输出的向量已经被归一化成了0~1范围内的值,当待测向量xi和数据库子向量y模长为1时,将余弦距离的计算转化成了矩阵xiy的乘法计算,待测向量和数据库向量的距离集合为
$d_{\cos }\left(\boldsymbol{x}_{i}, \boldsymbol{y}\right)=\boldsymbol{x}_{i} \boldsymbol{y}_{j}^{\mathrm{T}}$
式(9)中:i表示第i个待测向量;j表示数据库中第j个子向量。
得出相似度最大的前N向量,就是检索到的疾病特征最相似的间质性肺疾病影像,即
{d1,d2,d3,…}
实验所需的软硬件环境和设备信息如下:
CPU处理器:12th Gen Intel(R)Core(TM) i5-12400F;GPU处理器:NVIDIA GeForce RTX 3060 操作系统: Windows10 64 位;编程语言及编译环境:Python3.7;PyCham2021.1x64;深度学习框架:Pytorch。
本次实验以DICOM病例为单位,将每一类20个病例的间质性肺疾病数据集划分为16∶4的交叉训练验证集进行5折交叉验证。网络训练轮次为100,使用自适应矩估计(adaptive moment estimation,Adam)优化器,初始学习率设置为0.000 1,为了防止训练的过拟合,优化器中L2正则项设置为0.000 001,批处理量设置为8。
对于肺部CT影像的检索,取检索到的相似度最高的影像CT间质性肺疾病类别作为分类类别。以此来评估图像检索类别的精度。本次实验使用精确度、召回率以及F1分数进行评估,评估公式为
P= T P T P + F P×100%
R= T P T P + F N×100%
F1= 2 P R P + R×100%
式中:TP、FP、FN分别代表真正、假正、假负,以单个类别为正,其余类别为负,分别计算4个类别的精确度、召回率以及F1分数。
同时本次实验还计算了各个模型特征的类内距离Dintra(intra-class distance)、类间距离Dinter(inter-class distance)、和轮廓系数s(x)来量化各个模型特征的可区分度,具体计算公式为
Dintra= 1 N i n t r a i = 1 k x , y C i k‖x y‖
D i n t e r = 1 N i n t e r i = 1 k j = i + 1 k 1 N s a m p l e s x C i y C j x y
s(x)= D i n t e r - D i n t r a m a x { D i n t e r , D i n t r a }
式中:Ci为类别;‖x-y‖为特征向量的欧式距离; 1 N 为计算距离后对结果进行取均值处理。
为了验证本次研究所提出的多特征融合与对比学习方法(ILD-FFN+SCL)是否能很好地提高间质性肺疾病图像检索的精度,将多特征融合模型与单模态VGG[20](visual geometry group)AlexNet[21]、(Alex Krizhevsky network)、ResNet50、影像组学(radiomics-OMICS,Omic)、和主干网络为ResNet50的对比学习网络(ResNet50+SCL)进行对比实验。其计算结果如表2所示。可以看出所提多特征融合对比学习方法精确度比单一模态特征提升3.9%,召回率提升3.5%,F1分数提升4%。同时还计算了混淆矩阵,来查看本次实验的模型在各个类上的分类结果,如图3所示。混淆矩阵对角线的值代表模型分类正确的样本比例,对角线的值越大说明对每个类的分类效果越好。可以看出,特征融合检索方法对间质性肺疾病图像数据库分类效果最好,在ASS-ILD、PF-ILD类上,正确分类比例提升0.03、0.02,而在DM/RP-ILD类上只降低了0.01。
为了比较不同方法提取的特征有效性,将多特征方法与单模态方法获取的特征进行对比实验,取测试集中4个类别,随机400个肺部CT,用不同模型获取它们的多维特征,使用类内距离、类间距离、轮廓系数计算各个特征,其计算结果如表3所示。从轮廓系数计算结果可以看出,(ILD-FFN+SCL)方式获得的特征拥有最好的区分度,轮廓系数为0.482。Omic模型的特征图以及轮廓系数较差。这是因为组学模型仅代表各个特征的计算结果,如果需要进一步使用组学特征,则需要增加分类器对特征做进一步的聚合。(ResNet50+SCL)模型轮廓系数次之。从上述结果可以看出,多特征融合加对比学习方式能将类内距离减少,而类间距离放大,提升了间质性肺疾病各个类别的区分度。
为了可视化多特征融合对比学习能否有效的对间质性肺疾病特征进行提取,使用类激活图CAM(class activation map)对各个类别检索到的影像特征进行可视化实验,如图4所示,图4中绿色框内为检索图像,检索图像下方为检索到的相似度最高的图像。红色部分为特征响应,可以看出待测图片与检索到的图片拥有相似的特征响应。并且检索到的特征会更接近与影像的病理特征而不是仅仅是形状上相似,增加了间质性肺疾病诊断的可解释性。
对间质性肺疾病进行准确诊断是一个具有挑战的临床问题, 由于间质性肺疾病所包含的疾病特征的难以量化和统计,基于单一特征的分类方法没能取得很好的效果。基于以上问题,本文提出使用基于对比学习的医学影像组学特征和深度学习特征融合检索框架。使用更加全面的特征描述策略,不仅提高了间质性肺疾病的分类精度,而且以疾病相似度的方式量化出最具有参考性的间质性肺疾病图像。
为了能使拥有不同形状尺度的影像组学特征和深度学习特征融合,提出了多特征融合网络(ILD-FFN)。该模型通过计算两个特征的外积空间,获取特征的相关空间表达,和原特征一起形成新的特征空间,通过这种方式,消除了向量形状尺度带来的影响。
同时,本文将多特征融合检索的方法与单一模态的方法进行对比实验,展示了多特征融合方法在间质性肺疾病图像检索中的显著优势。ILD-FFN + SCL的方式在精确度、召回率和F1分数上优于其他单模态模型,效果提升3.9%、3.5%、4%,并在特征区分度方面表现出色,轮廓系数达到0.482,远高于其他模型。这表明多特征融合对比学习方法不仅能更好的捕获图像的信息,还能提升间质性肺疾病不同类别之间的区分度。这一研究成果对于改善间质性肺疾病的诊断和治疗具有重要意义,为未来开发更有效的医疗影像诊断工具提供了有力的理论和实证基础。
本项工作仍然存在一定的局限性。目前的间质性肺疾病检索系统只能适用于包含肺实质较大的部分,因为肺实质较小的部分疾病模式表征不明显且分布不均。模型容易更关注于其形状特征上,后续考虑为疾病模式添加注意力机制,使得每个肺部影像都能有准确的检索对应关系。
本研究提出一种基于多特征融合对比学习检索间质性肺疾病框架。与现有的基于图像的分类方法比较,图像检索的方法最大的优点是可以提供历史病例的参考。并且,本文使用深度特征和人工影像组学特征进行融合,并提出了有效融合多尺度特征的架构。大量的实验表明,基于对比学习融合特征的检索方案在检索典型间质性肺疾病数据库的任务上获得了最高的精度,比使用单一特征的方案效果更好。该方法提高了间质性肺疾病医学影像检索的精度,能够为临床医生提供有效的辅助诊断依据。
  • 国家重点研发计划(2022YFC2010000)
  • 国家重点研发计划(2022YFC2010006)
参考文献 引证文献
排序方式:
[1]
Wijsenbeek M, Suzuki A, Maher T M. Interstitial lung diseases[J]. The Lancet, 2022, 400(10354): 769-786.
[2]
Raghu G, Remy-Jardin M, Myers J L, et al. Diagnosis of idiopathic pulmonary fibrosis. An official ATS/ERS/JRS/ALAT clinical practice guideline[J]. American Journal of Respiratory and Critical Care Medicine, 2018, 198(5): e44-e68.
[3]
魏龙晓, 崔光彬, 魏经国, 等. 实验性肺纤维化HRCT表现及其病理学基础相关性研究[J]. 科学技术与工程, 2005, 25(12): 770-772.
Wei Longxiao, Cui Guangbin, Wei Jingguo, et al. Experiment study on the characteristics of pulmonary fibrosis by HRCT and its relationship with pathological changes[J]. Science Technology and Engineering, 2005, 25(12): 770-772.
[4]
Watadani T, Sakai F, Johkoh T, et al. Interobserver variability in the CT as-sessment of honeycombing in the lungs[J]. Radiology 2013, 266(3): 936-944.
[5]
Lynch D A, Al-Qaisi M A. Quantitative computed tomography in chronic obstructive pulmonary disease[J]. Journal of Thoracic Imaging, 2013, 28(5): 284-290.
[6]
柳洪亚, 祝洁, 刘晨, 等. 高分辨率CT影像组学分类诊断类风湿关节炎相关间质性肺疾病的效能[J]. 陆军军医大学学报, 2024, 46(8): 878-885.
Liu Hongya, Zhu Jie, Liu Chen, et al. Efficacy of high-resolution CT imaging radiomics classification for diagnosis of rheumatoid arthritis-associated interstitial lung disease[J]. Journal of Army Medical University, 2024, 46(8): 878-885.
[7]
Poynton C, Ash S, Putman R K, et al. Radiomics detects intersti-tial lung abnormalities (ILA) progression in high-risk patients[M]//TP026 Diagnosis, Assessment, and Prognosis of Fibrotic ILD. New York: American Thoracic Society, 2021: A1848.
[8]
Huang S, Lee F, Miao R, et al. A deep convolutional neural network architecture for interstitial lung disease pattern classification[J]. Medical & Biological Engineering & Computing, 2020, 58: 725-737.
[9]
Mei X, Liu Z, Singh A, et al. Interstitial lung disease diagnosis and prognosis using an AI system integrating longitudinal data[J]. Nature Communications, 2023, 14(1): 2272.
[10]
谷宇, 刘佳琪, 杨立东, 等. 融合注意力特征及动态卷积的肺结节辅助诊断[J]. 科学技术与工程, 2023, 23(16): 6834-6844.
Gu Yu, Liu Jiaqi, Yang Lidong, et al. Fusion of attention and dynamic convolution for computer aided diagnosis of pulmonary nodules[J]. Science Technology and Engineering, 2023, 23(16): 6834-6844.
[11]
中华医学会风湿病学会. 多发性肌炎和皮肌炎诊断及治疗指南[J]. 中华风湿病学杂志, 2010, 14(12): 828-831.
Chinese Medical Association Society of Rheumatology. Guidelines for the diagnosis and treatment of polymyositis and dermatomyositis[J]. Chinese Journal of Rheumatology, 2010, 14(12): 828-831.
[12]
Lynch D A, Sverzellati N, Travis W D, et al. Diagnostic criteria for idiopathic pulmonary fibrosis: a fleischner society white paper[J]. The Lancet Respiratory Medicine, 2018, 6(2): 138-153.
[13]
Targoff I N, Miller F W, Medsger Jr T A, et al. Classification criteria for the idiopathic inflammatory myopathies[J]. Current Opinion in Rheumatology, 1997, 9(6): 527-535.
[14]
Horii S C, Bidgood W D. DICOM: A standard for medical imaging[C]// Enabling Technologies for High-Bandwidth Applications. Richardson: SPIE, 1993: 87-102.
[15]
He K, Zhang X, Ren S, et al. Deep residual learning for image recognition[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2016: 770-778.
[16]
Van Griethuysen J J M, Fedorov A, Parmar C, et al. Computational radiomics system to decode the radiographic phenotype[J]. Cancer Research, 2017, 77(21): e104-e107.
[17]
Baltrušaitis T, Ahuja C, Morency L P. Multimodal machine learning: a survey and taxonomy[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2018, 41(2): 423-443.
[18]
Khosla P, Teterwak P, Wang C, et al. Supervised contrastive learning[J]. Advances in Neural Information Processing Systems, 2020, 33: 18661-18673.
[19]
Andoni A, Indyk P. Near-optimal hashing algorithms for approximate nearest neighbor in high dimensions[J]. Communications of the ACM, 2008, 51(1): 117-122.
[20]
Chhabra M, Kumar R. An advanced VGG16 architecture-based deep learning model to detect pneumonia from medical images[M]// Emergent Converging Technologies and Biomedical Systems: Select Proceedings of ETBS 2021. Singapore: Springer Singapore, 2022: 457-471.
[21]
Krizhevsky A, Sutskever I, Hinton G E. Imagenet classification with deep convolutional neural networks[J]. Advances in Neural Information Processing Systems, 2012, 25: 1106-1114.
2025年第25卷第11期
PDF下载
324
127
引用本文
BibTeX
文章信息
doi: 10.12404/j.issn.1671-1815.2406118
  • 接收时间:2024-08-15
  • 首发时间:2025-07-09
  • 出版时间:2025-04-18
补充材料
相关文章
文章信息
作者
出版历史
  • 收稿日期:2024-08-15
  • 修回日期:2024-11-24
基金
国家重点研发计划(2022YFC2010000)
国家重点研发计划(2022YFC2010006)
作者信息
    1 海南大学机电工程学院,海口570228
    2 医智源健康科技有限公司, 杭州 310000
    3 南京鼓楼医院呼吸与危重症医学科, 南京 210000

通讯作者:

* 蔡后荣(1961—),男,汉族,江西万安人,博士,主任医师。研究方向:间质性肺炎。E-mail:
参考文献
分享链接
https://castjournals.cast.org.cn/joweb/kxjsygc/CN/10.12404/j.issn.1671-1815.2406118
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

Family
属数
Number of
genus
种数
Number of
species
占总种数比例
Percentage of
total species (%)

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