34论著影像研究与医学应用2024年3月第8卷第5期多参数MRI影像基因组学对乳腺癌早期复发的预测研究周素菊1,廖锦元2(1桂林市中医医院放射科广西桂林541002)(2广西医科大学第一附属医院医学影像科广西南宁530021)MultiparameterMRIimaginggenomicsinbreastcancerpredictivestudiesofearlyrecurrenceZHOUSuju1,LIAOJinyuan21DepartmentofRadiology,GuilinHospitalofTraditionalChineseMedicine,Guilin,Guangxi541002,China;2DepartmentofMedicalImaging,TheFirstAffiliatedHospitalofGuangxiMedicalUniversity,Nanning,Guangxi530021,China【Abstract】ObjectiveToinvestigatethevalueofdynamiccontrastenhancedmagneticresonanceandhighb-valuediffusion-weightedimaging(DWI)inimaginggenomicsinpredictingearlyrecurrenceofbreastcancer.MethodsThisstudyretrospectivelyanalyzed100patientswithbreastcancerpathologicallyconfirmedbysurgeryorpuncturebiopsyadmittedtotheFirstAffiliatedHospitalofGuangxiMedicalUniversityandGuilinHospitalofTraditionalChineseMedicinefromJanuary2017toApril2019.Clinicalandimagingdataofpatients,includingageofonset,menopausalstatus,immunohistochemistry(includingER,PR,HER-2,Ki-67,etc.),OncotypeDX21-genetestresults,tumorsize,lymphnodemetastasis,recurrence,tumorenhancementpattern,tumormargin(noburrsor/lobed),tumormorphology,andtumorenhancementcurve.AccordingtotheresultsofOncotypeDX21-genetest,breastcancerpatientsweredividedintomedium-highriskgroup(RS≥18)andlow-riskgroup(RS<18).Diffuse-weightedimaging(highb-valueDWI)andcontrast-enhancedT1-weighted(T1+C)imageswereexportedfromthePACSsystem,andareasofinterest(ROI)weremanuallydelineatedbytwoexperiencedradiologists.HuiyiHuiyingsoftwarewasusedtoextracttheimageomicsfeaturesofdynamiccontrastenhancement(DCE)andhighb-valueDWIimagesfromtheselectedROI.DifferentmethodswereusedtoreducethedimensionofMRIfeatures.KNNmethodwasusedtoestablishthemodelofpredictingearlybreastcancerrecurrencewithimageomicstagsandOncotypeDX21-genetestresults.Theyweredividedintothetrainingset(70cases)andthetestset(30cases)tocomparetheAUC,95%CI,sensitivityandspecificityofthetwogroups.Then,theextract...