Multiparametric MRI-based Radiomics Model For Predicting HER-2 2+status Of Breast Cancer And HER-2 Positive Breast Cancer Under Different Status Of HR

Part 1.Multiparametric MRI-based radiomics model for predicting HER-2 2+status of breast cancerBackground:Breast cancer(BC)is the most common malignancy in women,seriously threatening female health and life.BC is a highly heterogeneous tumor,primarily classified into four major molecular subtypes:Luminal A type,Luminal B type,human epidermal growth factor receptor-2(HER-2)overexpression type and basal cell-like type.Among them,HER-2 positive BC is characterized by high recurrence,metastasis rate and poor prognosis.Since the use of targeted drugs such as trastuzumab,pertuzumab,lapatinib and neratinib in the clinic,the progression-free survival and overall survival of patients with HER-2 positive BC have been significantly improved.Therefore,accurate assessment of HER-2 status is highly critical for the clinical treatment and prognosis of BC.It is recommended by the guidelines for HER-2 detection in BC that immunohistochemistry(IHC)be adopted to detect the expression level of HER-2 protein.IHC 3+is determined to be HER-2 positive,and IHC 0 and 1+is determined to be HER-2 negative.Those with IHC 2+need to adopt fluorescence in situ hybridization(FISH)to detect HER-2 gene amplification status.FISH is achieved by fluorescently labeled DNA probes with DNA target sequences in the nucleus.The probe signal hybridized to the DNA target sequence in the nucleus is observed and analyzed under a fluorescence microscope,thereby obtaining information on the gene status on the chromosome(or chromosome fragment)in the nucleus.At present,the dual probes containing both the HER-2 gene and the chromosome 17 centromere(CEP17)where the gene is located or a single probe containing only the HER-2 gene is dominated in the HER-2 gene status detection,However,FISH testing is costly and time-consuming,and it requires high internal and external quality control in the pathological laboratory.Recently,a series of achievements have been obtained in the application of radiomics,showing great potential in the diagnosis and prognosis of a variety of tumors,including the identification of benign and malignant breast cancer,molecular subtype identification,neoadjuvant chemotherapy effect monitoring,and prognostic factor assessment.However,there are few reports on the assessment of HER-2 2+ status by multi-parameter radiomics.The present study aims to explore whether the multi-parameter MRI radiomics can predict HER-2 2+status,so as to provide a convenient and quick HER-2 2+assessment method to better serve clinical patients.Objective:To investigate the application value of the combined model based on multiparameter MRI radiomics and clinical features in predicting HER-2 2+status of BC.Materials and methods:The data of 104 patients pathologically diagnosed with HER-2 2+invasive BC after operation in our hospital from October 2015 to December 2019 and undergoing MRI before operation were collected.All patients underwent FISH.The inclusion criteria for subjects were as follows:(1)patients without receiving any other treatments before operation and MRI;(2)those who underwent breast MRI scan,with fat suppress T2WI,DCE-MRI and ADC images obtained within 2 weeks before operation;(3)those pathologically diagnosed with invasive BC of a single mass type;(4)those with complete pathology,IHC and FISH data,and the IHC indicated HER-2 2+.The exclusion criteria were as follows:(1)patients receiving any treatment before MRI,such as needle biopsy,operation,radiotherapy,chemotherapy or endocrine therapy;(2)those with a poor image quality(such as artifacts)that failed to be adopted for analysis.All patients were divided into training group(n=72)and validation group(n=32)at 7:3 by complete randomization.Clinicopathological data and conventional breast MRI image analysis:According to clinical surgical pathology,IHC and FISH results,all patients were divided into HER-2 2+FISH positive(HER-2 positive)and HER-2 2+FISH negative(HER-2 negative).Clinical and conventional MRI imaging data were also divided into corresponding groups.Patients' age,lesion location,maximum diameter,peritumoral edema,prepectoral edema,necrosis and cystic degeneration,ADC value,initial enhancement rate(IER),ki-67 expression level and axillary lymph node metastasis were included.Peritumoral edema is the high signal intensity around the tumor on T2WI images,and prepectoral edema is the high signal intensity in front of the pectoral muscle on T2WI images.When measuring ADC value,the region of interest(ROI)should avoid visible necrosis,cystic change,bleeding area and normal glandular tissue as far as possible.After DCE-MRI images were imported into the workstation,ROI was delineated in the most obvious part of tumor enhancement,and IER value was measured.IER=(peak signal intensity within 3 min after contrast injection-signal intensity before contrast injection)/signal intensity before contrast injection ×100%.The traditional quantitative index finally takes the average value of two doctors as the final result for follow-up analysis.If the results of peritumoral edema and prepectoral edema were different between two physicians independently,a consensus would be reached through consultation.Radiomics analysis:Two senior radiologists manually delineated the three-dimensional volume of interest(VOI)layer by layer on the MRI image,and performed intra-and inter-class consistency analysis of the radiomics features,thereby retaining stable features for subsequent modeling.Based on the T2WI-fat suppress,ADC and the most obvious lesion enhancement phase in dynamic contrast enhanced(DCE)of each patient(according to the DCE time-signal intensity curve,the most obvious lesion enhancement phase was found on the DCE sequence),1906 radiomics features were extracted,and the dimensionality was reduced by feature screening methods such as L1 based and F-test.A logistic regression(LR)classifier was adopted to establish a binary radiomics prediction model,i.e.T2WI model,ADC model,DCE model,T2WI+ADC model,T2WI+DCE model,ADC+DCE model and T2WI+ADC+DCE model,and they were validated in the verification group.Independent sample t test or Mann-Whitney test was used for measurement data,and Chi-square test was used for classification data.Univariate logistic regression analysis was used to compare the clinicopathological features,conventional MRI imaging features and the effect of radiomics score on predicting HER-2 2+status.The independent predictors were screened by multivariate logistic regression analysis to construct a radiomics nomogram.The area under the ROC curve(AUC),accuracy,sensitivity and specificity were calculated by the receiver operating characteristic(ROC)curve,so as to assess the predictive efficacy of the model.Results:All patients were divided into HER-2 negative and HER-2 positive according to clinical pathology,IHC and FISH results,among which 74 patients were HER-2 negative and 30 patients were HER-2 positive.There were 51 HER-2 negative patients and 21 HER-2 positive patients in the training group.In the validation group,there were 23 HER-2 negative patients and 9 HER-2 positive patients.There was no statistically significant difference in baseline data between training group and validation group(P>0.05).A total of 10 radiomics features were obtained after dimensionality reduction based on the combined features of T2WI,ADC and DCE,including 6 wavelet feature parameters,3 LoG feature parameters and 1 squareroot feature parameter.The AUC corresponding to the radiomics label established by the combination of the three sequences was 0.883,and its diagnostic efficiency was better than any single and two sequences,which was validated in validation group(0.816).In the training group,the predictors of HER-2 2+status were identified as maximum lesion size,peritumoral edema and radiomics signature by univariate logistic regression analysis.By multivariate logistic regression analysis,the independent predictors for identifying HER-2 2+status were combined radiomics signature and peritumoral edema.The radiomics nomogram established by combination of radiomics signature and peritumoral edema displayed good discrimination in training group and validation group,and the corresponding AUC value was 0.966 and 0.884,respectively.Conclusion:The multi-parameter MRI-based nomogram incorporating radiomicssignature and peritumoral edema can effectively assess the HER-2 2+status of breast cancer,providing a scientific basis for clinicians in clinical decision-making.Part 2.Multiparametric MRI-based radiomics model for predicting HER-2 positive breast cancer under different status of hormone receptors(HR)Background:HER-2 positive occurs in 20-30%of breast cancer(BC)patients.HER-2 positive BC is characterized by high malignancy,early recurrence and metastasis,and poor prognosis.HER-2 positive BC can be divided into two subtypes,i.e.,HR negative(HR-)and HR positive(HR+)according to the expression of hormone receptor(HR).Great differences can be found in the biological behavior,clinicopathological features,survival rate,recurrence,response to treatment and prognosis between different subtypes of HER-2 positive BC.Compared with HR+HER-2 positive BC,HR-HER-2 positive BC has a higher 5-year mortality rate,more prone to brain and liver metastasis,and has a better response to neoadjuvant chemotherapy.Chemotherapy combined with trastuzumab and other anti-HER-2 targeted drugs is recommended for HR-HER-2+BC,while endocrine therapy combined with targeted therapy is recommended for HR+HER-2+BC.Radiomics has been the hotspot and focus of research in recent years,which displays great potential in the differentiation of benign and malignant BC,molecular subtype identification,neoadjuvant chemotherapy effect monitoring and prognostic evaluation.Previous imaging studies on molecular subtypes of BC focused on Luminal A,Luminal B,HER-2 overexpression,and triple-negative BC,but there were few MR imaging studies on HER-2 positive BC under different HR status,and the treatment response,recurrence and metastasis types of HER-2 positive BC are quite different under different HR status.This paper,therefore,aims to explore the application value of multiparameter-based MRI radiomics model in identifying HER-2 positive BC under different HR status,so as to guide the clinical therapeutic regimen and prognostic evaluation.Objective:To investigate the application value of the combined model based on multiparameter MRI radiomics and clinical features in predicting HER-2 positive BC under different HR status.Materials and methods:The data of 98 patients pathologically diagnosed with HER-2 positive invasive BC after operation in our hospital from October 2015 to December 2019 and undergoing MRI before operation were collected.The age range was 23-73 years,with an average age of 47.12 ± 10.31 years.The pathological,immunohistochemical and FISH data of all patients were complete.All patients were divided into training group(n=68)and validation group(n=30)at 7:3 by complete randomization.Two senior radiologists manually delineated the three-dimensional volume of interest(VOI)layer by layer on the MRI image,and performed intra-and inter-class consistency analysis of the radiomics features,thereby retaining stable features for subsequent modeling.Based on clinical surgical pathology,IHC and FISH results,all HER-2 positive BC patients were divided into HR+and HR-groups.The clinical and imaging data were also divided into corresponding groups,including the patients' age,lesion site,size,peritumoral edema,prepectoral edema,necrosis and cystic change,apparent diffusion coefficient(ADC),early enhancement rate,Ki-67 and axillary lymph node metastasis.Based on the T2WI-fat suppress,ADC and the most obvious lesion enhancement phase in dynamic contrast enhanced(DCE)of each patient(according to the DCE time-signal intensity curve,the most obvious lesion enhancement phase was found on the DCE sequence),1906 radiomics features were extracted,and the dimensionality was reduced by feature screening methods such as L1 based and F-test.A logistic regression(LR)classifier was adopted to establish a binary radiomics prediction model,i.e.T2WI model,ADC model,DCE model and T2WI+ADC+DCE model,and they were validated in the verification group.Independent sample t test or Mann-Whitney test was used for measurement data,and Chi-square test was used for classification data.Univariate logistic regression analysis was used to compare the clinicopathological features,conventional MRI imaging features and the effect of radiomics score on predicting HR status.The independent predictors were screened by multivariate logistic regression analysis to construct a radiomics nomogram.The area under the ROC curve(AUC),accuracy,sensitivity and specificity were calculated by the receiver operating characteristic(ROC)curve,so as to assess the predictive efficacy of the model.Results:All patients were divided into HR positive and HR negative according to clinical surgical pathology,IHC and FISH results,of which 65 were HR positive and 33 were HR negative.In the training group,45 cases were HR positive and 23 cases were HR negative.In the validation group,20 patients were HR positive and 10 were HR negative.There was no statistically significant difference in baseline data between training group and validation group(P>0.05).A total of 10 radiomics features were obtained after dimensionality reduction based on the combined features of T2WI,ADC and DCE,including 1 original image feature,6 wavelet features,2 LoG features and 1 gradient feature.The AUC corresponding to the radiomics label established by the combination of the three sequences was 0.797,and its diagnostic efficiency was better than any single sequence,which was validated in validation group(0.750).Univariate logistic regression analysis showed that the predictors of HR status in the training group were peritumoral edema,prepectoral edema,necrosis and cystic change,and combined radiomics signature.By multivariate logistic regression analysis,the independent predictors for identifying HR status were combined radiomics signature and peritumoral edema.The radiomics nomogram established by combination of radiomics signature and peritumoral edema displayed good discrimination in training group and validation group,and the corresponding AUC was 0.815 and 0.805,respectively.Conclusion:The multiparameter MRI-based nomogram incorporating radiomics signature and peritumoral edema can effectively assess the HR status of HER-2 positive BC,providing a scientific basis for clinicians developing therapeutic regimens.