| PART Ⅰ Preoperative differentiation of benign and malignant breast lesions in breast imaging reporting and data system category 4 utilizing an intratumoral and peritumoral radiomics nomogram based on digital breast tomosynthesisObjective: To explore the value of digital breast tomosynthesis(DBT)based intralesional and perilesional radiomics nomogram in distinguishing the benign and malignant nature of breast lesions categorized as Breast Imaging Reporting and Data System(BI-RADS)4.Materials and Methods: A retrospective analysis was performed on clinical data collected from 312 patients with BI-RADS 4 lesions diagnosed through DBT.These patients were randomly divided into a training set(N=218)and a validation set(N=94)at a ratio of 7:3.A total of 948 radiomics features were extracted based on DBT for intra-lesional regions,while an equal number of features were extracted for perilesional regions.Additionally,1896 features were extracted for intra-perilesional regions.Utilizing independent sample t-test(or Mann Whitney U-test),Spearman rank correlation,and LASSO regression,the optimal subset of features was screened to construct three Radscore models: Radscore IL,Radscore PL,and Radscore Combine.Independent risk factors for malignant BI-RADS 4 lesions were determined through univariate and multivariate logistic regression analyses,and were subsequently used to develop clinical and nomogram models.The performance of the models was evaluated and validated using receiver operating characteristic(ROC)curve analysis,calibration curve analysis,decision curve analysis(DCA),net reclassification improvement(NRI),and comprehensive discriminant improvement(IDI)methods.Results: Among the three Radscore models,Radscore＿Combine exhibited the most superior predictive performance for malignant BI-RADS 4 lesions,achieving the area under curve(AUC)values of 0.935 and 0.888 in the training and validation sets,respectively.Moreover,the age of patients and subclassification within the BI-RADS category 4 were identified as two clinically independent predictive factors for malignant BI-RADS 4 lesions.These factors were then employed to construct a clinical model.A radiomics nomogram model was further constructed by integrating Radscore＿Combine,age,and subclassification within the BI-RADS category 4.The AUC values of this nomogram model in both datasets were 0.960 and 0.942,respectively,demonstrating excellent diagnostic performance.The NRI and IDI values exceed 0,while both P-values are below 0.05.These findings indicate that Radscore＿Combine holds the potential to become a crucial imaging biomarker,aiding in improving the efficiency of distinguishing between benign and malignant lesions within this category.Conclusion: The DBT-based intralesional and perilesional radiomics nomogram holds significant utility in distinguishing between benign and malignant BI-RADS 4 lesions.PART Ⅱ Utilizing grayscale ultrasound?based radiomics nomogram for preoperative identification of triple negative breast cancerObjective: This study aimed to develop a radiomics nomogram based on grayscale ultrasound to distinguish triple-negative breast cancer(TNBC)from nontriple-negative breast cancer(NTNBC)prior to surgery.Materials and Methods: A retrospective analysis of 454 breast carcinoma patients confirmed by pathology was conducted,with 317 patients in the training dataset(including 59 TNBCs)and 137 patients in the validation dataset(including 27 TNBCs).474 radiomics features were extracted from grayscale ultrasound images of breast tumors.Independent sample t-test(or Mann Whitney U-test),Spearman rank correlation,and LASSO regression were utilized to select the optimal feature subset and construct a Radscore model.Through univariate and multivariate logistic regression analysis,independent risk factors for TNBC were determined,and both a clinical model and a nomogram model were established.The nomogram models were assessed using the ROC curve analysis,calibration curve,DCA,NRI and IDI.Results: Multivariate logistic regression analysis revealed that tumor shape,margin,and calcification were independent clinical predictive factors for TNBC,which were then utilized to develop clinical models.Additionally,16 radiomics features were selected to construct the Radscore model out of a total of 474 extracted features.The radiomics nomogram model,which incorporated tumor shape,margin,calcification,and Radscore,achieved an AUC value of 0.837 in the training dataset and 0.813 in the validation dataset,outperforming both the Radscore and clinical models in terms of diagnostic efficacy.The NRI and IDI values exceed 0,while both P-values are below 0.05.These findings indicate that Radscore shows potential as a valuable imaging biomarker for distinguishing TNBC from NTNBC.Conclusion: The ultrasound-based radiomics nomogram has the potential to offer crucial guidance for preoperative distinguishing between TNBC and NTNBC.PART Ⅲ Preoperative assessment of lymphovascular invasion in invasive breast cancer utilizing an intratumoral and peritumoral radiomics nomogram based on digital breast tomosynthesisObjective: This study aimed to develop an intratumoral and peritumoral radiomics nomogram using DBT,so as to achieve preoperative prediction of the lymphovascular invasion(LVI)status in invasive breast cancer(IBC)patients.Materials and Methods: A retrospective analysis was conducted on clinical data obtained from 178 patients,who were randomly allocated into a training set(N=124)and a validation set(N=54)at a ratio of 7:3.A total of 948 radiomics features were extracted based on DBT for intra-tumoral regions,while an equal number of features were extracted for peritumoral regions.Additionally,1896 features were extracted for intra-peritumoral regions.Utilizing independent sample t-test(or Mann Whitney Utest),Spearman rank correlation,and LASSO regression,the optimal subset of features was screened to construct three Radscore models: Radscore IT,Radscore PT,and Radscore Combine.Independent predictive factors for LVI in IBC patients were determined through univariate and multivariate logistic regression analyses,and were subsequently used to develop clinical and nomogram models.The performance of the models was evaluated and validated using the ROC curve analysis,calibration curve,DCA,NRI and IDI methods.Results: Logistic regression analysis identified the tumor margin and DBT reported lymph node metastasis(DBT_reported_LNM)as two clinically independent predictive factors for LVI,which were subsequently utilized to develop a clinical model.Among the three Radscore models,Radscore_Combine exhibited the most superior predictive performance for LVI,achieving AUC values of 0.865 and 0.817 in the training and validation sets,respectively.Moreover,a radiomics nomogram model was constructed by integrating Radscore_Combine,tumor margin,and DBT_reported_LNM.The AUC values of this nomogram model in both datasets were 0.906 and 0.905,respectively,demonstrating excellent predictive performance.The NRI and IDI values exceed 0,while both P-values are below 0.05.These findings indicate that Radscore_Combine could serve as a valuable imaging biomarker for predicting the status of LVI.Conclusion: The radiomics nomogram based on intratumoral and peritumoral features extracted from DBT images has the potential to assist in preoperative prediction of LVI status in patients with IBC. |
PDF Full Text Request