Objective:To investigate the clinical value of multi-center digital mammography ra-diomics nomogram model in predicting histological grading of invasive breast cancer.Methods:Using pathological diagnosis as the gold standard,a total of 437 invasive breast cancer patients from the First Affiliated Hospital/Yijishan Hospital of Wannan Medical College were randomly divided into a train-ing group(305 cases,including 217 cases of grade Ⅰ/Ⅱ and 88 cases of grade Ⅲ)and a validation group(132 cases,including 94 cases of grade Ⅰ/Ⅱ and 38 cases of grade Ⅲ),following a ratio of 7:3.Additionally,291 invasive breast cancer patients(203 cases of grade Ⅰ/Ⅱ and 88 cases of grade m)in Fuyang People's Hospital(n=129)and Taihe County People's Hospital(n=162)were included in the external test group.The mediolateral oblique(MLO)and cranial cauda(CC)digital mammography images were compared,and those with larger lesion areas were selected.Image segmentation and ra-diomics feature extraction were performed using the Shenrui Medical Multimodal Research Platform.Dimension reduction of radiomics features was achieved through characteristic correlation analysis and the least absolute shrinkage and selection operator(LASSO),followed by the construction of ra-diomics models using logistic regression(LR).Clinical indicators were analyzed using univariate and multivariate binary logistic regression analysis to construct the clinical model.Radiomics scores were combined with clinical indicators to construct a nomogram.Model performance was evaluated using re-ceiver operating characteristic(ROC)curves and decision curve analysis(DCA),and the predictive performance between models was compared using the DeLong test.Results:Three radiomics features closely associated with the histological grading of invasive breast cancer were ultimately selected.The predictive performance of the nomogram for histological grading of invasive breast cancer in the train-ing,validation,and external test groups were 0.811,0.825,and 0.803,respectively.This diagnostic effi-cacy surpassed that of the individual models.Decision curve analysis(DCA)indicated that the net ben-efit of the nomogram in predicting the histological grading of invasive breast cancer was higher in the training,validation,and external test groups compared to the radiomics and clinical models when the probability value ranged from 20％to 60％.Conclusion:The radiomics model based on digital mam-mography demonstrates high efficacy in predicting the histological grading of invasive breast cancer,thereby holding significant clinical value for the development of personalized treatment plan and prog-nosis assessment of patients.
关键词
乳腺癌/数字乳腺X线/组织学分级/影像组学/列线图
Key words
Breast cancer/Digital mammography/Histologic grade/Radiomics/Nomogram
维普
万方数据