Prediction Study Of Malignant Risk And Molecular Subtypes Of BI-RADS Category 4 Calcifications In The Breast

Part Ⅰ:Predictive study of risk stratification for malignancy in BI-RADS category 4 calcificationsObjective:To investigate the value of the imaging features of Breast Imaging Reporting and Data System(BI-RADS)category 4 calcifications in mammography combined with clinical features to predict the risk stratification of malignancy in high-risk breast lesions,and to develop a Nomogram prediction model.Materials and Methods:Retrospective analysis was performed for 446 female patients with BI-RADS category 4 calcifications diagnosed by mammography with pathological findings at our hospital from January 2019 to May 2022.The morphology and distribution of calcifications were classified and given assessment category according to the fifth edition of BI-RADS,and relevant clinical factors of the patients were collected.Malignancy rate for each classification were calculated using pathological findings as the gold standard.Predictors of BI-RADS category 4calcifications were screened using binary logistic regression for imaging features and clinical factors,and prediction model for imaging features and combined imaging-clinical prediction model were constructed based on these predictors,and the predictive efficacy of the two models for the category 4 calcifications was compared,and Nomogram were plotted to visualize the results,and calibration curve and decision curve analysis were used to assess the calibration of the prediction model and clinical usefulness.Results:The overall malignancy rate was 49.1% in 446 patients,and the malignancy rates for morphology(32.5%-95.5%)and distribution(22.5%-90%)were generally consistent with the suggested BI-RADS categories.The BI-RADS categories changed after amorphous and coarse heterogeneous were combined with distribution features.Binary logistic regression incorporating imaging features and clinical features showed that menopausal status,personal history of breast cancer,morphology and distribution were predictors of the BI-RADS category 4 calcifications.The combined imaging-clinical prediction model(AUC=0.835)constructed based on these four factors had better discrimination(P=0.028)and better calibration and clinical utility than the imaging features prediction model(AUC=0.815).The Nomogram constructed by the combined prediction model showed better calibration and clinical usefulness.Conclusions:(1)The morphology and distribution of suspicious calcifications have good clinical application in the malignant risk stratification of BI-RADS category4 calcifications.(2)Compared with the imaging features prediction model,the combined imaging-clinical prediction model is more accurate in stratifying the risk of malignancy for BI-RADS category 4 suspicious calcifications,and has better discriminatory power and clinical utility.Part Ⅱ: Prediction of molecular subtypes of invasive carcinoma by mammographic imaging features of BI-RADS category 4 calcificationsObjective: To investigate the association between the imaging features of suspicious calcifications on mammography and the four molecular subtypes of invasive carcinoma,and to predict the molecular subtypes of invasive carcinoma using the imaging features of calcifications on mammography.Materials and Methods:Retrospective analysis of 100 female patients with BI-RADS category 4 calcifications diagnosed on mammography and pathological results of invasive carcinoma at our hospital from January 2019 to May 2022.The imaging features of calcifications on mammography,including morphology,distribution and range of calcifications,were analyzed.The correlation between calcification features and molecular subtypes was analyzed based on Chi-square test or Fisher’s exact test,and the independent predictors of each molecular subtype were analyzed using binary logistic regression to establish a prediction model.Results:In 100 patients with invasive breast cancer,the Chi-square test or Fisher’s exact test showed that morphology and range were significantly correlated between molecular subtypes(P<0.05),age and distribution were not significantly correlated between molecular subtypes(P>0.05).Binary logistic regression showed that morphology as fine pleomorphic or fine linear and fine branched was an independent protective factor for Luminal A subtype(OR=0.113,95%CI: 0.024-0.532),which was well discriminated by the model(AUC=0.722,95%CI: 0.623-0.807);morphology(OR=3.083,95%CI: 1.066-8.919)and range of calcification(OR=3.820,95%CI: 1.096-13.308)were independent risk factors for HER-2 enriched subtype,which was also well discriminated by the model(AUC=0.713,95%CI: 0.614-0.799).Calcification features were not significantly different in Luminal B and basal-like subtype.Conclusions:(1)The imaging features of BI-RADS category 4 calcifications on mammogram can predict the molecular subtype of breast invasive carcinoma,with significant differences in morphology and range among the different molecular subtypes;(2)Morphology as amorphous or coarse heterogeneous can be used to independently predict Luminal A subtype,morphology as fine pleomorphic or fine linear and fine branched and range≥2 cm can independently predict HER-2 enriched subtype,and both models have good discriminatory power.