Correlation Between X-ray Signs Of Breast Cancer And Molecular Subtypes

Objective To explore the X-ray characteristics of different molecular subtypes and receptor expression in breast cancer.Materials and methods We retrospectively analyzed 439 patients with breast cancer.The X-ray and clinical characteristics of different molecular subtypes and receptor expressions were analyzed based on pathological results.Results The proportion of spiculate mass and calcification score < 9 points in Luminal A type was higher than that of the other three molecular subtypes(P<0.0083);The proportion of HER2 overexpression expresses calcification and calcification score ≥9 was greater than that of the other three molecular subtypes(P< 0.0083),the tumor length was the largest,and there was a statistical difference with Luminal A and Luminal B;The values of triple-negative ki-67 were all greater than those of the other three molecular subtypes(P < 0.0083).The proportion of non-spiculate masses and high grade ductal carcinoma were the highest,and the Triple negative was statistically different from both Luminal A and Luminal B.The proportion of non-menopausal,spiculate mass and calcification score < 9 points in ER and/or PR receptor positive group was greater than that in ER and PR receptor total negative group(P<0.05);The proportion of high-grade ductal carcinoma,calcification,calcification score were greater than or equal to 9 points and tumor length ≥2cm in the HER2 receptor-positive group,was greater than that in the HER2 receptor-negative group(P<0.05);The proportion of high-grade ductal carcinoma,non-spiculate mass and calcification score ≥9 in ki-67 ≥20% group,was greater than that in ki-67 < 20% group(P<0.05).Conclusion The expression of different molecular subtypes and receptors in breast cancer has a certain correlation in X-ray and clinical characteristics,which is instructional for the selection of clinical diagnosis and treatment options.Purpose To explore the predictive value of quantitative radiomics features extracted from Full-field Digital Mammography(FFDM),Digital Breast Tomosynthesis(DBT),and the predictive value of combined with simple imaging features in molecular subtypes of breast cancer.Materials and methods We retrospectively analyzed 207 patients with breast cancer,including 30 cases of Luminal A,121 cases of Luminal B,30 cases of Triple negative,and 26 cases of HER2 overexpression.The lesions were outlined and marked in the X-ray image,and 100 quantitative radiomic features were extracted,including morphological,grey/scale statistics,and texture features.The data set was randomly assigned to the training and test sets at a ratio of 7: 3.Logistic backward regression method together with the optimal subset method were applied to select the radiomics features.The classification model was established by using multinomial logistic regression.Results The quantitative radiomics features extracted from FFDM have the best performance in predicting HER2 overexpression,and the area under the curve(AUC)of the training and test sets were 0.796 and 0.748 respectively;The quantitative radiomics features extracted from DBT have the best performance in predicting Triple-negative type,and the AUC of the training and test sets were 0.749 and 0.776 respectively;The FFDM+DBT comprehensive extraction of quantitative radiomics features to predict the AUC of Luminal A,Luminal B,Triple negative,and HER2 overexpression training set,they were 0.913,0.740,0.841,0.896,the AUC of test sets were 0.741,0.532,0.867,and 0.765.When the FFDM+DBT comprehensive extraction of quantitative radiomics features combined with simple image features was used to predict the AUC of Luminal A,Luminal B,Triple negative,and HER2 overexpression training sets,they were 0.956,0.803,0.889,0.931,however,the AUC of test sets were 0.746,0.569,0.808,0.800.Conclusion It is feasible to predict molecular subtypes of breast cancer by extracting quantitative X-ray radiomics features;FFDM + DBT imaging omics machine learning combine with simple image features can further improve Luminal A,Triple negative and HER2 overexpression predictive performance.