Breast cancer Ki67 expression prediction by DCE-MRI radiomics features

•Quantitative radiomics imaging features of DCE-MRI are extracted and analyzed. •A predictive model for breast cancer low- and high-Ki67 expression is proposed. •The model yields 0.773 AUC, 0.757 accuracy, 0.777 sensitivity and 0.769 specificity. Aim To investigate whether quantitative radiomics features extracted from dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) are associated with Ki67 expression of breast cancer. Materials and methods This institutional review board-approved retrospective study comprised 377 Chinese women who were diagnosed with invasive breast cancer in 2015. This cohort included 53 low-Ki67 expression (Ki67 proliferation index less than 14%) and 324 cases with high-Ki67 expression (Ki67 proliferation index more than 14%). A binary-classification of low-versus high- Ki67 expression was performed. A set of 56 quantitative radiomics features, including morphological, greyscale statistic, and texture features, were extracted from the segmented lesion area. Three machine learning classification methods, including naive Bayes, k-nearest neighbour and support vector machine, were employed for the classification and the least absolute shrink age and selection operator (LASSO) method was used to select most predictive features set for the classifiers. Classification performance was evaluated by the area under receiver operating characteristic curve (AUC), accuracy, sensitivity, and specificity. Resules The model that used naive Bayes classification method achieved the best performance than the other two methods, yielding 0.773 AUC, 0.757 accuracy, 0.777 sensitivity and 0.769 specificity. Three most predictive features, i.e., contrast, entropy and line likeness, were selected by the LASSO method and showed a statistical significance (p<0.05) in the classification. Conclusion The present study showed that quantitative radiomics imaging features of breast tumour extracted from DCE-MRI are associated with breast cancer Ki67 expression. Future larger studies are needed in order to further evaluate the findings. To investigate whether quantitative radiomics features extracted from dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) are associated with Ki67 expression of breast cancer. This institutional review board-approved retrospective study comprised 377 Chinese women who were diagnosed with invasive breast cancer in 2015. This cohort included 53 low-Ki67 expression (Ki67 proliferation index less than 14%) and 324 cases with high-Ki67 expression (Ki67 proliferation index more than 14%). A binary-classification of low-versus high- Ki67 expression was performed. A set of 56 quantitative radiomics features, including morphological, greyscale statistic, and texture features, were extracted from the segmented lesion area. Three machine learning classification methods, including naive Bayes, k-nearest neighbour and support vector machine, were employed for the classification and the least absolute shrink age and selection operator (LASSO) method was used to select most predictive features set for the classifiers. Classification performance was evaluated by the area under receiver operating characteristic curve (AUC), accuracy, sensitivity, and specificity. The model that used naive Bayes classification method achieved the best performance than the other two methods, yielding 0.773 AUC, 0.757 accuracy, 0.777 sensitivity and 0.769 specificity. Three most predictive features, i.e., contrast, entropy and line likeness, were selected by the LASSO method and showed a statistical significance (p<0.05) in the classification. The present study showed that quantitative radiomics imaging features of breast tumour extracted from DCE-MRI are associated with breast cancer Ki67 expression. Future larger studies are needed in order to further evaluate the findings.