The aim of this study was to develop and validate intratumoral and peritumoral radiomic models based on DCE-MRI to predict the human epidermal growth factor receptor 2 (HER2) low and zero states of breast cancer(BCa). The clinical data of 168 patients with BCa were retrospectively analysed and divided into a training set and a validation set. The intratumoral and peritumoral radiomic features were extracted from DCE-MRI. Variance analysis, univariate logistic analysis, and least absolute shrinkage and selection operator (LASSO) were used as selection operators for dimension reduction. Feature selection and radiomics model construction were performed using max-relevance and min-redundancy (mRMR) and the least absolute shrinkage and selection operator (LASSO) on the training cohort. Logistic regression (LR) was used as a classifier to construct the intratumoral, peritumoral and intratumoral combined peritumoral radiomic models of DCE-MRI. The optimal model was selected to construct the fusion model combined with the screened clinical independent risk factors, and the model was displayed as a nomogram. The performance of each model was evaluated via the area under the curve (AUC) value of the receiver operating characteristic (ROC) curve. In the end, a total of 12 features were retained, and the AUC values of the DCE-MRI intratumoral + peritumoral model in the training set and validation set were 0.845 and 0.836, respectively. Age was identified as an independent risk factor for predicting HER2 status in BCa through Univariate and multivariate analysis in the training set.The AUC values of the final constructed nomogram were 0.862 and 0.844 in the training set and validation set, respectively. Intratumoral and peritumoral radiomic methods based on DCE-MRI have good value in the identification of HER2-low and HER-zero BCa. Not applicable.
Sun et al. (Tue,) studied this question.