Abstract:
Objective
To evaluate the predictive value of intratumoral and peritumoral ultrasound imaging features for human epidermal growth factor receptor 2 (HER2) expression status in breast cancer patients.
Methods
A retrospective analysis was conducted on data from 403 breast cancer patients diagnosed at the Third People's Hospital of Chengdu between April 2021 and March 2024.The patients were randomly divided into a training set of 282 patients and a validation set of 121 patients.For conventional ultrasound, the maximum longitudinal axis of the tumor was used to delineate the intratumoral region of interest (ROI).A 3 mm extension outward from the tumor boundary was used to define the peritumoral ROI.Radiomics features were extracted from these ROIs.In the training set, features were reduced using the Mann-Whitney U test, Pearson correlation analysis,and the Least Absolute Shrinkage and Selection Operator (LASSO).A support vector machine (SVM) classifier was used to construct a radiomics model and calculate the radiomics score.Univariate and multivariate analyses were used to compare clinical pathological indicators between HER2-positive and HER2-negative groups to identify independent clinical risk factors and construct a clinical model.Radiomics scores and independent clinical risk factors were combined into a multivariate logistic regression model to create a combined model.The performance and clinical utility of the models were evaluated using ROC curve, calibration curve, and decision curve analyses.
Results
The area under the curve (AUC) values for the intratumoral model, peritumoral model, combined intratumoral and peritumoral model, clinical model, and combined model in the training set were 0.791, 0.786, 0.848,0.749, and 0.861, respectively.In the validation set, the corresponding AUC values were 0.733, 0.739, 0.839, 0.704,and 0.845, respectively.Decision curve analysis showed that the combined model had the highest clinical utility,and the calibration curve indicated that the combined model was close to the ideal curve.The Hosmer-Lemeshow test showed good goodness of fit for the combined model (P=0.209).
Conclusion
The ultrasound radiomics models considering both intratumoral and peritumoral features can non-invasively predict HER2 expression status in breast cancer patients before surgery.The combined model, which incorporates independent clinical risk factors,demonstrates higher predictive efficacy and has clinical application value.
Key words:
Ultrasound; Radiomics; Breast cancer; Human epidermal growth factor receptor
Quanyu Nie, Mingxing Li, Ying Liu, Hui Wang, Yuanyuan Chen, Hong Zhou, Yang Zhou. Ultrasound-based intratumoral and peritumoral radiomics for predicting HER2 status in patients with breast cancer[J]. Chinese Journal of Medical Ultrasound (Electronic Edition), 2024, 21(12): 1103-1110.