Dual-Modality Virtual Biopsy System Integrating MRI and MG for Noninvasive Predicting HER2 Status in Breast Cancer
Qian Wang1, Zi-Qian Zhang2, Can-Can Huang1
1Department of Radiology, The Affiliated Huai'an Clinical College of Xuzhou Medical University, Huai'an, Jiangsu Province, China (Q.W., C.-C.H., H.-W.X., G.-J.B.).
Academic Radiology
|March 11, 2025
Summary
A novel deep learning system (DM-VBS) accurately predicts human epidermal growth factor receptor 2 (HER2) status in breast cancer using MRI and mammography. This tool aids clinicians in making targeted therapy decisions for breast cancer patients.
Area of Science:
- Oncology
- Radiology
- Artificial Intelligence
Background:
- Accurate human epidermal growth factor receptor 2 (HER2) expression determination is crucial for guiding breast cancer targeted therapy.
- Current methods for HER2 assessment can be invasive and time-consuming.
- There is a need for non-invasive methods to predict HER2 status.
Purpose of the Study:
- To develop and validate a deep learning (DL)-based decision-making visual biomarker system (DM-VBS).
- To predict HER2 status using radiomics and DL features from MRI and mammography.
- To assist clinicians in making treatment decisions for breast cancer.
Main Methods:
- Radiomics features were extracted from MRI, and DL features were derived from mammography.
- Four submodels were constructed for distinguishing HER2-zero/low from HER2-positive cases and differentiating HER2-zero from HER2-low/positive cases.
- Submodels were integrated into an XGBoost model for ternary classification of HER2 status and validated on independent datasets.
Main Results:
- The DM-VBS achieved high performance in predicting HER2 status, with AUC values ranging from 0.793 to 0.850.
- Average accuracies for HER2-zero, -low, and -positive patients in validation cohorts were 85.42%, 80.4%, and 89.68%, respectively.
- Significant imaging features associated with HER2 status included lesion size, number of lesions, enhancement type, and microcalcifications.
Conclusions:
- The developed DM-VBS demonstrates high accuracy in predicting HER2 status from medical imaging.
- This AI-powered tool can assist clinicians in making informed treatment decisions for breast cancer.
- The system offers a non-invasive approach to HER2 status assessment, potentially improving patient management.
