Ki-67 Prediction in Breast Cancer: Integrating Radiomics From Automated Breast Volume Scanner and 2D Ultrasound
Wei Wei1,2, Fei Xia3, Wang Zhou1
1Department of Ultrasound, The First Affiliated Hospital of Anhui Medical University, Hefei, Anhui, People's Republic of China.
Breast Cancer (Dove Medical Press)
|October 17, 2025
Summary
A new model accurately predicts breast cancer Ki-67 expression before surgery using radiomics from Automatic Breast Volume Scanners (ABVS) and 2D ultrasound. This aids personalized treatment planning for breast cancer (BC) patients.
Area of Science:
- Radiology
- Oncology
- Medical Imaging
Background:
- Accurate preoperative assessment of breast cancer (BC) biomarkers like Ki-67 is crucial for personalized treatment.
- Traditional methods may lack precision, necessitating advanced imaging techniques.
Purpose of the Study:
- To develop and validate a predictive model for Ki-67 expression in BC using radiomics features.
- To integrate Automatic Breast Volume Scanner (ABVS) and 2D ultrasound data for enhanced preoperative assessment.
- To support personalized clinical treatment planning for BC patients.
Main Methods:
- Retrospective analysis of 426 BC patients' data.
- Extraction of radiomics and habitat radiomics features from ABVS and 2D ultrasound images.
- Development of prediction models using machine learning classifiers (Logistic Regression, ExtraTree, XGBoost, LightGBM).
- Construction of a comprehensive model integrating radiomics, habitat radiomics, and clinical factors (US-ALNs, T-stage).
Main Results:
- The combined radiomics model (Rad-Hab ABVS + 2D) achieved an AUC of 0.850.
- The comprehensive model (CM Clinical + Rad-Hab), integrating clinical factors, demonstrated superior performance with AUCs of 0.951 (training) and 0.884 (validation).
- The comprehensive model showed excellent calibration and clinical utility via ROC, calibration curves, and DCA.
Conclusions:
- The developed comprehensive model accurately predicts Ki-67 expression preoperatively in BC patients.
- This predictive capability facilitates personalized and precise treatment strategies.
- Integration of radiomics and clinical data offers significant potential for improving BC management.
