Breast cancer is a serious health concern, and predicting its response to neoadjuvant therapy is crucial for personalized treatment planning. This study aims to develop a predictive model by integrating radiomic and deep learning features from different phases of dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI). The model, which combines traditional radiomics and 3D deep learning, can accurately predict pathological complete response (pCR) to neoadjuvant therapy in breast cancer patients.
The study utilized a dataset of 234 breast cancer patients who underwent neoadjuvant therapy. Traditional radiomics and 3D deep learning features were extracted from early and peak phases of DCE-MRI, and the top ten discriminative features were selected for model training. The results showed that the combined model integrating early and peak phases of DCE-MRI provided the best pCR prediction. The performance was further enhanced by adding 3D deep learning features.
The optimal model, which integrated radiomics and deep learning features from early and peak phases of DCE-MRI, achieved high accuracy and AUC values. The DeLong test confirmed the statistical significance of the model's performance. SHAP analysis revealed that radiomics texture features contributed the most to the model.
This study highlights the importance of multi-phase imaging and diverse features in improving predictive accuracy. The constructed model has the potential to guide personalized treatment strategies for breast cancer patients, offering a promising approach to precision medicine.
