Development and validation of machine learning models for predicting HER2-zero and HER2-low breast cancers

Area of Science:

• Radiology and Medical Imaging
• Oncology
• Machine Learning in Medicine

Background:

• Accurate classification of human epidermal growth factor receptor 2 (HER2) status in breast cancer is critical for guiding targeted therapy.
• HER2-low expressing breast cancer, a distinct subtype, may benefit from HER2-targeted treatments, necessitating reliable prediction methods.
• Pre-treatment imaging features from Magnetic Resonance Imaging (MRI) hold potential for non-invasively assessing HER2 expression levels.

Purpose of the Study:

• To develop and validate machine learning models for differentiating HER2-zero and HER2-low breast cancer phenotypes.
• To utilize pre-neoadjuvant therapy (NAT) MRI features for predicting HER2 status.
• To evaluate the models' ability to predict pathological complete response (pCR) and disease-free survival (DFS).

Main Methods:

• Five hundred and sixteen breast cancer patients were included, with data split into training (n=362) and internal validation (n=154) sets.
• Various MRI features (e.g., tumour diameter, enhancement patterns, apparent diffusion coefficient) were analyzed.
• Machine learning algorithms including Logistic Regression (LR), Support Vector Machine (SVM), K-Nearest Neighbour (KNN), and Extreme Gradient Boosting (XGBoost) were employed.

Main Results:

• The XGBoost model demonstrated superior performance, achieving an Area Under the Receiver Operating Characteristic Curve (AUC) of 0.783 in the validation dataset.
• The XGBoost-derived HER2 score showed moderate accuracy in predicting pCR (AUC 0.708 training, 0.695 validation).
• A low HER2 score was significantly associated with shorter DFS in the validation cohort (Hazard Ratio: 2.748, P=.037).

Conclusions:

• The XGBoost model effectively distinguishes between HER2-zero and HER2-low breast cancers using pre-NAT MRI features.
• This model shows potential for predicting pCR and prognosis in breast cancer patients undergoing NAT.
• MRI-based machine learning offers a promising approach for identifying HER2-low breast cancer, guiding treatment decisions.