GBDT radiomics model with DCE-MRI predicts breast cancer molecular subtypes

A DCE-MRI radiomics model using gradient boosting decision tree (GBDT) and feature selection showed high predictive performance for breast cancer molecular subtyping.

• The GBDT model with rigorous feature selection outperformed RF, SVM, and LR classifiers for predicting five molecular subtypes.
• DCE-MRI radiomics provides a noninvasive alternative to histopathology for preoperative molecular classification.
• Manual ROI delineation and feature extraction were performed on DCE-MRI images before model building.

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.