Ultrasound Radiomics + ML Predicts Breast Cancer Hormone Receptor Status

A retrospective study (n=353 tumors) found that machine learning analysis of breast ultrasound images could predict estrogen receptor (ER) status with an AUC of 0.70 and accuracy up to 83%, suggesting a potential 'digital biopsy' approach, though external validation is needed.

• Design: Retrospective, single-center study of 311 patients (353 breast tumors) imaged between 2019-2024, using an 80/20 train/test split.
• Finding: For predicting ER status, the ExtraTrees model achieved an AUC of 0.70 and accuracy of 0.83. Prediction of progesterone receptor (PR) status was lower, with an AUC of 0.61.
• Limitation: Single-center, retrospective design without an external validation cohort; models require large-scale prospective study before clinical use.

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