QUANTITATIVE ultrasound enables accurate prediction of three-year survival in locally advanced breast cancer, offering a non-invasive biomarker to guide prognosis and treatment planning before therapy begins.
Quantitative Ultrasound in Breast Cancer Prognosis
Locally advanced breast cancer remains associated with poor outcomes despite advances in systemic therapy. As many cases are not suitable for immediate surgery, current practice supports the use of aggressive neoadjuvant chemotherapy. Conventional ultrasound imaging provides information on tissue echogenicity but is limited by variability in hardware and acquisition settings, making comparison across scans challenging.
Quantitative ultrasound addresses this limitation by using normalised power spectra to generate parameters that are independent of imaging systems. In this study, researchers developed a deep learning pipeline to evaluate whether quantitative ultrasound could predict survival outcomes prior to treatment initiation.
Deep Learning Pipeline and Predictive Performance
The integrated pipeline combined data scaling, oversampling, feature selection, and classification within a single framework, designed to reduce the risk of data leakage and improve analytical efficiency. The model was trained using five quantitative ultrasound maps obtained before treatment in patients scheduled to receive neoadjuvant chemotherapy.
Among the evaluated parameters, average acoustic concentration emerged as the most predictive feature. For identifying patients in the survivor group, the model achieved a recall of 95% and a precision of 91%. These findings indicate strong discriminatory performance in distinguishing between patients likely to survive and those at higher risk of poor outcomes over a three-year period.
Implications For Treatment Planning
The ability to predict survival before initiating treatment may have important clinical implications. Quantitative ultrasound offers a non-invasive method for risk stratification, potentially supporting more personalised treatment strategies and informed decision making.
Early identification of patients with a lower likelihood of survival could guide adjustments in therapeutic intensity or prompt consideration of alternative approaches. Furthermore, integrating such predictive tools into clinical workflows may enhance discussions between clinicians and patients with regard to expected outcomes and treatment goals.
Overall, the findings suggest that quantitative ultrasound, supported by deep learning, has potential as a clinically relevant biomarker in locally advanced breast cancer, enabling earlier and more precise prognostic assessment.
Reference
Falou O et al. Machine learning for the prediction of three-year survival in locally advanced breast cancer patients receiving neoadjuvant chemotherapy using quantitative ultrasound imaging. Sci Rep. 2026; DOI:10.1038/s41598-026-51550-7.
Featured image: ALAUDDIN on Adobe Stock.
