MRI deep learning model predicts axillary lymph node status to guide breast cancer surgery de-escalation

BCALN-Net, an MRI-based hierarchical multitask deep learning model, predicts sentinel and non-sentinel lymph node status in 6,271 breast cancer patients, outperforming clinical criteria for axillary surgery de-escalation (pooled n=4,081; exact AUC not reported in source).

• Retrospective multicenter development cohort of 6,271 breast cancer patients; model simultaneously predicts sentinel lymph node (SLN) metastasis, SLN metastatic burden, and non-SLN metastasis in a single pipeline.
• In pooled analysis of 4,081 patients, BCALN-Net showed superior performance over established clinical criteria for identifying patients who could safely omit axillary invasive procedures; specific AUC/sensitivity values not reported in the source abstract.
• Key limitation: external prospective validation not yet reported; generalizability across different MRI protocols and non-Asian populations remains unconfirmed.

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