Prediction of axillary lymph node metastasis with the ultrasound and Magnetic Resonance Imaging features in breast cancer

Abstract

Purpose The prediction of axillary lymph node metastasis (ALNM) in patient with breast cancer before surgery is of great value. We aim to develop a preoperative nomogram by integrating clinical-pathological variables with ultrasound (US)and magnetic resonance imaging (MRI) features to forecast axillary lymph node metastasis, and to evaluate whether the diagnostic performance of a combined US-MRI model outperforms that of standalone imaging modalities. Method In this retrospective study, 1481 women with breast cancer who underwent surgery were identified from the hospital between November 2009 and April 2022. According to inclusion and exclusion criteria,885women were classified at 6:4 ratio into training and validation set. MRI and US scans before surgery and clinical-pathologic data were reviewed. The prediction models were developed in the training set by using logistic regression and LASSO regression and then tested in the validation set. Result Our training set included 535 women, aged 52 ± 11 years, with 165 cases of axillary lymph node metastases. The validation set comprised 356 women, aged 54 ± 11 years, including 113 cases with metastases. Variables such as the number, size, and location of nodes, morphology, calcification, ADC, pathological patterns, molecular subtypes, and minor-axis dimensions, along with cortical thickness, were significantly associated with an increased risk of axillary lymph node involvement (all P < 0.05). Our predictive model, integrating ultrasound (US) and MRI-based clinical-pathological features (CPUM), showed better performance (AUC = 0.795 for ALNM) in predicting axillary lymph node metastases than the individual models based solely on US features (CPU) (AUC = 0.766 for ALNM; P = 0.0192) or MRI features (CPM) (AUC = 0.760 for ALNM; P = 0.0088) in the validation set. Conclusion The preoperative nomogram, in combination with clinical-pathologic variables and US and MRI features, demonstrated superior predictive performance for axillary lymph node metastasis in patients with breast cancer compared to the use of US or MRI features alone.