Sentinel Lymph Node Metastasis Prediction Based on Primary Breast Cancer US and CEUS Images of Clinical T1 Stage Breast Cancer Patients.

[PURPOSE] This study aimed to develop a preoperative logistic regression model to predict sentinel lymph nodes (SLN) metastasis risk in clinical T1 stage (cT1, diameter ≤ 2 cm) breast cancer patients using ultrasound (US) and contrast-enhanced ultrasound (CEUS) characteristics.

[METHODS] Consecutive patients with primary cT1 breast cancer from June 2018 to May 2024 who have undergone breast CEUS examination and subsequent breast surgeries with SLN biopsies were retrospectively enrolled. Histopathological results following surgical resection were considered the gold standard. The patients were randomly classified into training and validation sets in a 7:3 ratio for the development and validation of the logistic regression, respectively. Univariable analysis and multivariable logistic regression analysis were performed to identify independent indicators of SLN status. We developed Model_1 (solely based on conventional US characteristics) and Model_2 (integrating conventional US and CEUS characteristics) to predict SLN metastasis (present vs. absent) and further the number of metastatic SLN (≤ 2 vs. > 2). The additive prediction effect of CEUS characteristics was also discussed by comparing the predictive performance of Model_1 and Model_2.

[RESULTS] In the final analysis of 383 patients, multivariable analysis identified tumor size, hyperechoic halo, positive axillary nodes, perfusion defect, enhancement order, penetrating vessel, and crab claw-like enhancement as independent indicators of SLN status. In the validation set, for predicting SLN metastasis (present vs. absent), the AUCs of Model_1 and Model_2 were 0.70 and 0.80, respectively. For predicting SLN metastasis (≤ 2 vs. > 2), the AUCs of Model_1 and Model_2 were 0.75 and 0.88, respectively. Both models were well-calibrated, and the addition of CEUS features significantly improved the predictive performance of Model_2 compared to Model_1.

[CONCLUSION] The Model_2, using US and CEUS characteristics from cT1 breast cancer patients, effectively predicts SLN metastasis and the number of metastatic SLNs. This model aids clinicians in assessing SLN metastasis risk and making informed decisions about axillary surgery.