Objectives To determine the accuracy of MRI in detecting axillary lymph nodes (ALNs) metastases preoperatively and to define predictive characteristics of ALN involvement in patients with invasive breast cancer.

Methods Breast MR (3 Tesla) examinations of 169 patients with invasive breast cancer were reviewed at Hôtel-Dieu de France Hospital. Morphological parameters in addition to apparent diffusion coefficient (ADC) value were compared with pathological nodal status.

Results The sensitivity and specificity of MRI in detecting ALN involvement were 87.5% and 55.6% respectively. The negative and positive predictive value of MRI was 81.64% and 66.34% respectively. The mean size of metastatic ALN was larger than that of negative ALN (13.9 mm vs. 10.9 mm, p = 0.000). ALNs larger than 12 mm were associated with higher risk of metastases (p = 0.000). The asymmetry of size between ipsilateral and contralateral ALNs was more significant in positive ALNs on pathology (p= 0.008 vs. 0.043). In a univariate analysis, the round shape of ALN, loss of fatty hilum, irregular contours and hypo-intensity/heterogeneous intensity on T2-weighted sequence were significantly predictive of lymph node metastasis (p = 0.000 for the four characteristics). In a multivariate analysis, only the round shape of lymph node and the hypo-intensity/heterogeneous intensity on T2-weighted sequence were significantly associated with lymph node metastasis (p=0.01 and p=0,018 respectively). The ADC value of ALN did not aid the differentiation between benign and metastatic lymph nodes (p= 0.862).

Conclusions Conventional MRI using the ALN shape and the signal intensity in T2-weighted sequences can evaluate the axilla with high sensitivity.