BACKGROUND: We performed a retrospective study to establish the optimal radiological criteria for axillary lymph node metastases from breast cancer by measuring all dissected nodes, and to determine whether magnetic resonance imaging (MRI) could reliably reveal axillary involvement. METHODS: Pathological findings and MRI scans of 202 patients with invasive breast cancer were reviewed. The long- and short-axis dimensions of all level I an II lymph nodes were measured microscopically, and then the long-to-short axis (L/S) ratio of each node was calculated. These parameters were compared with pathological nodal status to define radiological criteria for axillary involvement. MRI was carried out using T1-weighted spin-eho sequences in the coronal and sagittal planes. On MRI, every detected lymph node was measured and the shape ofthe nodal cortex was also examined. Then the diagnostic ability of MRI was assessed using these morphologic criteria. RESULTS: On histopathological examinations of 4043 dissected lymph nodes, a long-axis dimension of 10 mm or larger combined with a long-to-short axis ratio of less than 1.6 was the most accurate criteria for predicting lymph node metastases. On MRI, eccentric cortical hypertrophy was seen in only metastatic axillae.When these morphologic features were used as criteria for malignancy, MRI had asensi-tivity of 79%, a specificity of 93%, and an accuracy of 88%. In 16 of 17 false-negative axillae, MRI showed normally sized lymph nodes (< 10 mm). CONCLUSION: Our study indicates that MRI is a useful diagnostic method for the evaluation of axillary nodal status, but is limited in the detection of small metastatic lymph nodes.