Currently, breast cancer patients without clinically suspicious lymph nodes are candidates for sentinel lymph node procedures (SLNPs). The aims of this study were to investigate whether preoperative axillary ultrasonography and fine-needle aspiration cytology (FNA) can reduce the number of the more time-consuming SLNPs, and to identify a subset of quantitative nodal features to predict metastatic involvement. 268 axillae were ultrasonographically examined. FNA was performed on suspicious nodes (smallest diameter > or =5 mm or atypical cortex appearance). SLNP was omitted if a tumour-positive node was found on FNA. Length, width, maximum cortex thickness and appearance of cortex and hilus were ultrasonographically established. In 93 axillae (35%), at least one node was detected with ultrasound. FNA was performed once per axilla on 66 nodes; 37 (56%) contained tumour cells. 31% of all tumour-positive axillae (macro-+micrometastases) was found by ultrasound and FNA (37/121). 41% of all axillae containing macrometastases was found by ultrasound and FNA (36/87). SLNPs were reduced by 14% (37/268). Maximum cortex thickness is the main feature to predict metastatic involvement (area under Receiver Operating Characteristic (ROC) curve (A(Z))=0.87).