The purpose of this study is to evaluate the accuracy of gray scale and Doppler US findings in the detection of axillary metastases in breast cancer patients with no palpable lymph nodes. One-hundred and ninety-eight lymph nodes detected in 83 women were evaluated. The size and longitudinal/transverse axis ratios of each node were documented. Absence of echogenic hilum, asymmetrical cortical thickening, and presence of peripheral flow were prospectively considered signs of malignancy. Histopathologically, there were 93 malignant and 105 benign nodes. The above criteria and a low longitudinal-transverse axis ratio were statistically significant for malignancy. In lymph nodes smaller than 1 cm, only asymmetric cortical thickening and presence of peripheral flow were significant. The sensitivity, specificity, positive predictive value, negative predictive value and accuracy of US were 86.49, 93.62, 91.43, 89.8 and 90.48%, respectively. In conclusion, US is successful and reliable in the determination of axillary metastatic involvement in nonpalpable and small lymph nodes. Inclusion of axillary US in the preoperative diagnostic evaluation would be complimentary to sentinel node biopsy, and also could eliminate the need for it in patients with positive US results, after confirmation with biopsy.