Purpose: To examine the diagnostic accuracy and the limitations of bre
ast sonography, sonographic diagnosis was retrospectively compared wit
h pathological findings and mammographic diagnosis. Methods: During a
2-year period, 304 women underwent breast biopsy and a sonographic exa
mination was performed before the surgery. In 202 patients we performe
d mammography in our hospital. These diagnoses were compared with the
pathological findings. In both of the examinations, the diagnoses were
classified into three groups ''cancer'', ''undetermined'' and ''benig
n''. Results: Breast sonography: In a total of 304 patients, cancer wa
s found in 126 cases (41.4%). Among these patients a diagnosis of ''ca
ncer'' was made sonographically in 96 women (76.2%). Twenty cases were
classified as ''undetermined'' and ten cases were diagnosed as ''beni
gn''. In 178 patients with benign histological findings a diagnosis of
''benign'' was made in 120, ''undetermined'' in 45 and a false diagno
sis of ''cancer'' was made in 13 cases (sensitivity: 92.1%, specificit
y: 67.4%). In patients with local recurrence of breast cancer (n = 24)
, no tumor was diagnosed incorrectly as ''benign''. Accuracy of the so
nography in case of breast cancer depends greatly on its size. If the
cancer was larger than 1 cm, 97.5% (79/81) of lesions were correctly d
iagnosed either as ''cancer'' or ''undetermined''. However, if the tum
or was 1 cm in size or smaller, this rate went down to 58.3% (7/12). T
he cell type of breast cancer had no influence on sonographic diagnosi
s. Comparison of breast sonography to mammography: In patients with be
nign histology (n = 114), both of the examinations showed no differenc
es on their accuracy. In case of breast cancer (n = 88), breast sonogr
aphy yields higher diagnostic accuracy than mammography. This is parti
cularly true in women with palpable masses (n = 66): a false negative
diagnosis was made in only one woman (1.5%) with sonography as compare
d to nine women (13.6%) with mammography (p < 0.01). However, this dia
gnostic power was completely different in women with no palpable masse
s (n = 22), in which a false diagnosis ''benign'' was made much more o
ften by sonography (8/22, 36.4%) as compared to mammography (2/22, 9.1
%, p < 0.025). Conclusion: Breast sonography with a ''state-of-the-art
'' equipment can give very useful information for the diagnosis of bre
ast cancer. It has, however, crucial limitations, particularly in case
of small and nonpalpable tumors, and should not be used as a screenin
g method.