Abstract: Objective 　To evaluate the accuracy and feasibility of intraoperative frozen section diagnosis in nonpalpable breast lesions (NPBLs) . Methods 　Diagnoses on frozen sections were performed on 158 consecutive wire2localized breast biop sies with mammographically detected nonpalpable breast lesions. Every specimen was diagnosed by introperative frozen resection. The initial frozen section diagnoses were compared with the diagnoses obtained on permanent paraffin sections to estimate it s accuracy. Results 　On the surgical specimen examination of the 158 NPBLs, macroscopically abnormal lesions could be found only in 80 (50. 6 %) cases, with a mean diameter of 1. 2cm. There were 15 invasive breast cancers, 15 microinvasive carcinomas, 12 carcinoma in situ and 5 atypical ductal hyperplasia altogether, which was 29. 7 %(47/158) of the all NPBLs, according to the permanent paraffin sections. The accuracy of frozen section diagnoses on invasive breast cancer, microinvasive carcinoma, carcinoma in situ and atypical ductal hyperplasia were 93. 3 %, 60 %, 58. 3 % and 60 %, respectively. The errors of frozen sections were false negative diagnosis and underestimation, while no false positive diagnosis and overestimation were found. Errors in frozen slicing and interpretation were main reasons of the false negative diagnosis and underestimation. Conclusion 　Frozen section is feasible in aiding decision-making for operation in the invasive breast cancer due to it s high accuracy. But the operation selection for microinvasive carcinoma, carcinoma in situ and atypical ductal hyperplasia should delay to the permanent paraffin sections diagnosis because of the f requent false negative and underestimation in the f rozen sections diagnosis.