ABSTRACT
Objective To investigate the accuracy and missed diagnosis rate of preoperative ultrasound in the diagnosis of thyroid carcinoma . Methods With the method of natural language recognition ,816 solitary thyroid nodules which received thyroidectomy were enrolled for the analysis of preoperative ultrasound examination and post-operative pathology results . The accuracy and the rate of missed diagnosis of ultrasonography for the diagnosis of thyroid carcinoma and lymph node metastasis were evaluated with the pathological results as the gold standard . Results There were 783 ( 96 .0％ ) malignant thyroid tumors ,6 ( 0 .7％ ) follicular tumor with uncertain malignant potential ,and 27 ( 3 .3％ ) benign thyroid tumors . Ultrasonography yielded an accuracy of 96 .1％ in the differential diagnosis between benign and malignant thyroid nodules and an accuracy of 94 .5％ in the quantification of nodule numbers . 5 .5％ of solitary nodules reported by ultrasound were proved to be 2 or more malignant nodules after the surgery . Regarding the diagnosis of central lymph node ,the preoperative ultrasound yielded an accuracy of 73 .0％and a misdiagnosis rate of 38 .7％ . For the lateral neck lymph node ,the accuracy was 88 .3％ ,and the misdiagnosis rate was 3 .4％ . Conclusions Natural language recognition method can assist clinical research . Ultrasonography is valuable in the diagnosis of thyroid carcinomas and neck lymph nodes . However , ultrasonography has a certain rate of misdiagnosis for the quantification of nodule numbers ,and for the diagnosis of lymph nodes in central and lateral necks .