人工智能S-Detect技术结合钙化特征对甲状腺结节的诊断价值

Value of artificial intelligent S-Detect technique combined with calcification characteristics in differential diagnosis of thyroid nodules

目的探讨人工智能S-Detect技术结合钙化特征在甲状腺结节良恶性诊断中的价值。方法选取2019年7月至2020年1月于安徽医科大学第二附属医院行甲状腺超声检查并行手术治疗的94例患者(94个甲状腺结节)。采用常规超声和S-Detect技术对94例患者的94个甲状腺结节进行检查。以术后病理结果作为金标准,分析超声医师、S-Detect及S-Detect结合钙化特征对甲状腺结节的诊断效能。结果经手术病理证实,94个甲状腺结节中,良性37个,恶性57个。超声医师诊断甲状腺结节良恶性的敏感度91.2%,特异度91.8%,准确性91.4%;S-Detect诊断甲状腺结节良恶性的敏感度96.4%,特异度81.1%,准确性90.4%;S-Detect结合钙化特征诊断甲状腺结节的敏感度98.2%,特异度81.8%,准确性92.5.%。超声医师、S-Detect技术及S-Detect结合钙化特征诊断甲状腺结节的ROC曲线下面积分别为0.879、0.864、0.890。S-Detect结合钙化特征的诊断效能优于单独应用常规超声和S-Detect技术(Z=2.020,P=0.043;Z=2.231,P=0.026)。结论 S-Detect技术结合钙化特征可提高甲状腺结节的诊断效能,值得临床推广应用。

Objective To assess the value of S-Detect combined with calcification characteristics in the diagnosis of benign and malignant thyroid nodules.Methods Ninety-four patients with 94 thyroid nodules were examined by conventional ultrasonography and S-Detect technique at the Second Affiliated Hospital of Anhui Medical University from July 2019 to February 2020,and the diagnostic efficacy of conventional ultrasound,S-Detect,and S-Detect combined with calcification characteristics were analyzed according to the postoperative pathological results.Results As confirmed by surgical pathology,of 94 thyroid nodules in 94 patients,37 were benign and 57 were malignant.The sensitivity,specificity,and accuracy of conventional ultrasound,S-Detect,and S-Detect combined with calcification characteristics in the differential diagnosis of benign and malignant thyroid nodules were 91.2%,91.8%,and 91.4%,96.4%,81.1%,and 90.4%,and 98.2%,81.8%,and 92.5%,respectively.The areas under the ROC curves of conventional ultrasound,S-Detect,and S-Detect combined with calcification characteristics were 0.879,0.864,and 0.890,respectively,with the value of S-Detect combined with calcification characteristics being significantly higher than those of conventional ultrasound and S-Detect(Z=2.020,P=0.043;Z=2.231,P=0.026).Conclusions The combination of S-Detect and calcification characteristics can improve the diagnostic efficiency for thyroid nodules.