Abstract: Background　S-Detect is a new computer aided diagnostic technique，which can automatically analyze the ultrasound images qualitatively and quantitatively，helping doctors to identify benign and malignant thyroid modules，but there is no reliable evidence-based practice to verify its differential diagnostic value. Objective　To evaluate the value of S-Detect in differential diagnosis between benign and malignant thyroid nodules. Methods　We did a literature review of PubMed，EMBase，Web of Science，the Cochrane Library，Wanfang Data，CNKI，VIP and SinoMed databases to identify studies published as of January 6，2021 in which the value of S-Detect in differentially diagnosing thyroid nodules was investigated. Two reviewers independently screened the literature，extracted data，and evaluated the risk of bias and quality of the included studies. Meta-analysis was conducted using Meta-Disc 1.4 and Stata 15. The pooled indicators include：sensitivity and specificity，positive likelihood ratio，negative likelihood ratio，diagnostic odds ratio and area under the curve（AUC）. Results　Sixteen studies were included，in which the reported pooled sensitivity，specificity，positive likelihood ratio，negative likelihood ratio，diagnostic odds ratio and AUC of S-Detect in differential diagnosis between benign and malignant thyroid nodules were 0.84〔95%CI（0.81，0.86），P=0.003 6〕，0.71〔95%CI（0.69，0.73），P<0.000 1〕，3.31〔95%CI（2.45，4.47），P<0.000 1〕，0.22〔95%CI（0.17，0.29），P=0.000 6〕，15.93〔95%CI（9.85，25.78），P<0.000 1〕，and 0.89〔95%CI（0.84，0.94）〕，respectively. Conclusion　The current evidence shows that S-Detect significantly contributes to distinguishing benign and malignant thyroid nodules，which may be used as an effective auxiliary tool for usual ultrasound examination.

Key words: Thyroid nodule；Ultrasonography；S-Detect；Diagnosis, computer-assisted；Meta-analysis

摘要： 背景　S-Detect是一种新兴的计算机辅助诊断技术，可以实现超声图像定性及定量自动分析，为医师鉴别甲状腺结节的良恶性提供参考，但目前尚缺乏可靠的循证医学依据验证其鉴别诊断价值。目的　评估S-Detect对甲状腺结节良恶性的鉴别诊断价值。方法　计算机检索PubMed、EMBase、Web of Science、the Cochrane Library、万方数据知识服务平台、中国知网、维普网和中国生物医学文献服务系统，搜集有关S-Detect鉴别诊断甲状腺结节良恶性的诊断准确性试验，检索时限为建库至2021-01-06。由2名评价者独立筛选文献、提取资料并评价纳入文献的偏倚风险和质量，采用Meta-Disc 1.4和Stata 15软件进行Meta分析，合并指标包括：灵敏度、特异度、阳性似然比、阴性似然比、诊断比值比和受试者工作特征曲线下面积（AUC）。结果　共纳入16篇文献。Meta分析结果显示，S-Detect鉴别诊断甲状腺结节良恶性的合并灵敏度、合并特异度、合并阳性似然比、合并阴性似然比、合并诊断比值比、AUC分别为0.84〔95%CI（0.81，0.86），P=0.003 6〕、0.71〔95%CI（0.69，0.73），P<0.000 1〕、3.31〔95%CI（2.45，4.47），P<0.000 1〕、0.22〔95%CI（0.17，0.29），P=0.000 6〕、15.93〔95%CI（9.85，25.78），P<0.000 1〕、0.89〔95%CI（0.84，0.94）〕。结论　当前证据显示，S-Detect对甲状腺结节良恶性的鉴别诊断价值较高，可以作为甲状腺常规彩超检查的有效补充手段。

关键词: 甲状腺结节；超声检查；S-Detect；诊断, 计算机辅助；Meta分析