人工智能B超影像计算机辅助诊断系统对成人甲状腺肿诊断价值的研究  

The diagnostic value of artificial intelligence B-ultrasound image computer-aided diagnosis system in adult goiter

作　　者：张泽旭 岳宗育 解鸿蕾 苏越 潘皓雯 李佳 车文静 侯鑫 赵蒙 刘岚春 李丹丹[2] 许娴[3] 李卫东[3] 孟凡刚 范丽珺 刘丽香 李明 刘鹏 Zhang Zexu;Yue Zongyu;Xie Honglei;Su Yue;Pan Haowen;Li Jia;Che Wenjing;Hou Xin;Zhao Meng;Liu Lanchun;Li Dandan;Xu Xian;Li Weidong;Meng Fangang;Fan Lijun;Liu Lixiang;Li Ming;Liu Peng(Institute for Iodine Deficiency Disorders Control,Center for Endemic Disease Control,Chinese Center for Disease Control and Prevention,Harbin Medical University,NHC Key Laboratory of Etiology and Epidemiology,Heilongjiang Provincial Key Laboratory of Trace Elements and Human Health,Key Laboratory of Etiology and Epidemiology,Education Bureau of Heilongjiang Province,Harbin 150081,China;Harbin Institute of Technology,Harbin 150001,China;Anhui Provincial Center for Disease Control and Prevention,Hefei 230601,China)

机构地区：[1]哈尔滨医科大学中国疾病预防控制中心地方病控制中心碘缺乏病防治研究所,国家卫生健康委病因流行病学重点实验室,黑龙江省微量元素与人类健康重点实验室,黑龙江省普通高校病因流行病学重点实验室,哈尔滨150081 [2]哈尔滨工业大学,哈尔滨150001 [3]安徽省疾病预防控制中心,合肥230601

出　　处：《中华地方病学杂志》2024年第11期922-927,共6页Chinese Journal of Endemiology

基　　金：黑龙江省自然科学基金(LH2021H015);国家重点研发项目(2023YFC2508301)

摘　　要：目的探讨人工智能B超影像计算机辅助诊断系统(以下简称智能超声系统)对成人甲状腺肿的诊断价值。方法2022年6月和2023年3月,分两期在安徽省4个市开展甲状腺疾病调查,每个市选取1个村,每个村选取250名成人作为调查对象。分别使用智能超声系统和常规超声扫查设备对成人双侧甲状腺区域进行扫描,以常规超声检查结果为标准,分析智能超声系统诊断甲状腺肿的效能。绘制受试者工作特征(ROC)曲线,采用Kappa检验分析智能超声系统与常规超声检查对甲状腺肿诊断结果的一致性;同时,采用Spearman相关性分析和Bland-Altman法评估两种方法对甲状腺容积测量的一致性。结果经筛选剔除异常值和缺失值后,共纳入910名成人,其中男性253名(27.80%)、女性657名(72.20%);年龄为(45.92±10.20)岁,范围为18~60岁。智能超声系统诊断成人甲状腺肿的灵敏度为80.00%、特异度为99.67%、准确度为99.56%,ROC曲线下面积(AUC)为0.996,与常规超声检查诊断甲状腺肿的结果一致性好(κ=0.67,P<0.001)。控制性别、甲状腺功能、甲状腺结节变量后,智能超声系统与常规超声检查对女性、甲状腺功能异常成人及无甲状腺结节成人的甲状腺肿诊断结果一致性好(κ=0.66、0.80、0.80,均P<0.001),对有甲状腺结节成人甲状腺肿的诊断结果一致性中等(κ=0.56,P<0.001)。Spearman相关性分析显示,智能超声系统与常规超声检查对成人甲状腺容积的测量结果呈高度正相关(r=0.88,P<0.001)。Bland-Altman法结果显示,成人中仅4.62%(42/910)的点在95%一致性界限以外,智能超声系统与常规超声检查对成人甲状腺容积测量的一致性好(<5%);男性、甲状腺功能异常成人、有甲状腺结节成人的95%一致性界限以外的点比例分别为6.72%(17/253)、5.83%(12/206)、6.45%(12/186)。结论智能超声系统对成人甲状腺肿具有较好的诊断价值,与常规超声检查对甲状腺容积测量结果Objective To study the diagnostic value of artificial intelligence B-ultrasound image computer-aided diagnosis system(hereinafter referred to as intelligent ultrasound system)in adult goiter.Methods In June 2022 and March 2023,two phases of thyroid disease survey were carried out in 4 cities in Anhui Province.One village was selected in each city,and 250 adults were selected as survey subjects in each village.Adult bilateral thyroid area was scanned by both intelligent ultrasound system and conventional ultrasound scanning equipment,and the effectiveness of intelligent ultrasound system in the diagnosis of goiter was analyzed based on the results of conventional ultrasound examination.Receiver operating characteristic(ROC)curve was drawn,and Kappa test was used to analyze the consistency between intelligent ultrasound system and conventional ultrasound examination in the diagnosis of goiter.At the same time,Spearman correlation analysis and Bland-Altman method were used to evaluate the consistency of the two methods in measuring thyroid volume.Results After screening and removing outliers and missing values,a total of 910 adults were included,including 253 males(27.80%)and 657 females(72.20%).The age was(45.92±10.20)years old,ranging from 18 to 60 years old.The sensitivity,specificity,and accuracy of the intelligent ultrasound system for diagnosing adult goiter were 80.00%,99.67%,and 99.56%,respectively.The area under the ROC curve(AUC)was 0.996,which was consistent with the results of conventional ultrasound examination for diagnosing goiter(κ=0.67,P<0.001).After controlling for variables such as gender,thyroid function,and thyroid nodules,the intelligent ultrasound system showed good consistency with conventional ultrasound examination in the diagnosis of goiter in females,adults with thyroid dysfunction,and adults without thyroid nodules(κ=0.66,0.80,0.80,P<0.001).The consistency in the diagnosis of goiter in adults with thyroid nodules was moderate(κ=0.56,P<0.001).Spearman correlation analysis showed a hig

关 键 词：甲状腺肿 超声检查 人工智能 诊断效能 

分 类 号：R581[医药卫生—内分泌]