超声AI、弹性应变率及TSH检测在甲状腺结节良恶性鉴别中的应用研究
Application of ultrasonic AI,elastic strain rate and TSH detection in differential diagnosis of benign and malignant thyroid nodules
刘江云 1李利霞 1刘新羽1
作者信息
- 1. 延安市人民医院超声诊断科 陕西 延安 716000
- 折叠
摘要
目的 探讨超声人工智能(AI)、弹性应变率及促甲状腺激素(TSH)检测在甲状腺结节良恶性鉴别中的应用价值.方法 本研究为回顾性研究,选取2022年1月至2024年2月延安市人民医院收治的甲状腺结节患者168例为研究对象,根据临床病理学检查结果分为良性组(n=106)和恶性组(n=62).两组患者均行超声AI、弹性应变率及TSH检测.记录患者的病理检查结果,并以此为金标准,计算超声AI诊断结果、弹性应变率诊断结果、TSH水平诊断结果诊断甲状腺结节良恶性的特异度、灵敏度、误诊率、漏诊率,比较两组患者的TSH水平,采用受试者操作特征(ROC)曲线分析不同检查方法及联合检查的诊断效能.结果 168例甲状腺结节患者中,共有206个甲状腺结节;经病理组织检查,良性组中20例淋巴细胞性甲状腺炎,28例增生性结节,58例甲状腺腺瘤;恶性组中52例甲状腺乳头癌,9例滤泡状癌,1例未分化癌.超声AI的诊断结果与病理检查结果比较,其灵敏度为90.32％,特异度为80.19％,漏诊率为9.68％,误诊率为19.81％.弹性应变率诊断结果与病理检查结果比较,其灵敏度为87.10％,特异度为87.74％,漏诊率为12.90％,误诊率为12.26％.恶性组患者的TSH水平为(2.54±0.52)mIU/L,显著高于良性组[(1.34±0.45)mIU/L],差异有统计学意义(P＜0.05).TSH诊断结果与病理检查结果比较,其灵敏度为88.71％,特异度为89.62％,漏诊率为11.29％,误诊率为10.38％.超声AI、弹性应变率、TSH及其联合诊断甲状腺结节良恶性的曲线下面积(AUC)值为0.809、0.824、0.841、0.903.结论 超声AI、弹性应变率及TSH水平检测对于甲状腺结节良恶性诊断的效能较高,其中联合检测的应用价值更高.
Abstract
Objective To investigate the application value of ultrasound artificial intelligence(AI),elastic strain rate and thyroid stim-ulating hormone(TSH)detection in the identification of benign and malignant thyroid nodules.Methods This study was a retrospective study.A total of 168 patients with thyroid nodules admitted to Yan'an People s Hospital from January 2022 to February 2024 were selected as the research objects.According to the results of clinical pathological examination,they were divided into the benign group(n=106)and the malignant group(n=62).Ultrasound AI,elastic strain rate and TSH were detected between the two groups.The pathological examination results of patients were recorded,and specificity,sensitivity,misdiagnosis rate,missed diagnosis rate of ultrasonic AI diagnosis results,elastic strain rate diagnosis results and TSH level diagnosis results were calculated based on the gold standard.The TSH level of patients in the two groups was compared,and the di-agnostic efficiency of different examination methods and combined examination was analyzed by receiver operating characteristic(ROC)curve.Results There were 206 thyroid nodules in 168 patients.And there were 20 lymphocytic thyroiditis,28 proliferative nodules and 58 thyroid ade-nomas in the benign group and 52 papillary carcinoma,9 follicular carcinoma,and 1 undifferentiated carcinoma in the malignant group.The sensi-tivity,specificity,missed diagnosis rate and misdiagnosis rate of ultrasonic AI compared with the pathological examination were 90.32％,80.19％,9.68％and 19.81％respectively.The sensitivity,specificity,missed diagnosis rate and misdiagnosis rate of elastic scrain rate were 87.10％,87.74％,12.90％and 12.26％respectively.The TSH level of the malignant group was(2.54±0.52)mIU/L,which was significantly higher than that of the benign group[(1.34±0.45)mIU/L],and the difference was statistically significant(P＜0.05).Comparison of TSH diagnosis results and disease test results,the sensitivity was 88.71％,the specificity was 89.62％,the missed diagnosis rate was 11.29％and the misdiag-nosis rate was 10.38％.The area under the curve(AUC)values of benign and malignant thyroid nodules detected by ultrasonic AI,elastic strain rate,TSH and their combination were 0.809,0.824,0.841,0.903.Conclusion Ultrasonic AI,elastic strain rate and TSH detection are more effective in the diagnosis of benign and malignant thyroid nodules,and the combined detection has higher application value.
关键词
人工智能/甲状腺结节/弹性应变率/促甲状腺激素/良恶性Key words
Artificial intelligence/Thyroid nodule/Elastic strain rate/Thyroid stimulating hormone/Benign and malignant引用本文复制引用
出版年
2024
NETL
NSTL
万方数据