人工智能算法在不同仪器游离甲状腺素检测结果可比性中的应用  

Application of artificial intelligent algorithm for mutual recognition of free thyroxine determination results among different instruments

作　　者：周鑫 刘在栓 武宇翔 陆小琴 武永康 ZHOU Xin;LIU Zaishuan;WU Yuxiang;LU Xiaoqin;WU Yongkang(Department of Clinical Laboratory Medicine,West China Hospital,Sichuan University,Chengdu 610041,Sichuan,China;The Secon People's Hospital of Yibin,Yibin 644000,Sichuan,China;UWE College of Hainan Medical University,Haikou 571199,Hainan,China;Jintang First People's Hospital,Chengdu 610499,Sichuan,China)

机构地区：[1]四川大学华西医院实验医学科,四川成都610041 [2]宜宾市第二人民医院,四川宜宾644000 [3]海南医科大学西英格兰学院,海南海口571199 [4]金堂县第一人民医院,四川成都610499

出　　处：《检验医学》2025年第3期299-305,共7页Laboratory Medicine

基　　金：四川省科技厅基金项目(2024YFFK0131);四川大学华西医院学科卓越发展1·3·5工程基金项目(2020HXFH038);四川省科技厅基金项目(24ZDYF1386)。

摘　　要：目的为解决不同仪器游离甲状腺素(FT4)检测结果的显著差异,实现不同医疗机构检验结果互认,设计一种结果转换算法,使各仪器间的检测数据可以相互转换。方法使用确定不同仪器间比对样本数量的算法分别计算2套化学发光检测系统在其检测线性范围内FT4的样本比对个数和浓度,进而建立2套系统间的对应转换区间关系。通过开发数学转换的算法,实现不同仪器间检测结果的相互转换。在其中1套系统的标准曲线上选取20个浓度均匀分布在其检测线性范围内的样本,采用美国临床实验室标准化协会(CLSI)EP15-A2文件比对规则和Passing-Bablok回归分析对2套系统的检测结果进行直接比对和结果转换后的比对。结果2套系统比对所需最少样本数量为8例,直接比对2套系统FT4检测结果的平均百分偏差为38.97%,Passing-Bablok回归分析结果显示,2套系统斜率为0.491;通过转换算法进行结果转换后,FT4的平均百分偏差为9.13%,Passing-Bablok回归分析结果显示斜率为0.955。在临床诊断中,通过算法转换后,转换结果的诊断正确率为95%。结论2套系统FT4检测结果存在着显著的差异,这一问题可以通过智能算法得到解决。该算法能够实现在不同参考区间、不同检测线性范围内,对2套系统的检测结果进行互相转换,使2套系统的检测结果具有可比性。Objective In order to solve the significant difference of free thyroxine(FT4)determination results among different instruments and realize the mutual recognition of determination results among different medical institutions,a result conversion algorithm was designed in this study,and the determination data among different instruments can be converted to each other.Methods Using the algorithm to determine the number of comparison samples among different instruments,the comparison number and concentration of FT4 samples of 2 chemiluminescence instruments were calculated respectively in their determination linear ranges,and then the corresponding conversion interval relationship between the 2 instruments was established.Through the development of mathematical conversion algorithm,the conversion of determination results among different instruments was realized.Totally,20 samples with uniformly distributed concentrations within the linear range were selected on the standard curve of one instrument.The Clinical and Laboratory Standards Institute(CLSI)EP15-A2 document comparison rule and Passing-Bablok regression analysis were used to compare the results of the 2 instruments directly and after conversion.Results The minimum number of samples required for comparison between the 2 instruments was 8 cases,and the average percentage deviation of FT4 determination results directly compared between the 2 instruments was 38.97%.Passing-Bablok regression analysis showed that the slope was 0.491.The average percentage deviation of FT4 after conversion by conversion algorithm was 9.13%,and the slope was 0.955.In clinical diagnosis,after algorithm conversion,the diagnostic accuracy of conversion results was 95%.Conclusions There are differences between the 2 instruments.This problem can be solved by the intelligent algorithm.The algorithm can convert the determination results of the 2 instruments in different reference intervals and different linear ranges,so that the determination results of the 2 instruments are comparable.

关 键 词：检验结果互认 医学检验 智能算法 游离甲状腺激素 

分 类 号：R446.1[医药卫生—诊断学]