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

doi:10.3969/j.issn.1673-8640.2025.03.017

检验医学 ›› 2025, Vol. 40 ›› Issue (3): 299-305.DOI: 10.3969/j.issn.1673-8640.2025.03.017

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

周鑫¹, 刘在栓², 武宇翔³, 陆小琴¹, 武永康⁴()

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

收稿日期:2024-02-07 修回日期:2024-05-31 出版日期:2025-03-30 发布日期:2025-04-10
通讯作者: 武永康，E-mail：vipwyk@163.com。
作者简介:周鑫，男，1998年生，学士，主要从事智慧检验和互联网医院研究。
基金资助:
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)

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

ZHOU Xin¹, LIU Zaishuan², WU Yuxiang³, LU Xiaoqin¹, WU Yongkang⁴()

1. Department of Clinical Laboratory Medicine，West China Hospital，Sichuan University，Chengdu 610041，Sichuan，China
2. The Secon People's Hospital of Yibin，Yibin 644000，Sichuan，China
3. UWE College of Hainan Medical University，Haikou 571199，Hainan，China
4. Jintang First People's Hospital，Chengdu 610499，Sichuan，China

Received:2024-02-07 Revised:2024-05-31 Online:2025-03-30 Published:2025-04-10

摘要/Abstract

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。

关键词: 检验结果互认, 医学检验, 智能算法, 游离甲状腺激素

Abstract:

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.

Key words: Determination result mutual recognition, Medical laboratory, Intelligent algorithm, Free thyroxine

中图分类号:

R446.1

周鑫, 刘在栓, 武宇翔, 陆小琴, 武永康. 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[J]. 检验医学, 2025, 40(3): 299-305.

ZHOU Xin, LIU Zaishuan, WU Yuxiang, LU Xiaoqin, WU Yongkang. Application of artificial intelligent algorithm for mutual recognition of free thyroxine determination results among different instruments[J]. Laboratory Medicine, 2025, 40(3): 299-305.

图/表 7

参考文献 21

[1]	王金金, 徐黎明, 俞宛君, 等. 医学检验结果同质化管理发展现状与趋势[J]. 中华预防医学杂志, 2023, 57(9):1504-1509.
[2]	ZHENG Z, YAO Z, WU K, et al. The diagnosis of SARS-CoV2 pneumonia:a review of laboratory and radiological testing results[J]. J Med Virol, 2020, 92(11):2420-2428.
[3]	BOZKAYA G, SISMAN A R. The comparison of analytical performances of Mindray CL-1000i and Beckman Coulter Access Ⅱ Troponin Ⅰ methods in the light of recent guidelines and the quality requirements[J]. Ann Transl Med, 2020, 8(19):1237.
[4]	王延群, 曹源, 胡成进. 不同型号血凝仪检测凝血酶原时间及部分凝血活酶时间的可比性评价[J]. 中华临床实验室管理电子杂志, 2015, 3(4):240-243.
[5]	李莎, 欧阳能良, 谢晋烨, 等. 一致性比对工作在推进检验结果互认中的应用[J]. 检验医学与临床, 2022, 19(3):382-385.
[6]	THELEN M H, VANSTAPEL F J, KROUPIS C, et al. Flexible scope for ISO 15189 accreditation:a guidance prepared by the European Federation of Clinical Chemistry and Laboratory Medicine(EFLM)Working Group Accreditation and ISO/CEN standards(WG-A/ISO)[J]. Clin Chem Lab Med, 2015, 53(8):1173-1180.
[7]	PANTEGHINI M, BRAGA F, CAMARA J E, et al. Optimizing available tools for achieving result standardization:value added by Joint Committee on Traceability in Laboratory Medicine(JCTLM)[J]. Clin Chem, 2021, 67(12):1590-1605.
[8]	ZHOU X, LIU Z, MA Y, et al. A mathematical algorithm to harmonize measurements for thyroid-stimulating hormone between instruments[J]. Clin Chim Acta, 2023, 548:117513.
[9]	Clinical and Laboratory Standards Institute, et al. User verification of performance for precision and trueness; approved guideline—second edition[S]. EP15-A2，CLSI, 2006.
[10]	HEGEDÜS L, BIANCO A C, JONKLAAS J, et al. Primary hypothyroidism and quality of life[J]. Nat Rev Endocrinol, 2022, 18(4):230-242. DOI PMID
[11]	GENO K A, REED M S, CERVINSKI M A, et al. Evaluation of thyroid function in pregnant women using automated immunoassays[J]. Clin Chem, 2021, 67(5):772-780. DOI PMID
[12]	徐清, 陈焱, 韩元元, 等. 游离甲状腺激素/游离三碘甲状腺原氨酸比值(FT4/FT3)对急性心肌梗死患者院内全因死亡风险的临床预测价值[J/OL]. 中国动脉硬化杂志,(2023-10-18) [2025-01-17]. http://kns.cnki.net/kcms/detail/43.1262.r.20231017.1715.004.html.
[13]	YIN L, TANG Y, CHEN X, et al. Measurement differences between two immunoassay systems for LH and FSH:a comparison of roche cobas e601 vs. abbott architect i2000sr[J]. Clin Lab, 2018, 64(3):295-301.
[14]	ZHANG H, HUANG Y, LIANG K. Comparison of three immunoassays systems for determining serum estradiol[J]. Clin Lab, 2019, 65(4).
[15]	THIENPONT L M, VAN UYTFANGHE K, VAN HOUCKE S, et al. A progress report of the IFCC committee for standardization of thyroid function tests[J]. Eur Thyroid J, 2014, 3(2):109-116. DOI PMID
[16]	DE GRANDE L A C, VAN UYTFANGHE K, REYNDERS D, et al. Standardization of free thyroxine measurements allows the adoption of a more uniform reference interval[J]. Clin Chem, 2017, 63(10):1642-1652. DOI PMID
[17]	MENG F, JONKLAAS J, LEOW M K. Interconversion of plasma free thyroxine values from assay platforms with different reference intervals using linear transformation methods[J]. Biology(Basel), 2021, 10(1):45.
[18]	PADOAN A, CLERICO A, ZANINOTTO M, et al. Percentile transformation and recalibration functions allow harmonization of thyroid-stimulating hormone(TSH)immunoassay results[J]. Clin Chem Lab Med, 2020, 58(10):1663-1672.
[19]	段秀群, 田欢, 龚国富. 两种仪器测定血清C反应蛋白的性能比对分析[J]. 中国医疗设备, 2017, 32(2):87-90.
[20]	AHN S, CHO Y, LEE S G, et al. Performance evaluation of thyroid function tests on a novel chemiluminescent microparticle immunoassay system[J]. J Appl Lab Med, 2021, 6(2):367-377. DOI PMID
[21]	张莹, 唐彦, 谭晶, 等. 重庆市渝中区2019、2020年检验结果互认新鲜血标本比对与分析[J]. 检验医学与临床, 2022, 19(8):1100-1103.

样本编号	雅培i2000SR/pmol·L^-1	发光值	西门子ADVIA/pmol·L^-1	发光值
1	5.15	357 215	6.89	82 565
2	9.85	241 988	13.35	56 830
3	14.59	155 765	24.39	36 932
4	25.93	76 541	39.91	24 991
5	30.71	65 183	52.14	20 116
6	41.02	53 498	81.21	14 131
7	49.92	49 104	94.84	12 549
8	64.35	42 310	134.42	9 752

样本编号	雅培i2000SR/pmol·L^-1	发光值	西门子ADVIA/pmol·L^-1	发光值
1	5.15	357 215	6.89	82 565
2	9.85	241 988	13.35	56 830
3	14.59	155 765	24.39	36 932
4	25.93	76 541	39.91	24 991
5	30.71	65 183	52.14	20 116
6	41.02	53 498	81.21	14 131
7	49.92	49 104	94.84	12 549
8	64.35	42 310	134.42	9 752

样本编号	雅培i2000SR/（pmol·L^-1）	西门子ADVIA/（pmol·L^-1）	百分偏差/%
1	11.04	17.59	37.24
2	13.03	18.59	29.91
3	12.87	17.90	28.10
4	16.07	21.85	26.45
5	14.09	21.58	34.71
6	38.36	88.09	56.45
7	10.56	14.35	26.41
8	44.67	101.49	55.99
9	5.68	8.77	35.23
10	10.00	16.12	37.97
11	46.43	82.51	43.73
12	47.27	94.80	50.14
13	47.84	95.23	49.76
14	25.77	41.63	38.10
15	31.35	60.89	48.51
16	11.46	16.57	30.84
17	10.59	16.37	35.31
18	12.42	19.80	37.27
19	6.68	10.83	38.32
20	24.44	40.02	38.93

样本编号	雅培i2000SR/（pmol·L^-1）	西门子ADVIA/（pmol·L^-1）	百分偏差/%
1	11.04	17.59	37.24
2	13.03	18.59	29.91
3	12.87	17.90	28.10
4	16.07	21.85	26.45
5	14.09	21.58	34.71
6	38.36	88.09	56.45
7	10.56	14.35	26.41
8	44.67	101.49	55.99
9	5.68	8.77	35.23
10	10.00	16.12	37.97
11	46.43	82.51	43.73
12	47.27	94.80	50.14
13	47.84	95.23	49.76
14	25.77	41.63	38.10
15	31.35	60.89	48.51
16	11.46	16.57	30.84
17	10.59	16.37	35.31
18	12.42	19.80	37.27
19	6.68	10.83	38.32
20	24.44	40.02	38.93

样本编号	雅培i2000SR转换后结果/（pmol·L^-1）	西门子ADVIA检测结果/（pmol·L^-1）	百分偏差/%
1	15.45	17.59	12.19
2	18.59	18.59	9.01
3	19.87	17.90	11.02
4	26.79	21.85	22.60
5	22.63	21.58	4.88
6	73.54	88.09	16.52
7	14.72	14.35	2.55
8	87.02	101.49	14.26
9	7.57	8.77	13.67
10	13.62	16.12	15.51
11	89.76	82.51	8.87
12	90.99	94.80	4.02
13	91.81	95.23	3.60
14	39.37	41.63	5.43
15	53.45	60.89	12.22
16	17.40	16.57	4.98
17	15.88	16.37	2.96
18	20.06	19.80	1.29
19	9.28	10.83	14.33
20	38.98	40.02	2.60

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

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 7

参考文献 21

相关文章 15

编辑推荐

Metrics

样本编号	雅培i2000SR 检测结果	西门子ADVIA检测结果	雅培i2000SR 转换后结果
1	-	-	-
2	-	-	-
3	-	-	-
4	-	-	↑
5	-	-	-
6	↑	↑	↑
7	-	-	-
8	↑	↑	↑
9	↓	↓	↓
10	-	-	-
11	↑	↑	↑
12	↑	↑	↑
13	↑	↑	↑
14	↑	↑	↑
15	↑	↑	↑
16	-	-	-
17	-	-	-
18	-	-	-
19	↓	↓	↓
20	↑	↑	↑

样本编号	雅培i2000SR 检测结果	西门子ADVIA检测结果	雅培i2000SR 转换后结果
1	-	-	-
2	-	-	-
3	-	-	-
4	-	-	↑
5	-	-	-
6	↑	↑	↑
7	-	-	-
8	↑	↑	↑
9	↓	↓	↓
10	-	-	-
11	↑	↑	↑
12	↑	↑	↑
13	↑	↑	↑
14	↑	↑	↑
15	↑	↑	↑
16	-	-	-
17	-	-	-
18	-	-	-
19	↓	↓	↓
20	↑	↑	↑

[1]	郭晓俊, 赵晓君, 蒋玲丽. 医学检验能力验证样品均匀性方差检验值远小于1的问题探讨[J]. 检验医学, 2024, 39(1): 75-78.
[2]	白建华, 曾宪飞, 郝晓柯. 区域医学检验中心模式探索与实践[J]. 检验医学, 2022, 37(12): 1109-1112.
[3]	阚丽娟, 李靖丽, 陈大洋, 韩心远, 汪之红, 蔡钦泉, 文启林, 柳震宇, 李明, 张秀明. 区域医学检验中心智慧互联网检验的探索与实践[J]. 检验医学, 2022, 37(12): 1113-1117.
[4]	陈洪卫, 郭海涛, 侯彦强. 后疫情时代区域医学检验中心建设新发展[J]. 检验医学, 2022, 37(12): 1118-1121.
[5]	范维肖, 曾宪飞, 郝晓柯, 徐修礼. 区域医学检验中心在提升基层医疗机构临床微生物检测水平中的作用[J]. 检验医学, 2022, 37(12): 1122-1124.
[6]	陈洪卫, 侯彦强. 区域医学检验中心运行中存在的问题与对策[J]. 检验医学, 2022, 37(12): 1125-1128.
[7]	杨雪, 居漪, 欧元祝, 朱宇清, 徐翀, 蒋玲丽, 王青, 周靖. 上海市检验结果互认定量项目室间质量评价结果分析[J]. 检验医学, 2022, 37(12): 1129-1134.
[8]	刘张玲, 李娟, 蒲然, 汤荣睿. PDCA循环管理在推进生化项目检验结果互认中的应用[J]. 检验医学, 2020, 35(9): 943-945.
[9]	于思远, 吴文娟. COVID-19方舱医院医学检验科建设和管理初探[J]. 检验医学, 2020, 35(5): 492-496.
[10]	郭晓俊, 朱岭峰, 徐翀, 居漪. 传统统计法与稳健统计法评价医学检验实验室能力验证结果比较[J]. 检验医学, 2017, 32(5): 435-439.
[11]	贾汝静, 赵炳南, 张传宝, 王华梁, 黄维纲, 曹实. 医学检验能力验证提供者认可评审不符合项分析[J]. 检验医学, 2016, 31(5): 423-425.
[12]	潘柏申. 积极推进医学检验结果一致性的工作[J]. 检验医学, 2016, 31(4): 343-344.
[13]	王寅, 王庆忠. 上海地区医学检验专业继续医学教育现状和发展的研究[J]. 检验医学, 2015, 30(1): 68-71.
[14]	黄维纲，贾汝静，张健，郭晓俊. 医学检验领域能力验证计划的比较[J]. 检验医学, 2013, 28(10): 937-943.
[15]	张莉;吴炯;郭玮;潘柏申. 医学检验检测系统应用前的性能评价[J]. 检验医学, 2006, 21(06): 560-563.

样本编号	雅培i2000SR 检测结果	西门子ADVIA检测结果	雅培i2000SR 转换后结果
1	-	-	-
2	-	-	-
3	-	-	-
4	-	-	↑
5	-	-	-
6	↑	↑	↑
7	-	-	-
8	↑	↑	↑
9	↓	↓	↓
10	-	-	-
11	↑	↑	↑
12	↑	↑	↑
13	↑	↑	↑
14	↑	↑	↑
15	↑	↑	↑
16	-	-	-
17	-	-	-
18	-	-	-
19	↓	↓	↓
20	↑	↑	↑

样本编号	雅培i2000SR 检测结果	西门子ADVIA检测结果	雅培i2000SR 转换后结果
1	-	-	-
2	-	-	-
3	-	-	-
4	-	-	↑
5	-	-	-
6	↑	↑	↑
7	-	-	-
8	↑	↑	↑
9	↓	↓	↓
10	-	-	-
11	↑	↑	↑
12	↑	↑	↑
13	↑	↑	↑
14	↑	↑	↑
15	↑	↑	↑
16	-	-	-
17	-	-	-
18	-	-	-
19	↓	↓	↓
20	↑	↑	↑

样本编号	雅培i2000SR 检测结果	西门子ADVIA检测结果	雅培i2000SR 转换后结果
1	-	-	-
2	-	-	-
3	-	-	-
4	-	-	↑
5	-	-	-
6	↑	↑	↑
7	-	-	-
8	↑	↑	↑
9	↓	↓	↓
10	-	-	-
11	↑	↑	↑
12	↑	↑	↑
13	↑	↑	↑
14	↑	↑	↑
15	↑	↑	↑
16	-	-	-
17	-	-	-
18	-	-	-
19	↓	↓	↓
20	↑	↑	↑

样本编号	雅培i2000SR 检测结果	西门子ADVIA检测结果	雅培i2000SR 转换后结果
1	-	-	-
2	-	-	-
3	-	-	-
4	-	-	↑
5	-	-	-
6	↑	↑	↑
7	-	-	-
8	↑	↑	↑
9	↓	↓	↓
10	-	-	-
11	↑	↑	↑
12	↑	↑	↑
13	↑	↑	↑
14	↑	↑	↑
15	↑	↑	↑
16	-	-	-
17	-	-	-
18	-	-	-
19	↓	↓	↓
20	↑	↑	↑