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

doi:10.3969/j.issn.1673-8640.2025.03.017

Abstract

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

CLC Number:

R446.1

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.

Figures/Tables 7

References 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