从目标分组到模糊参考区间：甲状腺功能测试的标准化机器学习方法。

IF 2.9 3区医学 Q2 MEDICAL LABORATORY TECHNOLOGY

Clinica Chimica Acta Pub Date : 2025-10-09 DOI:10.1016/j.cca.2025.120635

Semih Fazlı Kayahan , Muhammed Fatih Alaeddinoğlu , Mehmet Şeneş

{"title":"从目标分组到模糊参考区间：甲状腺功能测试的标准化机器学习方法。","authors":"Semih Fazlı Kayahan , Muhammed Fatih Alaeddinoğlu , Mehmet Şeneş","doi":"10.1016/j.cca.2025.120635","DOIUrl":null,"url":null,"abstract":"<div><h3>Background</h3><div>Accurate interpretation of thyroid function tests (TFTs) requires reliable reference intervals (RIs). Indirect methods based on retrospective laboratory data are increasingly used, but current strategies face major limitations, including rigid age cut-offs, inconsistent partitioning, and lack of objective evaluation of subgrouping criteria. To address these challenges, we developed a novel machine learning (ML)-based framework for the objective, data-driven determination of continuous and personalized RIs.</div></div><div><h3>Methods</h3><div>A total of 48,397 records (2019–2021) from individuals aged ≥18 years were retrieved from the hospital laboratory information system. After applying inclusion and exclusion criteria, 9455 individuals constituted the reference sample group. Partitioning was based on age, sex, thyroid-stimulating hormone (TSH), and free thyroxine (fT4). The Elbow method suggested the optimal number of clusters; K-means clustering was used to form subgroups, and the Extra Trees Classifier quantified the relative importance of partitioning criteria. RIs were estimated using the non-parametric percentile method, and Fuzzy C-Means clustering was applied to smooth sharp transitions between groups.</div></div><div><h3>Results</h3><div>Six subgroups were identified, with age as the dominant determinant (feature importance score = 0.96), whereas sex had a negligible effect. Fuzzy clustering generated continuous and personalized RIs. In clinical evaluation, applying fuzzy RIs reclassified 7 % of patients, predominantly shifting diagnoses toward hyperthyroidism.</div></div><div><h3>Conclusions</h3><div>This study offers a universal and adaptable ML-based framework for generating continuous, personalized RIs. This advance toward precision laboratory medicine can enhance diagnostic accuracy, reduce misclassification, and support more patient-centered care.</div></div>","PeriodicalId":10205,"journal":{"name":"Clinica Chimica Acta","volume":"579 ","pages":"Article 120635"},"PeriodicalIF":2.9000,"publicationDate":"2025-10-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"From objective grouping to fuzzy reference intervals: A standardized machine learning approach for thyroid function tests\",\"authors\":\"Semih Fazlı Kayahan , Muhammed Fatih Alaeddinoğlu , Mehmet Şeneş\",\"doi\":\"10.1016/j.cca.2025.120635\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<div><h3>Background</h3><div>Accurate interpretation of thyroid function tests (TFTs) requires reliable reference intervals (RIs). Indirect methods based on retrospective laboratory data are increasingly used, but current strategies face major limitations, including rigid age cut-offs, inconsistent partitioning, and lack of objective evaluation of subgrouping criteria. To address these challenges, we developed a novel machine learning (ML)-based framework for the objective, data-driven determination of continuous and personalized RIs.</div></div><div><h3>Methods</h3><div>A total of 48,397 records (2019–2021) from individuals aged ≥18 years were retrieved from the hospital laboratory information system. After applying inclusion and exclusion criteria, 9455 individuals constituted the reference sample group. Partitioning was based on age, sex, thyroid-stimulating hormone (TSH), and free thyroxine (fT4). The Elbow method suggested the optimal number of clusters; K-means clustering was used to form subgroups, and the Extra Trees Classifier quantified the relative importance of partitioning criteria. RIs were estimated using the non-parametric percentile method, and Fuzzy C-Means clustering was applied to smooth sharp transitions between groups.</div></div><div><h3>Results</h3><div>Six subgroups were identified, with age as the dominant determinant (feature importance score = 0.96), whereas sex had a negligible effect. Fuzzy clustering generated continuous and personalized RIs. In clinical evaluation, applying fuzzy RIs reclassified 7 % of patients, predominantly shifting diagnoses toward hyperthyroidism.</div></div><div><h3>Conclusions</h3><div>This study offers a universal and adaptable ML-based framework for generating continuous, personalized RIs. This advance toward precision laboratory medicine can enhance diagnostic accuracy, reduce misclassification, and support more patient-centered care.</div></div>\",\"PeriodicalId\":10205,\"journal\":{\"name\":\"Clinica Chimica Acta\",\"volume\":\"579 \",\"pages\":\"Article 120635\"},\"PeriodicalIF\":2.9000,\"publicationDate\":\"2025-10-09\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Clinica Chimica Acta\",\"FirstCategoryId\":\"3\",\"ListUrlMain\":\"https://www.sciencedirect.com/science/article/pii/S0009898125005145\",\"RegionNum\":3,\"RegionCategory\":\"医学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q2\",\"JCRName\":\"MEDICAL LABORATORY TECHNOLOGY\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Clinica Chimica Acta","FirstCategoryId":"3","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S0009898125005145","RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"MEDICAL LABORATORY TECHNOLOGY","Score":null,"Total":0}

引用次数: 0

摘要

背景：准确解释甲状腺功能检查（TFTs）需要可靠的参考区间（RIs）。基于回顾性实验室数据的间接方法越来越多地被使用，但目前的策略面临着主要的局限性，包括严格的年龄界限，不一致的划分，以及缺乏对亚分组标准的客观评价。为了应对这些挑战，我们开发了一种新的基于机器学习（ML）的框架，用于客观、数据驱动地确定连续和个性化的RIs。方法：从医院实验室信息系统中检索年龄≥18 岁的48,397份记录（2019-2021）。采用纳入和排除标准后，9455人构成参考样本组。根据年龄、性别、促甲状腺激素（TSH）和游离甲状腺素（fT4）进行划分。肘部法给出了最优聚类数；K-means聚类用于形成子组，Extra Trees分类器量化了划分标准的相对重要性。RIs采用非参数百分位法估计，模糊c均值聚类用于平滑组间的急剧过渡。结果：确定了六个亚组，年龄是主要决定因素（特征重要性得分 = 0.96），而性别的影响可以忽略不计。模糊聚类生成连续的个性化RIs。在临床评价中，应用模糊RIs对7 %的患者进行重新分类，主要将诊断转向甲状腺功能亢进。结论：本研究提供了一个通用的、适应性强的基于ml的框架，用于生成连续的、个性化的RIs。这种向精密检验医学的进步可以提高诊断的准确性，减少错误分类，并支持更多以患者为中心的护理。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

查看原文本刊更多论文

From objective grouping to fuzzy reference intervals: A standardized machine learning approach for thyroid function tests

Background

Accurate interpretation of thyroid function tests (TFTs) requires reliable reference intervals (RIs). Indirect methods based on retrospective laboratory data are increasingly used, but current strategies face major limitations, including rigid age cut-offs, inconsistent partitioning, and lack of objective evaluation of subgrouping criteria. To address these challenges, we developed a novel machine learning (ML)-based framework for the objective, data-driven determination of continuous and personalized RIs.

Methods

A total of 48,397 records (2019–2021) from individuals aged ≥18 years were retrieved from the hospital laboratory information system. After applying inclusion and exclusion criteria, 9455 individuals constituted the reference sample group. Partitioning was based on age, sex, thyroid-stimulating hormone (TSH), and free thyroxine (fT4). The Elbow method suggested the optimal number of clusters; K-means clustering was used to form subgroups, and the Extra Trees Classifier quantified the relative importance of partitioning criteria. RIs were estimated using the non-parametric percentile method, and Fuzzy C-Means clustering was applied to smooth sharp transitions between groups.

Results

Six subgroups were identified, with age as the dominant determinant (feature importance score = 0.96), whereas sex had a negligible effect. Fuzzy clustering generated continuous and personalized RIs. In clinical evaluation, applying fuzzy RIs reclassified 7 % of patients, predominantly shifting diagnoses toward hyperthyroidism.

Conclusions

This study offers a universal and adaptable ML-based framework for generating continuous, personalized RIs. This advance toward precision laboratory medicine can enhance diagnostic accuracy, reduce misclassification, and support more patient-centered care.

求助全文

通过发布文献求助，成功后即可免费获取论文全文。去求助

来源期刊

Clinica Chimica Acta 医学-医学实验技术

CiteScore

10.10

自引率

2.00%

发文量

1268

审稿时长

23 days

期刊介绍： The Official Journal of the International Federation of Clinical Chemistry and Laboratory Medicine (IFCC) Clinica Chimica Acta is a high-quality journal which publishes original Research Communications in the field of clinical chemistry and laboratory medicine, defined as the diagnostic application of chemistry, biochemistry, immunochemistry, biochemical aspects of hematology, toxicology, and molecular biology to the study of human disease in body fluids and cells. The objective of the journal is to publish novel information leading to a better understanding of biological mechanisms of human diseases, their prevention, diagnosis, and patient management. Reports of an applied clinical character are also welcome. Papers concerned with normal metabolic processes or with constituents of normal cells or body fluids, such as reports of experimental or clinical studies in animals, are only considered when they are clearly and directly relevant to human disease. Evaluation of commercial products have a low priority for publication, unless they are novel or represent a technological breakthrough. Studies dealing with effects of drugs and natural products and studies dealing with the redox status in various diseases are not within the journal''s scope. Development and evaluation of novel analytical methodologies where applicable to diagnostic clinical chemistry and laboratory medicine, including point-of-care testing, and topics on laboratory management and informatics will also be considered. Studies focused on emerging diagnostic technologies and (big) data analysis procedures including digitalization, mobile Health, and artificial Intelligence applied to Laboratory Medicine are also of interest.