Validation and interpretation of machine-learning models for rapid identification of active tuberculosis infection using routine laboratory indicators

Sep 28, 2026

Diagnosis of active Mycobacterium tuberculosis (Mtb) infection depends on symptoms, imaging, and molecular testing, yet these options are often expensive and slow. A rapid, accessible diagnostic approach is needed for early detection and to limit ongoing TB transmission. Six supervised machine-learning models were developed from routine laboratory indicators using a discovery cohort of 3,829 individuals and externally validated on 405.

LanguageEnglish

CategoryHealthcare

FormatPDF

Pages13

Reading Time33 min

AccessFree to download

Introduction

Methods

Results

Conclusion

Graphical abstract

Background

TYPE Original Research PUBLISHED 18 December 2025 DOI 10.3389/fcimb.2025.1718614 OPEN ACCESS EDITED BY Ying Luo, University of Texas Southwestern Medical Center, United States REVIEWED BY Elena Niculet, Dunarea de Jos University, Romania Fu Gao, Yale University, United States Panpan Gu, Fudan University, China *CORRESPONDENCE Liang Wang [healthscience@foxmail.com](healthscience@foxmail.com)[ ](healthscience@foxmail.com)Hui-Jin Chen [13699994292@163.com](13699994292@163.com)[ ](13699994292@163.com)Kang-Peng Du [kangpengdu@163.com](kangpengdu@163.com) †These authors have contributed equally to this work RECEIVED 04 October 2025 REVISED 27 November 2025 ACCEPTED 30 November 2025 PUBLISHED 18 December 2025 CITATION Liu Z-Z, Yuan Q, Zhang Y-D, Zhang X-D, Liu J, Yan J-W, Du K-P, Chen H-J and Wang L (2025) Validation and interpretation of machine-learning models for rapid identiﬁcation of active tuberculosis infection using routine laboratory indicators. Front. Cell. Infect. Microbiol. 15:1718614 . doi: 10.3389/fcimb.2025.1718614 COPYRIGHT © 2025 Liu, Yuan, Zhang, Zhang, Liu, Yan, Du, Chen and Wang. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY) . The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms. Validation and interpretation of machine-learning models for rapid identiﬁcation of active tuberculosis infection using routine laboratory indicators Zhan-Zhong Liu 1†, Quan Yuan 1,2,3†, Yu-Dong Zhang 4†, Xue-Di Zhang 1, Jian Liu 5, Jia-Wei Yan 1, Kang-Peng Du 5*, Hui-Jin Chen 2* and Liang Wang 3,6* 1thXuzhou Hospital, Beijing Ditan Hospital Afﬁliated to Capital Medical University, Xuzhou Infectious Diseases Hospital (The 7th People ’s Hospital of Xuzhou), Xuzhou, Jiangsu, China, 2 Department of Laboratory Medicine, Shengli Oilﬁeld Central Hospital, Dongying, Shandong, China, 3School of Medical Informatics and Engineering, Xuzhou Medical University, Xuzhou, Jiangsu, China, 4School of 1st Clinical Medicine, Xuzhou Medical University, Xuzhou, Jiangsu, China, 5 Department of Pharmacy, The 6th People’s Hospital of Xuzhou, Xuzhou, Jiangsu, China, 6 Laboratory Medicine, Guangdong Provincial People’s Hospital (Guangdong Academy of Medical Sciences), Southern Medical University, Guangzhou, Guangdong, China Introduction: Diagnosis of active Mycobacterium tuberculosis (Mtb) infection relies on clinical symptoms, imaging, and molecular testing, but these methods are often costly and slow. Consequently, there is an urgent need for a rapid and accessible diagnostic approach that can support early detection and reduce ongoing tuberculosis transmission. Methods: A discovery cohort of 3,829 individuals and an external validation cohort of 405 individuals were included. Six supervised machine learning models were trained using routine laboratory data, and model interpretability was assessed with SHapley Additive exPlanations (SHAP) . Results: Among the six models, XGBoost demonstrated the best diagnostic performance in the internal cohort (accuracy 97.49%; sensitivity 97.56%; speciﬁcity 97.42%) and maintained strong performance in the external cohort (accuracy 93 . 67%; sensitivity 91 . 56%; speciﬁcity 91 . 13%) . SHAP analysis indicated that key predictors reﬂected characteristic host-response patterns, including inﬂammation-related hypoalbuminemia, lipid metabolism suppression (HDL-Cand LDL-C), altered platelet activity (MPV), and lymphocyte reduction (LYM) . Conclusion: The study presents a high-performing and interpretable machine learning model capable of accurately identifying active Mtb infection using routine blood tests. This low-cost and no

FAQs

Why is there a need for rapid identification of active Mtb infection?

Active Mtb detection is essential for early treatment and for interrupting transmission chains, but conventional diagnostic approaches can be costly and slow.

How were the machine-learning models validated in the study?

Six supervised models were trained on routine laboratory data using a discovery cohort of 3,829 individuals and validated externally on 405 individuals.

Which model performed best and what did SHAP show?

XGBoost showed the best diagnostic performance in both internal and external cohorts, and SHAP indicated key predictors aligned with host-response patterns such as hypoalbuminemia, altered lipid metabolism, platelet activity changes, and lymphocyte reduction.

Sep 28, 2026

Uploaded by Quinn Holloway