Individual approach to the diagnosis of occupational bronchopulmonary pathology using machine learning algorithms

Introduction. The prevalence of occupational respiratory diseases remains quite high. Artificial intelligence technologies are successfully used to assess the risk of bronchopulmonary diseases. The novelty of the study lies in the simultaneous comparison of classical statistical methods and interpreted machine learning using SHAP analysis, which ensures both scientific validity and practical applicability of the results obtained in the occupational pathology system.

The study aims to build a machine learning model based on the results of clinical, functional and laboratory parameters in patients with occupational bronchopulmonary pathology and to determine predictors of disease progression in the post-exposure period.

Materials and methods. To study changes in clinical, functional, laboratory and radiological data, taking into account the duration of the disease, the authors divided the patients into 2 groups: the first (61 people) — patients with occupational bronchopulmonary pathology (the diagnosis was established during initial hospitalization while continuing to work in aluminum production); the second group (69 people) — persons with occupational bronchopulmonary pathology in the post-exposure period (observed in a clinic with an established diagnosis of an occupational disease for 5 years or more). All of the surveyed are former employees of the aluminum industry. The final database includes 130 observations with 58 features (binary and quantitative) and a binary target variable. To build a prognostic model, the gradient boosting algorithm on XGBoost decision trees (XGBClassifier) was used. The result is a predictive function f(x), which converts the vector of input features into a predictive value of the probability of assigning a patient to the first or second group.

Results. When conducting mathematical statistics using the nonparametric Mann–Whitney criterion after correction by the Benjamini–Hochberg method, the authors identified signs that differ statistically between the groups. These included: work experience in harmful conditions, the total number of points on the CAT scale (COPD assessment test), the severity of shortness of breath on the mMRs and Borg scales, the number of meters covered during the 6-minute walking test. Of the spirometric parameters, an increase in the vital capacity of the lungs turned out to be significant, and of the laboratory parameters, the level of transferrin in the blood serum. In addition, the model revealed the influence of the following laboratory parameters: the number of platelets, lymphocytes, rod-shaped leukocytes, high-density lipoproteins (HDL), triglycerides, atherogenicity index, creatinine, alkaline phosphatase, potassium, ceruloplasmin, as well as functional parameters — FEV1 and residual lung volume.

Limitations. The study has professional (aluminum production workers) and gender (men) limitations.

Conclusion. From the period of diagnosis to the post-exposure period of occupational bronchopulmonary pathology, shortness of breath rates on the MMRs and Borg scales, CAT 9 index, increase, with a simultaneous deterioration in 6-MSHT indicators, which manifests itself in maximum symptoms and has an extremely strong effect on the patient's health, quality of life and physical activity. At the same time, LDL levels increase with an increased level of platelets in the blood, which increases the risk of atherosclerosis, thrombosis and cardiovascular pathology. At the stage of diagnosis, there is a significant pro-inflammatory background, characterized by an increase in the number of rod-shaped leukocytes and blood ceruloplasmin. The XGBClassifier model demonstrated high prediction accuracy, supported by ROC-AUC values, sensitivity, and specificity.

Ethics. In accordance with the requirements of the Committee on Biomedical Ethics, the examination was conducted only with the written informed consent of the patients, the work did not infringe on the rights and did not endanger the well-being of the study subjects in accordance with the requirements of biomedical ethics approved by the Helsinki Declaration of the World Medical Association (2024) and the order of the Ministry of Health of the Russian Federation No. 200N (dated 04/01/2016). FGBNU VSIMEI has a license to curry out medical activities of the Territorial body of Roszdravnadzor in the Irkutsk region No. L041-00110-38/00355362 dated 07/28/2020.

Contributions:

Beigel E.A. — research concept and design, material collection and data processing, text writing, design and editing of the article;
Lakhman O.L. — concept and design of the research, design and editing of the article; Rozhkova N.Y. — statistical processing, text writing, design and editing of the article;
Peshcherova S.M. — statistical processing, text writing, design and editing of the article;
All co-authors — approving the final version of the article and ensuring the integrity of all parts of the article.

Funding. The funding was provided within the Framework of the research work "Study of the mechanisms of formation and progression of neurodegenerative and bronchopulmonary disorders under the influence of industrial toxicants" (state registration number No. 01201355913) and exploratory scientific research "Development of methods for diagnosis, treatment, rehabilitation and prevention of diseases aimed at prolonging the active longevity of the Siberian population" (number state registration No. AAAAA-A20-120100190008-8); "Development of approaches to the treatment and medical rehabilitation of patients with comorbid post-COVID syndrome and military personnel injured in combat operations" (state registration number 12303200011-5).

Conflict of interest. The authors declare no conflict of interest.

Received: 17.08.2025 / Accepted: 12.09.2025 / Published: 30.10.2025

1. Shpagin I.S., Kotova O.S., Kamneva N.V., Kuznetsova G.V., Logvinenko K.V., Lisova E.S. Professional chronic obstructive pulmonary disease in combination with heart failure — clinical and functional features. Russian Journal of Occupational Health and Industrial Ecology. [Med. truda i prom. ekol.]. 2019; 59(7): 388–394. https://doi.org/10.31089/1026-9428-2019-59-7-388-394 (in Russian).

2. Boykin A.S., Kurenkova G.V. Hygienic characteristics of incidence with time loss of employment of workers employed in the production of aluminum. Prikaspijskij vestnik mediciny i farmacii. 2021; 2(2): 17–22. https://elibrary.ru/inwhfw (in Russian).

3. Beigel E.A., Katamanova E.V., Efimova N.V. Risk factors for the development of comorbid diseases of occupational bronchopulmonary pathology. Hygiene and Sanitation, Russian Journal. 2022; 101(9): 1043–1048. https://doi.org/10.47470/0016-9900-2022-101-9-1043-1048 (in Russian).

4. Babanov S.A., Budash D.S. Clinical and laboratory features and prognosis of the course of chronic dust bronchitis, silicosis and pneumoconiosis from exposure to welding aerosols. Sanitarnyj vrach. 2016; 4: 37–47. https://clck.ru/3PQHun (in Russian).

5. Krylov A.P. Statistical methods in medicine. Terapevt. 2025; 3: 13–25. https://doi.org/10.33920/MED-12-2503-02 https://elibrary.ru/boeceb (in Russian).

6. Severtsev A.N., Anosov V.D., Domrachev S.A., Ismati A.O., Ovchinnikov S.V., Solovyev N.O., Shmakov A.V., Rogozhina L.S., Repin I.G., Muradyan T.G. Identifying key factors impacting outcomes in hospitalized patients with gastrointestinal bleeding and receiving antiplatelet and anticoagulant therapy: a machine learning approach. Kremlyovskaya meditsina. Klinicheskij vestnik. 2025; 1: 35–39. https://doi.org/10.48612/cgma/v83n-tmbe-g71t https://elibrary.ru/gkwsbp

7. Britvina P.V. Machine learning in medicine: a revolution in diagnostics, treatment and personalization. Vestnik nauki. 2024; 4(1): 442–444. https://clck.ru/3PQHsP

8. Mondal, Proloy Kumar and Jahan, Md. Khurshid and Bijoy, Mehadi Hasan and Rusho, Maher Ali and Rahman, Mohammad Hasibur. A Systematic Literature Review on the Application of Artificial Intelligence and Machine Learning in Personalized Medicine: Methodological Advances and Emerging Trends (August 16, 2024). https://ssrn.com/abstract=4928301

9. Luchinin A.S. Prognostic Models in Medicine. Klinicheskaya onkogematologiya. 2023; 16(1): 27–36. https://doi.org/10.21320/2500-2139-2023-16-1-27-36

10. Gusev A.V., Gavrilenko G.G., Gavrilov D.V. Development of a machine learning model to interpret the results of laboratory diagnostics in order to identify suspected diseases. Laboratory Service. 2022; 11(2): 9–17. https://doi.org/10.17116/labs2022110219 (in Russian).

11. Savushkina O.I., Astanin P.A., Kryukov E.V., Zaicev A.A. Prediction of pulmonary gas exchange disorders in patients with long-term COVID-19 using machine learning methods. Bulletin Physiology and Pathology of Respiration. 2023; (87): 18-28. https://doi.org/10.36604/1998-5029-2023-87-18-28 (in Russian).

12. Wang W., Li M., Fan P., Wang H., Cai J., Wang K., Zhang T., Xiao Z., Yan J., Chen C., Lv Q. Prototype early diagnostic model for invasive pulmonary aspergillosis based on deep learning and big data training. Mycoses. 2023; 66(2): 118–127. https://doi.org/10.1111/myc.13540

13. Jiang Z., Bakker O.J., Bartolo P.J. Industry 4.0-Compliant Occupational Chronic Obstructive Pulmonary Disease Prevention: Literature Review and Future Directions. Sensors. 2024; 24(17): 5734. https://doi.org/10.3390/s24175734

14. Si X.A., Xi J. Deciphering Exhaled Aerosol Fingerprints for Early Diagnosis and Personalized Therapeutics of Obstructive Respiratory Diseases in Small Airways. Journal of Nanotheranostics. 2021; 2: 94–117. https://doi.org/10.3390/jnt2030007

15. Kotlyarov S., Lyubavin A. Early detection of atrial fibrillation in chronic obstructive pulmonary disease patients. Medicina. 2024; 60(3): 352. https://doi.org/10.3390/medicina60030352

16. Trusov P.V., Tsinker M.Yu., Zaitseva N.V., Nurislamov V.V., Svintsova P.D., Kuchukov A.I. Assessing Spatial Distribution of Sites with a Risk of Developing Bronchopulmonary Pathology Based on Mathematical Modeling of Air-Dust Flows in the Human Airways and Lungs. Health Risk Analysis. 2024; 2: 141–152. https://clck.ru/3PQP9m

17. Egorov I.A., Smirnova S.S., Semenov A.V. Using inductive machine learning to identify risk factors for healthcare workers to get infected with highly contagious viruses (based on covid-19 model). Health risk analysis. 2024; 2: 122–131. https://doi.org/10.21668/health.risk/2024.2.11 (in Russian).

18. Pashkovskaya A.P. The study and comparative analysis of machine learning algorithms for evaluating their effectiveness in dementia classification task. Modern Innovations, Systems and Technologies. 2025; 5(1): 1041–1047. https://doi.org/10.47813/2782-2818-2025-5-1-1041-1047 (in Russian).

19. Amaral J.L.M., de Sá Sousa C.M., de Oliveira Ribeiro C., de Sá P.M., Lopes A.J., de Melo P.L. A diagnostic support system based on interpretable machine learning and oscillometry for accurate diagnosis of respiratory dysfunction in silicosis. bioRxiv. 2025. https://doi.org/10.1101/2025.01.08.632001

20. Merryn J. Baker et al. Rapid, non-invasive breath analysis for enhancing detection of silicosis using mass spectrometry and interpretable machine learning. Journal of Breath Research. 2025; 19(2): 026011. https://doi.org/10.1088/1752-7163/adbc11