Possibilities of using speech biomarkers for remote screening of bronchial asthma

Introduction. Bronchial asthma (BA) remains one of the main health problems. Therefore, the search for objective, effortless methods of ASTHMA control from the patient and special equipment remains an urgent task. The standard approach to controlling asthma symptoms is episodic and retrospective, preventing timely detection of signs of deterioration in patients. In this article, we identify the possibilities of analyzing voice biomarkers for remote screening of asthma.

Aim: To identify acoustic parameters of speech those differ statistically significantly in patients with bronchial asthma and practically healthy individuals.

Material and Methods. The data for the analysis were voice recordings of 38 men (group 1.1) and 60 women (group 1.2) with bronchial asthma, as well as voice recordings of control groups of 57 healthy men (group 2.1) and 106 women (group 2.2). The acoustic analysis was performed using the Praat v 6.4.35 program, which was used to calculate spectral and temporal features, and Python v.3.11.4, which was used to calculate the parameters of the randomness and complexity of the pitch frequency. Acoustic and prosodic parameters of speech were calculated. Classification and testing were performed on all input attributes for five classifiers (LogisticRegression, randomForest, SVM_RBF, MLP, XGBoost).

Results. The most informative features for the binary classification of healthy individuals are mfcc_01_mean, mfcc_10_mean, shimmer local for men and mfcc_01_mean, pitch_mean, approximantropy for women. The affiliation of the leading features to three different groups of parameters confirms the multidimensional nature of voice changes in asthma.Speech parameters have gender specificity in bronchial asthma, which necessitates the construction of separate classifiers for men and women. More rigorous validation using the LOPO scheme and the randomForest classification algorithm confirmed the high generalizing ability of the models. For men, F1 = 0.84 [0.75; 0.91] and for women, F1 = 0.94 [0.90; 0.98] with an AUC ROC of 0.92 and 0.96, respectively. Using only the three most informative features, the randomForest model in the LOPO scheme achieved an F1 measure of 0.89 [0.82; 0.94] for women and 0.77 [0.68; 0.85] for men. The shuffle test showed that the classification results are not random for both men and women. The difference in the F1 measure between the initial model and after random mixing of class labels was 0.24 for men and 0.47 for women.

Conclusions. The speech parameters of patients with bronchial asthma significantly differ from healthy individuals in temporal, spectral, and nonlinear features. The degree and nature of the differences are sex-specific, which requires separate analysis for men and women. In women, the changes cover a wider range of parameters, which is consistent with the greater severity of asthma caused by anatomical, hormonal and immunological factors. The results obtained confirm the applicability of voice analysis of free speech as a tool for passive monitoring of bronchial asthma.

1. Barnes P.J., Szefler S.J., Reddel H.K., Chipps B.E. Symptoms and perception of airway obstruction in asthmatic patients: Clinical implications for use of reliever medications. The Journal of Allergy and Clinical Immunology. 2019;144(5):1180–1186. DOI: 10.1016/J.JACI.2019.06.0402

2. Gibson P.G. Monitoring the patient with asthma: an evidence-based approach. J. Allergy Clin. Immunol. 2000;106(1 Pt 1):17–26. DOI: 10.1067/mai.2000.108307

3. Metting E., Dassen L., Aardoom J., et al. Effectiveness of Telemonitoring for Respiratory and Systemic Symptoms of Asthma and COPD: A Narrative Review. Life (Basel). 2021;11(11):1215. DOI: 10.3390/life11111215 EDN: PJPEAN

4. Reid D., Mehta J., Anis K., Mehta S. Impact of Remote Patient Monitoring Platform on Patients with Moderate to Severe Persistent Asthma: Observational Study. JMIR Form. Res. 2023;7:e51065. DOI: 10.2196/51065 EDN: BMLGYR

5. Antalffy T., De Simoni A., Griffiths C.J. Promising peak flow diary compliance with an electronic peak flow meter and linked smartphone app. NPJ Prim. Care Respir. Med. 2020;30(1):19. DOI: 10.1038/s41533-020-0178-y EDN: XASWLK

6. DeVrieze B.W., Goldin J., Giwa A.O. Peak Flow Rate Measurement. [Updated 2024 Oct 6]. In: StatPearls [Internet]. Treasure Island (FL): StatPearls Publishing. 2026 Jan. URL: https://www.ncbi.nlm.nih.gov/books/NBK459325/ (10.05.2026).

7. Larsen E., Song X., Joachim D., et al. Respiratory-Responsive Vocal Biomarker for Asthma Exacerbation Monitoring: Prospective Cohort Study. J. Med. Internet. Res. 2025;27:e68741. DOI: 10.2196/68741 EDN: IDESIF

8. Postnikova L.B., Simulin M.A., Kubysheva N.I., Grigorieva N.Yu., Pogrebetskaya V.A. Early diagnosis in comorbid patients. Vestnik sovremennoj klinicheskoj mediciny. 2024;17(3):44-52. (In Russ.). DOI: 10.20969/VSKM.2024.17(3).44-52 EDN: GGRNQF

9. Babanov S.A., Strizhakov L.A., Lavrentieva N.E., et al. Immunological, molecular and genetic features and prognosis in occupational bronchial asthma. Vrach. 2025;(7):41-47 (In Russ.). DOI: 10.29296/25877305-2025-07-06 EDN: KNTKAF

10. Tharwat A. Classification assessment methods. Applied Computing and Informatics. 2020;17(1):168–192. DOI: 10.1016/j.aci.2018.08.003 EDN: PQDEQV

11. Dogan M., Eryuksel E., Kocak I., et al. Subjective and objective evaluation of voice quality in patients with asthma. J. Voice. 2007;21(2):224–230. DOI: 10.1016/j.jvoice.2005.11.003

12. Dominelli P.B., Ripoll J.G., Cross T.J., et al. Sex differences in large conducting airway anatomy. J. Appl. Physiol. (1985). 2018;125(3):960– 965. DOI: 10.1152/japplphysiol.00440.2018

13. Harvey B.J., McElvaney N.G. Sex differences in airway disease: estrogen and airway surface liquid dynamics. Biol. Sex Differ. 2024;15(1):56. DOI: 10.1186/s13293-024-00633-z EDN: IOHKUG

14. Borrelli R., Brussino L., Lo Sardo L., et al. Sex-Based Differences in Asthma: Pathophysiology, Hormonal Influence, and Genetic Mechanisms. Int. J. Mol. Sci. 2025;26(11):5288. DOI: 10.3390/ijms26115288 EDN: SZBKDZ

15. Li T., Zhang J., Wu J., et al. Asthma Detection Research Based on Voice Signal Processing and Machine Learning. J. Vis. Exp. 2025;(221). DOI: 10.3791/67742 EDN: GZXMXU

16. Lyu Y., Jiang Q.C., Yuan S., et al. Non-invasive acoustic classification of adult asthma using an XGBoost model with vocal biomarkers. Sci. Rep. 2025;15(1):28682. DOI: 10.1038/s41598-025-14645-1 EDN: ODJGDZ