A value of respiratory muscle strength indicators to determine severity of chronic obstructive pulmonary disease using artificial neural networks

Abstract

The aim of the study was to analyze a diagnostic value of respiratory muscle (RM) strength indicators to assess severity of chronic obstructive pulmonary disease (COPD) using machine learning methods and artificial neural networks (ANN). Methods . One hundred and fifteen males with acute exacerbation of COPD were involved in the study. RM strength indicators (MEP, MIP, and SNIP), demographic parameters, spirometry, blood gases, dyspnea with mMRC and CAT scales were measured. Statistical analysis was performed using Mann-Whitney’s, Fisher’s and Tukey’s tests and correlation analysis. RM strength model was performed using linear and nonlinear regression analysis. COPD stratification model was performed using ANN. Results . RM strength models in healthy males and COPD patients allowed estimation the impact of different factors on the RM functional status. Comparison of COPD stratification for severity using the mathematical model or expert diagnosis showed that combination of FEV1 with other indicators could increase the accuracy of ANN model. MIP, the total body mass, partial CO2 tension in the arterial blood and serum fibrinogen concentration were the most valuable indicators. Moreover, MIP was considered as the universal predictor increasing the accuracy of all models. Conclusion . Practical application of ANN models in telemedicine projects is related to the improvement of ANN architecture and development of informational services which would allow a real-time assessment of the patient's condition.