| Popis: |
The respiratory muscles (RM) strength is the main indicator of their functional state. The study of RM strength is becoming increasingly prevalent in clinical pulmonology, especially in case of chronic obstructive pulmonary disease (COPD). However, RM strength is not used as criteria for COPD stratification. The aim of the study was to analyze the informative value of RM strength indicators in assessing of COPD severity on the basis of methods of by means of machine learning and artificial neural networks (ANN). 115 patients with different severity of COPD in exacerbation phase were examined. RM strength indicators (MEP, MIP and SNIP), 9 anthropometric parameters, spirometric and blood gas parameters, mMRC and CAT scales data were recorded. Processing data were carried out by means of Mann-Whitney, Fisher and Tukey tests and correlation analysis. RM strength models were performed by linear and nonlinear regression methods. COPD stratification models were performed by ANN. RM strength models of healthy men and COPD patients allowed to estimate the effects of various factors on the RM functional status. Comparative analysis of «model» results of COPD severity verification showed that ANN accuracy increased when FEV1 was combined with other indicators in models. The most informative indicators were MIP, total body mass, PaCO2 and fibrinogen. Moreover, MIP increased the accuracy of all the models. Practical application of ANN models in telemedicine projects is related to the improvement of ANN architecture and the development of information services which will allow real-time assessment of the patient’s condition. |
