Thoracal motion-based analysis of breathing patterns in individuals with a mild-moderate Covid-19 history using machine learning techniques: A single blinded multidisciplinary study on post-Covid

Abstract

BackgroundCovid-19 led to deaths worldwide and left significant sequelae in a lot of people. Thoracic movements are important for the proper functioning of the respiratory system. However, there is no study on how the thoracic mobility of individuals who have recovered fully from Covid-19 is affected. MethodsIn this study, the differences between thorax movements of healthy individuals and individuals with Covid-19 were investigated from a multidisciplinary perspective for the first time. Spontaneous and deep breathing data under two (at sitting- at standing) different conditions were collected and analyzed. In terms of engineering, using the Boruta feature selection method and various machine learning algorithms, discriminative features that will benefit clinically were determined. Clinically, the effect of Covid-19 was examined statistically in terms of respiratory biomechanics with thoracal motion-based analysis of 22 individuals. ResultsThe use of Boruta in sitting and standing positions during deep breathing increased the classification performance. In spontaneous breathing, using Boruta only in the sitting position provided an increase in classification performance achieving an accuracy of 95.45 %. The results of the study showed that respiratory movements of the thoracic cage in the anteroposterior and craniocaudal directions were more restricted and had weaker respiratory acceleration skills in individuals with a Covid-19 history (p < 0.05). ConclusionFrom a clinical point of view, it was observed that the respiratory acceleration movements were restricted in individuals with a Covid-19 history even though full recovery. Also, it was revealed that machine learning models can classify with high performance in situations requiring effort.