Machine learning techniques application for lung diseases diagnosis

Abstract

The article considers the basic methods of machine learning for applying them to the task of the lungs sounds classifying. A number of signal parameters were obtained on the basis of the lungs sounds set. The task of the study was to classify sounds using five different machine learning methods. It was also necessary to determine from a number of signal parameters those that give the highest accuracy. Thus the seven most diagnostically valuable parameters of lung sounds were found. The results showed that two methods of machine learning - the method of reference vectors and the decision tree method - have the best accuracy. Thus this classification technique can serve as an auxiliary tool for a pulmonary physician to diagnosis.

Ähnliche Arbeiten