Diagnosis of chronic obstructive pulmonary disease using empirical wavelet transform analysis from auscultation sounds

Along with the developing technology, great developments are seen in the scientific world. Clinical advances in clinical diagnosis and treatment have also been greatly influenced by these developments. The systems developed as a result of the studies in the field of health are used both in the diagnosis of the diseases and in the treatment of the diseases. The main purpose of these studies is to reduce the margin of error and to help physicians by developing fast, reliable and cost-effective systems. In this thesis, a new method has been developed for the classification of Chronic Obstructive Pulmonary Disease (COPD) disease by using auscultation method. There are many studies in the literature on auscultation-based analysis and diagnosis of respiratory diseases. However, the difference of this study from the others is that Empirical Wavelet Transform (EWT) based data are analyzed. EWT is a method used to divide a given signal into different modes. It is based on the design of an adaptive wavelet filter bank. The database used in this study has never been studied before using the EWT technique. The RespiratoryDatabase@TR database, which consists of 12-channel lung sounds and 4-channel heart sounds, can be used in clinical diagnosis and treatment applications based on artificial intelligence. The RespiratoryDatabase@TR database includes heart and lung sounds as well as chest X-ray, pulmonary function test measurements, and answers to the SGRQ-C questionnaire. After the necessary pre-process were made on auscultation sounds, different modes of the signals were obtained by applying EWT method. Statistical characteristics were calculated on the data obtained after the transformation. Finally, the success of the system was applied on 9 different classification models and their performance was evaluated.