System for Wheezing Detection in Lung Sounds

Abstract

A lung disease is mainly characterized by the variations that occur in breathing sound of human being. These variations are characterized as wheezes, crackles, etc. Wheezes are one of the most important adventitious sounds in pulmonary system. They are observed in asthma, chronic obstructive pulmonary disease (COPD) and bronchitis. The current method of detecting a lung disease for example asthma involves usage of spirometers and stethoscope. Spirometry proves to be difficult for patients with heart problems and children as it involves blowing air thrice in the spirometer. The results with stethoscope are not efficient. A wheezing detection system may help physicians to monitor patients over the long-term. This technique involves converting the recorded sound into image by applying short time Fourier transform. The image needs to be analyzed. The analysis involves extracting features that define the occurrence of wheezes. Image being in the raw form needs to be processed. The image processing involves noise removal. The preserved image components are then studied and classified based on their occurrence (frequency, total time, amplitude etc).

Keywords: Wheeze; Spectrogram; Image processing; Thresholding; MATLAB; Pattern recognition; Support Vector Machine (SVM).