| Summary: | Pneumonia currently accounts for 20.6 percent of total death and is ranked 2nd under the top 10 principal causes of death in Singapore [1]. Respiratory sounds are able to provide important information for the diagnosis of lung diseases such as Pneumonia and Chronic Obstructive Pulmonary Disease (COPD). Adventitious respiratory sounds such as stridor, crackle and wheeze are considered abnormal lung sounds. Physicians will usually conduct physical examination through the use of stethoscope to detect any abnormal lung sounds before conducting further evaluation through Magnetic Resonance Imaging (MRI), X-ray or Computed Tomography (CT) scan. Such examinations are often lengthy and rely heavily on the physician’s experience. In today’s world, the advancement in technology is disrupting and driving digital transformation in the healthcare industry. Hence, this prompts for research in automatic detection based method to analyse the characteristics of various lung sounds and classify them based on their distinctive features. Such automated detection based method aims to simplify and improve the accuracy of the diagnosis process for lung diseases. This study provides comprehensive analysis on relevant signal processing techniques and machine learning models to develop a multi-classification algorithm that can be implemented on a simple device. The algorithm will be able to classify and identify 4 different lung sounds namely healthy, stridor, crackle and wheeze lung sound. The development process for this project is carried out in 3 main stages; Feature Extraction, Feature Selection and Model Classification.
|
|---|