Automatic Algorithm for identifying Abnormal Lung Sounds through the Recognizing of Sound Patterns

摘要

Acute Respiratory Infections are caused by viruses, bacteria and fungi. The most serious is pneumonia, which is the leading cause of death in children and older adults around the world. Approximately 60.2% of cases of pneumonia in Peru from 2008–2016 are in children under 5 years old. For this reason, the prevention of pulmonary diseases is fundamental in the goal of reducing infant mortality in Peru. One of the main associated problems is the lack of medical personnel and equipment in remote areas of poor resources that is exposed to low temperatures, such as in Puno, Arequipa or Huancavelica. This study develops an algorithm to differentiate between normal and abnormal lung sounds, for this purpose a sample of 11 sounds was used. From each signal, 14 characteristics of the spectral signal that is determined by the lung sound were extracted. The developed model obtained an F value of 0.038, which shows that it is statistically significant. The R-square is 0.9744 which indicates that the model explains 97.44% of the variance of the dependent variable which is the abnormality in lung sounds based on the independent variables ASC, ZCR, normalized Slope, Kurtosis, centroid and spectral energy of the analyzed audio signal. This study is a proof of concept that provides interesting findings about the correct classification of lung sounds, in order to develop a platform to assess the risk of pneumonia at first for the start of a treatment at the correct time with the aim to reduce the mortality of pneumonia especially in children. Our group continues working for preventive health, reducing the gaps in health for favors to the most vulnerable population.