Diagnosis Model Based on Chest Correlation Map Derived from Lung Sounds

Abstract

The increasing number of respiratory diseases worldwide increases the need for accurate diagnostic methods. Today, stethoscopes and pulmonary function tests, which are widely used in clinical practice, may be insufficient in diagnostic accuracy. To overcome this deficiency, new approaches based on computerised analysis of lung sounds are being developed. In this study, a model based on chest correlation maps is proposed for the detection of respiratory diseases. Within the scope of the study, 14-channel lung sound recordings of 120 participants (60 healthy, 60 asthma) were used and the correlation coefficients between the microphones located in the middle and lower regions were calculated according to the reference microphones placed in the upper lung region. Various features were extracted by dividing the obtained correlation coefficients by different threshold levels, and these features were given as input to the Bayesian classifier. According to the classification results, the highest F1 scores were calculated as 0.793 for the upper left reference microphone and 0.754 for the upper right reference microphone. These findings show that the proposed model provides an effective approach for the detection of respiratory diseases.