Analysis of multichannel respiratory sounds under different respiratory maneuvers in healthy subjects and pulmonary disease patients

2019 edition

Jasna Nuhic

Different lung function tests are used to determine how well lungs work. One of the most widely used is spirometry, a standardized procedure that measures the lung volume and airflow. However, results of conventional lung function tests are highly affected and consequently limited by experience and hearing ability of clinical professionals. These limitations have been the subject of research for long time, and analysis of respiratory sounds (RS) has crucial place in it.

General research question is how RS in pulmonary disease patients can be used to reduce human error in the process of respiratory disease diagnostics and to predict the onset of it later. Sensors’ locations during signal acquisition that enables good distribution of RS remain an open question, as well.

Since RS are vital clinical tool to evaluate respiratory function and indicate presence of pulmonary disease, use of RS in lung functionality testing dates a long time ago, since the invention of stethoscope. This research proposes novel approach to multichannel RS recording, using five different microphones located on the commonly reported regions of chest surface and trachea. RS are analyzed according to different airflow levels, with expected additional contributions of RS at very high loads. Combination of different loading protocols and multichannel RS recording is the newest point of the study of RS.

The two datasets acquired at King College London are available for study, one includes RS recordings of healthy subjects, while another one COPD patients. Lung sound intensity (LSI), as the main pointer of robustness, is estimated and characterized by Root mean square, Average Power and Entropy. Inclusion of a tracheal microphone provides us with the RS from the cricoid cartilage which might contain complementary information about pulmonary function.


Old subjects refer to subjects older than 60 years old, while young refer to subjects younger than 40 years old. The comparison between young and old will be achieved by characterization of the RS and airflow signals. In young subjects, sample entropy estimated correctly LSI, whereas RMS and mean power overestimated LSI. LSI increased linearly with airflow levels, whereas load protocol showed an increment in airflow as the load increased, deriving in higher LSI. Spatial distribution analysis showed that all subjects were within a normal interval prefixed by literature (3 dB).

Variations among the healthy subjects and Chronic obstructive pulmonary disease (COPD) patients are expected to be much more significant than between old and young healthy subjects. Some of indicators of obstruction are variability of lung sound intensity (LSI) and presence of adventitious sounds.

The further objective of the RS analysis is investigating an early stage Amyotrophic lateral sclerosis (ALS) through the new database which is to be formed on ALS patients, in the following 3 months at the Hospital del Mar, Barcelona.