Respiratory Sounds Analysis for Monitoring the Severity of Pharyngeal Collapses during Sleep

Saha, Shumit 1, 2 ;  Yadollahi, Azadeh 1, 2

1. Institute of Biomaterials and Biomedical Engineering, University of Toronto; 2. Toronto Rehabilitation Institute-University Health Network

Background: Obstructive sleep apnea (OSA) is a chronic respiratory disorder characterized by the repetitive complete or partial collapses of the pharyngeal airway during sleep. OSA severity is determined by the apnea-hypopnea index (AHI) which represents the number of apneas (complete collapse) and hypopneas (partial collapse) per hour of sleep. However, AHI is not sensitive enough to capture mild pharyngeal collapses. The consequences of those mild pharyngeal collapses include excessive daytime sleepiness and long-term cardiovascular comorbidities. To overcome this problem, a breath-by-breath pharyngeal collapse monitoring system is needed. This can be performed using magnetic resonance imaging or endoscopy; however these techniques are costly and difficult to perform during sleep. Conversely, respiratory sounds, which are generated by the passage of airflow through the pharynx, can be recorded conveniently with a microphone placed on the neck. Previous studies from our group have shown that more narrowing in the pharyngeal airway is associated with louder snoring sounds. Thus, respiratory sounds can be a good modality for monitoring the severity of pharyngeal collapses. The aim of this study is to investigate the feasibility of using respiratory sounds to detect the severity of the pharyngeal collapses.

Methods: A full night polysomnography will be performed on 20 patients to access the AHI. A pressure sensor will be inserted through the nose up to 2 cm below the base of the tongue. The degree of the collapse in the pharynx will be identified by noting the continued decreases in pressure. Respiratory sounds will be recorded with a microphone attached to the neck. Respiratory sounds will be classified into inspiration, expiration and snoring segments. Feature extraction techniques will be performed to extract acoustic features (i.e. spectral power, bi-spectrum frequencies, formants and pitch). Subsequently the association between the acoustic features of respiratory sounds and pressure sensor data will be investigated to find the best features. Based on the association, a prediction model will be developed to estimate the severity of the narrowing in the pharyngeal airway.

Expected results: We expect to extract 3 to 5 features such as intensity, formant frequencies, or higher order features from the respiratory sounds that will have the ability to estimate the severity of pharyngeal narrowing in each breath.  

Conclusion: The results of this study will be used to develop a convenient portable acoustic technology to estimate the pharyngeal collapses. This will help physicians to identify the severity of pharyngeal collapses in each patient and determine adequate treatment.