MINNEAPOLIS – An automated analysis of tracheal breath sounds while awake was predictive of obstructive sleep apnea in a small study.
If validated in larger studies, the findings may streamline the obstructive sleep apnea (OSA) screening process, resulting in earlier diagnosis and treatment of severe cases, Zahra Moussavi, Ph.D., said at the annual meeting of the Associated Professional Sleep Societies. "The ability to predict the likelihood and severity of obstructive sleep apnea without performing overnight polysomnography is very appealing and would lead to significant reductions in health care costs, compared with full-night sleep assessments."
Obstructive sleep apnea (OSA) is highly prevalent in the general population, but only 30% of patients referred to a sleep lab for evaluation have severe OSA requiring treatment, said Dr. Moussavi, a professor in the department of electrical and computer engineering at the University of Manitoba, Winnipeg. With no fast, accurate, clinical or laboratory tools for predicting the severity of suspected OSA, full-night polysomnography is required to confirm the diagnosis and determine its severity. "Unfortunately, the demand [for full-night sleep studies] outweighs the available resources, resulting in appointment backlogs and long wait times, which can delay the initiation of potentially lifesaving care," she said.
Acoustic analysis has been used during sleep to evaluate the breathing and snoring patterns of suspected apnea patients. To examine wakeful breathing patterns associated with OSA, Dr. Moussavi and colleagues, recorded the tracheal breath sounds of 35 patients with varying severity of OSA and 17 age-matched controls.
"We recorded the tracheal breath sound in supine and upright positions during nose and mouth breathing," said Dr. Moussavi.
Spectral analysis of the respiratory signals indicated that variation in the average power of the tracheal breath sounds at different positions was a characteristic feature that discriminated the OSA and control groups.
Using the maximum relevancy/minimum redundancy method, the investigators reduced the number of sound features that were significantly different between the groups to two, "and unsupervised clustering of these showed an overall accuracy of 84%, with a sensitivity of 88% and a specificity of 80%," Dr. Moussavi reported.
"It is known that [OSA] patients have a smaller and more collapsible pharynx, which is compensated by increased dilator muscle activity during wakefulness. They tend to have more negative pharyngeal pressure, which can be detected via breathing sounds through the nose because of higher resistance," she said. Because breath sounds are directly related to pharyngeal pressure, "this method is sensitive to the severity of [OSA] even during wakefulness."
The study was supported by the National Sciences and Engineering Research Council of Canada and TRI Labs Winnipeg where Dr. Moussavi is an adjunct scientist.