A new neural network based on nocturnal pulse oximetry enables accurate assessment of sleep apnea severity in patients with cerebrovascular disease. Developed by researchers at the University of Eastern Finland and Kuopio University Hospital, the automated assessment provides a pathway for screening for sleep apnea in stroke units.
“Although screening of sleep apnea is recommended for patients with cerebrovascular disease, it is rarely done in stroke units due to complicated measurement devices, time-consuming manual analysis, and high costs,” says researcher Akseli Leino, the University of Eastern Finland, in a release.
In the new study, researchers developed a neural network to assess the severity of sleep apnea in patients with acute stroke and transient ischemic attack (TIA) by using a nocturnal oxygen saturation signal. When the researchers compared the results of manual scoring and those obtained using the new neural network, the median difference was 1.45 events per hour. The neural network was also 78% accurate in classifying patients into four different categories on the basis of sleep apnea severity (no sleep apnea, mild, moderate, severe). The neural network was able to identify moderate and severe sleep apnea in patients with acute stroke or transient ischemic attack with a 96% specificity and a 92% sensitivity.
“The neural network developed in the study enables an easy and cost-effective screening of sleep apnea in patients with cerebrovascular disease in hospital wards and stroke units. The nocturnal oxygen saturation signal can be recorded with a simple finger pulse oximetry measurement, with no time-consuming manual analysis required,” says medical physicist Katja Myllymaa from Kuopio University Hospital, in a release.
The findings are published in Sleep Medicine.
Photo credit: UEF / Raija Törrönen