Performance of an Automated Polysomnography Scoring Using Noxturnal Program versus Manual Scoring

Abstract

Background and Objectives A lack of investigators for polysomnography has risen due to increased demand since health insurance started to cover the cost of the test. We examined the reliability of the automated scoring of polysomnography, which has been deployed to improve this imbalance.Subjects and Method We analyzed the data of 20 patients who underwent level 1 polysomnography from April 1 to July 27, 2019. The software from Noxturnal (Nox Medical) was used for the scoring of the Polysomnography data. Each of the polysomnographic data was scored both by the automated scoring system and by a skilled technician.Results Twenty patients were analyzed. There was no significant difference between automated scoring and manual scoring in sleep latency, apnea index, and rapid eye movement sleep stage ratio. However, the concordance rate of the sleep stage by epoch was 83.32%, and there was a significant difference with regards to apnea-hypoapnea index (AHI) and respiratory disturbance index (RDI). Two obvious errors were noted in the automated scoring that could be easily fixed; the failure to recognize wakefulness during sleep and the fragmentation of respiratory events. When two errors were corrected, many differences in polysomnography parameters, including AHI and RDI, were eliminated.Conclusion It showed 80% coincidence of epoch in the sleep stage between the automated scoring and manual scoring. However, there was no difference in AHI and RDI when the fragmented respiratory events of the automated scoring were adjusted. Therefore, automated scoring is considered to be useful if only a little modification could be made.