The concordance of manuel (visual) scoring and automatic analysis in sleep staging

Abstract

Full night polysomnography (PSG) remains the gold standard diagnostic test for the evaluation of sleep and the detection of sleep disorders. The computer-assisted scoring methods have been developed to accelerate the scoring. It is said that there was a concordance up to 80% between these scoring softwares and manual scoring. According to our experiences, it is not matched with this belief. In this study, we intend to examine whether the results of automatic analysis match with manual (visual) evaluation. The PSG records of 30 cases with a diagnosis of obstructive sleep apnea syndrome (OSAS) are chosen randomly. We compare the results of automatic analysis with the results of two scorers who have a concordance of 8095% and at least 1000 PSG scoring experiences. We evaluated 21.060 epochs of 18 men with 48.83 +/- 13.51 ages, and 12 women with 44.56 +/- 14.28 ages. In automatic analysis; total sleep time (p= 0.003) and sleep efficiency (p= 0.004) were low. AHI (p= 0.802) and ODI (p= 0.193) values were high. The epochs scored differently were 8819 epochs (41.88%). The stage I (88.43%) scored mostly different, was allocated to be awake (572 epochs). Stage II and stage IV were scored as stage III in 2276 and 983 epochs respectively. REM epochs were allocated to stage II (574 epochs). The differences in recording times and sleep architecture of PSG tests which examed by automatic analysis will affect all other parameters. Thus, we believe that it will make mistakes in the diagnosis and treatment of sleep disorders.