Validation of automated sleep analysis in normal children

With the aim of determining normal reference values for our sleep laboratory and evaluating the reliability of automated analysis for scoring polysomnographic studies in children, we recorded polysomnograms in 16 healthy boarding-school children. Sleep recordings were obtained with a computer system (Medilog SAC, Oxford Instruments). Polysomnographic variables were monitored continuously on a 16-channel recorder equipped with a video. Data were acquired on optical disk for computer-assisted data interpretation. Sleep stages and respiratory events were also scored visually by operator. Comparison with visual scores showed that the computer system significantly overscored wakefulness (W) (p<0.02) and stage IV (p<0.001) and underscored stage II (p<0.001) and rapid eye movement (REM) sleep (p<0.001). It also assigned respiratory events a higher score than did visual scoring, as shown by the higher apnoea index (AI) and hypopnoea index (HI) (AI p<0.03; HI p<0.001). Regression analysis showed a significant correlation between visual and automated scores for central (r=0.679; p<0.004) and obstructive apnoea (r=0.631; p<0.008). Computer apnoea scores did not correlate with visual scores. Much remains to be done before computer-based scoring systems can be relied upon, without visual scoring, for polysomnographic sleep studies in children. Their main advantage at present is that they offer a convenient means of saving paper, space and time.