Computer detection and analysis of periodic movements in sleep

Standard polysomnography (PSG) and electromyographic (EMG) recording of the right and left anterior tibial muscles by Medilog 9000 recorder was carried out in 10 patients with periodic movements in sleep (PMS). The analog EMG signals and the sleep data from the Oxford ss90III sleep stager were fed into an IBM compatible personal computer for automatic detection and analysis using specific criteria to define the leg events, the interevent interval, the number of epochs (greater than 30 jerks), and the number of bursts (4 to 29 jerks). The results of automatic analysis were compared with data obtained from visually scored paper PSG. The total number of leg jerks detected by visual scoring was 2,812 and by automatic scoring was 2,789. The mean duration of the leg jerks 2.77 s [standard deviation (SD), 1.22] and the mean inter-interval event duration was 27.3 s (SD, 14.7). The total numbers of epochs scored visually was 35 and scored automatically was 33. The correlation co-efficient to the line of identity was 0.8 (p less than 0.005). The sensitivity of automatic scoring was 94% and the specificity was 85%. Automatic detection and analysis of PMS can provide an objective method for the study of several aspects of this disease that are still not yet fully understood.