Journal of Neuroscience Methods 148 (2005) 43–48 A visual aid for computer-based analysis of sleep–wake state in rats Viara R. Mileva-Seitz, Rhain P. Louis, Richard Stephenson ∗ Department of Zoology, University of Toronto, 25 Harbord Street, Toronto, Ont., Canada M5S 3G5 Received 12 August 2004; received in revised form 21 October 2004; accepted 5 April 2005 Abstract Computer-based sleep scoring systems are often calibrated by reference to a conventional visual analysis of electroencephalographic (EEG) and electromyographic (EMG) traces. However, these types of data place high demands on digital storage capacity which may limit the duration or feasibility of some studies. The present paper describes an approach to visual analysis that involves reconstruction of a waveform (termed a “pseudopolygram” (PPG)) from conditioned data derived from the EEG and EMG. The PPG is the sum of three sine waves, each of which has a distinct frequency (non-REM sleep (NREM), 3 Hz; rapid eye movement sleep (REM), 7 Hz and wakefulness (WAKE), 60 Hz) and amplitude proportional to the value of a state-specific scoring variable. Thus, in NREM sleep the wave depicting the NREM quantifier has high amplitude and produces a PPG with dominant 3 Hz frequency. In REM sleep, the wave depicting the REM quantifier has high amplitude and produces a PPG with a dominant 7Hz frequency, and in WAKE the PPG is dominated by 60Hz. Thus, the PPG provides a means for visual discrimination of the three behavioural states. Validation studies found an overall reliability of 94% compared with conventional visual analysis of EEG and EMG. The PPG was also found to remain accurate in rats after 24 h of sleep deprivation. © 2005 Elsevier B.V. All rights reserved. Keywords: Wistar rats; Sleep; Electroencephalogram; Electromyogram; Pseudopolygram 1. Introduction The assignment of behavioural state is conventionally based on the visual interpretation of a “polysomnogram” (PSG), which consists minimally of electroencephalographic (EEG) and electromyographic (EMG) records, sometimes in combination with other physiological and behavioural vari- ables (Robert et al., 1999). PSG recordings from rats are usually divided into epochs of 5–30 s in duration and each epoch is usually assigned one of three behavioural states; wakefulness (WAKE), rapid eye movement sleep (REM), or non-REM sleep (NREM), although additional states or stages are sometimes defined (Gottesmann, 1992). The monitoring and recording of sleep–wake states in mammals is labour- intensive and can be error-prone because assessment of the PSG is not entirely objective, involving a process of pattern recognition guided by rules based on established correlations between behaviour and electrophysiology (Costa-Miserachs ∗ Corresponding author. Tel.: +1 416 978 3491; fax: +1 416 978 8532. E-mail address: rstephsn@zoo.utoronto.ca (R. Stephenson). et al., 2003; van Luijtelaar and Coenen, 1984). Many epochs are difficult to score with certainty because, for many reasons, the EEG and EMG waveforms may not appear characteris- tic of any one state, or they may contain transitions between states. Satisfactory resolution of this problem requires a sub- stantial amount of training and experience on the part of the rater. It also requires sustained concentration in a task that is inherently tedious, leading inevitably to fatigue and increased risk of error over time. Thus, in studies where large quantities of data are to be acquired, and especially where analysis is required in real-time, a computer-based sleep scoring system is a practical necessity. Numerous computer algorithms have been developed for on-line real-time sleep scoring in human and animal studies (Robert et al., 1999). Many of these have been shown in val- idation studies to be as accurate as human raters (Louis et al., 2004; Neckelmann et al., 1994). The majority of these computer systems are semi-automated, in that they compare incoming data, on an epoch by epoch basis, with predeter- mined threshold values, and assign state according to whether the data fall above or below the thresholds. The threshold 0165-0270/$ – see front matter © 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.jneumeth.2005.04.004