Graphic visualization of the transition of sleep phases in humans: methodsof 27 tridimensional representation.

Sleep evaluation is usually done by visual inspection of the electroencepha logram (EEG) raw recordings. This approach is the most widely used until no w. Many electrical characteristics can be easily detected and measured usin g this approach, but others not. This is because the simple visual inspecti on is a timedomain method in which the embedded frequency-and-phase relatio ns are not so evident. The frequency-domain characteristics of the raw EEG signals can be measured in the paper but only in an elemental way, that is, the frequency and shape of a particular wave. Subtle phases and frequency shifts are usually completely hidden for the observer. Another concern of t he visual EEG inspection is the amount of EEG recording that can be seen at a time. During a visual EGG evaluation only one or two pages at a time are displayed difficulting the precise tracking of fine frequency changes. We developed and tested a method that generates and displays as much as 8 h ours of continuous EEG recording on a single image. We can create an "abstr action" of the recording in a sngle image that shows the real evolution in the frequency domain. The program we designed, computes the power spectrum of the whole recording and creates a tridimensional graph showing the spect rum as a solid object. This image is comapred with the relative or absolute power of a particular frequency band on other channels allowing the visual identification of some sleep stages and transitions. We used this program in humans during whole-night sleep recordings. The program was written usin g a GNU CH compiler running in an ULTRA SPARC creator I computer.