Prof. Dr. Wilfried Dimpfel, Justus-Liebig-University Giessen, Germany
Sleep is a physiological status that still can be regarded as a mystery. Even today we do not know what sleep is good for. But one thing is for sure and that is: we cannot live without sleep. The common view is that sleep provides restfulness. I think that the first way to solve this question with respect to the importance of sleep is to find surrogate parameters of sleep and to measure them as accurate as possible. The next step then is to find correlative features between actions of daily life and parameters derived from the detailed measurement of the sleep process represented by its surrogate parameters. For this kind of approach an objective quantitative description of sleep must be recommended. Historically sleep has been described in a more subjective manner preferably on the base of individual recordings of the electrical activity of the brain during the night.
Fig. 1 Polyhypnographic recording in the sleep laboratory of NeuroCode AG in Wetzlar, Germany. Not only EEG is recorded but also breathing, electrooculogram, ECG, and snoring (“poly” = “many” parameters). An example of raw signals is given in Fig. 2.
First measurement of electric brain activity in humans has been achieved by the German Hans Berger from Jena, Germany. Results were published in 1929 and Berger named the documentation an “electroencephalogram" (EEG). An example of the potential changes as might have been seen by Hans Berger (who recorded only from one location) is shown in Fig. 2 for 17 electrode positions.
Fig. 2 Electric activity of the brain (EEG traces) during alertness and sleep, respectively. On the ordinate: 17 electrode positions recorded against so-called “common average” (median of potential differences of each electrode against the central position Cz). On the abscissa: total time of 4 seconds.
Only three years later Berger showed the path to a mathematically founded quantitative evaluation (frequency analysis) of the raw signals (kind of field potentials) recorded from the scalp of subjects. The manuscript containing the first frequency analysis of human EEG was published together with Dietsch in 1932. At that time it took him about 3 weeks to perform the necessary calculations by hand. Not before the sixties this type of mathematical quantification of the EEG came into broader use because of the aid of computers. Since then the analysis progresses with the speed of computer development. Today`s frequency analysis has been named after the French mathematician as “Fast Fourier Transformation” or “FFT”. This method is not only used in sleep research but also in pharmacology, neurology, anaesthesiology, sport medicine and physiology. Last not least, the method is currently also used in the field of psychology and neuro-marketing by us. We regard the electrical activity of the brain as its language that seems now to be ready to be decoded.
When I started this business in 1979 computers still needed a whole night for the computation of several channels of EEG activity recorded for more than one hour. Since begin of the nineties 99 channels of EEG activity are managed on line in real time in my labs. Frequency analysis is performed in many disciplines of natural sciences, astronomy, material testing and so forth. The basic principle behind it is the question: what kind of sine waves (frequency and amplitude) do I need in order to construct the present course of the signal. This is done in periods of 4 seconds during EEG analysis, which can then be averaged over longer time periods to give values of 1 minute up to one hour or more.
Fig. 3 Power density spectrum for electrode position C4. Frequency is given on the abscissa in Hz, electric power on the ordinate. Please remark dominant alpha frequencies (yellow and green) under the condition of alertness, dominant delta power (red and orange) during deep sleep.
Since the number of sine waves needed for construction of such a complicated curve like the EEG, are numerous, frequencies are grouped together to give so-called frequency bands or frequency ranges. Unfortunately, there is no common sense how to define these ranges according to physiology. There are frequencies that dominate during particular physiological states and they vary according to location to a significant degree. The electric power spectrum (EPS) gives a good impression of the variations. The slowest waves are called delta waves (1-4.5 Hz, red), followed by theta waves (4.75-6.75 Hz, orange) and alpha1 waves (7-9.5 Hz, yellow). Faster waves appear within the alpha2 frequency range (9.75-12.5 Hz, green), beta1 range (12.75-18.5 Hz, light blue) and beta2 band (18.75-35 Hz, dark blue). Examples of electric power spectra for alertness and deep sleep are given in Fig. 3 for one-minute periods at one recording location (electrode position C4).
EEG During Sleep
Evaluation of the EEG by most practising doctors is performed by visual inspection. For example, according to a set of rules published by Rechtschaffen and Kales in 1968, recordings of the electroencephalogram on one or more electrode positions are evaluated in successive time periods of 20 or 30 seconds all over the night in order to discriminate between alertness, 4 different sleep stages (1-4) and so-called rapid-eye-movement (REM) sleep. In addition time latencies have been defined to reach a particular sleep stage. But interpretation of these latencies has turned out to be quite complicated. Correlations of such surrogate parameters with subjectively experienced quality of sleep have not been found. But there can be no doubt that the electrical activity of the brain contains a great deal of information on sleep and its meaning for man. The main question therefore is how to define parameters that can be correlated with the subjective experiences of a subject. The resulting profile is called a hypnogram (from the Greek word “hypnos” for sleep and “graphein” for writing). It is quite clear that this kind of analysis of sleep is time consuming (about 2 hours per subject) and asks for a considerable degree of experience. A major disadvantage of this kind of sleep evaluation is also seen in the more or less subjective assignment of single sleep periods to one of the defined stages within the given classification. There has also been a considerable effort to automate this process by computer aid. Historical staging and a computerized version of it are shown in Fig. 4.
Fig. 4 Comparison between visual staging by an experienced “sleep stager” and an automatic computer aided staging (based on the same rules) (Todorova et al., 1999).
Further progress was reached by frequency analysis of the electric activity during sleep over night. As has been shown above, deep sleep is dominated by delta and theta waves, whereas during alertness other frequencies like alpha1 and alpha2 dominate the spectrum. Following these spectra over the night, fluctuations of single frequency ranges can be seen as depicted in Fig. 5.
Fig. 5 Depiction of continuous frequency analysis (using FFT) over night for one electrode position (F3). Six frequency ranges from top down are shown: beta2, beta1, alpha2, alpha1, theta, delta. Please remark also increases of beta1 spectral power in addition to delta, theta and alpha1 indicating so-called spindle sleep (light sleep). Time is given on the abscissa in hours; electrical power is given on the ordinate as μV2/ω.
Definition of the Electrohypnogram
The electrohypnogram is defined as the time course of a special algorithm derived from spectral frequency analysis of the electric activity over night. The particular algorithm consists in the addition of theta and delta activity, divided by the beta2 activity at centro-parietal locations of the scalp. This patented algorithm has been called spectral frequency index (SFx) and has been validated against the criteria of Rechtschaffen and Kales within a clinical study (Dimpfel et al., 1998). It quantitatively describes the depth of sleep in consecutive one-minute periods. The electrohypnogram depicts the time course of depth of sleep throughout the whole night (Fig. 6)
Fig. 6 Comparison of the electrohypnogram of a healthy subject (upper panel) with that of an insomnia patient (lower panel). The ordinate gives the depth of sleep in percent of maximal alertness (spectral frequency index (SFx), the abscissa gives the time (Hammer et al., 2001). Transition from alertness to sleep occurs at a spectral frequency index (SFx) of about 80%. Deep sleep without cardiovascular reactions starts at 68% (Renz et al., 1999).
Definition of Sleep Quantity
Having the true continuous process of changing depth of sleep over the night it becomes possible to take a certain depth of sleep and to determine how much time the subject spends below this particular depth of sleep (i.e. below a SFx value of 83%). This parameter is called the “quantity of sleep (<83%)”. An example is given in Fig. 7
Fig. 7 Comparison of a good sleeper (upper panel) to a bad sleeper (lower panel) with respect to “sleep quantity (<83%)” defined as time spent below an SFx of 83%.
This parameter can now be used for further analysis. Within a sleep study comparing placebo with a valerian/hops extract mixture it was found that this fluid extract produced deeper sleep in so-called bad sleepers. This result is depicted in Fig. 8 as an example of practical use of this new parameter “sleep quantity”. The final evaluation consists in relating the “sleep quantity” to a parameter that is subjectively experienced by the patient. The common way to analyse this subjective feeling is to have the patient fill out a questionnaire. Subsets of parameters like that reflecting the restfulness of sleep (termed “quality of sleep”) can now be taken for correlation analysis. From the study mentioned above the following graph is taken, which has been accepted for publication (Dimpfel and Suter, 2008).
Fig. 8 Depiction of a result of a clinical study comparing Placebo (red) with Dormeasan® (green), a fluid extract mixture of valerian and hops. Columns represent time spent below a particular SFx value given above the column. Under Dormeasan® more time is spent in deeper sleep at all SFx values (Dimpfel and Suter, 2008).
Fig. 9 Correlation between sleep quantity (ΔT(SFx<74%) and sleep quality ΔSFA-SQE, derived from the sleep questionnaire SF-A (Görtelmeyer, 1996). Pearson correlation coefficient of r=0.477 at an error probability of p<0.0001 was reached.
Thus, the criteria for quantitative evaluation of sleep by a surrogate parameter related to sleep quality could be fulfilled for example by “sleep quantity (SFx<74%)”. Further clinical studies are necessary in order to proof the general validity of this new parameter derived from the electrohypnogram.
Berger H (1929) Über das Elektroenkephalogramm des Menschen. Arch Psychiatr Nervenkr 87: 527-570.
Dietsch G (1932) Fourier-Analyse von Elektroenkephalogrammen des Menschen. Pflüger´s Arch ges. Physiol. 230: 106-112.
Dimpfel W, Hofmann H-C, Schober F, Todorova A (1998) Validation of an EEG-Derived Spectral Frequency Index (SFx) for continuous monitoring of sleep depth in humans. Eur J Med Res; 3: 453-460.
Dimpfel W and Suter A (2008) Sleep improving effects of a single dose administration of a valerian/hops fluid extract. A double blind, randomized, placebo-controlled sleep-EEG study in a parallel design using Electrohypnograms. Eur J Med Res. 2008; 13: 1-5.
Görtelmeyer R (1996) SF-A and SF-B Schlaffragebogen A und B. In CIPS (Hrsg.), Internationale Scalen für Psychiatrie (4. überarbeitete Aufl. 1996). Göttingen: Beltz Test.
Hammer N, Todorova A, Hofmann H-C, Schober F, Vonderheid-Guth B, Dimpfel W (2001) Description of healthy and disturbed sleep by means of the spectral frequency index (SFx) – a retrospective analysis. Eur J Med Res 2001; 6: 333-44.
Rechtschaffen A, Kales A (1968) A manual for standard terminology, techniques and scoring system for sleep stages of human subjects. Public Health Service 1968; US Government, Printing Office, Washington DC.
Renz D, Dimpfel W, Schober F, Karliczek GF (1999) Bedeutung des Spektralen Frequenzindex (SFx) zur Quantifizierung der Schlaftiefe während der Narkoseeinleitung. Acta Anaesthesologica Helvetia 1999; 2: 9-11.
Todorova A, Hofmann H-C, Dimpfel W (1999) A new frequency-based automatic sleep analysis – Description of the healthy sleep. Eur J Med Res 2: 185-197
Prof. Dr. Wilfried Dimpfel c/o NeuroCode AG, D 35578 Wetzlar, Germany