J.H.K.C. Psych. (1991) 1, 16-21


Peter B.C. Fenwick

pdf Full Paper in PDF


Until the 1980s, EEG technology had remained virtually unchanged since the time of Hans Berger. It was conventional for the changes in potential measured on the surface of the scalp to be written out as a voltage fluctuation on a sheet of moving paper. The advantages of this method are that it is simple, and that there is a permanent record of all that has gone on during the EEG recording session. The disadvantage is that the analysis that is available is limited to obtaining measurements from the paper trace. It is not surprising that there has been a search for a method of transforming the EEG waveform into a more flexible format for further analysis.

With the advent of the digital computer in the 1960s, early attempts were made to convert the EEG into numbers and these numbers were then used in statistical calculations to analyse the EEG (Reymond 1967, Lehman 1971). Since these early days, the field of EEG analysis has grown dramatically, and the computer analysis of EEG traces is no longer restricted to specialist units. Modern advances in computing technology and the reduction in unit cost, has led to the development of digital EEG machines. This advance is in the process of revolutionising EEG recording, analysis and display. It has also led to a split amongst those working in the field. This split is mainly by age. Those with a long established EEG practice prefer the traces written out on paper, and argue that any other method of display is liable to mislead. The younger amongst us recognise the power of the digital computer and believe that although there is still a place for the permanent recording of EEG traces on paper, these traces can usually be held in a computer memory and reviewed visually. They also maintain that modern computers come with a graphics package and that this allows for different formats of display to highlight different aspects of the signal.

The debate about brain mapping can thus be seen to be artificial. The new generation of EEG machines will automatically have a graphics package and thus will of necessity provide the opportunity for an infinite variety of displays. I suspect that within another five years all new EEG machines will be digital, and so all new machines will have the ability to brain map and the added facility of transforming EEG data according to the whim of the investigator.



Before the raw EEG can be processed by computer, it has to be turned into numbers. This process is called analogue to digital conversion (A/D). This conversion process samples all the EEG channels at an instant in time, and records their voltage levels. These voltage levels are then transformed into numbers and stored in the computer. This process is repeated at the next instant in time and so a continuous series of numbers is built up in the computer memory. It is usual to sample the EEG 128 times a second, which will give a good pictorial resolution of the signal up to a frequency of about 25 Hz. However, for special purposes the sampling rate may need to be increased to 256 samples per second, or occasionally, on very special situations (such as recording the brain stem auditory response) up to 10 kHz per channel. Large amounts of data are thus rapidly collected and so there must be adequate storage within the computer.


Once the EEG is turned onto numbers, then it is possible to illustrate any aspect of it. Brain mapping machines initially display the EEG channels on the screen in the conventional format. However, because the traces are stored as numbers, there is a flexibility which cannot be achieved in a conventional EEG machine. The traces may be expanded or reduced, they can be highlighted, overlaid one on another, turned upside down or manipulated in any number of ways. This makes it much easier to get detailed information from them. Colour aids the eye, and so brain mapping machines allow the traces to appear in colour on the screen.

Early workers in the field attempted to map the distribution of the potential field on the surface of the scalp. With modern graphics packages this is now very simple. On the head map, points of equal potential are joined together into contours. The contour levels are usually coloured, so that extremes of potential are coloured red if they are positive and blue if they are negative. Intermediate levels have intermediate colour shades (see Fig. 1). This makes it very easy to see what the potential field is at that moment. By moving the computer cursor, subsequent moments can be mapped, and thus a picture of the evolution of cerebral potentials across time can be mapped. Colours do, howev­ er, introduce errors. The eye can very easily pick out a colour boundary and will allocate to a boundary a special meaning. The colour boundaries may not have any significance of their own, but will appear to do so because of the colour change. Some brain mappers take this into account, and provide special colour scales with blurred edges, so that the change from one shade to another is not so abrupt.



The new time maps open up a totally different field of analysis. The evolution of scalp potentials can be followed over short epochs of time. It is thus possible to map the earliest potential field change and its location on the head as an epileptic spike is starting to arise. This change can be followed moment by moment so that it is possible to see where the potential is arising, and how it spreads. This may give important information about the locus of the epileptic discharge, and the areas into which it is spreading. If surgery is contemplated, this can be of great importance. If, for example, there are two temporal lobe spikes, it is often impossible to tell from the routine paper EEG from which side of the head they are arising, and whether they are linked together by conduction (Fig. 2).

The evolution of potential fields on the scalp can give further information into the functioning of the brain. Lehman et al (1987) has drawn attention to the way the EEG shows micro-segmentation. This segmentation is shown by potential field patterns which tend to last for a few seconds before being replaced by a different pattern. Lehman suggests that each state may be physiologically different, and in support of this, he has produced some data to show that there is a link with cognition. One pattern may correlate with visual imagery, another sound imagery. He suggests that the stability of these global field patterns may correlate with individual thoughts!

A possible future use for these instantaneous brain maps would be to set up a pattern recognition programme to detect particular map configurations. When a specific map pattern appears. then the patient could be stimulated or asked to do a task. It could be predicted that learning may be more easily carried out in oi:e brain state than in another. More recently, Binnie (1986) has shown small transient cognitive impairments (TCis) following epileptic spikes in different brain areas. It is thus quite clear that the microstructure of the potential fields on the surface of the scalp can reflect different underlying physiological mechanisms, and that the mapping of these states can give significant insight into brain functioning.


The EEG waveforms can be analysed in the frequency domain. This means analysing the frequency component of each trace. This is done by the fast Fourier transform, which finds the power in single Hz steps up to the highest frequency, which is usually about 40 Hz for EEG work. These individual frequencis are then commonly combined into the EEG bands (delta, < 4Hz, theta 4-8 Hz, alpha, 8- 13 Hz, and the beta bands, > 13 Hz.). To make the map, the contours of equal power points are plotted out on the map, and thus a colour map is produced which at times looks similar to that of the instantaneous potential amplitude mentioned above, but which contains quite different information. The banded power maps are the maps which are most commonly used. The earliest work on banded power maps was carried out by Grey Walter in the UK during the 1950s. One of the first groups to make brain mapping popular was that of Duffy et al (1979) at Harvard.


A simple example of the use of banded maps is given by the asymmetries which are normally seen in the alpha map of a healthy subject. It has been known for many years that the alpha activity is highest over the non-dominant hemisphere. Indeed, it is routine in EEG departments to ask the handedness of the patient so that this can be taken into account. This information is seldom used when reading an EEG record, and discrepancies between handedness and side of alpha power are not routinely discussed. The reason for this is that it is quite difficult to be certain where the maximum alpha amplitude is to be found. Brain maps of the alpha band, however, show the asymmetries of alpha activity very clearly. It is thus easy to decide whether the alpha brain map that you are looking at comes from a left or right handed subject (Fig. 3). A left hemisphere prominent alpha map in a right handed subject raises the question of possible right hemisphere damage, either recent or during childhood. This should certainly alert the clinician to possible organic factors in this patient's illness.

In the laboratory, on a day to day basis, banded frequency maps are helpful in diagnosis. They are usually used in conjunction with interpretation of the routine EEG recording. Many laboratories are, however, beginning to use brain mapping systems as their major diagnostic device. Dr. Hamburger of the Children's Hospital, Amsterdam, in a lecture reviewing the work of his department over the previous year, concluded that brain mapping contributed additional information in over 70% of the cases.

One further use of the mapper is to detect the changes in spectral maps which are due to a treatment procedure. This has been successful following drug therapy, ECT, and psychotherapy. The baseline banded frequency brain maps are recorded, and stored in the machine. The treatment procedure is then carried out, and this map is also stored. The control map is then subtracted from the post-treatment map and the resultant map shows the net effect of treatment. It is thus possible to say whether specific treatments can alter brain activity, and if they do, whether this is related to therapeutic outcome.

A 16 year old boy of impeccable previous character and behaviour spent an evening drinking half a bottle of martini with a friend. He became drunk, but then had an alteration in behaviour. He had an alcoholic blackout, during which he assaulted a 30 year old woman and battered her to death. Following the attack he was clearly in an abnormal state of consciousness that could not be accounted for by the alcohol. Routine EEG was normal. However, brain mapping of the EEG under an alcohol load showed clear evidence of a right anterior temporal delta focus, suggesting the presence of right temporal damage. This was confirmed on neuropsychometric testing and he showed a dilated right temporal horn on MRI scan.

The limitation of brain mapping in determining the presence of underlying structural brain disorder has yet to be defined. Poimann et al (1989) in a study of 14 patients showed a high correlation between the site of an epileptic focus, the presence of an underlying tumour as detected by the CT scan, and an equivalent abnormality in the brain map. Clearly, further studies are required.

Gunther (1989) used brain mapping in an interesting way, following on from the work of Morstyn et al (1983), by attempting to activate an abnormal response in the schizophrenic brain. He asked both treated and untreated schizophrenics, Type 1 and Type 2, to make repetitive hand movements. He was able to show that for the Type 1 schizophrenics, in the resting condition compared to normals there was increased delta activity in all electrodes, but with predominance in the frontal regions. The theta band showed increased power values bifrontally and in the regions, while the alpha band showed lower power with the greatest difference being over the left hemisphere posteriorly. With activation, there were changes in the delta, theta and beta bands which discriminated them from the normals. For the Type 2 schizophrenics, after activation there was a decrease in delta power, the theta, alpha and beta bands remained virtually unchanged. Untreated schizophrenic patients' maps became normalised with treatment.

Numerous studies have shown differences in the frequency profiles of patients with schizophrenia, affective, and dementing illnesses. A major review is given in Maurer, (1989). However, it is not the purpose of this article to review these studies in detail, but to draw attention to the contribution that brain mapping has already made, and will continue to make, to the diagnosis and detection of psychiatric illness.


Following a stimulus, the brain shows a series of potentials. The early potentials, up to about 20 milliseconds, show information about the integrity of the tracts leading to the brain. The late potentials, from about 100-230 milliseconds, indicate the arrival and ·elaboration of the stimulus in the cortex. The late cognitive potentials are all those waves which occur after 230 milliseconds, and which are DEPENDENT on the psychological set and task performance of the patient.

It is these late cognitive potentials which have been mainly used in psychiatry. Latency measurements are derived from stimulus onset, and the polarity of the wave is either negative or positive. Thus, the P300 which is the most widely used potential in psychiatry is a positive wave at about 300 milliseconds.

Brain mapping an evoked potential is essentially similar to creating an instantaneous time map. An instant in time, usually the peak of a wave, is defined by the user, and the instantaneous amplitude profile is mapped. The brain map then shows how the amplitude of the wave at this point is distributed over the head.


It rapidly became apparent that brain maps gave considerably more information about the P300 than the routine EEG, by easily showing up asymmetries in the distribution of the peak. The P300 is usually symmetrical and centered in the parietal region (Fig. 1. a P300 evoked potential). However, in schizophrenia it is frequently asymmetrical and moves forward, although the exact change depends on the patient population and the type of schizophrenia. Another complexity is that there may be a difference in latency between the potential maxima on different parts of the scalp. For example, the P300 may have a latency of 350 milliseconds in the central parietal region, whereas it develops earlier at 300 milliseconds in the frontal region. This difficulty is not easily resolved as there is no clear way of deciding which is the most important latency. Lehman (1971) has suggested that evoked potential peaks should no longer be measured singly. He has suggested that an average of the activity at all electrode sites should be calculated (the golbal field power), and that this value should be used to determine the latency of the main peaks. The advantage of calculating the global field power is to produce a point of maximum power for the whole head, thus minimising spatial differences; but the disadvantage is to conceal the information which is contributed by the differential development of the potential in different brain areas.

The P300 has been used in the diagnosis of schzophrenia, affective illness, Alzheimer's disease, brain damage, epilepsy and in detecting the children of alcoholics, who may later be at risk of alcoholism. Brain mapping has much to contribute in these areas, as a routine brain map of a P300 potentail quickly shows up the asymmetry (Morstyn et al, 1983; Pfefferbaum et al, 1984; Maurer et al, 1989). Sequential maps plotted every few milliseconds throughout the development and evolution of the potential allow a more precise location of the peak, both in time and in space.

The routine use of P300 as a screening measure can provide evidence of cortical dysfunction. Grossly asymmetrical responses should warn the clinician that further investigation may be necessary. In my unit we routinely use the P300 to indicate the side of the epileptic focus in patients being prepared for epilepsy surgery. Of 20 such patients examined, all of whom had temporal lobe lesions with unilateral foci, the asymmetry of the P300 accurately predicted the side of the lesion in 19. All these patients had mesial temporal sclerosis in their pathological specimen.

With Dr. Holland and his team we are at present conducting a P300 study into patients with Down's syndrome who were identified in a larger survey. The study is still in progress; however, early results show that there is a close correlation between the latency of the P300 and whether or not the Downs patient has demented. Many of those patients with dementia also show grossly asymmetrical responses.


It is important to be able to test the maps of a patient population against those of a normal population. Several different ways of doing this have been proposed. The standard way is to calculate the T statistic for each point on the map between the two populations and then print out a new map which shows the value of T at each point on the map (Fig. 4). Although this is the accepted method, considerable doubt has recently been cast on its validity. At least 20 T tests are carried out on each map, and as there may be many maps being tested in any study, large numbers of T tests will be calculated and thus the probability of significance by chance will be greatly increased. It can be estimated that a clinical study may produce about 680 variables being tested. Thus at the 0.05 probability level, around 25 variables will be significant by chance alone, and around 7 variables at the 0.01 level. When such large numbers of comparisons are made, an alteration in the probability level of the T statistic is required.

In a clinical setting, it is important to know whether a single individual is abnormal when compared to the normative group. In this case, the patient's Z score is used, which indicates the number of standard deviations the observation differs from that of the mean of the reference set. The Z value for each point on the map is calculated and plotted in a similar way to that of the T statistic. Thus the colour of the map indicates the number of standard deviations that that point on the map differs from the mean of the reference population. (Duffy et al 1981, John et al 1977, John et al 1980). As Duffy (1989) and others have pointed out, the clinical usefulness of this technique depends on the statistical distribution of the reference population. Often, a specific normal individual may lie within the distribution of an abnormal population, and so cannot be distinguished as different. Thus the Z scores should be interpreted with caution.

Most studies using brain mapping which have been carried out since the mid- l980s make use of this method of statistical evaluation. Exciting advances have been made in the diagnosis of schizophrenia, affective illnesses, Alzheimer's disease, learning disorders, and disorders of behaviour. There is little doubt that computerised EEG with the graphics package of brain mapping and the powerful statistical methods that can be applied will rapidly advance the field even further. The 1950s were an exciting time, when EEG was thought to provide a way to increase our understanding of brain function in psychiatric illness. The 1960s showed that the crude traces produced by the current technology was too inaccurate to give us the detailed knowledge that we required. The advent of the digital EEG machine has changed all this, and there is little doubt that brain mapping and digital EEG are the techniques of the 20th Century.


Defining the potential field maps of brain activity is a necessary first step in the location of the underlying cerebral generators. There are considerable difficulties as all modelling strategies require precise details concerning the size and configuration of the electrical generator and the characteristics of the medium through which the electrical fields pass. This knowledge is only availabel for superficial sources and thus considerable inaccuracy results for the deeper generators. At the Northern European Brain Mapping Conference in 1990, Dr. Velus from Hemsteeder and the group of Dr. Lopes de Silva have shown that in patients with depth electrodes implanted in different brain areas, the accuracy of location of generators at the tips of the electrodes could only be achieved to within 2-4 ems. The accuracy depended on how close the generators were to the cerebral ventricles, or the air-filled cavities of the skull sinuses. Thus, precise generator location in the depth by means of the EEG is unlikely to achieve the accuracy that may be required for precise localisation of discharging foci.


The new generation of brain mappers has simple modelling programmes which allocate precise locations within the head to the generators of electrical discharges whether spontaneous or evoked. More complex modelling programmes will be available in the next few years, and at that time it may be more convenient to refer to the presumed generator location of an evoked response such as a P300 rather than to the maximum of its field pattern. However satisfactory the model, there will always be some doubt as to the precise location, because of the unknown characteristics of that individual's brain. Magnetic technology may hold the answer.

The measurement at the scalp of magnetic fields arising from the brain adds a further dimension to the analysis of brain activity. Magnetic fields pass through the brain substance, skull and scalp, unaccounted, and furthermore, the measurement of their surface field patterns allows the location of the generating source within the depth of the brain. The technology for measuring magnetic fields is still in its infancy, and requires superconducting technology which at present involves the use of liquid helium (Bath et al 1984, Sato et al 1985). However, the new generation of high temperature superconductors may avoid the necessity for helium, and so make the technology simpler. The EEG laboratory of the future will combine both EEG and MEG measurements to give a truly detailed picture of brain activity and function.


Bath, D.S., Sutherling, W., Engel, J., Beatty, J. (1984). Neuromagnetic evidence of spatially distributed sources underlying epileptic spikes in the human brain. Science. 223: 293-296.

Binnie, C., (1986) The inter-ictal EEG. In: Trimble M.R., Reynolds,

E.H. (eds). What is epilepsy? Churchill Living stone, Edinburgh. Duffy, F.H., Burchfiel, J.L., Lombroso, C.T. (1979). Brain electrical activity mapping (BEAM): a method for extending the clinical utility of EEG and evoked potential data. An. Neurol. 5: 455- 462.

Durry, F.H., Bartels, P.H., Burchfiel, J.L. (1981). Significance probability mapping: an aid in the topographic analysis of brain electrical activity. Electroenceph. Clin. Neurophysiol. 51: 455- 462.

Duffy, F.H. (1989). Topographic mapping of brain electrical activity: clinical applications and issues. In: Topographic Brain Mapping of EEG and Evoked Potentials. Ed: K. Maurer. pp. 19-52. Springer-Verlag, Berlin.

Gunther, W., Steinberg, R., Petsch, R., Strech, P., Kugler, J. (1989). EEG mapping in psychiatry: studies on type I/II schizophrenia using motor activation. In: Topographic Brain Mapping of EEG and Evoked Potentials. Ed: Maurer, K. pp. 438-450. Springer Verlag, Berlin.

John, E.R., Carmel, B.Z., Comigh, W.C. et al. (1977). Neurometrics. Science 196: 1393-1410.

John, E.R., Ahn, H., Princhep, L. (1980). Developmental equations for the electroencephalogram. Science 210: 1255-1258.

Lehman, D. (1971). Multichannel topography of human alpha EEG fields. Electroenceph. Clin. Neurophysiol. 31: 439-449.

Lehman, D., Ozaki, H., Pal, I. (1987). EEG alpha map series: brain microstc1tes by space-oriented adaptive segmentation. Electroenceph. Clin. Neurophysiol. 67: 271-288.

Maurer, K., Dierks, T., Ihl, R., Laux, G. (1989). Mapping of evoked potentials in normals and patients with psychiatric diseases. In: Topographic Brain Mapping of EEG and Evoked Brain Potentials. Ed: Maurer, K. pp. 458-473. Springer-Verlag, Berlin.

Morstyn, R., Duffy, F.H. McCarley, R.W. (1983). Altered P300 topography in schizophrenia. Arch. General Psychiatry. 40: 729-734.

Pfeberbaum, A. Wenegrat, B.G., ford, J.M., Roth, W.T., Kopell, B.S. (1984). Clinical application of the P3 component of eventrelated potentials: II. dementia, depression and schizophrenia. Electroenceph. Clin. Neurophysiol. 59: 104-124.

Poimann, H., Maurer, K., Dierks, T. (1989). Mapping of EEG in patients with intercranial structural lesions. In: Topographic Brain Mapping of EEG and Evoked Potentials. Ed: K. Maurer. pp. 278-284. Springer-Verlag, Berlin.

Reymond, (1969). Chronotopograms. EEG Clin. Neurophysiol. Sato, S., Douglas, R., Roger, P. (1985). Single magnetic spike mapping. In: Weinberg, H., Stroink, G., Catilla, T. (Eds). Biomagnetism: applications and theory. Pergamon, New York, pp. 261-265.

Peter B.C. Fenwick MBBChir (Cantab), FRCPsych. Consultant Psychiatrist, Neuropsychiatry and Epilepsy Unit, The Maudsley Hospital, Denmark Hill, London, SES BAZ., United Kingdom.

View My Stats