WO1999053836A1

WO1999053836A1 - Method and device for diverting an electroencephalogram in a nuclear spin tomograph

Info

Publication number: WO1999053836A1
Application number: PCT/DE1999/001149
Authority: WO
Inventors: Alexander Hoffmann
Original assignee: Alexander Hoffmann
Priority date: 1998-04-17
Filing date: 1999-04-16
Publication date: 1999-10-28
Also published as: DE19980653D2; DE19817094A1; EP1071368A1

Abstract

A method and device for diverting an electroencephalogram in nuclear spin tomographs, whereby a static high magnetic field is generated in order to align the nuclear spin, the nuclear spin is excited in a circular polarized field in the radio frequency range and three switched magnetic gradient fields are produced for local coding of the nuclear resonance signal in three directions in space. The EEG signals that are diverted from the head of the patient inside a nuclear spin tomograph tube (1) by means of EEG diverter electrodes (4) are subjected to spectral analysis and are filtered in a digital signal processing path in order to eliminate disturbances caused by the switched field gradients in the active nuclear spin tomograph.

Description

Method and device for deriving an electroencephalogram in an MRI scanner

For some years now, efforts have been made to derive an electroencephalogram (EEG) in an MRI scanner. The clinical areas of both brain research and brain diagnostics are interested in the simultaneous use of EEG and MRI. This is due to the fact that EEG and MRI can optimally complement each other when examining and displaying processes in the brain.

The functional imaging with the MRT provides an unsurpassed spatial resolution of the active regions of the brain, but has the disadvantage that the imaging takes place with a time delay and therefore fast processes in the brain can only be insufficiently examined. This is also due to the fact that the processes to which MRI imaging responds, namely the blood flow to active regions in the brain occur with a time delay when activated.

The EEG, on the other hand, provides a real-time signal with which even fast processes can be detected instantaneously, but the spatial resolution, especially the resolution of the depth of origin, is clear compared to the spatial resolution achieved with the MRT due to the signal derivation with surface electrodes limited. The EEG also enables functional examinations in which evoked potentials can be recorded, that is, response potentials to a generated external stimulus (e.g. a noise). - 2 -

Therefore, the two major diagnostic methods of neurology with MRI and EEG could complement each other if an EEG could be derived in MRI. So far, however, it has not been possible to obtain useful results with the simultaneous use of EEG and MRI. This is due to the way the MRI and EEG work, which will be briefly outlined for a better understanding.

For imaging, the MRT uses a static magnetic field to align the nuclear spins, a circularly polarized alternating field in the radio frequency range to excite the nuclear spins, and finally three switched magnetic gradient fields for the spatial coding of the nuclear magnetic resonance signal in all spatial directions. The static high magnetic field is also present when the MRI is inactive.

The EEG arises from voltage fluctuations in the range of 50 microvolts on the head surface, which are generated by the synchronous activity of many nerve cells in the cerebral cortex. These are derived and fed to an amplifier. Because of the extremely small size of these EEG signals, they are of course extremely sensitive to external interference fields.

On the other hand, in order to achieve a sufficient signal-to-noise ratio, the MRI can only work in a perfect Faraday's cage, which in no case may be injured by electrical cables passed through it. For this reason, all signals from the EEG amplifier are transmitted from the shielded room through light guides and not via electrical cables. The electrodes for deriving the EEG must not contain any ferro- or diamagnetic substances so that they do not interfere with the homogeneous magnetic field in the MRI. But even eddy currents, which are induced by the switched gradient fields of the MRT in electrically conductive surfaces, i.e. in EEG electrodes, already generate an opposing field, which also disturbs the homogeneous field and leads to artifacts in the data line of the corresponding gradient. The derivation of an EEG in MRI has so far been unsatisfactory even in inactive MRI. This is due to the fact that the static high magnetic field is also present in the inactive MRT and the pulsating blood currents generate interference signals which superimpose and disrupt the EEG signals. These disorders are known as pulse artifacts. These can initially occur in the form of movement artifacts, since the pulsating blood flow triggers corresponding pulsating head movements of the patient in the MRI, which can be expressed in a slight pulse-synchronous nod. This in turn requires a corresponding slight movement of the EEG electrodes attached to the head with the associated lines in the static magnetic field, as a result of which interference voltages are induced and appear in the EEG signal in a pulse-synchronous manner. In addition, pulse artifacts also occur in a second form, which will be described in detail below, due to ion separation processes in the blood flow, which fluctuates due to the pulse. Due to the superimposed pulse artifacts, the EEG signal is largely unusable. However, the derivation of an EEG in inactive MRT is of essential diagnostic importance, as will be illustrated below using an example.

In the case of an additional MRI measurement, however, the EEG was previously completely disturbed because the switched gradients during the imaging by the MRI lead to the complete illegibility of the EEG.

The object of the invention is therefore on the one hand to create a possibility to eliminate the interference of the measuring MRI on the quality of an EEG if possible, and on the other hand to eliminate the influence of pulse artifacts on an EEG if possible, in order to thereby use MRI and EEG simultaneously to bring and to be able to derive an EEG of diagnostic quality even in inactive MRI.

Two examples should clarify the importance of this task:

Severe and often occurring epilepsy are often focal - 4 -

jump, i.e. they start from a certain point in the brain and can usually only be treated by determining the focus in the brain as precisely as possible and then reducing it surgically. These focus regions generate spikes typical of epilepsy in the EEG and can thus be identified. An epileptic seizure announces itself earlier in the EEG than could be identified by other external characteristics. In order to increase the spatial resolution of the origin of these signals, 64 electrodes and more are used. A combination of EEG with simultaneous functional imaging in MRI can make the localization of an epileptic focus even more reliable and, above all, determine the time for the seizure to be recorded. Attempts have already been made to do this, but the EEG was completely disrupted by the switched gradients during data acquisition by the MRI.

From the just mentioned fact that an epileptic seizure announces itself early on in the EEG, the importance of the derivation of a diagnostically usable EEG in the inactive MRT becomes clear. An EEG derived in the still inactive MRI can be used to precisely determine the onset of an epileptic seizure and then switch on the imaging by means of the MRI, in order to then determine the epileptic focus with great precision during the epileptic seizure by means of combined measurement with MRI and EEG and thus the basis for to be able to win the exact surgical treatment. In the case of pulse artifacts superimposed on the EEG signal, on the other hand, the beginning of an epileptic seizure would not be recognizable from the EEG signal. - 5 -

Another example is brain research. In brain processes, for example in thinking processes, in response to any external stimuli of the sensory organs etc., the activated brain regions are supplied with blood considerably more. With the EEG, the processes taking place in the brain can be recorded and localized approximately; With MRI, the areas of increased blood flow can be localized very precisely with high spatial resolution. A combined use of MRI and EEG can provide detailed information about brain areas responsible for certain functions and processes.

The invention achieves the object presented by a method and a device as specified in the patent claims. Accordingly, the invention works by eliminating the interference in the EEG by digital signal processing with spectral analysis and filtering of the EEG signals and with triggered subtraction of the pulse artifacts in order to eliminate them and thereby create an EEG in MRI with diagnostic quality.

The invention will be explained in more detail below with reference to the accompanying drawings, in which:

1 shows a schematic representation of the apparatus for deriving an EEG in the MRI,

2 shows some pulse shapes and their discrete spectra,

3 shows a spectrum of a trapezoidal gradient that is often used,

4 shows a comparison of an unfiltered and a filtered EEG signal from the MRT, Fig. 5 signal images that illustrate the reduction of pulse artifacts.

1 shows in a very schematic representation a preferred arrangement for performing the EEG derivation in the MRT. The MRI tube 1 is arranged in a shielded tomography room 2. A static high magnetic field acts in the MRT tube 1, which is indicated by an arrow. The patient's head in the MRT tube is located within an RF head coil 3, via which the high-frequency alternating field is generated to excite the nuclear spins. Further coils in the MRT tube 1, which are not shown for the sake of simplicity of illustration, are used to generate the switched gradient fields for the three spatial coordinates for spatial coding of the nuclear magnetic resonance signals in the three spatial directions. The lines of the individual EEG lead electrodes 4 on the head surface of the patient are, for example, fixed on the head with a hood and bundled centrally on the head and routed to an EEG amplifier 5 as close as possible to the center of the MRT tube 1. From the EEG amplifier 5, the signals are passed via an optical fiber 6 out of the shielded tomography room to a light receiver 7, which converts the light signals back into electrical signals and feeds them to an EEG recorder 8, to which a PC 9 is connected.

The bundling of the cables at the head and the avoidance of conductor loops reduce the area into which the gradients or the magnetic part of the high frequency can couple. If you move the cables away from the edge of the MRT tube, the influence of the switched gradients is less.

The EEG amplifier 5 must behave magnetically neutral in the static high magnetic field of the MRT tube 1, ie the field lines must be able to pass the amplifier essentially unchanged and must not be distorted. Therefore, ferro- or diamagnetic substances in the components of the - 7 -

starters 5 can be avoided. Likewise, the electrodes must not contain any ferromagnetic or diamagnetic substances in order not to disturb the homogeneous magnetic field, because by superimposing a defined inhomogeneous gradient field, the location of the nuclear magnetic resonance is encoded via the changed resonance frequency. Since eddy currents, which are induced by the switched gradient fields in conductive surfaces, also generate an opposing field that disturbs the homogeneous magnetic field, the electrodes are preferably formed from amorphous sintered Ag-AgCl with low electrical conductivity in order to make the formation of eddy currents more difficult.

The inputs of the EEG amplifier 5 are very high-resistance due to the use of an FET, so that there is practically no current flow. The FET enables a high resistance with low thermal noise. The potentials derived at the electrodes are amplified accordingly in the EEG amplifier 5 and fed to an A / D converter using the multiplex method, the digital output signal of which is fed into the light guide 6 and fed to the light receiver 7.

As explained at the beginning, there are two basic interferences on the EEG derivation in the MRT, which render the EEG measurement result unusable.

The one interference factor is the field gradients switched in the active MRT for spatial resolution in the xyz coordinate space, which cause interference signals superimposed on the EEG output signal.

The second disruptive factor is pulse artifacts due to the pulsating blood flows in the patient's head, the cause of which was previously unknown. These pulse artifacts have so far largely made the EEG results unusable even in inactive MRI. It has been assumed that these pulse artifacts are actually only movement artifacts caused by ballistocardial movement of the head due to the pulsation of the blood flow in the large arteries. However, it has been shown that even when using artificial handles real Although motion artifacts could be greatly reduced, the pulse artifacts continued to occur.

This further cause of the pulse artifacts, which have nothing to do with a head movement, lies in the fact that ion separation occurs in blood vessels with a flow velocity component perpendicular to the static high magnetic field in the MRI due to the high magnetic field, whereby voltages are induced which are dependent on the vessel diameter, current density, Charge carrier concentration in the blood stream and the volume of blood transported per unit of time is dependent. For example, a maximum voltage value of approximately 150 microvolts occurs between the vessel walls of a blood vessel with an inner diameter of approximately 1 mm and an assumed maximum flow rate of blood of 10 cm / s in a magnetic field of 1.5 T, which generates pulse artifacts.

Pulse artifacts are very similar to an alpha activity in the EEG, because this alpha activity lies in the same frequency and amplitude range as a pulse artifact and is only longer pronounced. Therefore, pulse artifacts in the EEG signal cannot be accepted if the EEG derivation is to be informative.

With the present invention, the two explained interference factors in the EEG derivation are eliminated by digital signal processing of the derived EEG signals, thereby making the pure, unadulterated EEG signal visible.

The disturbances impressed on the EEG signals by the switched field gradients in the active MRT are eliminated by spectral analysis and filtering. First of all, you have to determine the frequencies that are coupled into the EEG by the switched gradients. The Nyquist theorem states that the maximum visible frequency is determined by half the sampling rate. The EEG sampling rate is, for example, 500 Hz. A harmonic oscillation whose frequency is higher than half the sampling rate reappears as a result of the sampling - 9 -

as a harmonic vibration of lower frequency. This effect is avoided by a low-pass filter at the input of the EEG amplifier 5, the cut-off frequency of which is below half the sampling rate. With spectral analysis it is also important that the time segment to be analyzed is folded with a window function. As a result, the section can be continuously continued periodically, which is a condition for a Fourier transformation. The resolution of the frequency is

4 Δ f Δ T where lT is the analyzed time range.

Common gradient shapes are triangles, trapezoids and harmonic vibrations. Since the circuits are periodic, they have a second spectrum. The distance between these peaks is determined by the repetition rate of the individual pulses; the amplitude results from the repeated pulse shape, whereby individual peaks can be without any power component.

2 shows some pulse shapes and their discrete spectra. Here ö is the frequency with which the pulse shape is repeated periodically. The respective pulse shape (left representation) is compared with the associated spectral analysis (right representation) in FIG. 2.

3 shows a spectrum of a trapezoidal gradient that is often used. The share of the EEG is visible in the range from 0 to 40 Hz; the proportion of alpha activity around 12 Hz can be clearly seen. The discrete spectrum consists of multiples of 33.3 Hz; so the individual gradient pulses are repeated periodically at 33.3 Hz. The trapezoidal pulse shape is defined by the amplitudes of the individual peaks. In practice, three gradients are switched, the contributions of which add up. In order to make an EEG visible in its actual form by filtering, the repetition rate should not be below 30 Hz. - 10 -

If only a low pass is used at the input of the EEG amplifier to filter out the interference frequencies impressed on the EEG signals by the switched field gradients in the active MRT, the cutoff frequency of the low pass must be chosen to be relatively low. But that also has its disadvantages. Because although the relevant range of brain waves to be recorded with the EEG is in the frequency range from 0 to 40 Hz, it is not desirable to cut off all higher frequencies by a low pass. This has the consequence that interesting higher-frequency areas in the EEG signal are suppressed and that the spikes, which are typical for epileptic seizures and are characterized by their peaks, appear blurred, because their characteristic peaks then appear rounded in the filtered signal and no longer appear sufficiently , so that an EEG signal filtered in this way has only a significantly lower diagnostic value.

It is therefore desirable to be able to select the highest possible limit frequency for a low-pass filter arranged at the EEG amplifier input, but without falsifying any other falsification of the EEG signals by MRI-specific interference frequencies.

This is achieved by using a combination of a low-pass filter with a higher cut-off frequency and very narrow-band bandstopper, which can be achieved with digital data processing. For this purpose, the interference frequencies caused by the MRI operation must be determined. The interference frequencies are determined by the sequence and nesting of the program that determines the measurement sequence of the MRI. The program loops for the scanning layers, the image lines and the number of image acquisitions are of particular importance. For example, the above-mentioned frequency of 33 Hz and multiples thereof arise as interference frequencies from the frequency program and from 7 Hz through switching processes within the program. - 11 -

The filter problem can therefore be solved by using additional bandstoppers with narrow blocking frequency bands around the respectively identified interference frequencies within the pass band of the low pass.

Instead of several bandstops, other filters can be used, the characteristics of which suppress the multiple of a basic interference frequency.

For example, 10th-order or higher-order Butterworth filters can be used for filtering. Butterworth filters have a low ripple in the frequency response, but must be used in a higher order in order to cut the interference as sharply as possible at the limit of the frequency range of the EEG. The filters used in conventional EEG amplifiers are of a low order and cannot suppress the high power of the interference frequency close to the EEG.

All channels of the EEG are simultaneously filtered by an identical filter.

FIG. 4 shows an example of filtering, the upper diagram showing the unfiltered EEG signal and the lower diagram showing the filtered EEG signal. The effect caused by the filtering is striking.

In principle, Butterworth filters with a lower order can of course also be used, but generally higher results are obtained with a higher order of the filters.

Fourier filters can also be used instead of Butterworth filters.

The elimination of the interference frequencies can also be done directly in the

Fourier spectrum of the signal take place. This will be the spectrum - 12 -

the recording sequence previously determined. The interfering frequencies can be eliminated directly and with an accuracy that is proportional to the length of the Fourier transformed signal in the Fourier spectrum. The signal is obtained through a reverse transformation without interference.

The use of filters is cheaper for online signal processing. In signal post-processing, where a time delay is not important, the latter possibility of eliminating the interference frequencies in the Fourier spectrum is cheaper because of the even better filter quality.

The interference factor based on the pulse artifacts is eliminated by triggered subtraction of the pulse artifacts. In order to be able to eliminate a pulse artifact, it must first be clear which part of the signal is the EEG and which part is the disturbance. The temporal waveform of the artifact is determined by averaging.

This is due to the fact that the EEG signal is an undetermined stationary signal. It is made up of many harmonic vibrations, the power components of which change non-periodically and influenced by the respective situation (not determined), but the overall power remains largely constant (stationary). Because of this indeterminacy, the influence of the EEG on the wave train is reduced with n averaging of the pulse artifacts

with the factor 1 FΓ

When averaging, it is important that the individual wave trains are added up in time. Because the pulse artifacts are in amplitude - 13 -

and frequency are similar to the EEG, the exact time of the event cannot be determined from the superimposition of both signals. A second channel must therefore be used, with which only the pulse with a temporally sharp signal is registered. An EKG is derived at the same time. The pointed R-wave in the EKG, which can also be seen clearly on the MRI, has proven itself. Since the heart's pulse rate fluctuates slightly, the minimum distance between two R-waves must be determined, which gives the length of the area to be averaged. This time range is placed over each pulse artifact so that it lies in the middle. These areas of equal length are then subjected to a coherence analysis and the coherence efficiency is output. In this way, it can be checked whether a subsequent averaging is useful or whether it is due to strong EEG activity or through

Movement artifacts differentiate the individual areas. After averaging over a sufficient number of, for example, approximately 15 events, the waveform obtained is then folded using a window function, and thus both the beginning and the end of the pulse artifact are reduced to zero.

This means that no stages are created during the subsequent subtraction. The areas that were used for averaging are then subtracted from the averaged wave train and the EEG signal remains. Alpha activities and other forms of EEG can be clearly recognized again, which was not possible before subtraction.

In practice, pulse-triggered subtraction of pulse artifacts takes place in two steps, namely a calibration routine and an online subtraction.

In the calibration routine, EEG signals from all channels and an EKG signal with the patient's eyes open are recorded and displayed for a period of, for example, 30 seconds. It must be filtered with a high pass filter of 0.1 Hz or more. The user can then - 14 -

finally decide for which channels (depending on the interference of pulse artifacts, the pulse-synchronous portion is to be determined and later subtracted online).

The trigger event is the signal increase in the ECG signal recorded parallel to the ECG at the R wave. The average (D) and maximum (M) of the ECG signals recorded over the period of 30 seconds are determined. The mean (T) is then calculated from the maximum (M) and the average (D) of the ECG section. T is the triggering amplitude of the R wave of the EKG. For this purpose, the ECG is compared with T from the beginning of the 30-second period. As soon as a value is greater than T, this point in time is a trigger event (E). For the next 300 milliseconds, the EKG is not compared to T; then the comparison with T is continued again and the next trigger event is determined. The first and the last determined trigger event are discarded.

The pulse-synchronous interference component in the EEG is now determined by calculating the minimum time interval between two trigger signals (E). This time will be reduced by another 10%. This determines the maximum length (L) of the pulse-synchronous interference component that can be subtracted.

In each EEG channel selected by the user, the pulse sections of length L are compared with each other with a coherence analysis from each trigger event (E). If the variance exceeds a certain value, the calibration routine must be repeated. Otherwise, the pulse sections are added and divided by the number of E determined. The averaged pulse section is folded using a window function. The interference signal (P) determined in this way is used for later online subtraction.

The flow of the calibration routine is clearly shown in the flow chart below: - 15th

Calibration routine

30 s EEG and EKG willows recorded

1

Determine the speed rush from the EKG

Search for the times of the day

Discard the first and last point in time

1

Determination of the minimum time interval between z triggers

KoharenzanaK se dei interference signals

Averaging the pulses

Folding with window function

The on-line subtraction of the pulse artifacts now takes place.

As soon as oneline subtraction is permitted after the calibration routine, a trigger signal is generated whenever the signal value of the EKG exceeds the value T that was determined in the calibration routine. From this point in time, data point by data point is subtracted from each selected EEG channel signal P over the length L and the cleaned EEG signal is obtained. - 16 -

Instead of using a calibration routine carried out separately and before the actual measurement, the ECG-synchronous pulse artifacts can also be determined by means of ongoing calibration. For this purpose, the last 10 to 20 pulse artifacts in each case are averaged by means of an ECG trigger and subtracted from the pulse artifact that then follows. The area used for averaging thus migrates with the EEG signal and thus always gives the current form of the artifact.

In this method for the ECG-synchronous determination of the pulse artifacts, the length L of the pulse sections is of course variable.

This latter possibility of determining the pulse artifacts also adjusts itself automatically to any changes in the heart rate and thereby achieves greater accuracy. Instead of carrying out a separate calibration routine method before the start of the measurement, it is also possible to work with a preliminary run over a period of 10 to 20 pulse beats during which the pulse artifact is determined using the latter method at the start of the measurement, over which period these are recorded before the The actual measuring process begins, in order to then be able to carry out the online subtraction of the pulse artifacts from the EEG signal immediately during the measurement.

A variation of the pulse artifact with the heart rate can also be taken into account. The time of the heart's blood output is linked to the beat frequency by a formula. The heartbeat frequency is determined via the R-wave distance in the EKG. If the pulse artifacts are stretched or compressed by means of interpolation using this formula before averaging, a temporal smearing of the artifact can be avoided. To subtract the EEG signal, the averaged artifact is interpolated again to the appropriate length. - 17 -

5 shows an example of a reduction in pulse artifacts. It shows

diagram a) the EEG signal with superimposed pulse artifacts,

diagram b) an EKG derived on MRI,

- the diagram c) the interference signal component to be subtracted, and

- the diagram d) the pure EEG signal after elimination of the pulse artifacts.

The invention shows the causal relationship between the switched gradients and the frequency components in the spectrum of the disturbed EEG. The interference by the switched gradients can be eliminated by the signal processor during the recording. Corresponding low-pass filters or filter functions can be programmed in accordance with the above description for the respective sequences which are to be used for imaging in the MRT. Since the causes and the relationships are shown, a sequence for imaging can also be taken into account when deriving the EEG. This means that the EEG can be evaluated immediately even during the long data acquisition times of the MRI.

The causes of the pulse artifacts are also shown. These arise from the ion separation of the pulsating blood in the high magnetic field of the MRI. If one does not want to restrict the positioning of the electrodes for EEG recording to blood vessel-free or blood vessel-poor head zones, the pulse artifacts can only be soothed up by digital signal processing. With the pulse-triggered subtraction shown according to the invention, an effective and reliable method is offered to eliminate the pulse artifacts. As a result, the EEG regains diagnostic quality in MRI. - 18 -

As far as the filtering of the derived EEG signals in the MRT is concerned, the complex digital processing described above only needs to take place in order to eliminate the interference from the MRI operation when the MRI is actually in operation. In the non-measuring MRI, as explained at the beginning, only the pulse artifacts represent disturbances which have a considerable effect due to the static high magnetic field which also prevails in the non-measuring MRI.

The arrangement can thus be made such that the digital filtering of the EEG signal only takes place while the MRT is in the measuring mode, that is to say the MRI functions with the switched field gradient etc. generate an interference spectrum. This can be done automatically as part of the signal evaluation by evaluating the EEG signal accordingly in order to recognize the time range of the disturbance due to the activity of the magnetic resonance tomograph, or the period during which the magnetic resonance tomograph is active can also be done by an external one Trigger during EEG recording can be determined. This allows the EEG signal filtering to be switched on and off automatically.

Since the digital filtering of the EEG signal to eliminate the interference caused by the MRT operation naturally weakens the EEG signal, appropriate gain control or the like can be used to ensure that after processing the EEG signal by filtering in the period in question or Fourier space, the energy content of the processed EEG signal is adapted again to the energy content of an undisturbed and unprocessed EEG signal. The EEG signal is thus raised by a corresponding factor in this period or Fourier space. This gain control can also be carried out automatically in the same way. The result of this is that an EEG signal derived continuously over inactive and active periods of the MRT appears, which is cleaned of disturbances and runs continuously without jumps in intensity.

Claims

- 19 patent claims

1.Procedure for deriving an electroencephalogram in a magnetic resonance tomograph, in which a static high magnetic field is generated for aligning the nuclear spins and in imaging operation a circularly polarized alternating field in the radio frequency range for exciting the nuclear spins as well as three switched magnetic gradient fields for the spatial coding of the nuclear magnetic resonance signal in the three Spatial directions are generated

characterized,

that the EEG signals derived from the patient's head inside the nuclear spin tomograph tube via the EEG lead electrodes are subjected to spectral analysis and filtering by means of digital signal processing in order to eliminate the interference imprinted in the active magnetic resonance tomograph by the switched field gradients.

2. The method according to claim 1, characterized in that the filtering of the EEG signals in a low-pass filter less

Waviness and higher order, for example a Butterworth filter.

3. The method according to claim 1, characterized in that for

Filtering the EEG signals digital bandstop or digital

Filters are used which, due to their characteristics, suppress the multiples of a frequency by which the

Hide spectral analysis determined interference frequency ranges.

4. The method according to claim 1, characterized in that digital Fourier filters are used to filter the EEG signals in order to effect the elimination of the interference frequencies directly in the Fourier spectrum of the signal.

5.Method for deriving an electroencephalogram in an MRI scanner, in which a static high magnetic field is generated for aligning the nuclear spins and in imaging operation a circularly polarized alternating field in the radio frequency range for exciting the nuclear spins as well as three switched magnetic gradient fields for the spatial coding of the nuclear magnetic resonance signal in the three Spatial directions are generated

characterized in that

a) an EKG signal is recorded as a trigger signal parallel to the EEG signals,

b) a periodically occurring, time-sharp signal element is selected as a trigger event from the EKG signal,

c) a time length L corresponding to the minimum time interval between successive trigger events is determined,

d) in each EEG channel selected by the user, EEG signal sections of time length L from each trigger event are compared with one another with a coherence analysis and an average pulse artifact interference signal is determined,

e) the averaged pulse artifact interference signal obtained is subtracted by means of the oneline subtraction from each trigger event in each selected EEG channel signal over the time length L in order to obtain a pulse artifact-adjusted EEG signal. - 21 -

6. The method according to claim 5, characterized in that the signal increase of the R wave in the EKG signal is selected as the trigger event.

7. The method according to claim 6, characterized in that the trigger event is determined in that the average and maximum of the amplitude are determined from the ECG signals recorded over a certain period of time and then the mean is calculated from the maximum and the average and the mean of the amplitude is selected as the trigger event.

8. The method according to any one of claims 5-7, characterized in that step c) either by means of a separate calibration routine method before the start of the EEG derivation or continuously during the EEG derivation from the ECG signal recorded in parallel and one each certain number of recent trigger events is executed.

9. Device for performing the method according to claim 1, with a magnetic resonance scanner with means for generating a static high magnetic field within a tomography tube (1), an RF head coil (3) for generating a high-frequency alternating field for exciting the nuclear spins and coils for generating switched Gradient fields for the three spatial coordinates for the spatial coding of the nuclear magnetic resonance signals, further with EEG lead electrodes (4), an EEG amplifier (5) arranged inside the tomograph tube (1), and signal processing devices (7, 8, 9) located outside the shielded tomograph room , which are connected to the EEG amplifier via an optical fiber (6), characterized by a low-pass filter at the input of the EEG amplifier (5), the cut-off frequency of which is below half the EEG sampling rate, and by a Butterworth filter of at least tenth order, through which all EEG signal channels are routed.

10. Device for performing the method according to claim 1, with a magnetic resonance tomograph with means for generating a static high magnetic field within a tomograph - 22 -

tube (1), an RF head coil (3) for generating a high-frequency alternating field for exciting the nuclear spins and coils for generating switched gradient fields for the three spatial coordinates for spatial coding of the nuclear magnetic resonance signals, further with EEG lead electrodes (4), one inside EEG amplifier (5) arranged in the tomography tube (1), and with signal processing devices (7, 8, 9) located outside the shielded tomography room, which are connected to the EEG amplifier via an optical fiber (6), characterized by a low pass in conjunction with digital bandstops to mask out the interference frequency ranges for filtering the EEG signal or by a filter, the characteristics of which suppress the multiples of a frequency, for masking out a fundamental interference frequency with its multiples.

11. Device for carrying out the method according to claim 1, with a magnetic resonance scanner with means for generating a static high magnetic field within a tomograph tube (1), an RF head coil (3) for generating a high-frequency alternating field for exciting the nuclear spins and coils for generating switched Gradient fields for the three spatial coordinates for the spatial coding of the nuclear magnetic resonance signals, further with EEG lead electrodes (4), an EEG amplifier (5) arranged inside the tomograph tube (1), and with signal processing devices (7, 8, 9) located outside the shielded tomograph room, which are connected to the EEG amplifier via an optical fiber (6), characterized by a Fourier filter for eliminating the interference frequencies in the Fourier spectrum of the EEG signal.

12. Device for carrying out the method according to one of claims 5 to 8, with a magnetic resonance scanner with means for generating a static high magnetic field within a tomography tube (1), an RF head coil (3) for generating a high-frequency alternating field for exciting the nuclear spins and Coils for generating switched gradient fields for the three spatial coordinates for spatial coding of the nuclear magnetic resonance signals, further with EEG lead electrodes (4), one inside - 23 -

the EEG amplifier (5) arranged in the tomography tube (1), and with signal processing devices (7, 8, 9) located outside the shielded tomography room, which are connected to the EEG amplifier via an optical fiber (6), characterized in that an additional Channel for recording an EKG signal is provided and a digital signal processing device is provided for executing the method measures b) to e).

13. The method according to any one of claims 1 to 4, characterized in that the filtering of the EEG signal takes place only during those periods during which the magnetic resonance tomograph is active.

14. The method according to claim 13, characterized in that the periods of active magnetic resonance imaging are automatically detected by appropriate observation and evaluation of the derived EEG signal or by means of an external trigger and the start and end of the filtering of the EEG signal are each automatically switched.

15. The method according to claim 13 or 14, characterized in that a gain control of the derived EEG signal is carried out in such a way that the energy content of the filtered EEG signal weakened by the filtering process by raising to the energy content of an undisturbed and unfiltered derived with inactive magnetic resonance imaging EEG signal is adjusted.