WO2003084605A1

WO2003084605A1 - Method and device for the prevention of epileptic attacks

Info

Publication number: WO2003084605A1
Application number: PCT/EP2003/003543
Authority: WO
Inventors: Oliver Holzner
Original assignee: Oliver Holzner
Priority date: 2002-04-05
Filing date: 2003-04-04
Publication date: 2003-10-16
Also published as: EP1492593A1; AU2003226786A1; JP2005528141A; DE10215115A1

Abstract

The invention relates to a method and a device for the automatic non-invasive controlled or regulated electromagnetic prevention of epileptic attacks in vivo, based on attack models. The method firstly comprises, in addition to continuous extracranial measurement of electromagnetic fields ( in particular those connected with brain activity), the continuous calculation of early warning indicators from the measured data and secondly, in the case of critical indicator values, in addition to the continuous calculation of attack-preventing interventions based on an attack model, the continuous application of said interventions by means of extra-cranial generation of suitable magnetic fields.

Description

METHOD AND DEVICE FOR PREVENTING EPILEPTIC SEASONS

The invention relates to a method and a device for the automatic non-invasive controlled or regulated electromagnetic prevention of epileptic seizures in vivo.

The relevant technologies include the following approaches:

1 Major research efforts are focused on the neurophysiological development of epileptic seizures, and there on the level of cell preparations [8], [9].

2 Early warning methods based on extracranial EEG data have been described in principle, e.g. [1], [21].

3 Diagnostic use of TMS (transcranial magnetic stimulation) has been described in principle, also coupled with EEG, e.g. [2]

4 The use of TMS for intervention in epilepsy consists of finding an epileptic focus based on the medical experience of the doctor using it, imaging procedures, or trying it out, and then trying to induce epileptic seizures with single or double coil systems (e.g. [5], [6 ], [10]).

5 seizure models, the prerequisites and properties of collective excitement describing development processes exist as exemplary examples of recent physical theories, such as synergetics (for an overview, see [7]).

WO 98/18394 describes a method with which magnetic stimulation is carried out on a subject, at the same time his brain activity is measured by means of an EEG. This known method is used for diagnosis.

WO 01/21067 discloses a method for the early detection of an impending epileptic seizure. This procedure is designed to predict an impending epileptic seizure for hours or days. This procedure measures a patient's brain activity at different locations before, during and after epileptic seizures. With the help of various nonlinear methods, sensor pairs are determined for this patient that predict the seizure particularly well in the context of a training phase consisting of seizures. Signal pairs are adapted at regular intervals, for which further attacks are necessary. The training and adaptation contained in this procedure prevent complete prevention, since the data must be updated again and again with new seizures.

The object of the present invention is to create a method and a device for the prevention of epileptic seizures.

The object is achieved by a method with the features of claim 1 and by a device with the features of claim 8. Advantageous embodiments of the invention are specified in the respective subclaims.

Using a seizure model reliably prevents epileptic seizures. The invention is based on the knowledge that the processes leading to epileptic seizures can be made quantifiable by naming suitable control parameters, so that reliable prevention is possible.

The invention is explained below using the drawings as an example. The drawings show: 1 shows a transmitter in a sectional view, FIG. 2 shows the transmitter from FIG. 1 in a view from below, FIG. 3 shows a planar projection of openings for sensors and transmitters according to their arrangement on a helmet,

4 shows a helmet and a carrier axis together with a chin rest,

5 shows a further planar projection of openings for sensors and transmitters according to their arrangement on a helmet, and FIG. 6 shows an example of a time series of measured values of an EEG sensor, FIG. 7 shows a section of the time series from FIG. 6 in a phase space representation 8, and FIG. 8 shows a typical course of the SNR (signal-to-noise ratio).

contraption

On the input side, the device according to the invention comprises a measuring system with apparatuses for electromagnetic measurement data acquisition, preprocessing and forwarding, for example in an advantageous embodiment comprising an EEG cap with its sensors, connections to the amplifier, amplifier, connections to the AD converter, ADC Converter, connections to the computer unit, electricity supplier for the apparatus, and connections.

On the output side, the device comprises an actuating system with apparatus for the extracranial generation of magnetic fields, referred to as “transmitter”, and a device for converting the digital control or regulation specifications originating from the computer unit into transmitter signals, for example in an advantageous embodiment comprising current-carrying coils, power suppliers , Connections, D / A converter, together with connections.

Furthermore, the device between the input and output side comprises a computer unit (PC or workstation) with software for implementing the method explained in more detail below. Suitable sensors are EEG or MEG sensors. The MEG sensors are formed, for example, from a SQUID sensor element with a suitable evaluation device for detecting a magnetic field and a cooling device. The EEG sensors have, for example, two electrodes for measuring an electrical potential difference.

A sensor can have an electrical and / or magnetic shield from its surroundings, provided that its function is not hindered (for example, no shield in the direction of the patient's crane, but very well shield in the direction of other transmitters and / or sensors and / or connecting cable).

The part of the input side near the head can have a multiplicity of sensors which are distributed over the surface of the head near the brain; this multiplicity of sensors is referred to as a sensor grid.

The sensor grid has a fixing device for fixing the same with respect to the patient's crane, so that when the sensor grid is put on and taken off several times, the sensors return to their respective relative positions, for example by fitting the sensor grid into a helmet, the inside of which has the cranial shape of the respective patient replicates. The fixation can also be carried out with the aid of a camera, the position of the patient's head in the room and the sensors with respect to the head being recorded by several cameras and converted in real time into 3D data.

An advantageous embodiment of the input side comprises its partially outpatient form, in which the measurement data is obtained via a portable sensor grid, which is connected to devices for measurement data preprocessing to be carried by the patient in a backpack or as part of the clothing, and in which the data transfer to the computer unit is advantageously wireless he follows.

A transmitter 5 comprises a current-carrying coil 6 with a para-, dia-, or ferromagnetic core 7, as shown in a sectional view in FIG. 1, the arrow directions symbolizing the directions of the current flow. The transmitter 5 has essentially have a cylindrical shape, the outer surface and an end face of the cylinder forming the rear side being clad with a shield 8. The coil 6 and the core 7 directly adjoin the side of the transmitter which is free of the shielding, and with this side the transmitter 5 is aligned with the cranium during operation to emit exogenous magnetic fields. On the back of the transmitter 5, a holding element 9 is arranged, with which the transmitter 5 can be fixed in a home.

The extracranial transmitter 5 can be protected against deformation, for example by pouring the live parts into suitable resin or embedding the live parts in stable insulating material.

The transmitter 5 can be provided with a cooling device.

In a further embodiment, intracranially implanted electrodes are used as sensors and / or transmitters, via which both EEG measurements can be carried out and currents can be conducted into the brain. Lines leading to these electrodes and / or their interfaces to the computer unit and / or further lines and / or further measuring devices and / or the associated computer unit and / or the energy supplier of electrodes and / or computer unit can also be implanted, thereby permitting outpatient operation.

An advantageous embodiment of the parts of the positioning system close to the head comprises a plurality of transmitters which are distributed intracranially or extracranially; this arrangement of transmitters is referred to as a transmitter grating.

An advantageous embodiment of an extracranial transmitter grating includes fixing it with respect to the crane of the respective user, so that when the transmitter grille is put on and taken off several times, the transmitters assume their respective relative positions again, for example by fitting the transmitter grille into a helmet, the inside of which is the cranial shape of the respective user replicates. Another advantageous embodiment of the transmitter grid comprises implanted electrodes.

An advantageous embodiment of the parts of an extracranial measuring and positioning system close to the head comprises a helmet 10 on its inside which reproduces the cranial shape of the respective user, with connecting cables running through a support axis 11 and a chin rest 12.Sensor and transmitter grids in the interior of the helmet are fixed in such a way that both grids overlap - ie there are sufficient transmitters in the vicinity of each sensor and vice versa. FIG. 3 shows a planar projection of the superimposition of the transmitter with the sensor grid (openings 13 for sensors are shown as circles and openings 14 for transmitters 5 as squares). In the embodiment described, the user sits on an armchair with a neck support below the helmet 10.

In an alternative embodiment, the sensor grid is intracranial and the helmet contains the extracranial transmitter grid.

In an alternative embodiment, the transmitter grid is intracranial and the helmet contains the extracranial sensor grid.

In an alternative embodiment, both sensor and transmitter gratings are intracranial.

In an advantageous embodiment, the sensor density or sensor configuration of an extracranial sensor grid can be set. In a further advantageous embodiment, this change is automated, controlled or regulated via the intermediate unit.

In an advantageous embodiment, the transmitter density or transmitter configuration of an extracranial transmitter grid can be set and / or the angle of inclination of each individual transmitter to the patient's cranium can be changed. FIG. 5 shows a planar projection of a mechanical holder of this embodiment. Openings 13 for sensors are circles and openings 14 for Transmitter shown square. Here it is possible to anchor transmitter 5 in the openings 14 of the holder and / or to tilt transmitter 5 with respect to the holder. In this embodiment, among other things, all conventional coil configurations can be represented with their arrangement, orientation and field direction.

In an advantageous embodiment, the device is provided with conventional protection against power failures and / or voltage fluctuations.

In an advantageous embodiment, the computer unit runs real-time and automatically: i) ongoing calculation of the seizure early warning indicator from the input data, ii) if the indicator exceeds thresholds, calculation of an intervention instruction to prevent seizure, and implementation of the intervention in question using the transmitter Magnetic fields, iii) If the indicator returns to normal and / or a time limit is exceeded, shutdown of the intervention, iv) Conventional algorithms for removing artifacts by artificially generated magnetic fields (see, for example, [2]), as well as for other artifact removal (e.g., due to muscle twitching) ,

In an advantageous embodiment, in addition to i-iv, algorithms for optimizing the density and positioning of sensors and transmitters also run automatically and real-time on the computer unit.

Method:

1. The measurement data acquisition by EEG, measurement data preprocessing and measurement data transfer in digital form along with possible artifact elimination are carried out continuously with conventional methods. The measurement data are automatically processed according to the empirically validated early warning indicator used to a value of this early warning indicator. 2. When the early warning indicator responds, the automatic intervention instructions for seizure prevention, which are compatible with the seizure model used, are calculated, and their ongoing implementation via magnetic field generation (B-field generation) is carried out with the help of the transmitters. The specifics of magnetic field generation (for example location, strength, direction, frequency pattern, and / or others) result from the intervention instruction. The B-field changes cause intracranial induction voltages. The digital control of the magnetic field generation takes place with conventional methods. Current health recommendations for extracranial generated electromagnetic radiation are known, and compliance with them is automated.

3. When the early warning indicator returns to its normal range and / or a time limit for the intervention is exceeded, the intervention is shut down.

An early warning indicator is a quantity calculated from electromagnetic brain activity data that changes significantly before an epileptic attack. Early warning indicators are preferred for the present invention, which are changed at least a few minutes before the attack.

A suitable early warning indicator is the correlation of similarity indices of a predefined proportion of sensors, with falling similarity indices. The similarity index is known from [1] and a number of previous publications, for example [21]. The average early warning time given here is 325 seconds.

In another advantageous refinement of the method, the early warning indicator is the mutual information of similarity indices of a predefined portion of sensors, with decreasing similarity indices. "Mutual information" is known as a binary logarithm of "probability of the occurrence of two random variables together divided by the product of their individual probabilities".

In another advantageous embodiment of the method, the early warning indicator is the mutual information of similarity indices of a predefined portion of sensors, in the case of falling similarity indices, linked to activation indicators (for example, characteristic changes in body temperature when waking up, muscle movements, characteristic EEG patterns, and / or others). This minimizes the possibility of false alarms due to simultaneous changes in the patient's state of wakefulness for many sensors, with additional requirements for the device depending on the additional indicator (for example, ongoing EMG measurement).

The examples given above for calculating early warning indicators do not require training phases that contain epileptic seizures. The early warning indicators are calculated on the basis of a phase space representation of the normal state of the patient concerned.

The early warning indicators given above are robust against noise and artifacts. For other non-robust early warning indicators, filtering or artifact removal methods must be interposed.

An example of phase space embedding is given in FIGS. 6 and 7, FIG. 6 showing an EEG time series of 8 seconds for a single channel, with a sampling rate of 128 measuring points per second (x-axis time, y-axis voltage between Electrode and reference electrode in freely selected units), FIG. 7 shows a section of the time series from FIG. 6 comprising 32 measuring points starting with measuring point 128 in phase space representation (x-axis measured value at time t, y-axis measured value at time t-20). The method of embedding in a phase space is described in detail in [13], for example. It is assumed here that the one-dimensional signal (as in FIG. 6) is a projection of a higher-dimensional signal which is to be restored. This higher-dimensional signal is shown in two dimensions in FIG.

As a preferred embodiment of an early warning system before an epileptic attack, a detection module can be specified with

1) a device for calculating the similarity of the current measurement series with the measurement series representing the normal state for each measurement channel. The normal condition of each individual patient is assessed before the public use of the detection module.

2) Device for emitting a local warning signal for each measuring channel: if the above-mentioned similarity falls below a threshold value. \

3) Device for emitting a global warning signal if several measurement channels promptly lead to local warning signals.

Seizure models are used to make the intervention reliable. The following models can be used as seizure models: oscillator seizure model, chaos seizure model, synergetic seizure model, stochastic oscillator seizure model, stochastic chaos seizure model, stochastic synergetic seizure model

These seizure models describe the parameters relevant for an epileptic seizure, which are calculated from the electromagnetic activities of neurons and / or neuron populations. These parameters are e.g. Chaoticity of the potential difference time series measured by means of an EEG electrode and its reference electrode, expressed by their maximum Lyapunov exponent [12]. Typical further parameters are critical slowdown, critical fluctuations, similarity to a normal state in the (meta) phase space, etc. These parameters are expressed by concrete numerical parameters. For example, instead of the Lyapunov exponent, the chaoticity can alternatively be represented by the embedding dimension [13], correlation dimension, Kullback-Leibler entropy, etc.

An oscillator seizure model is based on [3]. The neuron populations described here are so-called neural limit cycle oscillators, which means that they can oscillate or rest depending on the parameters. The interaction of neural oscillators with each other is described with an interaction equation. This interaction presupposes the development of seizures. Preventing seizures is based on decoupling the neural oscillators.

In the following, "neural oscillator" is used synonymously with "limit cycle oscillator". A distinction must be made here between the special case of phase oscillators (see for example [22]), in which the amplitude and phase are decoupled and only the phase of an oscillator is considered. In the phase space, the limit cycle is bige closed curve, the phase oscillator as a circular path. A corresponding seizure model is based on the increased occurrence of 1 clusters compared to other clusters. This special case and the associated intervention procedures (resetting plus entrainment, see for example [22]) do not lead to success in the case of general limit cycle oscillators, although not even a hard reset with high amplitude, which is often repeated (problematic anyway) leads to seizure prevention under the rTMS health limits). General interventions for limit cycle oscillators, however, also work for phase oscillators.

A suitable interaction for the oscillator seizure model is the specific weak coupling between neural oscillators. Seizures are accompanied by an increase in the number of oscillating neural oscillators as well as increased mutual information between the oscillation frequencies of these weakly coupled neural oscillators. A neural oscillator is a localized ensemble of neurons that is capable of oscillating and non-oscillating behavior. The dynamics of each neural oscillator interacting with other neural oscillators is through

^Z , - = (z, -) + ^£ ∑ " ₌ Λ * ^Z ; ε« l

given. Here, for each i between 1 and n Zj neural oscillator, gj is given by the Wilson-Cowan equations known from [3] for the i-th neural oscillator, hy is the strength of the connection from Zj to Zj. The coupling strength epsilon is empirically between 0.04 and 0.08. If one assumes the coupling strength and connection strengths to be slowly changing compared to the time scale of a seizure, there remains primarily an intervention via the function g-. It is known from the theory of neural oscillators that they only interact in the case of oscillations, and only at commensurable oscillation frequencies.

An advantageous embodiment of an instruction for preventive action that is compatible with the “seizure model with specific weak coupling between neural oscillators” on epileptic seizures is:

First, neural oscillators that are adjacent and secondly that oscillate with the same and / or commensurable frequencies prior to the intervention are forced to incommensurable frequencies that are contained in their original frequencies or to nearby incommensurable frequencies (example: neighboring oscillators have the frequencies 3 Hertz and 15 Hertz , therefore force the second oscillator to the frequency 5 Hertz. Another example: both have the frequency 8 Hertz, therefore force one of them to 7 Hertz). The vibrations are forced by means of magnetic fields of these frequencies at high amplitudes. Since oscillating neural oscillators and neighborhoods on the same and / or commensurable frequencies indicate the possible existence of physiological connections, the forced incommensurability, i.e. Modification of the gj, the possible and even more factual interaction between the respective zones is interrupted, which minimizes mutual information, and thus prevents the onset of seizures. The complexity of the process allows the continuous real-time calculation of all required sizes.

However, whether neighboring oscillators succeed in making them incommensurable depends on their diversity and the minimality of their mutual coupling. In extreme cases, neighboring, almost identical, strongly coupled oscillators in the area of influence of a transmitter cannot be forced to different incommensurable frequencies. Here, however, it is sufficient to move groups of oscillators in the area of influence of various transmitters to incommensurable frequencies in order to prevent the attack. This also has the effect of preventing the formation of 1 clusters in the area of influence of several transmitters and preventing the formation of "traveling waves".

Another advantageous embodiment of an instruction for the prevention of epileptic seizures that is compatible with the “seizure model with specific weak coupling between neural oscillators” is: In step 1, first stimulate the neural oscillators to chaotic behavior [14] (for example by time-delayed feedback with systematic error ), and then stabilize the neural oscillators in step 2 depending on the influence of the respective transmitter first orbits with incommensurable frequencies that reach them using conventional methods. As is known from [4], step 2 of this method has already proven sufficient in cell preparations to prevent the spread of seizures. However, the algorithm used there (“OGY method”) is unsuitable for the in vivo real-time case because of its requirements for computer speed and storage capacity.

In the stochastic oscillator seizure model, certain parameters are assumed to be randomly variable compared to the seizure model mentioned above. The proposed methods can also be used (e.g. [15]).

The chaos seizure model assumes that normal brain activity, as captured by each sensor, has a minimum of chaoticity. The seizures are accompanied by a simultaneous decrease in this chaos for all sensors. Seizures are prevented by maintaining a certain degree of chaos ([4] and [16]).

In the stochastic chaos attack model, high-dimensional influences supplement the low-dimensional deterministic change in the electromagnetic variables. The seizure prevention strategies correspond to those of the chaos seizure model.

In the synergetic attack model, it is assumed that brain activity can be described by a small number of degrees of freedom, so-called order parameters [17]. Circular causality applies: the order parameters are caused and determined by the cooperation of neurons, but at the same time the order parameters determine the macroscopic system behavior. An epileptic seizure corresponds to a phase transition. This goes hand in hand with critical slowdown and critical fluctuations. Prevention of seizures takes place by preventing the phase transition (eg by checking bifurcation points according to [18]). In the stochastic synergetic attack model, the so-called Langevin approach is used to introduce stochastic forces into the synergetic model in a phenomenological way. To prevent seizures, in addition to the above-mentioned method, there is the possibility of stochastic resonance [20], and its opposite, noise-drowning: It is known that, for example, it depends on a noise amplitude (for example for Gaussian White Noise) in systems signals are generated with stochastic components ("coherence resonance"), or the signal-to-noise ratio (SNR) is amplified ("stochastic resonance") or weakened (the latter is to be referred to here as "noise drowning"). The typical course of the SNR is shown in FIG. 8 (x-axis noise amplitude, y-axis SNR).

An intervention instruction describing the magnetic field to be generated is calculated on the basis of these models. This description is e.g. by location, strength, direction, frequency pattern, and / or other parameters of the magnetic field (B field). With this magnetic field, the electromagnetic activities of neurons and / or neuron populations are changed in a suitable manner and an impending epileptic attack is thus prevented.

The use of one or more seizure models reliably prevents epileptic seizures. The invention is based on the knowledge that the processes leading to epileptic seizures are made quantifiable by naming suitable control parameters, so that reliable prevention is possible.

The preferred embodiment of the invention comprises an intervention module, which is suitable in a variety of models to prevent the attack, for example in the case of high transmitter density, the transmitters are to be divided into three classes, class 1 for chaotization, class 2 for incommensurable stabilization, - class 3 for Noise-Drowning, in such a way that in every neighborhood of each transmitter of one class there are transmitters of the other classes. Among other things, class 1 seizure models are satisfied, class 2 oscillator model seizures, class 3 models with stochastic components. The satisfaction of synergetic models results automatically from devaluation of the master modes (by frequency shifts) while at the same time preventing the ascent from slave modes to master modes (by noise-drowning). Satisfaction of phase oscillator seizure models also results automatically, since 1 cluster states are prevented (incommensurability prevents phase locking, noise drowning prevents higher modes). Satisfaction of chaos seizure models also arises automatically due to class 1 and class 3 (roughness = high-dimensional chaos).

In the methods described above, the brain activity can be measured either during or immediately after an intervention, which results in a closed control loop, since the early warning indicator and, if necessary, a further intervention instruction are calculated from the measured brain activity.

An advantageous embodiment of the retraction of the intervention is the sliding simultaneous retraction of all generated magnetic fields.

An advantageous embodiment of the retraction of the intervention is the sliding thinning of the transmitters producing magnetic fields (thinning as a spatially uniformly distributed retraction and / or switching off a percentage of all transmitters).

An advantageous embodiment of the retraction of the intervention is the spatially localized retraction and / or shutdown of several transmitters with gradual expansion of the area in which the retraction and / or shutdown takes place.

It is not necessary to operate the interventions with the high field strengths of 1-2 Tesla per coil that are common in TMS. "Running" is defined as "continuous" or as "at suitable time intervals". Monitoring of compliance with electromagnetic exposure limit values is carried out automatically on an ongoing basis. The use of the invention is not aimed at curing epilepsy, but rather enables seizure freedom during the period of use without the need for medical personnel or the use of pharmaceuticals. This not only results in the minimization of the consequences of the disease and the side effects of the treatment, but also a significant reduction in the running costs. Furthermore, an outpatient use of the invention is possible, which, in addition to a further reduction in costs, improves the freedom of movement of the patients to a considerable extent.

The procedure described above prevents epileptic seizures. In addition, with a similar, but more general, proactive procedure, instead of reactive prevention of epileptic seizures on the basis of seizure models, other behavioral goals, preferably for healthy people, can be achieved on the basis of appropriate behavior models as well as general brain activity models. The general procedure comprises the interactive determination of the non-observables of the models used for the person concerned, based on this, and based on the models the calculation of an a priori unknown intervention instruction, as well as the selective implementation of the instruction while at the same time preventing undesired propagation effects. The aim of this modification is, at the request of a person, to reliably stabilize or change their behavior and / or to stabilize this change. This method can be carried out with a device similar to the device described above.

Literature:

[1] "Anticipation of epileptic seizures from standard EEG recordings", Le

Van Quyen M & Martinerie J & Navarro V & Boon P & D'Have M & A- dam C & Renault B & Varela F & Baulac M, The Lancet 2001 Jan 20; 357: 183-8

[2] WO 98/18384 A1

[3] "Excitatory and inhibitory interactions in localized populations of model neurons", Wilson HR & Cowan JD, Biophysical Journal 1972; 12: 1-22

[4] "Controlling chaos in the brain", ship SJ & Jerger K & Duong DH &

Chang T & Spano ML & Ditto WL, Nature 1994 Aug 25; 370: 615-620

[5] "Transcranial magnetic stimulation in patients with epilepsy", Dhuna A &

Gates J & Pascual-Leone A, Neurology 1991 July; 41: 1067-1071

[6] "Epileptic seizures triggered directly by focal transcranial magnetic

Stimulation ", Classen J & Witte OW & Schlaug G & Seitz RJ & Holthausen A & Benecke R, Electroencephalography and clinical Neurophysiology 1995; 94: 19-25

[7] "Epilepsy: multistability in a dynamic disease", Milton JG, in "Self-organized biological dynamics and nonlinear control", ed. Walleczek J, Cambridge University Press 2000

[8] "Electric field suppression of epileptiform activity in hippocampal slices",

Gluckman BJ & Neel EJ & Netoff Tl & Ditto WL & Spano ML & Schiff SJ, Journal of Neurophysiology 1996 Dec; 76 (6): 4202-4205

[9] "Coupled intra- and extracellular Ca ²⁺ dynamics in recurrent seizure-like events", Szilägy N & Koväcs R & Kardos J, European Journal of Neuroscience 2000; 12: 3893-3899 [10] "Deliberate Seizure Induction With Repetitive Transcranial Magnetic

Stimulation in Nonhuman Primates ", Lisanby SH & Luber B & Sackeim HA & Finck AD & Schroeder C, Arch Gen Psychiatry 2001; 58: 199-200

[11] WO 01/21067

[12] "How to Extract Lyapunov Exponents from Short and Noisy Time Series", Banbrook M & Ushaw G & McLaughlin S, IEEE Transactions on Signal Processing 1997; 45: 1378-1382

[13] "Determining embedding dimension for phase-space reconstruction using a geometrical construction", Kennel MB & Brown R & Abarbanel

HDI, Physical Review A 1992, 45 (6): 3403-3411

[14] "Chaotification via arbitrary small feedback controls: theory, method, and applications", Wang XF & Chen G; International Journal of Bifurcation and Chaos 2000; 10 (3): 549-570

[15] "Controlling Nonchaotic Neuronal Noise Using Chaos Control Technologies", Christini DJ & CoHins JJ, Physical Review Letters 1995 Oct 2; 75 (14): 2782-2785

[16] "Anticontrol of chaos in continuous-time systems via time-delay feedback", Wang XF & Chen G & Yu X; Chaos 2000 Dec; 10 (4): 771-779

[17] "A derivation of macroscopic field theory of the brain from quasi-microscopic neural dynamics", Jirsa VK & Haken H, Physica D 1999, 99: 503-526

[18] "Controlling Bifurcation Dynamics via Chaotification", Wang XF & Chen

G; proposed paper for CDC 2001 [19] "Impact of noise on a field theoretical model of the human brain", Frank

TD, Daffertshofer A, Beek PJ & Haken H, Physica D 1999, 127: 233-249

[20] "Functional Stochastic Resonance in the Human Brain: Noise Induced

Sensitization of the Human Baroreflex System ", Hidaka I & Nozaki D & Yamamoto Y, Physical Review Letters 2000 Oct 23; 85 (17): 3740-3743

[21] "Anticipating epileptic seizures in real time by a non-linear analysis of similarity between EEG recordings", Le Van Quyen M & Martinerie J &

Baulac M & Varela F, Neuroreport 10 (1999): 2149-2155

[22] "Effective desynchronization with a stimulation technique based on soft phase resetting", Tass P, Europhysics Letters 57 (2), 2002: 164-170

Claims

claims

1. Method for the non-invasive controlled or regulated electromagnetic prevention of epileptic seizures in vivo, comprising the following steps:

- automatic extracranial electromagnetic measurement of brain activity,

- automatic calculation of an early warning indicator for epileptic seizures, - automatic calculation of an intervention instruction for seizure prevention when the early warning indicator responds based on a seizure model and the measured brain activity, and

- Automatic implementation of the intervention instruction via controlled or regulated extracranial magnetic field generation.

2. The method according to claim 1, characterized in that the parameters relevant to the epileptic seizures from electromagnetic activities of neurons and / or neuron populations are used as the seizure model.

3. The method according to claim 2, characterized in that a seizure model from the group oscillator seizure model, stochastic oscillator seizure model, chaos seizure model, stochastic chaos seizure model, synergetic seizure model, stochastic synergetic seizure model is used.

4. The method according to any one of claims 1 to 3, characterized in that the measurement of brain activity and calculation of the early warning indicator is carried out continuously.

5. The method according to any one of claims 1 to 4, characterized in that an intervention instruction is implemented by extracranial generation of magnetic fields.

6. The method according to any one of claims 1 to 5, characterized in that brain activity is measured either during or immediately after an intervention.

7. The method according to any one of claims 1 to 6, characterized by controlled or regulated automatic shutdown of the intervention when the early warning indicator returns to its normal range and / or a time limit is exceeded.

8. Device for automatic non-invasive controlled or regulated electromagnetic prevention of epileptic seizures in vivo, in particular for carrying out a method according to one of claims 1 to 7, comprising: a measuring device with at least one sensor for measuring electromagnetic brain activity,

- a device for determining an early warning indicator for the early detection of epileptic seizures,

a device for calculating an intervention instruction based on a seizure model and the measured brain activity, and

- A device for implementing the Inten / entionsanleitung with at least one transmitter for generating a magnetic field.

, Apparatus according to claim 8, characterized in that the measuring device has a plurality of extracranial sensors which form a sensor grid.

10. The device according to claim 8 or 9, characterized in that the sensors are arranged in an EEG cap.

11. The device according to one of claims 8 to 10, characterized in that the device for implementing the intervention instruction has a plurality of transmitters which form a transmitter grid.

12. Device according to one of claims 8 to 11, characterized in that a computer unit is provided in which a software module for implementing the method according to one of claims 1 to 7 is stored.

13. Device according to one of claims 8 to 12, characterized in that an electrical and / or magnetic shield is provided for each sensor and each transmitter.

14. Device according to one of claims 8 to 13, characterized in that a fixing device of the measuring device is provided with respect to the crane of the respective patient, so that the sensors assume their respective relative positions again when the sensor grid is put on and taken off several times.

15. The device according to one of claims 8 to 14, characterized in that that the measuring device can be mechanically decoupled from the rest of the device, so that the measuring device can be carried by a patient.

16. The device according to one of claims 8 to 15,. characterized in that a fixing device is provided for the device for implementing the intervention instruction with regard to the patient's crane.

17. The device according to one of claims 8 to 16, characterized in that the sensors and transmitters are arranged on an inside of the cranial shape of the respective patient's helmet.

18. Device according to one of claims 9 and 11 to 17, characterized in that the transmitter grating and the sensor grating are interleaved in such a way that each transmitter and sensor transmitter are adjacent.

19. Device according to one of claims 11 to 18, characterized in that holders are provided in the transmitter grid for receiving additional transmitters, so that the transmitter density of a transmitter grid can be changed locally and / or so that the angle of inclination of individual transmitters to the patient's cranium can be changed.