WO2018009991A1

WO2018009991A1 - Epileptic seizure prediction method and device configured for the prediction of an epileptic seizure

Info

Publication number: WO2018009991A1
Application number: PCT/BR2017/000070
Authority: WO
Inventors: Paula Renata Cerdeira GOMEZ; Hilda Alicia GOMEZ
Original assignee: Gomez & Gomez Ltda
Priority date: 2016-07-13
Filing date: 2017-06-29
Publication date: 2018-01-18
Also published as: BR102016016259A2

Abstract

The invention consists in an epileptic seizure prediction method which uses as input data the difference in electric potential between two non-invasive electrodes 2 arranged on the outer surface of the head of a user 4, said method comprising five sequential steps. The invention also consists of a device 1 configured for predicting epileptic seizures, comprising two non-invasive electrodes 2, an electronic processor, a battery and a wireless transmission component. Said device 1 being able, to calculate in real time the imminence of an epileptic seizure in a user 4. Among the main advantages of the invention are: the real-time calculation of the probability of a seizure; the low response time in the data computation; the detection of the imminence of the seizure with sufficient time in advance; the use of only two non-invasive electrodes; and portability of the device.

Description

EPILEPTIC SEIZURE PREDICTION METHOD AND DEVICE CONFIGURED FOR THE PREDICTION OF AN

EPILEPTIC SEIZURE

Field of invention

[01] The present invention belongs to the field of methods and devices for early detection of epileptic seizure.

State of the art description

[02] Epilepsy is the second leading cause of existing neurological dysfunction, second only to stroke. Approximately 1% of the world's population suffers from this disease. Nevertheless, drugs currently available in the market are not absolutely effective in treating epilepsy - they have no effect in at least 30% of the cases and when they mitigate the seizures they bring several side effects. This situation makes many patients suffer daily without the availability of a method of prevention and control of seizures that is practical, efficient and effective in predicting these events.

[03] Predicting an epileptic seizure with a certain amount of time may, for example, enable a worker to switch off or store his work tools in advance, a cyclist get off his bicycle or a swimmer to get out of the pool before the seizure occurs. Another utility of prevention is to avoid falling and compromising the physical integrity of adults and children afflicted with this disease. In possession of the information that a child may have an epileptic crisis within a few minutes, the individual's parents may lay him in bed, or keep him away from blunt objects, for example. Another utility is to alert the user to take a medication against epilepsy immediately after receiving the alert for an impending epileptic seizure.

[04] There are some techniques comprised by the state of the art which reveal methods of premeditated detection of epileptic seizures. None of them, however, reveals a practical method, which allows the identification of an epileptic seizure without any inconvenience, constraint or restriction to the patient. [05] On the following lines, there are a series of documents extracted from the academic literature that reveal algorithms and distinct methods of premeditated assessment of epilepsy.

[06] Chaovalitwongse in Performance of a Seizure Warning Algorithm Based on The Dynamics of Intracranial EEG, Epilepsy Research, vol. 64, 2005, discloses an epileptic seizure prediction method employing the estimation of the maximum short-term Lyapunov Exponent, a measure of the chaos of a dynamic system, to quantify the dynamics of each electrode in an intracranial EEG.

[07] Tetzlaff in Automated Detection of a Preseizure State: non-linear EEG analysis in epilepsy by Cellular Nonlinear Networks and Volterra systems. Int. J. Ore. Theor. Appl, Vol. 34, 2006 discloses a method of detection of epileptic seizure precursors from EEG data based on deterministic modeling of neurons to capture the behavior of neural networks, seeking to find a time series with the same characteristics of the pre-seizure, which should be previously known.

[08] Carney in Seizure prediction: Methods. Epilepsy & Behavior, vol. 22, 2011 presents a review of methods of predicting epileptic seizures from EEG, dividing them into univariate and multivariate. Among the univariate are the short-time Fourier transform, accumulated energy, autocorrelation and self-regression modeling, discrete wavelet transform, statistical moments, correlation dimension, correlation density, Kolmogorov entropy, dynamic similarity index, loss of recurrence and local flow and Lyapunov exponent. Among the multivariate are the simple synchronization measure, correlation structure, phase correlation, autoregressive measures of synchrony, index of maximum short term Lyapunov and phase synchronization,

[09] Ramgopal in Seizure detection, seizure prediction, and closed-loop warning systems in epilepsy. Epilepsy & Behavior, vol. 37, 2014 also review methods for predicting epileptic seizures from EEG, emphasizing non-linear processing methods, which analyze the spontaneous formation of spatial, temporal and spatiotemporal patterns of brain waves. The methods of entropy of permutation, Kolmogorov entropy, correlation dimension, relative energy wavelet and approximate entropy are cited. [010] Moghim and Came in Predicting Epileptic Seizures in Advance, 2014 present the ASPPR method (Advance Seizure Prediction via Pre-ictal Relabeling), divided into three components: i) selection of 14 characteristics out of 204 for each patient according to the ReliefF criterion from EEG data; ii) preparing the data by separating a part of the data for training the prediction algorithm and iii) each predictive model is trained using a multi-class support vector machine.

[011] Golestani and Gras in Can We Predict the Unpredictable? Scientific Reports, vol. 4, issue 6834, 2014 present an epileptic seizure prediction method from EEG data, using chaotic values using the P&H method in fixed time span lengths (20 minutes) and sliding time slots (the time slots moves every 20 seconds). During the seizure, a peak of P&H is obtained and if it exceeds a determined value based on previous data of the studied sample, there is an epileptic seizure detection up to 17 minutes in anticipation of the event.

[012] There are aiso patent documents dealing with the subject. However, none of them reveals a technique capable of allowing the prior detection of epilepsy, without constraints, restrictions or inconveniences to the patient.

[013] US patent document US5857978 relates to a method and apparatus for anticipated and automatic prediction of epileptic seizures from EEG brain wave analysis. The method uses linear metrics (standard deviation, absolute mean deviation, asymmetry, kurtosis) and nonlinear metrics (time-cycle steps, Kolmogorov entropy, first minimum in the mutual information function and correlation dimension) calculated from data from 16 channels of EEG and using four versions of the dataset with different mathematical transformations.

[014] US2015282755 deals with a system and method for detecting the occurrence of seizures from EEG signals combined with electrocardiogram (ECG) signals. The method described does not detect the seizure with anticipation. It is proposed a probabilistic model that classifies an event as a seizure or non-seizure.

[015] Document KR101375673 is an epileptic seizure alert method and device that applies epileptic brain wave data to an excitatory-inhibitory model based on neuronal chaos, calculating an optimal value for the connection coefficient, comparing it with the population data of brain waves of healthy people.

[016] KR20140119315 discloses an early epileptic seizure detection algorithm from the analysis of variation of a content rate metric of a given frequency band of EEG signals of the subject with epilepsy.

[017] Document KR20130068972 presents an algorithm for early detection of epileptic seizure using an autoregressive adaptive model of mandatory access, from a fractal dimension.

[018] Document RU249S769 discloses an apparatus for detecting and preventing epileptic activity by means of a microelectrode system, containing a processor, a preamp, a filter, an information stimulation unit and a power source. The microelectrode system is surgically implanted at the site of epileptic activity and functions as a diagnostic recorder or as a neurostimulator. It has a transceiver and micro antenna for transferring and receiving the recorded data. There are no details on data processing, which suggests the use of a conventional technique for this.

[019] The document TW20117770 addresses the brainwave event prediction method by transforming EEG data into a Poincare's surface chart, on which density distribution diagram and standard deviation curve are plotted for detection of epileptic seizures.

[020] WO2008057365 deals with systems and methods for predicting neurological events from sensor with various electrodes implanted in the epileptic brain and detecting peaks at certain time periods.

[021] US2008021342 relates to a method and apparatus for predicting epileptic seizures from EEG data and a probabilistic unified multrresolution framework that requires a learning period on actual data containing previous events to find the prediction strategies for each individual.

[022] US20O62OOO38 and WO2004023983 address noninvasive methods and devices for nonfinear epileptic seizure prediction from EEG data. They are specialized for patients who have focal type epilepsy, and the electrodes should be positioned strategically at the focal point for anticipation of the event.

[023] The documents US2004267152 and US2013231580 deal with methods and systems that can predict epileptic seizures by automatically training algorithms based on actual EEG data of each individual and the labeling of neurological events.

[024] WO20101416G3 relates to a real-time epileptic seizure prediction system, which includes an implantable electrode configured to transmit an analog neuro-electrophysiological signal from a subject, an analog-to-digital signal converter and a processor performing the operations of the algorithm predicted from the calculation of a plurality of autocorrelation coefficients of the neuro-electrophysiological signal in a given number of data samples.

[025] CA2383218 deals with an epileptic seizure alert method using a Lyapunov exponent- based algorithm, an indicator of chaoticity obtained from EEG data.

[026] The following problems in the state of the art are noted: inadequacy of proposed devices for continuous use in daily life, either by the number of EEG electrodes required (at least 16 electrodes) or by the need of intracranial surgical implants; need for machine learning period from previous EEG data of the specific user or from a given population to determine algorithm parameters before their practical application; finite-time statistical fluctuations and noise may fundamentally confound the predictive power of the Lyapunov exponent of EEG time series; part of the methods cited are applicable only to certain types of epilepsy that have a focal point of occurrence in the brain; all existing techniques require a large amount of input data, which overloads data processing, preventing real-time computing.

Advantages of the invention

[027] The epileptic seizure prediction method of the present invention requires only data from two non-invasive EEG electrodes without the need for implantation or surgical procedure for the application of said electrodes. [028] The epileptic seizure prediction method of the invention requires no prior database of user EEG signals or a population containing neurological events for the effectiveness of the algorithm.

[029] The epileptic seizure prediction method of the invention requires no learning period with actual user data for effectiveness of the algorithm.

[030] The epileptic seizure prediction method of the invention needs to accumulate data for only a brief period of time prior to processing. Therefore* the method of the invention functions practically in real time.

[031] The epileptic seizure prediction method of the Invention does not require image processing for the detection of the seizure.

[032] The epileptic seizure prediction method of the invention has sensitivity (or precision degree) of about 85%.

[033] The epileptic seizure prediction method of the invention reveals a short response time when compared to methods comprised by the state of the art,

[034] The epileptic seizure prediction method of the invention predicts the epileptic seizure with an average of 25 minutes in advance.

[035] The epileptic seizure prediction method can send message, visual alert, sound and / or vibration to mobile devices of the user and/or the caregiver of the user, warning about the imminence of the seizure.

[036] The method of the invention may be used in a daily-use device without, however, causing limitations to the user's well-being.

Brief description of the figures A brief description of the attached figures follows:

Figure 1: shows a front view of a first embodiment of the device of the present invention in use, positioned over the front of a user's ear- Figure 2: illustrates a perspective view of a user of the device of the invention in a second embodiment of the invention, applied to a hospital room.

Figure 3: shows a front view of the face of a user making use of a third embodiment of the device of the invention, in said embodiment, the sensors provided with a wireless system.

Figure 4; illustrates a front view of the face of a user making use of a fourth embodiment of the device of the invention, in said embodiment, a digital data processing platform is available on the web.

Figure 5: shows a flow chart of the epileptic seizure prediction method of the present invention.

Figure 6: shows a graph of electroencephalogram measurement as a function of time, associated to three windows of points, said graph being plotted with lines.

Figure 7: shows a graph of the electroencephalogram measurement as a function of time, associated to three windows of points, said graph being plotted-with points.

Figure 8: shows the same graph of figure 7, illustrating what occurs after step 3, when no abrupt, positive and simultaneous change over the same time point is detected and after step 5, the three windows of T points are shifted before repeating the method calculation steps.

Figure 9: shows an actual electroencephalogram chart at the time of an epileptic seizure and its corresponding pre-seizure, minutes before the seizure.

Figures 10 to 13: show graphs formed after the application of the first pre-seizure equation of the method of the invention.

Figure 14: shows a graph having a single peak formed after the application of the second pre- seizure equation of the invention, disqualifying the point studied as a pre-seizure. Figure 15: shows a graph with two peaks formed after the application of the second pre-seizure equation of the invention, qualifying the point studied as a pre-seizure.

Figure 16: shows an actual electroencephalogram graph at the time of an anomaly on the electroencephalogram and how the graphs of the first pre-seizure equation behave (charts in the upper row of the table) and second pre-seizure equation (charts in the lower row of the table) In case the anomaly is not confirmed as a pre-seizure (column to the left of the table) and in ease the anomaly is confirmed as a pre-seizure (column to the right of the table).

Figure 17: shows an actual electroencephalogram graph at the time of a pre~sei2ure and the respective charting behaviors of the first pre-seizure equation (major plot) and second pre-seizure equation (the two smaller plots).

Detailed description of the invention

[037] The present invention consists of an epileptic seizure prediction method and a device configured for the prediction of epileptic seizures.

[038] The epileptic seizure prediction method of the invention has as input the EEC (electroencephalogram) data of an user 4 prone to epileptic seizures, The great advantage of the epileptic seizure prediction method of the invention is the fact that this method is able to accurately identify the occurrence of a so-called "pre-seizure" event when this method analyzes an £EG data sequence 9.

[039] Pre-seizure 14 is the event preceding the epileptic seizure 15 (see electroencephalogram of figure 9). Most patients with epilepsy exhibit certain peculiar features in their electroencephalogram shortly before the occurrence of an epileptic seizure. It is not easy to identify these features, since the electroencephalogram data are extremely complex and chaotic, only when they are observed with the aid of computational tools it is possible to identify certain patterns in these signals.

[040] As described in the state of the art description of the present invention, there are already methods capable of detecting epileptic seizures, none of which, however, reveals a high degree of accuracy and efficacy, being able to detect a pre-seizure event 25 minutes before the seizure, using electronic signals sent by only two non-invasive electrodes 2, revealing at least 85% success in predicting the seizure and processing data in "real time",

[041] The epileptic seizure prediction method of the invention can be understood from the appreciation of Figure 5 of this report. Figure 5 shows a flowchart with five sequential steps: step 1 - make use of at least three windows of sequential points of different sizes, each one of them terminating the sequence at a same point, and applying to these windows of sequential points a first pre-seizure equation; step 2 - compare the three signals obtained in step 1 with each other; step 3 - when identifying a simultaneous abrupt and positive change over the same time point, skip to step 4, otherwise shift the windows on time to read new data and return to step 1; step 4 - applies the second pre-seizure equation in the data series of one of the three sequential window of points; step 5 - when two peaks are identified, report on the high probability of seizure imminence, and regardless of the identification of two peaks, after the end of this stage shift the windows on time and the method returns to step 1.

Each of these steps is explained in detail below.

[042] Step 1 - In step 1 three windows of sequential points of varied sizes are separated and for each one of them a same calculation algorithm is applied.

[043] To understand what it is meant by "window of points", the reader should refer to the graphs in figures 6 and 7 of this report. In these two graphs the difference in electric potential between two electroencephalogram electrodes 2 is plotted as a function of time. The graph of figure 6 is of the continuous type, facilitating the visualization of peaks and valleys. In the graph of figure 7, after a sampling process, the points are plotted without any kind of connection between them. The signal from these graphs is sampled at a frequency F, in other words, the interval between each one of the points plotted on these graphs Is 1/F. From figure 7 it is understood: a given quantity P of points defines a window of P points.

[044] Under the time coordinate, three windows of points are revealed.

[045] "Windows of points" means: the range of points detected between time X and Y.

[046] A "point" is: a value of a sample of the difference in electric potential indicated in the graph. The point is equivalent to a moment in time, and like all values plotted on a Cartesian plane, has two coordinates, in the present case they are (EEG,time(s)), "EEG" being the difference in electric potential of the measurement by electroencephalogram and "time(s)ⁿ being the time at which this signal was collected.

[047] Note that the three windows of points finish at the same point and the three windows have different sizes.

[048] Exemplifying, if the first window comprises the points collected between second # 0 and second # 8; the second window comprises the points between the second # 4 and # 8; and the third window comprises the points between the second # 6 and the second # 8.

[049] Preferably, the size of the windows of points (windows 1, 2 and 3) is of an order of magnitude which defines a geometric progression of ratio 2. This feature facilitates the processing of data (due to the binary character of digital processing) but it is not essential to the execution of this method.

[050] For example, if the first window of points comprises 16384 points, the second window comprises 8192, and the third window comprises 4096 points.

[051] Alternatively, instead of three windows of points, four or more windows of points may be used. The only mandatory feature here is the fact that these windows of sequential points must reveal distinct sizes and each of them must finish its series of points at the same end point (see Figures 6 and 7). [052] Preferably, the first window of points comprises 2048 points, Note that this number of points is not mandatory, it would be reasonable to use any value between 8 to 30720 points in the first window of points.

[053] For each of these windows of points (windows 1, 2 and 3), the following algorithm is applied.

[054] Definition of calculation variables:

[055] The variables and formulas of this study were developed by analogy with thermodynamic systems. By analogy it is understood that each variable and formula used in this report is not necessarily linked to the same physical property (e.g. free energy or partition function) in a thermodynamic system.

[056] There is a time series X = {x (t)}, of which the elements relevant to the method are the increments Δχ (t, T) = x (t + T) - x (t), and the values of land t can take any numbers proportional to 1/f. By keeping the same units of t, we define a new variable, the measure

which represents the probability that a given increment Δχ (t,T) occurs along the electroencephalogram, and

[057] We define the moments of order q, such that is the measure

raised to the power q. This function in which the normalized measures appear elevated to different powers q,

allows the study of the distribution of the measurements in the different moments. These quantities μ are handled by the methods specified below.

[058] Dividing the series x(t) of total length L into / segments of length N, we define the function Z{q_f N)

where

[059] This function, as demonstrated by P. Meakin et. A|., Phys. Rev. .A 34, 3325 (1986), is inversely proportional to the volume of the system, whether it is an entire system, a fractal or multifractal. [060] We define the multifractal dimension D_q, through the equation:

[061] We similarly define the function, C(q), as the second derivative of the function x{q), which can be continuous of discrete:

[062] We thus find the first pre-seizure equation: the rate of change of area under the curve

Where A(n) is the area under the curve C(q), and ζ(η) we call the first pre-seizure equation.

[063] The result of applying the algorithm used in step 1, henceforth "first pre-seizure equation" is plotted in three overlapping sequential plots, as shown in Figures 10 to 13.

[064] Step 2 - Graphs plotted as a result of step 1 are compared to each other. It should be noted that as a result of the application of the first pre-seizure equation abrupt changes in the vertical coordinate of the graph may appear:

- figure 10 reveals only one abrupt change 10 in window 2;

- figure 11 reveals no abrupt change 10;

- figure 12 reveals three abrupt changes 10, each one of them on a different point in the horizontal coordinate of the graph;

- figure 13 shows three abrupt changes 10 on the same point of the horizontal coordinate of the graph. [065] Step 3 - With a perfect synchrony between abrupt changes 10 in the graph formed from step 1 (see Figures 10 to 13), the execution of the method progresses to step 4, otherwise it reverts to step 1 and shifts the windows on time.

[066] "Shifting the window on time" means: shifting one or more points to the right of the graph (see windows lb, 2b and 3b in figure 8) and re-establishing a new window of points with the same number of points of the window used in step 1.

[067] An example of perfect synchronization between abrupt changes 10 is shown in Figure 13, wherein the three abrupt changes 10 identified in windows 1, 2 and 3 appear on the same point in the horizontal coordinate of the graph.

[068] Step 4 - a second pre-seizure equation is applied over any of the sequential windows of points (window 1, 2, or 3).

[069] The second pre-seizure equation is defined by the following mathematical formula:

[070] The result of applying the second pre-seizure equation is the production of a curve analogous to the curves shown in Figures 14 and 15 of this report.

[071] It is preferred that the second pre-seizure equation be applied over window 3, because it is less likely to contain data with large fluctuations of events prior to the point being analyzed.

[072] Step 5 - Identifying a curve analogous to the curve of Figure 14, which comprises only one peak 9, the system shifts the windows on time and returns to step 1.

[073] Identifying a curve analogous to that of figure 15, comprising at least two peaks 9, the system communicates to user 4 or to interested third parties (caregivers or physicians, for example), about the high likelihood of an imminent epileptic seizure and then shifts the windows on time and go back to step 1. [074] Figure 16 shows an actual electroencephalogram plot during the occurrence of an anomaly. From the electroencephalogram data, the graphs of the first pre-seizure equation (line Z of the table) and second pre-seizure equation (line X of the table) were plotted. In a certain window of points examined, a pre-seizure event was detected (column B of the table), in another window of points extracted from the same graph, moments before, no pre-seizure event was detected {column A of the table). Note that in column B, the graph of the first pre-seizure equation (B2 cell) reveals three abrupt changes 10 on the same point in the horizontal coordinate of the graph; the graph of the second pre-seizure equation (BX cell), by its turn, reveals more than one peak 9.

[075] Figure 17 shows an actual electroencephalogram plot at the time of a pre-seizure event. Note that in the graph of the first pre-seizure equation three superimposed abrupt changes 10 are detected, then in the second pre-seizure equation, no anomaly (only a peak 9) is detected first, and then when the windows are shifted a pre-seizure event (two peaks 9) is detected.

[076] The first pre-seizure equation consists of an electroencephalogram signal transform. This transform enables the view of details not seen in the original signal. It varies strongly and positively when the encephalogram has large fluctuations. These fluctuations are associated with the probability of occurrence of a pre-seizure, that is, of a future seizure alert. The second pre- seizure equation is analyzed if the first pre-seizure equation shows positive signals and indicates a possible alert. It is used to eliminate judgment errors that may have surfaced with the first equation, reducing the risk of false positives.

[077] The device 1 cf the present invention is configured to calculate in real time the imminence of an epileptic seizure. This device 1 comprises at least two non-invasive electrodes 2, an electronic processor, a battery and a wireless transmission component. By "non-invasive electrode" it is meant: an electrode applied directly on the epidermis of user 4, which is independent of execution of surgical method to enable its utilization.

[078] Compared to previous techniques, the possibility of having only two non-invasive electrodes 2 is a great technological advantage, which makes it possible to use device 1 permanently. In other words, since it comprises only two non-invasive electrodes 2, user 4 can make use of device 1 in any situation of his daily life (at work, driving, grocery shopping or sleeping, for example).

[079] Of course, in addition to comprising only two non-invasive electrodes 2, another feature that allows the user 4 to make use of the device 1 anywhere that is of interest is the fact that the device 1 is of a small size which enables its portability.

[080] Device 1 may be embodied in various forms, some of these embodiments are disclosed below:

[0&L] Option 1 (shown in Figure 1) discloses a portable electronic processor embedded in an anatomical compartment 11, which is compatible with the anterior portion of the ear 3 of a user 4. Said anatomical compartment 11 is analogous to the rear portion of a hearing aid.

[082] The electronic processor is associated with two non-invasive electrodes 2 and communicates wireless!y with one or more smartphone 6 devices. By smartphone 6 it is meant: any electronic device for communication or presentation of information, such as: a tablet, a telephone, cellphone, smartwatch, among other similar devices.

[083] When identifying a "pre-seizure" event, the device communicates with the smartphone 6 of user 4, which in turn communicates with the mobile device of its physician or his legal guardian (when the user 4 is Underage) informing about the occurrence; The communication may be carried out via SMS text message, app communication, a beep in sound, a visual alert on the device screen or any other kind of communication transmissible by a smartphone device 6. Preferably, it is necessary that the smartphone 6 of user 4 comprises a smartphone application configured exclusively for communication with device 1.

[084] Option 2, shown in figure 2, is preferably configured for use in hospital admission beds. For purposes of measuring the scope of protection of the present application claims, it is understood that the term "hospital admission bed" covers various forms of hospitalization such as Intensive Care Units (iCU), conventional hospitalization and home care. [085] In this embodiment, the device 1 comprises a parallelepiped-shaped casing 13 (with an external profile similar to the profile of a pack of cigarettes) associated with at least two noninvasive electrodes 2 and provided with a means of communication with a digital monitor 5. In this modality, it is also possible to use more electrodes 2, as this does not disturb the user's daily life and esthetics issues usually are not a major concern of the patient in these situations.

[086] Option 3» exibited in Figure 3 reveals two special electrodes 2', each of them comprising: a fiat adhesive surface, a battery and a wireless communication antenna. The electrodes 2' being configured to communicate directly with a remote hardware or smartphone device 6,

[087] This embodiment of the invention is interesting because it is even more discreet and easy to use than option 1,

[088] Option 4, see figure 4, consists of the provision of a digital data processing platform 8 accessible via WEB, which may be accessed by any interested user, anywhere in the world.

[089] For example, if a citizen in Japan suffers from epilepsy but does not detain the device 1 of the invention, he may plug his conventional electroencephalogram apparatus 13 to a desktop computer 7 which in turn communicates with the digital data processing 8.

[090] When the digital data processing platform 8 identifies a "pre-seizure" event it sends a signal to the user's desktop 7, alerting him/her about the imminence of an epileptic seizure.

[091] In an alternative configuration within this mode, instead of the desktop 7, the intermediate means between the digital data processing platform 8 and the conventional electroencephalogram apparatus 13 is a smartphone 6.

[092] With some preferred embodiments of the invention having been described, it is to be noted that the scope of protection provided by this document encompasses all other alternative forms applicable to the execution of the invention, which is defined and limited only by the content of the attached claims.

Claims

1. Epileptic seizure prediction method which uses as input data the difference in electric potential between at least two non-invasive electrodes (2) arranged on the external surface of a user's head (4), characterized in that, the method is subdivided into the following sequential steps: step 1 - makes use of at least three windows of sequential points of varied sizes, each one of them terminating the sequence at a same point, and applying to these windows of sequential points a first pre-seizure equation;

step 2 - compares the three signals obtained in step 1 with each other;

step 3 - identifying an abrupt, positive and simultaneous change over the same time point, skip to step 4, otherwise return to step 1;

step 4 - applies the second pre-seizure equation in the data series of one of the three windows of sequential points;

step 5 - Identifying two peaks, reports on the high probability of seizure imminence.

2. An epileptic seizure prediction method according to claim 1, characterized in that the three windows of sequential points have sizes that obey a geometric progression of ratio 2.

3. An epileptic seizure prediction method according to claim 1, characterized in that it uses between 8 and 30720 points for the largest window of points.

4. An epileptic seizure prediction method according to claim 1, characterized in that it uses 2048 points for the largest window of points.

5. An epileptic seizure prediction method according to claim 1, characterized in that the first pre-seizure equation is:

where: A(n) is an integral in the variable q of the function of the second pre- seizure; q represents the order of the powers of μ_t ; and μ_t is a normalized variation of the increment between electroencephalogram points spaced apart by a time T.

6. An epileptic seizure prediction method according to cfaim 1, characterized in that, the second pre-seizure equation is: where: q represents

the order of the powers of μ_t ; μ_t is the normalized variation of the increment between electroencephalogram points separated by a time T; and τ (q) is the free energy of the system.

7. An epileptic seizure prediction method according to claim 1, characterized In that, in step 5, the communication is performed by a visual or audible signal sent to a caregiver responsible for the user.

8. An epileptic seizure prediction method according to claim 1, characterized in that, in step 5, the communication is carried out by means of a sound or visual signal sent to the user

9. An epileptic seizure prediction method according to claim 1, characterized in that in step 5 the determination of the two peaks is performed by calculating the derivative of the curve.

10. A device (1) configured for predicting epileptic seizures, characterized in that it comprises at least two non-invasive electrodes (2), an electronic processor, a battery, a wireless transmission component, characterized in that it is able to calculate in real time the imminence of an epileptic seizure.

11. A device (1) configured for predicting epileptic seizures according to claim 10 characterized in that it is capable of detecting the imminence of an epileptic seizure with an average advance comprised between 20 and 30 minutes.

12. Device (1) according to claim 10, characterized in that it is of the portable type.

13. Device (1) according to claim 10, characterized in that it makes use of only two noninvasive electrodes (2).

14. Device (1) according to claim 10, characterized in that it comprises a portable electronic processor embedded in an anatomical compartment (11) compatible with the anterior portion of the ear (3) of the user (4).

15. Device (1) according to claim 10, characterized in that it comprises a portable electronic processor embedded in an anatomical compartment (11) compatible with the anterior portion of the ear (3) of the user (4), said hardware being able to communicate wirelessly with a smartphone device (6).

16. A device according to claim 10, characterized in that it comprises two non-invasive electrodes (2') provided with an adhesive surface; a battery and a wireless communication antenna, the electrodes (2') being configured to communicate directly with a remote hardware or smartphone device (6).

17. Device according to claim 10, characterized in that it comprises two non-invasive electrodes (2) associated with a hardware (13), which in turn associates directly to a digital monitor (5) of an intensive care unit.

18. Device according to claim 10, characterized in that it comprises two non-invasive electrodes (2), communicated directly or indirectly with a digital data processing platform (8) available on the web.