AT506113B1 - METHOD FOR CONVERTING EEG MEASUREMENT OF ELECTRIC BRAIN STREAMS TO A SQUARE FIGURE - Google Patents

Abstract

The subject invention relates to a method in which the brain waves measured by electroencephalography (EEG) are converted so that they are listed by curve symmetry and frequency change ordered in a square figure and the frequencies of the measured events by color or gray shades, different size of Squares and a number assignment from 1 to 100 come to the representation, with line by line curves appear with approximately the same symmetry, which have approximately the same rise and fall steepnesses in the middle line, the upward, correspondingly steeper descent and down, according steeper expectant Change steepness and column by column, the curves are ordered by frequency change, in the middle column appear almost constant frequencies, which are to the left, according to slowing frequency change and to the right, accordingly change faster becoming frequency change.

Description

Austrian Patent Office AT506113B1 2010-11-15

Description [0001] The invention relates to a method for the optical representation of brain waves measured by EEG according to the preamble of patent claim 1.

Such methods are used in the field of medical technology, in particular in the measurement of electrical brain waves with an EEG device, an electroencephalograph, to achieve a simple and clear square figure.

From the prior art, the detection of potential fluctuations in the cerebral cortex with electrodes of the scalp in waveform in defined frequency band ranges as conventional EEG is known.

Under continuous recording with a conventional EEG, a stimulus (click sound or light stimulus) is usually presented several hundred times. In a computer analysis, the sensory stimulus following EEG sections are superimposed and ave-raging. This eliminates the irrelevant spontaneous exchanges, while the always the same event-related potential becomes more and more visible as the number of averaging processes increases. The event-related potential has a characteristic shape: duration: at most 0.5 seconds, amplitude at most 10 mV, short positives. This becomes "P1" and is marked in the EEG by a deviation down; this is followed by a negative. It is represented by an upward deviation and is called N1 or N100 because it occurs after approximately 100 milliseconds. After further rashes there is a clear positivity. It is called a P3 or P300 wave because it occurs after about 300 milliseconds.

Known analysis methods (RU 2 029 492 C1 and WO 1990/11718 A1) are analyzes of conventional EEG waves recorded in the usual bandwidths. In the Russian patent, the bands are analyzed individually according to their frequency / amplitude characteristics and their frequency / phase relationships. In the international patent, conventional EEG waves are processed into data sets and compared with records from clinical trials. Diagnoses are derived from these comparisons. Both patents graphically display the analysis results that are based on unaltered EEG waves.

In Patent AT 502 014 B1, the voltages (potential differences) measured by an EEG are not displayed on a continuous measuring strip as a curve, but on a single image, the voltages on the x-axis and the steepnesses on the y-axis two consecutive measurements shows.

Imaging techniques such as magneto-encephalography (MRI) and positron emission tomography (PET) detect brain activity at a macroscopic anatomical level. This means that they can be used to measure the level of activity in different brain regions by measuring oxygen consumption or labeled biomolecules. The microscopic-functional level can not be detected. This level is about how nerve cells or neural networks communicate with each other. By means of the objective method, accelerations and decelerations in the direction of the action potential or in the direction of resting membrane potential can be identified and quantified in brain electrical activities.

All cognitive brain services such as attention, perception, thinking, memory and affects are based on electrical brain activity. The smallest unit of these electrical activities is the nerve cell (the neuron). The neurons are connected to a neural network. The neural network consists of subnetworks.

The technical task that results is to capture these brain electrical activities in as much detail as possible and to visualize well, to obtain electrophysiological representations for brain research and diagnosis, therapy and therapy monitoring in mental and neurological diseases. The inventive method solves the technical problem with the feature of the characterizing part of claim 1 and does not use the presentation of the directly measured by an EEG voltage values, but calculated from curve turning points frequencies and curve symmetries ,

According to the invention, it is provided that a change in the frequency in the form of a frequency change rate of the recorded signal and an asymmetry of the recorded signal in the form of the relative position of a minimum or maximum of the signal between the two nearest inflection points is determined.

Both for the change in frequency and for the asymmetry, a classification is specified with a plurality of classes, namely a plurality of frequency change classes for the classification of the change in frequency and several asymmetry classes for the classification of the asymmetry.

At any time, the change in frequency compared to a previous time is determined and assigned to exactly one frequency change class and assigned according to its asymmetry exactly one asymmetry class.

A matrix is created, wherein each row of the matrix corresponds to an asymmetry class and each column of the matrix corresponds to a frequency change class.

The occurring frequency changes and asymmetries are counted separately for each Frequenzänderungs- and asymmetry class at any time and entered in that field of the matrix whose row and column of the respective frequency change class and asymmetry class of the determined frequency change and asymmetry. The matrix is displayed.

Due to the arithmetic frequency determination curve analysis in the kilohertz area are possible. The representation is a square with five by five squares, where each square gives information about the frequency of certain waveforms and the speed of frequency changes.

Thus, it differs from the patent AT 502 014 B1 both in the type of data used for display, as well as in the representation itself.

It may further be provided that the frequency is determined according to the formula f = 1 / (t2-t1), where t1 and t2 are the times of two consecutive minima or maxi-ma or according to the formula f = 2 / (tw2-tw1), where tw1 and tw2 are the times of two consecutive inflection points, the change in frequency being determined by forming the differences of two adjacent minima, maxima or inflection points.

This enables a particularly efficient and computationally simple processing of the incoming EEG data.

Furthermore, it can be provided that the asymmetry of the signal according to the formula a = (te - ((tw2 + tw1) / 2)) * 100 / (tw2-tw1) is determined, where tw1 and tw2 the times of occurrence two successive turning points denote and where te denotes the respective time of lying between the two turning points extreme value.

By this procedure, the calculation is further simplified and achieved a more accurate result.

Further, to find a turning point, the curvature k at a time tk2 can be determined according to the formula k = w3 - ((tk3-tk1) / (tk2-tk1) * (w2-w1) + w1), where tk1 and tk3 are instants immediately before and after the time tk2 and w1, w2 and w3 are the signal values at the times tk1, tk2 and tk3, a turning point being detected at the time when the curvature is either equal to or equal to zero two immediately successive times tk1, tk2, tk3 is detected as present, for whose first time the curvature is smaller / greater than zero and for whose second time the curvature is greater / less than zero is.

This procedure is particularly efficient and requires little memory even with a large number of incoming EEG data.

Furthermore, it can be provided that the two classifications for the change of the frequency as well as for the asymmetry each contain the same number of, in particular five, classes. This allows a clear presentation of the processed data.

Another aspect of the invention provides that all matrix entries are multiplied by the factor H / m, where m corresponds to the maximum matrix entry and H is a predetermined constant, in particular 100. This improves the comparability of EEG data of different origin or different length.

Finally, it can be provided that a classification with respect to the change of frequency in five classes, with a reduction / increase of the frequency by more than 50% causes an entry in the first or fifth of the five classes, a constant frequency makes an entry in the third of the classes and a reduction / increase of less than 50% causes an entry in the second or fourth of the five classes. As a result, a particularly effective and meaningful subdivision of the data can be made, with the recorded patterns are particularly easy to differentiate for the viewer.

In the figures, an example embodiment of the subject invention is shown.

Fig. 1 shows a visualized representation of the result (a matrix) of the method according to the invention. Fig. 2 shows schematically typical waveforms that arise in an EEG recording.

Fig. 2 shows schematically typical waveforms that arise in an EEG recording. The individual curve progressions are arranged in the form of a matrix, with the respective curves shown in the matrix effecting an entry in the result matrix shown in FIG.

The 25squares method is a way to graphically represent the amount of data provided by an EEG. Instead of an endlessly long strip with a curve, you have a square grid with 25 fields in front of you, making it easier to evaluate measurement results and quickly diagnose changes.

The EEG device EMS-Biomedical delivers a maximum of 400,000 readings per second with a resolution of 16 bits. This sequence of numbers represents several superimposed waves with different frequencies.

If one were to print this curve according to the conventional method, one would arrive at the following result: All data points would have to be displayed on a strip of paper and connected by lines. Let's assume a very dense presentation method with one reading per millimeter. That would be a measurement of one second already a strip with a length of 400 meters. If the measurement takes only one minute, 24 kilometers of paper strips should be taken and viewed. Imagine a 5-minute measurement and think of a curve to be analyzed with a length of 120km.

With the method described here, the individual frequencies are determined and sorted according to waveform, frequency change and frequency of occurrence.

The 25 squares: This is a statistical method in which similar events are assigned to the same fields. Imagine a square with 25 fields. It therefore contains 5 rows and 5 columns. Accordingly, one could imagine a coordinate system with the axes x and y which each run from -2 to +2.

Lines: In the middle line (y = 0) all symmetrical curves are entered. As described below, there are two types of symmetry. Asymmetric curves go up or down in the coordinate system depending on the type of asymmetry.

Columns: Frequency changes also occur in EEGs. A constant frequency is now reflected in the middle column (x = 0). If a frequency change takes place, then the remaining columns are occupied. If the frequency is increased, the event is entered to the right of the middle; if the frequency is reduced, the left columns are filled.

Middle: The middle field with the coordinates (0/0) counts all events in which symmetric curves occurred at constant frequency. A pure sine wave would only lead to entries in this field. All other 24 fields remained unoccupied.

Diagonals: In 16 fields (x <> 0 and y <> 0) events are recorded or counted, in which both the frequency changes, and there is an asymmetry of the curves.

Intensity: There are two levels of intensity available. Let's call the events "strong". and "weak". If there is a frequency increase, for example, the event is entered in column 4 or 5, depending on the strength of the frequency change. A strong frequency increase by e.g. More than 50% of the output frequency is therefore entered in column 5 at the edge. The same applies to strong and weak asymmetries.

Colors, shades of gray: In order to facilitate the evaluation of the images, the fields are colored according to a predetermined color spectrum from red to violet or marked by predetermined gray shades and additionally by different sized squares and numbers from 1 to 100.

In order to be able to assign events in the EEG to the individual fields, a frequency analysis of the curves and the exact analysis of the symmetry values are required.

Frequency determination: First, it will be described here how the frequency is determined from the measured values.

This will first be described for non-superimposed curves. The frequency of a classical sinusoidal curve can be determined from the time difference between two consecutive maximum values. If the times (t1, t2) of the occurrence of two adjacent maximum values are known, the frequency (f) of the curve becomes: f = 1 / (t2-t1).

Between two extreme values of the curve (high point, low point) there is always also an inflection point in which the curvature of the curve is 0 or the curvature changes in its sign (for example from + to -). Also, the time interval between two turning points can be used for frequency determination. However, since there are always two inflection points between two peaks, here the formula would be f = 2 / (tw2-tw1), where tw1 and tw2 are the times of occurrence of the inflection points.

In the method described here, it is assumed that this finding can also be transferred to EEG curves which have superimposed curve courses.

It has been found that the exact position of the turning points is easier to identify than the position of the extreme values. For an EEG curve, it can be assumed that it is the superimposition of several curves of different frequencies. For a single classical sine curve, a peak can always be easily identified because it is always surrounded by lower values to the left and right. In EEG curves, however, this is no longer true in particularly steep sections. In the 25squares method, therefore, the location of the inflection points is used for frequency determination.

The method used will now be described: Let tk1, tk2 and tk3 be the times of three consecutive measurements with the measured values w1, w2 and w3, respectively, e.g. in millivolts. According to the following formula, a value k can then be determined which is directly proportional to the curvature of the curve at time t2.

K = w3 - ((tk3-tk1) / (tk2-tk1) * (w2-w1) + w1) The value of k may be positive or negative. If k changes from positive to negative or vice versa, then there is a turning point. In the turning point itself, a curvature of 0 is assumed.

Furthermore, the following is taken into account: If the value 0 for ka occurs several times in succession, then it is a straight section of the curve. The point of inflection t (w) is assumed to be exactly in the middle of this sequence according to [0055] t (w) = (t1 + t (n)) / 2 [0056] Will be in a sequence of measured values w1..w4 at the times t1 ..t4 determines a positive curvature for the times t2 and a negative curvature for the time t3, then the turning point lies exactly between t2 and t3.

For classical sinusoids, the construction is symmetrical in that an extreme value (high or low point) always lies exactly in the middle between two turning points. The extreme values are at the same time the curve parts with the strongest curvature between each two turning points with curvature 0.

For EEG curves it can be assumed that both symmetric and asymmetric curves can occur. A curve is called asymmetric if an extreme value is not exactly in the middle between two inflection points. The asymmetry can be e.g. This causes states that the rise of a curve to the extreme value (high point) is steeper than the subsequent descent to the next turning point.

It is therefore necessary to calculate the symmetry of the position of the extreme values necessarily. Here, too, it is assumed that extreme values in the EEG are revealed by a greater curvature in the course of the curve. Between two turning points, therefore, exactly one extreme value must be found. Let tl.tn be a sequence of times with the curve curvature values kl.kn, then the following facts can be assumed: The values of k1 and kn are 0.

All intervening values are necessarily different from 0, otherwise there would be an additional inflection point.

There are one or more values kx that are furthest from 0.

From this, the following knowledge can now be obtained: If a value kx is found whose absolute value is greatest (which is farthest from 0), then there is also the strongest curvature and thus an extreme value. If this largest value occurs several times, then the position of the extreme value in the middle is assumed.

Thus, a maximum value can now lie exactly in the middle between two inflection points and then applies according to convention as a symmetrical curve segment.

However, if a maximum value is closer than one of the two inflection points, then the curve is considered asymmetric since e.g. there are gradients or gradients of different degrees before and after the maximum value.

But since, depending on the location of the peak, the steeper section may occur before or after this, we speak of two types of asymmetry.

In order to allow a statistical evaluation, it is necessary to express the nature and strength of the asymmetry in numbers. That's fine with the following formula.

Be tw1 and tw2 the times of the emergence of two turning points and te the time point of the occurrence of lying between tw1 and tw2 extreme value. Then, a number a can be determined according to the following formula: a = (te - ((tw2 + tw1) / 2)) * 100 / (tw2-tw1) [0070] = 0 then there is a symmetric curve A always lies between -100 and +100 If a is not equal to 0, then an asymmetric curve is present For all positive a it is valid that they describe a similar asymmetry ,

The above formula applies to lows between two inflection points. It provides a positive result for a when the curve in the ascent after the inflection point is steeper than the descent before the trough.

For highs or maxima, a slightly modified formula is used: a = (((tw2 + tw1) / 2) -te) * 100 / (tw2-tw1) Superimposed curves can be accomplished as follows: An EEG consists of several superimposed curves of different frequency and amplitude. When determining the frequencies according to the method described above, the highest level is obtained, which corresponds to the highest ascertainable frequencies. The next curve below is longer wavelength and contains lower frequencies. Imagine two different frequencies, which are connected by addition to a mixing curve. If you create a new curve exactly through the inflection points - which are determined by our method - then this new curve corresponds exactly to the course of the lower of the two frequencies. If the position of the inflection points is calculated again for this new curve, its frequency can also be determined exactly.

This process could now be continued and repeated to infinity. It can be assumed that by this method almost all addition of the curve parts can be separated from each other. Lost only curves that cancel each other out by opposing vibrations, or curves that have exactly the same amplitude at slightly different frequencies.

When a high follows a lower frequency, there is a frequency change. The same applies, of course, in the reverse case. In the 25squares method, all events with the property "Frequency remains the same". entered in the middle column (y = 0). All events at which frequency increases come to the right columns 4 and 5, depending on intensity of an event. If the intensity is> 50%, the fifth column will be filled.

Likewise, events occur in which the frequency decreases in the left columns 1 and 2, again the intensity is crucial.

FIG. 2 schematically shows typical curves which occur during an EEG recording. The individual curve progressions are arranged in the form of a matrix, with the respective curves shown in the matrix effecting an entry in the result matrix shown in FIG.

Line Z1 to Z5, curves are detected, each with a similar rise and descent slope. In the middle row Z3 are curves with approximately the same rise and descent slope. Upwards, the descent increases, downwards, the steepness of increase increases.

Line Z1 contains highly unbalanced curves. The peaks are shifted to the right relative to the turning points. The test values are accordingly far left. Line Z2 contains slightly unbalanced curves. The maximum values are slightly shifted to the right. This corresponds to a slightly shallower rise and slightly steeper descent of the 6/10

Claims (7)

Austrian Patent Office AT506113 B1 2010-11-15 Curves. Line Z5 contains highly unbalanced curves. The maximum values are shifted strongly to the left relative to the turning points. The test values are accordingly far left. Line Z4 contains slightly unbalanced curves. The highs are slightly shifted to the left. This corresponds to a slightly flatter rise and slightly steeper descent of the curves. Columnwise, S1 to S5, the speed of the frequency changes, i. the rate of change of frequency, recorded. In the middle column S3 no frequency change takes place. To the left is an increasingly strong change from fast frequencies to slow ones; to the right from slow to fast. The mean square S3 / Z3 shows e.g. Curves with constant rise and descent steepness with constant frequency; the square S5 / Z1 Curves with steep descent, where a strong change from low to high frequencies occurs. A significant advantage of the invention is the comparability of the measurements obtained. With the methods described here, it is now possible to enter each event occurring in the EEG in one of the 25 fields. Differently long measurements, however, require varying degrees of filling of the 25 fields. In order to be able to compare two measurements of different lengths, it is necessary to adapt the numerical values in the fields. Scaling is done for this purpose. This is achieved by scaling to 100 in the 25squares method. First the field with the most entries is searched. Experience has shown that this is in the middle row e.g. in column 2 or 3. Let the value m be the largest number in the 25-field raster. The following formula determines a factor f. This serves as a multiplication factor to scale all fields on a scale from 0 to 100 or to transform them into a range from 1 to 100. F = 100 / m If one now multiplies all 25 fields by the factor f, then one finds only values between 0 and 100, whereby the field with most entries is always reduced or scaled to 100, one red color is assigned and represented by the largest square. In FIG. 1, the result of a measurement as a matrix comprising the 25 fields is shown graphically by way of example. For purposes of illustration, squares of different size and shading are used, with the respective value of the respective matrix element inscribed in the respective square. The more often an event (to be described) occurs, the larger the square, the darker the shading, the higher the assigned number, which can be between 1 and 100. The more frequently the respective event takes place, the larger the square and the higher the number in the respective field. The fields are colored according to a color spectrum (not shown), with equal clusters (numbers in the field) being represented by the same colors. In addition, the time interval of the recording, the considered frequency band and a file name are shown. 1. A method for the optical representation of brain waves measured by EEG, wherein the brain waves are recorded and in the form of a signal, characterized in that - a change in frequency in the form of a frequency change rate of the recorded signal and an asymmetry of the recorded signal in the form of The relative position of a minimum or maximum of the signal between the two nearest inflection points is determined - that for the change of frequency as well as for the asymmetry a classification with in each case several classes is given, namely several frequency change classes for the classification of the change of the frequency and several asymmetry classes for the classification of the asymmetry, - whereby the change of the frequency compared to a previous time is determined at any time and assigned to exactly one frequency change class and according to its asymmetry is assigned to exactly one asymmetry class, - a matrix is created, each row of the matrix corresponding to an asymmetry class and each column of the matrix corresponds to a frequency change class, - the occurring frequency changes and asymmetries at each time point separately for each frequency change and asymmetry class counted and entered into that field of the matrix whose row and column correspond to the respective frequency change class and asymmetry class of the ascertained frequency change and asymmetry, and - the matrix is displayed. 2. The method according to claim 1, characterized in that the frequency is determined according to the formula f = 1 / (t2-t1), wherein t1 and t2 are the time points of two consecutive minima or maxima or according to the formula f = 2 / ( tw2-tw1), where tw1 and tw2 are the times of two consecutive inflection points, the change in frequency being determined by forming the differences of two adjacent minima, maxima or inflection points. 3. The method according to any one of the preceding claims, characterized in that the asymmetry of the signal according to the formula a = (te - ((tw2 + tw1) / 2)) * 100 / (tw2-tw1) is determined, wherein tw1 and tw2 The times of occurrence of two successive inflection points denote te and designated the respective time of the lying between the two inflection points extreme value. 4. The method according to any one of claims 1 to 3, characterized in that for determining a turning point, the curvature (k) at a time tk2 according to the formula k = w3 - ((tk3-tk1) / (tk2-tk1) * (w2 w1) + w1), where tk1 and tk3 are instants immediately before and after the time tk2 and w1, w2 and w3 are the signal values at the times tk1, tk2 and tk3, respectively, with a turning point at the time being recognized as present to which the curvature is either equal to or equal to zero between two immediately consecutive times tk1, tk2, tk3, for whose first time the curvature is less than / greater than zero and for the second time the curvature is greater / less than zero is. 5. The method according to any one of the preceding claims, characterized in that the two classifications for the change of the frequency as well as for the asymmetry each contain the same number of, in particular five, classes. 6. The method according to any one of the preceding claims, characterized in that all matrix entries are multiplied by the factor H / m, where m corresponds to the maximum matrix entry and H is a predetermined constant, in particular 100. Method according to one of the preceding claims, characterized in that a classification with respect to the change of the frequency takes place in five classes, whereby a reduction / increase of the frequency by more than 50% causes an entry in the first or fifth of the five classes, if frequency remains the same, it will result in entry into the third of the classes and a reduction / increase of less than 50% will result in entry into the second or fourth of the five classes. For this 2 sheet drawings 8/10