JP5028579B2 - Arrhythmia signal detector and defibrillator - Google Patents

Description

The present invention relates to non Seimyaku signal detection apparatus and a defibrillator.

Conventionally, an apparatus for identifying an arrhythmia using a neural network is known (for example, see Patent Document 1).
The apparatus described in Patent Document 1 converts a detected digital QRS waveform into a frequency power spectrum, outputs a power spectrum value divided into a plurality of components in a specific frequency region, and outputs it to a neural network. Given, the pattern of the detected QRS waveform is recognized, and normal sinus node rhythm and arrhythmia are discriminated. In this device, normal sinus node rhythm and arrhythmia can be obtained by using an improved technique in which the processing target of the neural network is made information by frequency analysis of the electrocardiogram signal without directly applying the QRS waveform to the neural network processing. Has been identified.
Japanese Patent No. 2918472

  However, a weak point peculiar to a frequency analysis method such as Fourier transform may be included in this case, and there is a disadvantage that a signal to be extracted is weak or noise is present. In addition, if there is a waveform with a large signal intensity that covers the waveform to be detected, it is difficult to extract the information to be detected, and appropriate pre-processing that requires ingenuity in accordance with the waveform is required. It is hard to say that it can be applied.

  In practice, in order to use the neural network method, there is the same trouble as other neural networks that it is necessary to learn while processing various large amounts of data and to create data for that. The configuration of the FFT processing and the neural network processing with heavy processing load is very complicated, and there is still room for improvement for application in an in vivo implantable device. In addition, there is a weak point that cannot detect arrhythmia because there is no learning data, such as when an untrained arrhythmia symptom is encountered.

The present invention was made in view of the above circumstances, without the need for pretreatment and learning of large amounts of data, it allows to identify arrhythmia signal to any electrocardiographic waveform not Seimyaku An object is to provide a signal detection device and a defibrillation device.

In order to achieve the above object, the present invention provides the following means.
The present invention detects an electrocardiogram signal from a plurality of electrodes arranged in the heart, calculates a mutual information amount of the detected plural electrocardiogram signals, and the calculated mutual information amount is a predetermined threshold value. Provided is a method for detecting an arrhythmia signal that is determined to be an arrhythmia signal when it is smaller.

  In the above invention, the mutual information amount and the simultaneous occurrence frequency of a plurality of electrocardiogram signals are calculated, and when the variance value is smaller than a predetermined threshold value, it is determined that the fibrillation waveform is present, and the variance is When the value is greater than a predetermined threshold value, it may be determined that the waveform is a ventricular tachycardia waveform.

Further, in the above invention, the mutual information amount may be calculated by determining the number of samples within a specific time width.
In the above invention, the time width and the number of samples within the time width may be changeable according to the patient's condition and symptoms and the detection accuracy of the electrocardiogram signal.

Further, in the above invention, one mutual information amount may be calculated by determining the number of samples within a specific time width, and the mutual information amount within the next time width may be calculated by shifting the time width. .
Further, in the above invention, the shift amount of the time width may be changeable.

Moreover, in the said invention, it is good also as equally dividing the level of the electrocardiogram signal detected from the said some electrode.
Moreover, in the said invention, it is good also as carrying out the nonlinear division | segmentation of the level of the electrocardiogram signal detected from the said several electrode.

The present invention also detects electrocardiogram signals from a plurality of electrodes arranged in the heart, calculates Pearson's χ ² statistic of the detected plurality of electrocardiogram signals, and the calculated χ ² statistic A method for detecting an arrhythmia signal that is determined to be an arrhythmia when smaller than a predetermined threshold value is provided.
In the above invention, the co-occurrence frequency of the χ ² statistic and a plurality of electrocardiogram signals is calculated, and when the variance value is smaller than a predetermined threshold value, it is determined that the fibrillation waveform is present, When the variance value is larger than a predetermined threshold value, it may be determined that the waveform is a ventricular tachycardia waveform.

Moreover, in the said invention, it is good also as calculating the said (chi) ² statistic by determining the sample number within a specific time width.
In the above invention, the time width and the number of samples within the time width may be changeable according to the patient's condition and symptoms and the detection accuracy of the electrocardiogram signal.

In the above invention, the number of samples within a specific time width is determined to calculate one χ ² statistic, and the time width is shifted to calculate the χ ² statistic within the next time width. Also good.
Further, in the above invention, the shift amount of the time width may be changeable.

Moreover, in the said invention, it is good also as equally dividing the level of the electrocardiogram signal detected from the said some electrode.
Moreover, in the said invention, it is good also as carrying out the nonlinear division | segmentation of the level of the electrocardiogram signal detected from the said several electrode.

  The present invention also includes a plurality of electrodes arranged in the heart, a mutual information amount calculation unit that calculates a mutual information amount based on a plurality of electrocardiogram signals detected by the electrodes, and the mutual information amount calculation unit. There is provided an arrhythmia signal detection device comprising: a determination unit that determines whether or not the calculated mutual information amount is smaller than a predetermined threshold value.

The present invention also provides a plurality of electrodes arranged in the heart, a χ ² statistic calculation unit for calculating χ ² statistics based on a plurality of electrocardiogram signals detected by the electrodes, and the χ ² statistic calculation. An arrhythmia signal detection device comprising: a determination unit that determines whether or not the χ ² statistic calculated by the unit is smaller than a predetermined threshold value.

  In addition, the present invention is provided with the above arrhythmia signal detection device, and when an arrhythmia signal is detected by the arrhythmia signal detection device, generates an electrical stimulation pulse to be applied through an electrode arranged in the heart Provided is a defibrillator including a unit.

  According to the present invention, an arrhythmia can be identified for any electrocardiographic waveform without requiring preprocessing or learning of a large amount of data.

A method for detecting an arrhythmia signal according to an embodiment of the present invention will be described below with reference to FIGS.
As shown in FIG. 1, the method for detecting an arrhythmia signal according to the present embodiment acquires an electrocardiogram signal from the right ventricle and the left ventricle using electrodes arranged in the right ventricle and the left ventricle, respectively. S1, a mutual information calculation step S2 for calculating the mutual information amount of the two types of acquired electrocardiogram signals, and an arrhythmia determination step for determining whether the signal is an arrhythmia signal based on the calculated mutual information amount S3.

  The mutual information amount is a quantification of how much the other signal can be predicted from one signal for two events (random variables), and can be calculated by the following equation (1).

Where A | a _i , i = 1, 2,..., M and B | b _j , j = 1, 2,..., N are random variables, H (A) is the uncertainty of A before knowing B , H (a | B) is uncertainty of a after knowing _{B, p (a i),} p (b j) is _a i, the probability of occurrence of _{b j, p (a i,} b j) is a _This is the co-occurrence probability of _i and b _j .

The upper and lower limits of the mutual information amount satisfy the following expression (2).
0 ≦ I (A, B) ≦ H (A), H (B) (2)
Here, I (A, B) = 0 becomes p (a _i , b _j ) = p (a _i ) p (b _j ), that is, when A and B are independent of each other. The equal sign on the right holds when one is completely subordinate to the other.

  Further, the mutual information amount in the case of three dimensions can be calculated by the following equation (3).

In the mutual information calculation step, the electrocardiographic signals acquired from the right ventricular and left ventricular electrodes as shown in FIG. 2A are regarded as random variables A and B and shown in FIG. Histograms are generated, and components a _i (i = 1, 2,..., M) of the random variable A and components b _j (j = 1, 2,..., N) of the random variable B are included. The occurrence probabilities p (a _i ) and p (b _j ) are calculated by the following expression (4) using the number of elements k _{i ·} , k _{· j} in the bin. Here, m and n are the numbers of bins of A and B, respectively _, and the subscript “·” in k _{i ·} and k _{· j} means that the subscripts corresponding thereto are i and j.

p (a _i ) = k _{i ·} / N
p (b _j ) = k _{· j} / N (4)
Here, N is the total number of samples.

FIG. 2 illustrates, for example, an electrocardiogram signal acquired from an electrode of the right ventricle, and an illustration of the electrocardiogram signal acquired from the electrode of the left ventricle is omitted.
Next, as shown in FIGS. 3A and 3C, the electrocardiographic signals acquired from the right ventricular and left ventricular electrodes are arranged on the vertical axis and the horizontal axis, and FIGS. As shown in (d), histograms are respectively generated, and as shown in (e) of the figure, the number of elements in the box containing (a _i , b _j ) is set as k _ij and the co-occurrence probability p (a _i , b _j ) is calculated by the following equation (5).

Then, the occurrence probabilities p (a _i ), p (b _j ) and the co-occurrence probabilities p (a _i , b _j ) calculated in this way are substituted into the equation (1), and the mutual information of the equation (6) is obtained. The amount is calculated.

  In order to calculate a sufficient number of samples and a mutual information amount at a relatively short time interval, as shown in FIG. 4, a time window of N samples is cut out from each electrocardiogram, and the time window is, for example, N / While shifting by 10 each time, the calculation of the mutual information amount is repeated. In this case, the histogram is calculated by dividing the maximum value and the minimum value into n equal parts in each time window.

  In the arrhythmia determination step, it is determined whether or not the mutual information amount calculated in the mutual information amount calculation step is lower than a predetermined threshold value.

An example of the arrhythmia signal detection method according to the present embodiment will be described below with reference to FIGS.
The electrocardiograms shown in FIGS. 5 to 8 used in the arrhythmia signal detection method according to this example are dog electrocardiograms acquired in an arrhythmia induction experiment. ECGs were obtained from electrodes placed in the right and left ventricles, respectively. Sampling frequency was acquired at 200 Hz and analyzed after upsampling to 250 Hz.

And the mutual information amount was computed from the acquired biventricular electrocardiogram signal.
Here, for example, the number of bins n = 5, the number of blocks n × n = 5 × 5, and the total number of samples N = 250. As shown in FIG. 4, the mutual information amount was obtained in each time window while shifting the time window of the number of samples N = 250 by 50 samples.
Further, within the same time window as the above time window, Pearson's correlation coefficient was calculated by the following equation (7) and compared.

FIG. 5 shows a comparison between mutual information and correlation coefficient during ventricular tachycardia (VT).
FIG. 6 shows a comparison between mutual information and correlation coefficient during ventricular fibrillation (VF).
Further, FIG. 7 shows a comparison between mutual information and correlation coefficient during atrial tachycardia (AT).
8 shows a comparison between the mutual information amount and the correlation coefficient in normal (SR). In FIGS. 5 to 8, (a) is an electrocardiogram signal from an electrode arranged in the right ventricle, and (b) is An electrocardiogram signal from an electrode arranged in the left ventricle, (c) shows a mutual information MI (solid line) and a correlation coefficient PCC (dashed line).

5, 6, and 8, when ventricular tachycardia occurs and when ventricular fibrillation occurs, both the mutual information amount and the correlation coefficient greatly change from the normal time. The arrhythmia can be detected.
On the other hand, referring to FIGS. 7 and 8, when an atrial tachycardia occurs, the mutual information changes differently from the normal time, but the correlation coefficient does not change much from the normal time, and the atrial tachycardia. Cannot be detected. However, in the correlation coefficient, there is a possibility that the atrial tachycardia signal is erroneously detected as a ventricular arrhythmia signal.

In order to further clarify the above results, the box whisker chart shown in FIG. 9 is used for evaluation.
The box-and-whisker plot is created by first sorting each data and then plotting it according to the quartile range.
The interquartile range is the interval between the first quartile value (25% from the smallest value) and the third quartile value (value from the smallest to 75%), and the second quartile value is Refers to the median. The lower whiskers are then drawn from the first quartile to the lower quartile range × 1.5 in length. At this time, if there is a minimum value in the range, the line is drawn to the minimum value. Similarly, the upper whiskers are drawn from the third quartile value upward by a length of quartile range × 1.5. If the maximum value is within this range, draw a line up to the maximum value. Outliers (indicated by +) refer to those that exceed the whiskers both vertically and horizontally.

  According to the box-and-whisker diagram created in this way, the mutual information amount and the correlation coefficient are lower during ventricular arrhythmia (VT, VF) than when normal (SR). On the other hand, mutual information increases during atrial tachycardia (AT), but the correlation coefficient may decrease. Therefore, the correlation coefficient may not distinguish between ventricular arrhythmia and atrial tachycardia.

  According to the mutual information in the box plot of FIG. 9, ventricular arrhythmia is separated from the quartile range of sinus rhythm and atrial tachycardia (box range). What is sinus rhythm and atrial tachycardia? It can be clearly distinguished and determined. On the other hand, in the correlation coefficient, there is an overlap in the interquartile ranges of ventricular arrhythmia, sinus rhythm and atrial tachycardia. Further, the correlation coefficient has a wide range of outliers (+) of ventricular arrhythmias, and the data varies, making it difficult to make a qualitative judgment.

  Further, according to the box-and-whisker diagram of FIG. 9, it is impossible to distinguish between sinus rhythm (SR) and atrial tachycardia (AT), but because of the tachycardia state in the AT interval, the QRS wave peak The part corresponding to the value is larger than the SR section for each time window of FIG. Accordingly, since the histogram corresponding to the peak value in the time window changes in the SR section and the AT section, the discrimination is possible.

  Further, according to the box-and-whisker diagram of FIG. 9, it is difficult to distinguish between ventricular tachycardia (VT) and ventricular fibrillation (VF), but a QRS wave is emitted although it is tachycardia in the VT interval. Therefore, the co-occurrence frequency of the left ventricular electrocardiogram and the right ventricular electrocardiogram at the peak portion and the baseline portion of the electrocardiogram waveform is counted concentrated on the baseline portion. Also, calculate the variance of the co-occurrence frequency. Since the frequency distribution in the VF section is smooth, the variance is small. On the other hand, in the SR, AT, and VT sections, the variance is large because the frequency of the base line portion and the peripheral portion (peak portion of the electrocardiogram) is greatly different. This makes it possible to distinguish between VT and VF.

That is, differences between SR, AT (atrial tachycardia), VT (ventricular tachycardia), and VF (ventricular fibrillation) can be detected by a change in the mutual information MI. The MI value is small during VT and VF, and the MI value is large during SR and AT.
Further, the simultaneous occurrence frequency of the left ventricular electrocardiogram and the right ventricular electrocardiogram is greatly different between VT and VF, and the difference between VT and VF can be detected from the variance value.

  As described above, according to the arrhythmia signal detection method according to the present embodiment, learning and preprocessing using a large amount of data as in the neural network technique is not required, and noise processing is performed without performing heavy processing load FFT processing. Even if the signal is relatively large or based on an electrocardiographic signal that is difficult to detect, arrhythmia can be identified more reliably. In addition, a simple apparatus can be realized by processing with a light load. Furthermore, there is an advantage that not only ventricular tachycardia and ventricular fibrillation but also atrial tachycardia can be identified.

In sinus tachycardia, the excitement of the heart is normally performed, so both ventricles excite almost simultaneously via the His-Purkinje system. Even if ectopic excitement occurs in the atria of supraventricular tachyarrhythmia, both ventricles are excited almost simultaneously because the transmission of excitement between the atria is via the His-Purkinje system.
On the other hand, since ventricular arrhythmia is caused by ectopic excitement in the ventricle, the excitability of both ventricles is not synchronized without going through the His-Purkinje system.

  As shown in Table 1, the mutual information amount calculated from the electrocardiograms of both ventricles is large because sinus tachycardia and supraventricular arrhythmia are dependent on each other. On the other hand, in the case of ventricular arrhythmia, the independence of the electrocardiogram waveform is increased, so the value becomes small. Therefore, there is also an advantage that the inconvenience of erroneous determination in signal detection can be avoided without misidentifying an arrhythmia signal of non-lethal supraventricular tachycardia as a ventricular arrhythmia signal.

Here, as shown in FIG. 10, the accuracy of the test using the mutual information and the correlation coefficient as indices as the threshold for detecting the lethal arrhythmia signal is expressed as ROC (receiver-operator).
characteristics) Evaluate by curve.
FIG. 10A is a distribution of diagnostic indices, and FIG. 10B is a schematic ROC curve.
Here, using the frequencies TP, FN, TN, and FP shown in Table 2, sensitivity = TP / (TP + FN), specificity = TN / (TN + FP), false positive rate = 1−specific Degree.

The ROC curve is used for evaluation of accuracy such as screening inspection and comparison between a conventional inspection and a new inspection, and indicates in which range the cut-off point should be taken.
The cut-off point can visually indicate the ability of the test to distinguish between the disease state group and the normal group in the diagnostic index. When determining superiority of different examinations in the ROC curve, it is judged that this curve is more excellent as it is located at the upper left.

An ROC curve area (ROCA) is used as an accuracy index.
As a result, ROCA = 0.987 for the mutual information test and ROCA = 0.816 for the correlation coefficient test. Moreover, the ROC curve rises sharper in the test based on the mutual information amount. This indicates that the sensitivity can be improved without decreasing the specificity.

Next, the criteria for detecting VF based on the frequency of simultaneous occurrence of the left ventricular electrocardiogram and the right ventricular electrocardiogram will be described.
The electrocardiogram waveform at the time of SR, AT and VT is mainly composed of a base line portion and a QRS group having a steep rise. Therefore, the co-occurrence frequency of the left ventricular electrocardiogram and the right ventricular electrocardiogram is counted concentrated on the base line portion. On the other hand, since an irregular waveform is continuously observed in the electrocardiogram during VF, the co-occurrence frequency is smoothly distributed around the base line portion. FIG. 12 shows the co-occurrence frequency. In order to quantify such differences, the variance of the co-occurrence frequency is calculated. Since the frequency distribution of VF is smooth, the variance is small. On the other hand, SR, AT, and VT have a large variance because the frequencies of the base line portion and the peripheral portion (peak portion of the electrocardiogram) are greatly different. Therefore, the VF can be detected also by the difference in the dispersion value.

In this embodiment, each electrocardiogram is analyzed using the mutual information amount MI. However, the same analysis can be performed using Pearson's χ ² statistics.
Similarly to the mutual information calculation, the co-occurrence probability p (a _i , b _j ) of the electrocardiogram as shown in FIG. 2 counts the number k _{ij of the} box containing (a _i , b _j ) from FIG. Next, based on the counted k _ij , the m × n contingency table shown in FIG. 13 is created, and the χ ² statistic T is calculated by the following equation (8).

The algorithm for arrhythmia classification is shown in the following procedures 1 to 3. FIG. 14 shows a flowchart of arrhythmia classification.
(step 1)
SR or AT (hereinafter referred to as SR / AT) and VT or VF (hereinafter referred to as VT / VF) are separated by the mutual information (MI) or χ ² statistics (T) of the right ventricular electrocardiogram and left ventricular electrocardiogram. To do.
SR / AT if (MI or χ ² )> α,
If (MI or χ ² ) ≦ α, VT / VF.
Here, α is a classification threshold.

(Step 2)
SR and AT are separated by mutual information MI or χ ² statistic T between right atrial electrocardiogram and right ventricular electrogram. At this time, the right atrial electrocardiogram is shifted in the past by Td in order to correct the conduction delay Td between the atrioventricles. This synchronizes the P wave of the normal right atrial electrocardiogram and the R wave of the right ventricular electrocardiogram.
SR if (MI or χ ² )> β.
If (MI or χ ² ) ≦ β, it is AT.
Here, β is a classification threshold.

(Procedure 3)
The variance V of the co-occurrence frequency obtained from the co-occurrence frequency k _{ij of the} right ventricular electrocardiogram and the left ventricular electrocardiogram is calculated by the following equation (9), and VT and VF are separated.

ここで、μは、同時生起頻度の平均値である。

Here, μ is an average value of the co-occurrence frequency.

V VT if> γ,
V If ≦ γ, it is VF.
Here, γ is a classification threshold.

FIG. 15 shows a box-and-whisker plot based on the χ ² statistic T calculated from the left ventricular electrocardiogram and right ventricular electrocardiogram.
It can be seen that the use of the χ ² statistic T in the electrocardiogram analysis is as effective as the analysis based on the mutual information MI, and the difference between SR, AT and VT, VF can be detected.

Next, according to the procedure 2, the χ ² statistic T of the right atrial electrocardiogram and the right ventricular electrogram is calculated. FIG. 16 shows an example in which the mutual information MI and χ ² statistics T at the time of SR are obtained. Similarly, FIG. 17 shows an example in which the mutual information MI and χ ² statistics T at the time of AT are obtained. It can be seen that the mutual information MI and the χ ² statistic T are small in the AT section.
FIG. 18 shows box whiskers of the right atrial electrocardiogram and the left ventricular electrocardiogram.

  Next, for the VT section and the VF section, the variance of the co-occurrence frequency is calculated according to the flowchart of FIG. As shown in FIG. 12, since the frequency distribution of VF is smooth as shown in FIG. 12, the variance is small. On the other hand, SR, AT, and VT have a large variance because the frequencies of the base line portion and the peripheral portion (peak portion of the electrocardiogram) are greatly different. Therefore, VF can be detected from the difference in the dispersion value.

  In the present embodiment, when the amplitude of the detected electrocardiogram signal is divided, the linear division that equally divides between the maximum value and the minimum value has been described as an example. For example, nonlinear division may be adopted in which points whose absolute values of amplitude are, for example, maximum value × 0.2 or less are collected in one box and the other levels are linearly divided.

  In this method, the maximum value of the absolute value of the electrocardiogram waveform is obtained, and those below 20% are regarded as the zero level. Except for the level of 20% or less, the other levels (20% or more) are equally divided to determine the dividing points and set the values in the corresponding boxes. Thereby, there exists an advantage that the noise etc. which arise in a base line part can be reduced.

Next, a defibrillator 1 according to an embodiment of the present invention using the arrhythmia signal detection method will be described below with reference to FIGS. 19 and 20.
As illustrated in FIG. 19, the defibrillator 1 according to the present embodiment includes an arrhythmia signal detection device 2, a treatment control unit 3, and a stimulation pulse generation unit 4.

  The arrhythmia signal detection device 2 is detected by the three electrodes 5a, 5b, 5c arranged in the heart A, the switch unit 6 that switches connection of these electrodes 5a, 5b, 5c, and the electrodes 5a, 5b, 5c. An information processing unit 7 that calculates a mutual information amount based on an electrocardiogram signal and an analysis / diagnostic unit 8 that analyzes whether the information is an arrhythmia based on the mutual information amount calculated by the information processing unit 7 are provided.

The electrodes 5a, 5b, and 5c are disposed, for example, in the coronary sinus of the right atrium, right ventricle, and left ventricle, respectively.
As shown in FIG. 20, the information processing unit 7 includes a switch 9, an amplifier 10, a filter 11, an A / D converter 12, a level division unit 13, a ring memory 14, a histogram unit 15, And a mutual information calculation unit 16.

The switch 9 selects an electrocardiogram signal from any two of the three electrodes 5a, 5b, 5c.
The amplifier 10 amplifies two electrocardiographic signals selected by the switch 9.

The filter 11 removes noise contained in the two electrocardiographic signals amplified by the amplifier 10.
The A / D converter 12 converts an electrocardiogram signal into a digital signal.
The level dividing unit 13 equally divides an electrocardiogram signal which has been amplified, noise-removed and A / D converted from two electrodes (for example, the electrode 5a and the electrode 5b in FIG. 20) selected by the switch 9 (for example, The level conversion is performed by dividing into five levels. That is, the maximum value and the minimum value in the sample are equally divided.

  The ring memory 14 is a memory arranged so that the ECG signals level-converted by the level dividing unit 13 can be sequentially written and read in accordance with the sampling timing. Thus, recording is sequentially performed according to the sampling timing. Reading of the memory arranged in a ring shape is performed at a different timing so as to read a predetermined arrangement width (time window). The reading start position is also read out by shifting the memory in the arrangement direction by a predetermined number of memories (number of steps).

  The histogram section 15 includes an X array histogram section 15a, a Y array histogram section 15b, and a two-dimensional array histogram section 15c. Each of the X and Y array histogram sections 15a and 15b receives memory values corresponding to the time window recorded in a part of the ring memory 14, is converted into a histogram array corresponding to the memory values, and is recorded. It is like that.

The two-dimensional array histogram unit 15c is a two-dimensionally arranged memory, which is read out in synchronization with the read timing of the two ring memories 14, and converted into a histogram array in a two-dimensionally arranged memory as one point for each X value and Y value. And is now recorded.
In order to simplify the configuration, the values of the Y array histogram section 15b and the X array histogram section 15a are read out from the two-dimensional array histogram section 15c for each X array and added or read out for each Y array and added. It is good.

The mutual information calculation processing section 16 is recorded in the memory value k _{i ·} recorded in the X array histogram section 15a, the memory value k _{· j} recorded in the Y array histogram section 15b, and the two-dimensional array histogram section 15c. After reading the stored memory value k _ij , the mutual information MI is calculated, and the value is transferred to the analysis / diagnosis unit 8.
In this operation, signals from the heart A are sequentially detected by the electrodes 5a, 5b, and 5c, and the value of the mutual information MI is calculated and output while shifting the detection window by a predetermined number of steps. Yes.

Further, FIG. 22 shows an arrhythmia signal detection apparatus including a χ ² statistic calculation processing unit 16 ′ instead of the mutual information calculation processing unit 16. The χ ² statistic calculation processing unit 16 ′ stores the memory value k _{i ·} recorded in the X array histogram unit 15a, the memory value k _{· j} recorded in the Y array histogram unit 15b, and the two-dimensional array histogram unit 15c. After reading the recorded memory value k _ij , the χ ² statistic T is calculated, and the value is transferred to the analysis / diagnostic unit 8.
In this operation, signals from the heart A are sequentially detected by the electrodes 5a, 5b and 5c, and the value of the χ ² statistic T is calculated and output while shifting the detection window by a predetermined number of steps. ing.

Analysis and diagnosis section 8, from the value of the mutual information MI sent from mutual information processing unit 16, or from the value of chi ² statistic operation processing unit 16 'is sent from the chi ² statistic T Whether or not the signal is an arrhythmia signal is determined, and the necessity of outputting a stimulation pulse is determined based on the determination result. When the stimulation pulse needs to be output, a signal is output from the analysis / diagnosis unit 8 to the treatment control unit 3. Thereby, the stimulation control pulse generator 4 is controlled by the treatment controller 3, and the switch is switched so that the stimulation pulse is transmitted to the electrodes 5a, 5b, 5c, and then the stimulation pulse is applied to the heart A. . After the stimulation pulse is given, the switch 6 is switched so that two of the three electrodes 5a, 5b, and 5c and the sensor unit 7 are connected to detect the signal from the heart A again.

According to the defibrillator 1 according to the present embodiment configured as described above, the mutual information MI or χ ² statistics of the plurality of electrocardiographic signals detected by the plurality of electrodes 5a, 5b, 5c attached to the heart A. Based on the quantity T, there is an advantage that a lethal arrhythmia signal such as ventricular tachycardia, ventricular fibrillation or atrial tachycardia can be easily detected and a stimulation pulse can be output.

  In the present embodiment, the two-dimensional array histogram unit 15c is employed as the histogram unit, but instead, a three-dimensional array histogram unit 15c ′ may be employed as shown in FIGS. Good. In this case, the third Z array histogram portion 15d is necessary. In addition, three amplifiers 10, filters 11, A / D converters 12, level division units 13, and ring memories 14 are also provided. Further, the switch 9 is deleted and the switch 9 'is employed.

Then, reading is performed from the three ring memories 14 while synchronizing the reading timing, and each X value, Y value, and Z value is converted into a histogram array and recorded in a three-dimensionally arranged memory as one point.
In order to simplify the configuration, the X-array histogram is obtained by reading and adding from the three-dimensional array histogram unit 15c ′ for each X array, reading and adding for each Y array, or reading and adding for each Z array. The values of the section 15a, the Y array histogram section 15b, and the Z array histogram section 15d may be used.

It is a flowchart which shows the detection method of the arrhythmia based on one Embodiment of this invention. It is a figure which shows an example of the amplitude value of the electrocardiogram signal for calculation of occurrence probability in the detection method of arrhythmia of FIG. 1, and (b) the histogram. (A) one example of the amplitude value of one electrocardiogram signal for calculating the co-occurrence probability in the arrhythmia detection method of FIG. 1, (b) its histogram, (c) one example of the amplitude value of another electrocardiogram signal, (D) is a diagram showing the histogram and (e) the number of co-occurrence elements. It is a figure which shows the time window for calculation of the mutual information amount in the detection method of the arrhythmia of FIG. It is a figure which shows two electrocardiogram signals at the time of ventricular tachycardia (VT), its mutual information amount, and a correlation coefficient, (a) Right ventricular electrocardiogram, (b) Left ventricular electrocardiogram, (c) MI: Mutual information Quantity (solid line), PCC: correlation coefficient (dashed line). It is a figure which shows two electrocardiogram signals at the time of ventricular fibrillation (VF), its mutual information amount, and a correlation coefficient, (a) Right ventricular electrocardiogram, (b) Left ventricular electrocardiogram, (c) MI: Mutual information Quantity (solid line), PCC: correlation coefficient (dashed line). It is a figure which shows two electrocardiogram signals at the time of atrial tachycardia (AT), its mutual information amount, and a correlation coefficient, (a) Right ventricular electrocardiogram, (b) Left ventricular electrocardiogram, (c) MI: Mutual information Quantity (solid line), PCC: correlation coefficient (dashed line). It is a figure which shows two electrocardiogram signals at the time of normal (SR), its mutual information amount, and a correlation coefficient, (a) Right ventricular electrocardiogram, (b) Left ventricular electrocardiogram, (c) MI: Mutual information amount ( Solid line), PCC: correlation coefficient (broken line). It is a box-and-whisker diagram showing variation in (a) mutual information and (b) correlation coefficient. It is a figure which shows typically (a) distribution of a diagnostic parameter | index for accuracy evaluation, such as a screening test, and (b) ROC curve. It is a figure which shows the ROC curve of a mutual information amount and a correlation coefficient based on actual data. It is a figure which shows the average value of the simultaneous occurrence frequency of each electrocardiogram signal of FIGS. FIG. 9 is a modification of the arrhythmia detection method of FIG. 1 and is a diagram illustrating an m × n partition table used for calculating Pearson's χ ² statistics. It is a flowchart which shows the arrhythmia classification method. FIG. 14 is a box-and-whisker diagram illustrating variations in Pearson's χ ² statistics calculated using the m × n contingency table of FIG. 13. It is a figure which shows ^two electrocardiogram signals at the time of normal (SR), its mutual information amount, and χ ² statistics, (a) right ventricular electrocardiogram, (b) left ventricular electrocardiogram, (c) MI: mutual information amount (Thin line), χ ² statistics (thick line). It is a figure which shows ^two electrocardiogram signals at the time of atrial tachycardia (AT), its mutual information amount, and χ ² statistics, (a) right ventricular electrocardiogram, (b) left ventricular electrocardiogram, (c) MI: mutual Information amount (thin line), χ ² statistic (thick line). It is a box-and-whisker diagram showing χ ² statistics between the right atrial electrocardiogram and the right ventricular electrocardiogram. 1 is a block diagram illustrating a defibrillator according to an embodiment of the present invention. It is a block diagram which shows the information processing part of the defibrillator of FIG. It is a block diagram which shows the modification of the information processing part of FIG. It is a block diagram which shows the other modification of the information processing part of FIG. It is a block diagram which shows the other modification of the information processing part of FIG.

Explanation of symbols

A heart 1 defibrillator 2 arrhythmia signal detector 5a, 5b, 5c electrode 7 information processing unit (mutual information calculation unit)
8 Analysis and diagnosis unit (determination unit)

Claims (3)

A plurality of electrodes placed in the heart;
A mutual information amount calculation unit for calculating a mutual information amount based on a plurality of electrocardiographic signals detected by the electrodes;
An arrhythmia signal detection device comprising: a determination unit that determines whether or not the mutual information amount calculated by the mutual information amount calculation unit is smaller than a predetermined threshold value. A plurality of electrodes placed in the heart;
A χ ² statistic calculating unit that calculates χ ² statistic based on a plurality of electrocardiogram signals detected by the electrodes;
Chi ² statistic calculated by the chi ² statistic calculation unit arrhythmia signal detecting device and a determining unit configured to determine whether or not smaller than a predetermined threshold value. Comprising the arrhythmia signal detection device according to claim 1 or 2 ,
A defibrillation apparatus comprising a stimulation pulse generator that generates an electrical stimulation pulse to be applied via an electrode disposed in the heart when an arrhythmia signal is detected by the arrhythmia signal detection apparatus.

Images