JPS6354381B2 - - Google Patents

Description

[Detailed description of the invention]

The present invention relates to an electrocardiogram graphic processing device, and more specifically, it applies the concept of Lorentz plot (hereinafter abbreviated as LP) to electrocardiogram graphic processing to enable simple graphical display of electrocardiogram rhythm. The present invention also relates to an electrocardiogram processing device that makes it possible to perform quantitative and detailed analysis using the display. Conventionally, automatic electrocardiogram analysis has been performed using one of the following methods. (1) Using a standard 12-lead electrocardiogram that lasts for several minutes, a computer performs waveform diagnosis and rhythm diagnosis using the time intervals of P waves, QRS waves, and T waves, and the diagnostic results are coded and output on recording paper. . (2) In coronary artery disease treatment departments, etc., heart rate and premature ventricular contractions such as ventricular tachycardia are detected, and if either exceeds a set value, an alarm is generated. (3) Record a long-term continuous electrocardiogram on magnetic tape, play it back at high speed with a scanner, and observe the electrocardiogram waveform with a cathode ray tube oscilloscope. At the same time, the heart rate and the number of supraventricular and ventricular extrasystole are displayed, and one of the following operations (a) to (d) is performed for each scan. After performing these steps, a doctor with specialized knowledge combines some of the results (a) to (d) with the waveform observation results to diagnose arrhythmia and waveforms. (a) If the RR interval or QRS wave width is shorter or wider than the set value, it is considered to be an extrasystole, and when these are detected, a part of the ECG waveform is automatically recorded on recording paper in real time. Output. (b) Measure the average heart rate trend and ST level trend and output it to recording paper. (c) Reduce the entire electrocardiogram waveform and output it to recording paper. (d) R on T (T
A phenomenon in which QRS waves are superimposed on waves, which is one of the dangerous arrhythmias), is detected, and the waveform and generation circuit are displayed or output on recording paper. (4) Taking advantage of the fact that the pattern of the serial correlation diagram of the RR interval differs depending on the type of arrhythmia, the serial correlation diagram was divided into nine regions, and the combination pattern of presence/absence of points in each region was calculated. Recognize it with a computer,
Broadly categorize arrhythmias. However, these methods have the following drawbacks. First, method (1) allows detailed diagnosis with a high degree of accuracy, but it takes time to process and cannot even be processed in real time, and the measurement time is short, so it is difficult to detect steady abnormalities. Its application is not effective, and its use is mainly limited to screening for early detection of heart disease. In (2), the purpose is to generate an alarm, so it is not used for diagnosing waveforms or arrhythmia. The feature of (3) is that it is also possible to analyze long-term continuous electrocardiograms of daily life, but since electrocardiograms of daily life have significant changes in rhythm and waveforms and are also noisy, the results of automatic analysis are unreliable. Therefore, in practice, this is done using a method that requires skill and time, such as an operator finding an abnormality while observing the waveform, and outputting the detected part of the waveform to recording paper in real time. Detailed classification of arrhythmias, their frequency of occurrence, R on
The existence of T and its frequency cannot be determined unless an expert examines the result of (c) carefully. By the way, the specialist
If you look at (c) and the output results of the abnormal area, you can quickly and accurately diagnose the abnormality, but when you look at the whole, it is impossible to objectively show the abnormal area, normal heart rate, and the relationship between the abnormal areas. Have difficulty. Therefore, in the end, it is just a list of abnormal parts,
It is not possible to know how things change over a long period of time, which will become clear from the detailed description of the invention. (4) is an attempt to automatically classify arrhythmias by using differences in basic patterns, and the types of arrhythmias included can be categorized in many cases if the data is processed for a short period of about 100 seconds. Since it is single, it can be classified. However, when long-term electrocardiograms are often complicated due to multiple measurements, serial correlation diagrams often contain arrhythmias of unknown type, and are limited to merely determining the presence or absence of arrhythmias. In addition, since waveform diagnosis is not performed only for rhythm diagnosis, when using it for short-term electrocardiogram processing,
Inferior to (1). Furthermore, if the electrocardiogram is processed for a long time,
Since the type of arrhythmia and waveform abnormalities can be determined more quickly if the results of method (3) (c) are examined by an expert, this method cannot be said to be a very good method from a practical standpoint. Further, specific examples of the differences between (4) and the present invention are summarized as follows. 1. In a serial correlation diagram, there is no idea of sequentially following each point, so if the diagram is complicated, detailed classification of arrhythmia is difficult. 2 For the same reason as 1 (that is, there is no dynamic property), the multi-Lorentz plot (hereinafter referred to as
Abbreviated as MLP. ) cannot be performed, making detailed analysis impossible. 3. For the same reason as in 1, it is not possible to obtain the heartbeat state transition matrix (which represents the frequency of transitions between heartbeat states in a matrix format, hereinafter abbreviated as TM), which will be described later, and it cannot be quantified. 4 I don't understand R on T. 5 The RR interval should be processed using absolute values, but since the serial correlation diagram uses values normalized by the average value, it is difficult to understand the pattern for each measurement and detailed changes in connection time. First, there is no information available regarding the effects of drug administration or the refractory period of the atrioventricular node. An object of the present invention is to provide an electrocardiogram processing method that eliminates the above-mentioned drawbacks of conventional methods. The present invention will be explained in detail below. Since the central concept of the present invention is LP, we will first explain LP. Here, LP refers to creating time series data of coordinate points by sequentially pairing the preceding and succeeding values from time series data such as feature values in a signal, such as peak values and peak intervals, and converting this two-dimensional time series data into time information. A method of graphically representing the dynamical laws of a system's behavior by plotting them implicitly on a two-dimensional plane, or a method obtained by this method (Journal of
the Atmospheric sciences Vol.20 March,
1963 P130-141). When creating an LP, local maximum (minimum) values of signals, their time intervals, time intervals of level crossings, etc. are used as time series data, but in order to extract regularity, such features are used. It is necessary to choose the quantities wisely. Regarding electrocardiograms handled in the present invention, such quantities include electrocardiogram rhythms such as RR, PP, PQ, and QT intervals, among which the RR interval rhythm will be used as an example for explanation. Now, the time series of RR intervals X ₀ , X ₁ , X ₂ , ..., X _o ,
Assume that X _o+1 ,... are measured. Based on this, the value X _o at time n and the value X _{o+1 at time n+1} (n=
Coordinate points (X _o , X _o+1 ) with pairs of 0, 1, 2, ...)
Time series of (X ₀ , X ₁ ), (X ₁ , X ₂ ), ..., (X _o ,
X _o+1 ), make... The time series of this coordinate point is expressed as a two-dimensional (X _o , X _o+1 ) with X _o as the horizontal axis and X _o+1 as the vertical axis.
LP by sequentially plotting on a plane
is obtained. Next, in order to understand the relationship between LP and premature contractions, an explanation will be given using a model of a single compensatory premature contraction during constant sinus rhythm. Sinus rhythm, premature systole (triangle mark), and immediately following RR interval are respectively X _R ,
If expressed as X _S and _XL ,...X _R , X _R , X _S , X _L , X _R , X _R ,
Time series data (Figure 1a) was obtained, and based on this, (X _R , X _R ), (X _R , X _S ), (X _S , X _L ),
Time series data of coordinate points such as ( _XL , _XR ), ( _XR , _XR ), etc. can be obtained. If this is plotted on a two-dimensional plane, an LP as shown in Fig. 1b can be obtained. The coordinate point (black circle) moves from point R (X _R , X _R ) to point S (X _R , X _S ) due to the occurrence of extrasystole, and moves to point L (X _S , X _L ), then move from point R' (X _L , X _R ) to point R, where sinus rhythm continues again. When considering the movement between two successive coordinate points in LP, the X _{o +1} axis component of the previous coordinate point becomes the X _o axis component of the next coordinate point.
By the way, draw a line parallel to the X _o axis from the previous coordinate point.
Since the intersection with the 45° line is a point where both coordinate axes components are equal, a point on a line parallel to the X _o+1 axis from this point becomes the next coordinate point. This situation is true for all successive coordinate points and can be easily seen in FIG. 1b by following the arrows in numerical order from 1 to 6. Such dynamic properties are a major feature of LP. Also, since point R is a point on the 45° line that satisfies X _o+1 =X _o , it can be seen that the same interval is continuous. In LP, the coordinate points are also considered as a time series, so from the relationship between the points before and after, it can be determined whether each point precedes or follows sinus rhythm, premature contractions, or premature contractions. Figures 2a to 5a show intermittent extrasystole, block,
The two-stage pulse shows a model of the rhythm of atrial fibrillation, and the second
Figures b to 5b show their LPs. In Fig. 2b, the coordinate point (black circle) moves from point R to point S with the extrasystole X _S , and then the interval becomes X _R ″ ( _{X R} − and return to continuous sinus rhythm again. The characteristic _of this _case is that X _R
Along with this, it moves from point R to point L and returns to point →R'→R. The characteristic in this case is about 2X _R directly above the R point.
The L point is located at the position of . In Figure 4b, the coordinate point moves from R to S point with the first extrasystole,
After that, if it is compensated, it moves to point L, but if premature contraction occurs again, point L moves to point S' instead of R'. If we extract only _{the second} stage pulse, the time series becomes X _{S ,} This is a characteristic of veins. Figure 5b shows a wide random distribution, which is characteristic of atrial fibrillation. Since the example given here is only a single arrhythmia, arrhythmia classification can be performed based on static pattern differences. By the way, when a plurality of arrhythmias coexist and the LP becomes complex, it becomes difficult to classify the arrhythmia from the superposition of typical static patterns. However, I was able to follow the movement of each point even if it became complicated.
Classification can be easily performed based on the LP concept. Furthermore, as will be seen below, it is very important to utilize the dynamic properties of LP for detailed analysis. In actual long-term electrocardiograms, several of a wide variety of arrhythmias are mixed together and are often complicated, but basically, the method explained below is used to
A fairly detailed analysis can be performed. Here, we will select and explain typical examples. First, we will define the symbols used below. For convenience, the RR interval due to premature contraction and sinus rhythm before and after it is classified as follows. Ri = previous interval S = connection time L = immediately subsequent interval Re = interval two times after the extrasystole If these are shown in chronological order, X _o = R _i , X _o+1 = S,
X _o+2 = L, X _o+3 = R _e . FIG. 6 is an example of fully compensated premature ventricular contraction. The portion distributed along the 45° line is the RR interval due to normal sinus rhythm. Comparison with FIG. 1b makes it easy to see that this example is a compensatory premature contraction superimposed on changes in sinus rhythm. By the way, as the tendency for sinus irregularity becomes stronger, the disorder in the RR interval becomes larger, so the distribution in this area becomes considerably wider. By the way, the width parallel to the vertical axis in this part |
X _o+1 −X _o | is the difference between the front and rear intervals, and the difference between the maximum and minimum values of Xn or X _o+1 represents the fluctuation range of sinus rhythm, so if the movement status during measurement is known, , LP can also be used as one of the indicators indicating the state of cardiac function. Cardiac mechanism measurement is generally performed by a stress test using a treadmill. This is a certain period of time (3
This is done by increasing the angle of the treadmill every minute (minutes) to increase the load, and then removing the load when the heart rate appears to be at its maximum. Cardiac function is evaluated by examining how the heart rate changes during this process, such as how the heart rate increases at each stage and how the heart rate returns to normal when the load is removed. The advantage of using LP in this case is that by performing real-time processing, changes in heart rate can be seen on the spot, and if necessary, detailed analysis of all heartbeats can be performed quickly and accurately after completion. Things like that. By the way, in LP that uses each heartbeat continuously, the changes between each heartbeat are small, so it is difficult to see the changes described above. Therefore, by drawing the LP at intervals of about 10 heartbeats, it is possible to easily determine changes in the average heart rate, and quantitative evaluation can also be performed by curve fitting. By the way, it is difficult to tell whether the example in Figure 6 is fully compensated or incompletely compensated just by looking at the LP, but this distinction can be made by using the dynamic properties of the LP as follows. It can be done easily. In the case of complete compensation, X _o+1 +X _o+2 2Xn …(1). For fixed connection, X _o+1 C (C is a constant)
(Figure 6 can be considered to satisfy equation (1).) If there is no sinus irregularity, ReRi(X _o+3 X _o )
holds true, so equation (1) becomes as follows. X _o+3 1/2X _o+2 +C/2 (2) This means that the slope of the Re portion is 1/2. On the other hand, in the case of incomplete compensation, Xn<
Since X _o+1 +X _o+2 <2Xn, the slope takes a value between 1 and 1/2. Figure 6 satisfies equation (2) and is determined to be fully compensated. In this way, regularity can be extracted from LP. Examination of such inclinations and movement of points can be done by using a cursor made of transparent plastic, etc., as shown in Figure 7.
It can be done quickly. Furthermore, the sinus arrhythmia line in the figure provides a guideline for the aforementioned sinus arrhythmia, and if points due to sinus rhythm are distributed outside this line, there is a tendency for sinus arrhythmia. By the way, for more detailed analysis, it is necessary to know the relationship between Ri and L. As can be seen from Figure 6, in the case of fixed connectivity, that is, when S is constant, it is not possible to know which point in L the point in S will go to next using only LP, so Ri and L It is not possible to know the relationship between the two, making detailed analysis impossible. However, by using MLP, which is an extension of LP, this becomes possible as shown below. MLP is a method for clarifying the regularity of ununique parts in LP, and corresponds to a special filtering of points on a two-dimensional plane. For example, time series data..., X _o-1 , Xn, X _o+1 , X _o+2 , X _o+3 ,...
If X _o+1 is an extrasystole in
Draw a new time series LP excluding X _o+1.
MLP is required. Therefore, the time series data of the coordinate points are... (X _o-1 , Xn), (Xn, X _o+2 ),
(X _o+2 , X _o+3 ),..., and (X n, X _o+1 ) and (X _o+1 , X _o+2 ) in LP are replaced by (X n, X _o+2 ). It becomes something. This situation is shown in Figure 6 and its
The broken line shown in Fig. 8 which is MLP and X _o ,
It can be easily understood by tracing X _o+1 and X _o+2 . Furthermore, even when intercalary and compensatory contractions coexist, there is a clear difference in the RR interval after extrasystole between the two, so MLP can be used to determine at what heart rate each beat occurs. You can know. By the way, as mentioned above, the law of complete or incomplete compensation in the case of ReRi is
When Re and Ri are different, it is necessary to use MLP to distinguish between the two. In the case of fixed connectivity, equation ₍ 1) becomes ).
MLP can be used to extract regularities that cannot be captured by LP, and is very effective. So far, we have talked about extracting regularity, but in order to have objectivity, it is necessary to quantify it. If the type of arrhythmia is a single type and the connection time and sinus rhythm parts are clearly separated as shown in Figure 6,
The frequency of occurrence can be easily determined from the LP and histogram. However, if there are multiple arrhythmias, for example, a combination of single and triple heart rhythms, LP
It is not possible to determine the frequency of occurrence of each event from MLP or MLP. The frequency of arrhythmia occurrence is an important quantity, as it provides very important clues in searching for dangerous arrhythmias in treatment and an index of drug efficacy. Such quantification is performed by dividing time-series data into several states (m) according to their respective values, and determining the n-th TM for these states. The degree here indicates that up to n previous states are taken into consideration when transitioning to a certain state. Therefore, there are m ⁿ states as states before the transition.
For electrocardiogram rhythm, as described below, if n = 2 and conditions are classified into three (m = 3) based on LP, the type of arrhythmia and its frequency can be determined by TM.
This can be determined from the matrix elements and the arrhythmia index. In other words, the time series in which the RR interval is classified into three states, normal (R), short (S), and long (L) based on LP, is regarded as a second-order Markov chain (third-order, fourth-order ,...but the second order is easier and can be analyzed sufficiently.) Obtain the TM, and calculate the index of each arrhythmia and its part.
Extract the elements of TM. Next, the classification method for R, S, and L will be explained. R = normal sinus rhythm (Xn), S = extrasystole (X _o+1 )
and the interval following S (X _o+2 ) is set as Ra = (X _o+1 +
Classify as follows according to X _o+2 )/Xn. Ra＜1.1 and X _o+2 ＞Maximum value of S area (Smax)
Then, X _o+2 = R, if Ra>1.1, then X _o+2 = L, and if X _o+2 Smax, then X _o+2 = S. Further, if Xn is the maximum value (Rmax) in the region where Xn>R, then Xn=L. Classification of RR intervals according to this logic can be easily performed by using a computer, as it only requires calculation of inequalities.
Table 1 shows the relationship between each index and arrhythmia. Each frequency is the frequency of the TM element indicated by each index,
For example, in a 3-stage pulse, LR → S, and in a 2-stage pulse, SL → S
Or, the frequency is SR→S, or RR→S in the case of a single episode.
Table 2 is the result of application to Figure 6. Examining the above factors, we find that 3-stage pulse = 72, 2-stage pulse = 0, single-stage pulse =
It turns out that it is 275 times. The feature of TM is that in addition to quantification, states are classified based on LP, so Smax
and Rmax values (in Figure 6, Smax = 550 msec,
Rmax = 1100 msec) and how many intervals are included in each category are clear, making it possible to obtain a reliable frequency. Furthermore, by increasing the order of n, it is possible to know the correlation over a longer period of time. So far, we have described the detailed classification of arrhythmias and measurement of their frequency using LP, MLP, and TM, but the use of LP is not limited to this, and applications that are directly linked to prevention and treatment are possible. These will be explained below.

【table】

[Table] R on T is said to be one of the important phenomena with a high possibility of transition to ventricular fibrillation. Since it is difficult to automatically measure the QT interval, the presence or absence of R on T can be investigated without directly measuring the QT interval by using QT interval = QTc × √ interval (Bazett's formula). If each person's QTc has been measured in advance, R
The frequency of on T can be calculated. Specifically, each QTc
The calculation shown in FIG. 9 can be performed for . In order to know the dependence of R on T on the RR interval, a curve for QTc arbitrarily selected during LP (QTc curve)
is displayed together with the frequency of R on T. In the treatment of arrhythmia, it is very important to know the effects of drugs because their therapeutic range is narrow. The effect of drug administration is manifested as a change in coupling time, as shown in the examples in Figures 10a and 10b. Both figures show the LPs of the same patient who was diagnosed as having had sufficient effects from administration of a certain type of antiarrhythmic drug, with and without medication. The average value of the connection time is about 420 ms when no medication is administered as shown in Figure 10a, but it is clearly longer to about 500 ms when medication is administered as shown in Figure 10b. This shows that the effect of the drug and the extension of the binding time are closely related, and real-time processing makes it possible to quickly determine the drug's efficacy from changes in the binding time after administration. It is expected that the patient will be able to adjust dosage and select drugs accurately. In atrial fibrillation, the RR interval is said to be completely random. Reflecting this, Section 11a
In the example shown in the figure, the distribution is random and almost uniform. However, in this example, RR
The lower limit of the interval depends on the preceding RR interval in such a way that it rides a certain smooth curve like the broken line in the figure. This curve probably reflects the refractory period of the atrioventricular node. If this is confirmed, it is expected that real-time stress testing will enable us to easily and quickly determine the refractory period of the atrioventricular node and investigate the effects of drugs on the refractory period in a short period of time. and will be very useful. FIG. 11b shows an example of atrial fibrillation mixed with premature ventricular contractions with fixed connectivity. Atrial fibrillation often causes aberrant intraventricular conduction, but it is often difficult to differentiate between aberrant intraventricular conduction and premature ventricular contractions. In atrial fibrillation, it is very important to distinguish between the two in terms of treatment. 11b
As shown in the figure, if the connection time of ventricular premature contraction is shorter than the refractory period of the atrioventricular node, the two can be easily distinguished. As mentioned above, the new analysis method based on LP has many advantages, so many applications can be considered depending on the purpose, and more effective applications can be achieved by systematizing them. . Examples of the apparatus of the present invention will be described below. Since the heart rate is at most a few beats/second, the 12th
Real-time processing is possible with the configuration shown in the figure. In FIG. 12, the electrocardiogram waveform from the electrocardiograph is sent to a CRT display device or an analog data recorder for displaying it as necessary, and is also sent to a preprocessing device. This preprocessing device is for converting characteristic parts of the electrocardiogram waveform into signals, and for example,
It has a configuration as shown in the figure. 13th
In the figure, the differentiator is R
A low-pass filter is used to facilitate the detection of waves, and a low-pass filter is used to remove high-frequency noise.These filters can remove effects such as baseline fluctuations and hum. The AD converter has an accuracy of 8 bits or more and a conversion time of 1 m.
It is sufficient if it is less than sec. Also, A.D.
An amplifier is installed in front of the converter to ensure sufficient accuracy even with small signals. The R wave is detected by the interval measuring device shown in FIG. 12 from the differential electrocardiogram waveform AD-converted in the preprocessing device, and the RR interval is determined from the differential electrocardiogram waveform. Conventionally, automatic measurement of the RR interval has been carried out by passing an absolute value circuit after a differentiator, and determining the intervals such as the time when a set threshold value is exceeded, the time when the value crosses zero, etc. using a circuit or a computer. However, in daily clinical practice, even the interval between the rises of the QRS complex has an RR.
Spacing is measured manually. In the present invention, in order to match this, the following equation is used to measure the rising point of the QRS wave and the RR interval using this using a computer. Assume that the voltage of the differential electrocardiogram wave is V(t), the time of the starting point of the QRS wave is t _R , and the first sampling time after the time when the threshold value is exceeded is t _T (see FIG. 14). Since the rise of the differential QRS wave is generally linear, t _R
is given by the following equation. t _R = t _T −V(t _T )/｜V(t _T )−V(t _T −Δt)｜・Δt
...(4) Here, Δt is the sampling interval of AD conversion. When processing the differential electrocardiogram wave and taking its absolute value, if the T wave becomes sharp, there are cases where even the T wave is triggered. However, in general, in the differential waveform, the amplitude of either the + side or the - side of the T wave is often smaller than the other side. Therefore, we set two trigger levels, +V ₁ and -V ₂ , and detected only waves that exceed both levels within a refractory period of approximately 180 msec.
Recognizes it as a QRS wave and prevents T wave triggering.
In addition, since the trigger level and refractory period change greatly with changes in the electrocardiogram, highly accurate measurements can be made by monitoring the differential waveform with a CRT to find the optimal value in advance. By using a body surface electrode for the RR interval and an esophageal guidance electrode for the PP interval, the RR and PP intervals can then be measured in exactly the same manner. That is, in the case of esophagus guidance, equation (4) detects the rising point of the P wave. Also
The PQ interval can be measured by using both body surface and esophageal guidance, performing two-channel processing, and subtracting the time of the R wave start point from the P wave start point. R on T has meaning only in the RR interval, so it is omitted in other cases. FIG. 15 shows an aspect of processing in the above-mentioned interval measuring device. The RR intervals measured in this way are displayed one after another in real time on a display device, are stored in the memory of the computer, and become stored data by being written to an external memory. To save RR interval data,
It is also beneficial from the point of view of data compression, since it requires less storage capacity. Furthermore, if the electrocardiogram waveform is recorded on an analog recorder at the same time, waveform diagnosis can also be performed, which is very beneficial. The interval data measured by the interval measuring device is sent to the next-stage LP processing device, and coordinate point data (Xn, X _o+1 ) is created. FIG. 16 shows the mode of processing in the LP processing device. Further, FIGS. 17 and 18 show aspects of processing in the MLP processing device and the heartbeat state transition analysis processing device (hereinafter abbreviated as TM processing device). Using a microcomputer to perform detailed analysis of large amounts of data takes time and is inefficient, but if it is used for displaying LPs when quick response is required, the versatility of the microcomputer system can be utilized. It is optimal. Therefore, microcomputer systems can be used for early detection and prevention through relatively simple tests in general or large hospitals, bedside monitors, and stress test monitors for measuring medication effects and cardiac function. Such monitoring applications do not require expert knowledge or special techniques for electrocardiogram analysis, unlike conventional methods, and are therefore very practical. In the case of systemization using a mini- or large-sized computer, it is suitable for use mainly as a dedicated machine for detailed analysis because of its high speed and large processing capacity. FIG. 19 shows its configuration. For detailed analysis, we make full use of image processing technology to display LP and MLP in a visually easier to understand format using screen splitting and color display, and perform TM calculations to improve quality. perform high-quality analysis. In addition, by combining online with a scanner and external memory, displaying the electrocardiogram waveform of a designated area, LP, MLP, etc. on a cathode ray tube oscilloscope at the same time, and performing waveform diagnosis at the same time, it becomes possible to perform a quick and almost complete diagnosis. . According to the present invention described in detail above, the following effects can be expected. A large amount of time-series data can be expressed as a two-dimensional pattern on a single drawing. Data processing time can be as long or as short as you like. Arrhythmia classification is possible. More detailed analysis can be performed using MLP and TM. TM allows for quantification and objectivity. Since the logic of LP is simple, it can be processed in real time even on a microcomputer, making it easy to put into practical use. Depending on the purpose of use, a micro, mini, or large computer can be selected for processing, allowing for flexible systemization. Combining this with conventional methods, especially waveform diagnosis, enables almost complete diagnosis.

[Brief explanation of the drawing]

Figures 1a to 5a and 1b to 5b are explanatory diagrams of time series data models and their LPs,
Figure 6 is an electrocardiogram and LP diagram for an example of fully compensated ventricular premature contraction; Figure 7 is a front view of a cursor suitable for use in examining LP; Figure 8
The figure is an MLP diagram for the example shown in Figure 6, Figure 9 is a flowchart showing the method of calculating the frequency of R on T, Figures 10a and 10b are LP diagrams of the same patient when no medication is administered and when medication is administered. Figures 11a and 11b
The figure is an LP diagram of an example in which atrial fibrillation and premature ventricular contraction are mixed therein, FIG. 12 is a block configuration diagram showing an embodiment of the graphic processing device of the present invention, and FIG.
A block configuration diagram showing an example of the configuration of the preprocessing device in the device shown in FIG. 2, FIG. 14 is an explanatory diagram of symbols related to equation (4), and FIGS. 15 and 16 show the interval measuring device and LP processing in FIG. 12. Explanatory diagrams showing the mode of processing in the apparatus, FIGS. 17 and 18 are similar to those in FIG. 12.
FIG. 19 is an explanatory diagram showing the mode of processing in the MLP processing device and the TM processing device, and is a block configuration diagram showing another example of the configuration of the graphic processing device of the present invention.

Claims (1)

[Claims] 1. An interval measuring device that sequentially measures time intervals between specific waveform points of the same type or different types in an electrocardiogram waveform;
a Lorentz plot processing device that creates time series data of coordinate points by sequentially pairing preceding and following values from the time series data output from the interval measuring device;
An electrocardiogram processing device comprising: a device for displaying the coordinate points on a dimensional plane. 2. In the device according to claim 1,
A multi-Lorentz plot processing device that extracts unambiguous laws by removing unambiguous data from the original data and performing Lorentz plot processing on newly created time series data. An electrocardiogram processing device characterized by comprising: 3. In the device according to claim 1,
An electrocardiogram processing device comprising a heartbeat state transition analysis processing device that divides time-series data of coordinate points in a Lorentz plot processing device into several states according to each value, and calculates transition matrices for the states. .