JPH11318842A - Method and device for judging t wave marker point during analyzing qt dispersion - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a method and device for analyzing QT dispersion in an ECG guiding signal. SOLUTION: The method and device for analyzing QT dispersion in the ECG guiding signal obtains plural ECG guiding signals, filter-processes the signal (20), removes noise without distorting the form of a T wave, judges a T wave marker critical point to calculate QT dispersion and its pulse wave corrected value from a decided (22), (24) and (26) T wave marker. In addition, the device includes a processor programed to execute these functions.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION The QT interval in a surface electrocardiogram is the time interval from the beginning of a QRS to the end of a T-wave. QT dispersion is a measure of the difference during the QT interval measured at different ECG leads. One measure of the QT dispersion is the difference between the longest and shortest QT intervals, and another measure is the standard deviation of the QT intervals measured from all ECG leads.

[0002]

BACKGROUND OF THE INVENTION Recent studies have shown that increased QT dispersion is a marker for malignant arrhythmias.
Clinically, QT dispersions are used to monitor the effects of arrhythmic drugs and ventricular repolarization time. In most cases, the QT dispersion is measured manually,
There, the end of the T wave is determined by measuring the point at which the T wave returns to the equipotential line. However, some studies have found differences in QT dispersion between mutual and internal observers. It is more difficult to obtain consistent measurements from ECGs with noise or irregular T-wave shapes. For this reason, computerized and automated methods for measuring T-wave endpoints and T-wave peaks have been developed.

The markers proposed in these studies for objectively and consistently measuring the T-wave end point include (a) the point at which the T-wave crosses the equipotential line plus the threshold, and (b) the T-wave. (C) the intersection of the final slope of the T-wave with a threshold that is part of the amplitude of the final peak of the T-wave derivative; (D) There is a point where the area of the T wave reaches a certain percentage (90%) of the entire area of the T wave. The method (a) has the same problem as the manual method, and the methods (b) and (c) have difficulty in finding a point where the maximum slope is stable when determining the threshold. (D)
Has not been shown to be useful in differentiating low-risk patients from high-risk patients.

[0004] There is a need to provide a T-wave marker so that the QT dispersion measurement has good reproducibility and can well differentiate between high and low risk groups. It is well known that T-waves are not as clearly defined morphologically as QRS in both time and frequency domains. Therefore, any measurement method that depends on one point is susceptible to noise and changes in the shape of the T wave.

[0005]

An object of the present invention is to provide an ECG.
It is to provide a new and improved method and apparatus for determining a QT dispersion in a waveform. It is a further object of the present invention to provide a method and apparatus for determining the peak and endpoint of a T wave portion in an ECG waveform.

Another object of the present invention is to provide a method and apparatus for determining highly reproducible T-wave markers in QT dispersion measurements. Yet another object of the present invention is to provide a method and apparatus for providing a T-wave marker in which QT dispersion measurements can be highly differentiated between high risk and low risk groups.

It is another object of the present invention to provide a method and apparatus that can perform QT dispersion measurement independent of a single reference point. It is a further object of the present invention to provide a method and apparatus for measuring a QT dispersion whose accuracy is not affected by noise or changes in the shape of the T-wave.

[0008]

The above and other objects and advantages of the present invention will become more apparent from the detailed description taken in conjunction with the accompanying drawings. In general, the method comprises configuring a method for analyzing a QT dispersion in an ECG-derived signal, obtaining a plurality of ECG-derived signals, filtering the signals and removing noise without distorting the T-wave form. , Determining the critical marker point of the T wave, and calculating the QT dispersion and its heartbeat correction value from the determined marker.

In a more specific interpretation of the invention, the method of filtering is such that each of the ECGs described above is arranged in a column-by-column matrix.
Each ECG signal is reconstructed by accumulating the induced signals and decomposing the matrix by the dimension recurrence method. Further interpreted, the present invention relates to at least some reconstructed ECGs.
Generating a detection function from the absolute difference between successive samples of the signal and determining a T-wave peak from a valley point of the detection function. According to one embodiment, the valley points of the detection function are determined by taking the center of the region.

According to yet another embodiment of the present invention, the step of determining the T-wave marker critical point comprises the step of matching a least square value around the maximum slope of the reconstructed ECG signal after the T-wave peak. And determining the end point of T from the intersection of the non-linear correction function based on the threshold plus the amplitude of the T wave. According to another embodiment, determining the T-wave marker critical point comprises generating a template of at least some of the ECG-derived signals and matching the measured ECG to the template.

The present invention also includes an apparatus for analyzing a QT dispersion in an ECG-derived signal, means for obtaining a plurality of ECG-derived signals, filtering the signals, and distorting the T-wave form. And a means for determining a T-wave marker critical point and a means for calculating a QT dispersion and its heartbeat correction value from the determined marker.

According to a more detailed interpretation of the invention, the filtering process reconstructs each ECG signal by means of accumulating each signal of the ECG leads in a matrix for each column and decomposing the matrix by a dimensional recurrence method. Means. According to another interpretation, the present invention provides a means for generating a detection function from the absolute difference between successive samples of at least some reconstructed ECG signals, and a method for generating a T-wave peak from a valley point in the detection function. It includes means for determining.

The present invention also contemplates that the means for determining the T-wave peak includes means for centering the valley point region. With further interpretation of the invention, the means for determining the T-wave marker critical point is based on the intersection of the maximum slope of the reconstructed ECG signal after the T-wave peak and the threshold plus the non-linear correction function based on the amplitude of the T-wave. Includes a method to determine the end point of the wave.

In another embodiment, the means for determining a T-wave marker critical point includes generating a template of at least some of the ECG-derived signals, and matching the measured ECG signal to the template.

[0015]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Preferred embodiments of the present invention will be described in detail. FIG. 1 shows an electrocardiographic monitor according to the present invention and
1 schematically illustrates a 12-lead analysis system. The system includes a conventional data acquisition module 12 connected by lead wires to 10 to 14 electrodes placed on the patient's body. In the illustrated embodiment, the electrodes are 12
Set to provide inductive ECG signals. In particular, the electrodes are the right arm electrode RA, the left arm electrode LA, the chest electrodes V1, V2, V3, V
4, V5, V6, right foot electrode RL and left foot electrode LL. For a data acquisition module connected to 14 electrodes, the location of the additional thoracic lead may be combined with V3R, V4R, V8 or any location on the right ventricle and posterior chest. Data acquisition module 12 includes conventional in-phase rejection and filters to remove respiratory and muscle artifacts of the patient. The acquisition module converts the analog lead signals to digital signals and obtains the normal ECG leads I, II, V1, V2, V3, V4, V5, V6, V3R, and the normal ECG leads obtained directly on the patient leads.
Generates V4R, V8 and III, aVR, aVF, aVL derivations derived by Eindhoven's law. The combination of electrode signals used to determine or derive the ECG induction signal is well known in the art and will not be discussed here for the sake of brevity. The digitized ECG signal is provided to a processor 14 for 12 or 15 lead and QT dispersion and repolarization analysis.

Processor 14 is programmed to perform the method shown in FIG. 2 and forms part of QT dispersion analyzer 15. The analyzer is a personal computer running either Microsoft Windows 95 or Microsoft Windows NT operating system. The main purpose of the analyzer is to analyze a number of ECGs and thereby review these ECGs in the first window 16 of the analyzer 15. The system also automatically detects T-wave offset points and T-wave peaks, calculates the QT dispersion, the QT peak dispersion, and its heart rate correction, and displays them in the second window 18. In addition, QT parameters are automatically detected for groups of ECG files in batch mode and these results are stored in a Microsoft Access database. Also, the 12-lead ECG
The QT map is also displayed in the window 19. In addition, there is a review function that allows the user to review the processed ECG data in the window 17 and correct the T-wave offset point in the T-wave peak if necessary. Ultimately, users can copy these ECG files from floppy diskettes to their hard disks without overlapping, and print the QT dispersion results report.

FIG. 3 is a block diagram illustrating a method for determining a QT dispersion according to a preferred embodiment of the present invention, wherein the processor 14 is programmed to perform the method or function illustrated in FIG. This method or program
Filtering the ECG-induced signal to reduce noise without distorting the T-wave morphology. The filtered signals are divided into different groups and are classified into groups such as rising, inverting, biphasic or TU-wave coupling type. Critical markers such as T-wave peaks and T-waves are determined. Next, a corrected heart rate value is calculated based on the QT dispersion and the determined marker. Markers are reference points across all 12 leads, QRS onset points and QRS
It is determined by measuring an offset point or the like. 12
It has been shown that the variation of the QRS onset point across the leads is much smaller than the variation of the T-wave end point. Thus, the same QRS onset point can be used for all 12 leads. The global offset point is calculated as a first criterion to search for the T-wave markers of the individual leads. These markers are shown in FIG.
The programmed computer is means for performing the functions shown in the block diagram of FIG. 2, in particular means for filtering the ECG signal and removing noise without distorting the form of the T-wave 20; Means for categorizing the T-wave marker, and means 2 for determining the T-wave marker critical point.
4 and means 2 for calculating the QT dispersion
6 inclusive.

The step of the filter processing is Singular-Value
-Based on the dimension recurrence method of -Decomposition (SVD). This technique is described in the Cambridge University Press edition, William H. Press, Saul A.
Toukolsky, Brian P. Flannelly,
NumericalRec, by William T. Wettering
ipes in C, The Art of Scienti
fic Computing (Scientific Computer Computing Technology) (19
1988), pp. 60-71 and 534-539, incorporated by reference in the text. SVD
Is based on the algebra of linear algebra that in every Ux S matrix, the number of rows U is greater than or equal to the number of columns S, the product of the U x S column orthogonal matrix U1, Having a U x S diagonal matrix S 1, and
It can be represented as an arrangement matrix of a U x S orthogonal matrix V. The form of these matrices is illustrated below.

[0019]

(Equation 3)

The U and V matrices have columns that are orthonormal or orthogonal in the sense of a table.

[0021]

(Equation 4)

[0022]

(Equation 5)

Since V is a square, it is also orthonormal in the row, and V · V ^T = 1. The 12 or 15 lead ECG signal is accumulated, the first row constituting the digital signal representing the first sample in each lead, the first column constituting the sample of the first lead, and so on. Exists in each matrix. SVD dimension recurrence method uses A matrix
Decompose into three matrices.

[0024]

(Equation 6)

Where U and V are orthogonal matrices in the sense that their columns are each orthonormal, S is a diagonal matrix called a singular matrix, and [S
_1, S _2, ... can be expressed as S _n] diag ', the element values monotonically decreasing, i.e. S _1> S _2> ...> with S _n.

After more than a hundred ECG analyses, the first three diagonals were found to contribute more than 98% of the total energy. That is,

[0027]

(Equation 7)

Thus, the ECG can be reconstructed using the first three elements of each matrix. In this way, U1
Is a subset of the U matrix using the first three columns,
V1 is understood as a subset of V using the first three rows, and S1 is understood as a subset of S using the first three elements. The new 12 or 15 lead ECG is reconstructed as follows.

[0029]

(Equation 8)

The two reconstructed T waves are shown in FIG. 4 together with the original wave. The reconstructed T-wave is smoother than the original wave, but lacks the phase distortion and extension that often occur when ordinary digital filters are used. Also,
It has been found that the QRS portion of the 12-lead ECG signal can be reconstructed with high accuracy by using the first four components.

The form of the T-wave can be categorized into several large patterns using the T-wave detection function. These patterns include a monophasic rising T-wave (most normal ECGs), a monophasic reversal T
There is a combined type of wave, biphasic T wave and TU wave. To form the detection function, SVD uses the difference between successive samples of the reconstructed T-wave and the absolute value. Two examples of T-wave detection functions are shown in FIG. It can be seen as follows. (1) All values are positive in the T-wave detection function,
(2) The peak in the detection function corresponds to the maximum slope (rising slope or falling slope) of the original T wave, and (3) the valley point in the detection function corresponds to the peak of the original T wave. This T-wave detection function eliminates possible baseline drift and enhances T-wave characteristics. Since the peak of the T wave can be irregular or spurious on a relatively flat plateau, the peak of the T wave can be more easily identified from the valley points in the detection function.

Patterns based on the form in the detection function for pattern recognition can be summarized as follows. (1) Both T-waves with rising and reversing morphologies have similar characteristics in the detection function. That is, there are two peaks and one valley, and (2) the biphasic T wave has three or more peaks and two peaks.
(3) U wave adds more peaks and valleys to the main peak in the detection function.

As can be seen, the peak of the T wave can be determined from the valley point in the detection function. But things are more complicated. This is because the peak of the T wave is not as clearly defined as the peak of the QRS waveform. In many cases, there is a plateau around the T wave, especially for T waves with small amplitude. To this end, a step of centering the area is employed to overcome this problem. The area around the valley point of the detection function is as follows.

There

[0035]

(Equation 9)

Ptx is a sample point of the T wave, Tp is a peak point defined by the detection function, Th is a threshold value of the region, and Yx is the amplitude of the T wave around Tp. The center point of this region is defined as the last T-wave peak. This method uses the valley point of the detection function directly as the T wave peak,
It has been found that the reproducibility is higher.

Another T-wave peak detection method is a pattern matching method based on the SVD technique. SVD analysis was performed on T waves confirmed to be monophasic. The T-wave template is the first column of the U matrix, as shown in FIG. 6, and is the vector of the first principal element. This template is used to match the T wave in the moving window of the display 15 of FIG. The T-wave peak is defined as the point where the template and T-wave have the largest correlation function. This method is more accurate than the area center method. This is because an overall template is used instead of a method that depends only on a few points.

After locating the T-wave, the maximum slope point (MSP) after the last T-wave is used to find the end of the T-wave. The MSP tangent is more likely to be affected by noise, so instead of using it, we analyze some neighbors around the MSP. As shown in FIG. 7, the region is MS
Two points 30 and 32 before P, MSP itself, 3 after MSP
Includes 4, 36, 38, and 40 points. To fit a straight line to this part as shown in FIG. 7, a least squares method is used. Another line is drawn using the threshold defined in the TP section, which is the baseline in the window plus one standard deviation. The intersection of these two lines is defined as the first T-wave endpoint. Next, a non-linear correction based on the amplitude of the T wave peak is applied to the end of the first T wave. This correction is shown in FIG. 8, where the millisecond correction is shown in coordinates for the T wave. For the combined TU wave,
The end of the T wave is further extended to the lowest point between the T wave and some U waves.

Based on the T-wave peak and the end of the T-wave, Q
The T dispersion is calculated. First, a lead with any of the following conditions is deleted. (a) very noisy, (b) very flat T-wave, (c)
T-wave with an undefined pattern. For T-wave peak variability, leads with biphasic T-waves are eliminated.

The QT interval in milliseconds is the number of beats per minute (bpm)
, The variation of the T-wave peak also changes with the number of bpm. Variations between the 12 leads (whole) and the chest leads (chest) are corrected as follows according to the heart rate using the well-known Bazette equation.

[0041]

(Equation 10)

Where QTC = corrected QT interval QT = calculated QT interval HR = heart rate variation due to beats per minute For the variation between the maximum and minimum QT intervals, ie the end point of the maximum QT Or QT peak) and the minimum QT end point or QT peak. QT for variation
As another measure of, it is possible to base on the standard deviation of all QT values.

[0043]

It has been found that the results obtained in the determination of the QT dispersion according to the invention are more reproducible, that is to say the average difference in variability is smaller than with the known method.

[Brief description of the drawings]

FIG. 1 schematically illustrates an ECG monitor and a 12-lead analysis system according to the present invention.

FIG. 2 is a block diagram illustrating a method of determining a QT dispersion according to a preferred embodiment of the present invention.

FIG. 3 shows ECG markers.

FIG. 4 shows a T wave reconstructed by the method according to the invention.

FIG. 5 shows a reconstructed T-wave and a detection function derived therefrom according to the invention.

FIG. 6 shows six chest lead ECGs and their first main vectors as pattern matching templates.

FIG. 7 shows a least-squares method for determining the end point of a T-wave.

FIG. 8 shows a correction factor for determining the end point of the T wave.

Continued on the front page (72) Inventor Shankara Lady 53012, Wisconsin, United States, Seedberg, Tower Avenue Dove 75

Claims (42)

[Claims] Acquiring a plurality of ECG-derived signals; filtering the signals to reduce noise without distorting the shape of the T-wave; determining a T-wave marker critical point; A method for analyzing a QT dispersion in an ECG induction signal, comprising calculating a QT dispersion and a heartbeat correction value thereof from the determined T wave marker points. 2. The method of claim 1 including the step of classifying the T waves into different groups. 3. The method of claim 2, wherein the filtering step includes storing the respective ECG derived signals in a column-by-column matrix and reconstructing each ECG signal by decomposing the matrix by a dimensional recurrence method. The method of claim 1. 4. The filtering step includes accumulating each signal of the ECG in a matrix A for each column, and converting each ECG signal into a matrix A by a dimension recurrence method. , Where U and V are each orthogonal matrices, and S is a diagonal matrix: [S ₁ , S ₂ ,. . . S _k ] _diag whose elements are monotonically decreasing values, ie S ₁ > S
_2. The method of claim 1, wherein ₂ >...> S _k and have k <n and are of original dimension. 5. At least some reconstructed ECGs
5. The method of claim 4, comprising generating a detection function from absolute differences between successive samples of the signal and determining a T-wave peak from a valley point of the detection function. 6. The method according to claim 5, comprising determining a valley point of the detection function by centering the region. 7. The step of determining a T-wave critical marker point comprises: determining a T-wave from the intersection of the maximum slope of the reconstructed ECG signal after the T-wave with a threshold plus a non-linear correction function based on the T-wave amplitude. The method of claim 6, comprising determining a wave endpoint. 8. The method of claim 7, including the step of ignoring the noisy, flat T-wave or T-wave ECG signal with an undefined pattern. 9. The method of claim 8, including the step of classifying the T waves into different groups. 10. The method of claim 5, wherein determining a T-wave critical marker point comprises generating a template for at least some of the ECG-derived signals, and combining the measured ECG signal with the template. The described method. 11. The method of claim 10 including the step of ignoring the noisy, flat T-wave or T-wave ECG signal with an undefined pattern. 12. The method of claim 11, including the step of classifying the T waves into different groups. 13. At least some reconstructed ECs
2. The method of claim 1, further comprising: determining an absolute difference among successive samples of the G signal; generating a detection function from the absolute difference; and determining a T-wave peak from a valley point of the detection function. the method of. 14. The method of claim 13, wherein determining the valley point of the detection function comprises centering the region. 15. The step of determining a T-wave marker critical point comprises: adding a line coincident with a least-squares value around a maximum slope of the reconstructed ECG signal after the peak of the T-wave to a threshold plus an amplitude of the T-wave. 14. The method of claim 13, comprising determining the end point of the T wave from the intersection with the based non-linear correction function. 16. The method of claim 15, including the step of ignoring T-wave ECG signals having a noisy, flat T-wave or with an undefined pattern. 17. The method of claim 13, wherein determining a T-wave marker critical point comprises generating a template for at least some of the induced signals of the ECG, and combining the measured ECG signal with the template. The described method. 18. The method of claim 17 including the step of ignoring ECG signals for T-waves with noisy, flat T-waves or with an undefined pattern. 19. A means for acquiring a plurality of ECG-derived signals, means for filtering the signals to reduce noise without distorting the form of the T-wave, means for determining a T-wave marker critical point, and An apparatus for analyzing a QT dispersion in an ECG-guided signal, including means for calculating a QT dispersion and a heartbeat correction value thereof from the marker. 20. The apparatus according to claim 19, further comprising means for classifying the T waves into different groups. 21. A means for filtering comprises means for accumulating the ECG-derived signals in a matrix for each column, and means for reconstructing each ECG signal by decomposing the matrix by a dimensional recurrence method. 20. The device of claim 19, wherein 22. The filter processing means accumulates the ECG signal in the matrix A for each column, and the matrix A is divided into the following three partial matrices by a dimension recurrence method. To construct a method for reconstructing each ECG signal, where U and V are each orthogonal matrices, and S is a diagonal matrix: [S ₁ , S ₂ ,. . . S _k ] _diag whose elements are monotonically decreasing values, ie S ₁ > S
_2> ...> with S _k, apparatus according to claim 19. 23. At least some reconstructed ECs
20. The apparatus of claim 19, comprising means for generating a detection function from absolute differences between successive samples of the G signal and means for determining a T-wave peak from a valley point of the detection function. 24. The apparatus of claim 23, wherein the means for determining the peak of the T-wave includes means for centering the region of the valley points. 25. The means for determining a T-wave marker critical point,
The maximum slope of the reconstructed ECG signal after the T-wave peak,
24. The apparatus of claim 23, further comprising means for determining a T-wave end point from an intersection with a non-linear correction function based on a threshold plus a T-wave amplitude. 26. The apparatus of claim 23, further comprising means for ignoring the noisy, flat T-wave or T-wave ECG signal with an undefined pattern. 27. The means for determining a T-wave marker critical point,
24. The apparatus of claim 23, comprising means for generating a template of at least some of the ECG-derived signals and means for associating the template with the measured ECG. 28. The apparatus according to claim 27, further comprising means for ignoring the noisy, flat T-wave or T-wave ECG signal with an undefined pattern. 29. At least some reconstructed ECs
20. The apparatus of claim 19, further comprising means for generating a detection function from the absolute difference between successive samples of the G signal and means for determining a T-wave peak from a valley point of the detection function. 30. The apparatus of claim 29, wherein the means for determining the peak of the T-wave includes means for centering a region of valley points. 31. A means for determining a T-wave marker critical point,
T from the intersection of the maximum slope of the reconstructed ECG signal after the T-wave peak and the threshold plus a nonlinear correction function based on the T-wave amplitude
32. The apparatus of claim 31, including means for determining a wave endpoint. 32. The apparatus of claim 31, further comprising means for ignoring the noisy, flat T-wave or T-wave ECG signal with an undefined pattern. 33. The apparatus according to claim 32, further comprising means for classifying the T waves into different groups. 34. A means for determining a T-wave marker critical point,
31. The apparatus of claim 30, comprising means for generating a template for at least some of the ECG-derived signals, and means for combining the measured ECG signal with the template. 35. The apparatus of claim 34, further comprising means for ignoring a noisy, flat T-wave or T-wave ECG signal with an undefined pattern. 36. The apparatus according to claim 35, further comprising means for classifying the T waves into different groups. 37. The method of claim 1, wherein determining the T-wave marker critical point comprises determining a T-wave peak and a T-wave end point. 38. The means for determining the T-wave marker critical point is
20. The apparatus of claim 19, comprising means for determining a wave peak and a T-wave end point. 39. A method comprising: determining a T-wave marker critical point of an acquired ECG-guided signal; and calculating a QT dispersion and its heartbeat correction value from the determined T-wave marker. How to analyze QT dispersion in guidance signal. 40. The method of claim 37, wherein determining a T-wave marker comprises determining a T-wave peak and a T-wave endpoint. 41. A means for analyzing a QT dispersion obtained from an ECG signal to determine a T-wave critical marker, and a means for calculating the QT dispersion and its heartbeat correction value from the determined T-wave marker. Including equipment. 42. The method of claim 41, wherein the means for determining a T-wave marker comprises means for determining a T-wave peak and a T-wave endpoint.