JPH03268740A - Processing method for electrocardiogram waveform - Google Patents

Abstract

PURPOSE:To classify the form of an electrocardiogram waveform by an accurate reference and to improve the accuracy of the classification by classifying a waveform of an electrocardiogram, based on a multi-dimensional vector of N dimensions. CONSTITUTION:A tape cassette 1 in which an electrocardiogram is recorded magnetically by a digital modulating signal is loaded to a data reproducing device 2 and reproduced. In the data reproducing device 2, a DSP is provided, and from the reproduced digital modulating signal, an original digital signal is demodulated, and thereafter, signal processings such as filtering, error connection, de-interleave, etc., are performed. This digital data is supplied to an analysis computer 3, and the form of an electrocardiogram waveform is classified by a similarity SI. The similarity is calculated, based on an expression 1. In said expression, (xi) denotes a value at the time of sampling a QRS waveform QRS of one electrocardiogram waveform in the electrocardiogram waveform which is decided to be form abnormality and registered at every 500ms. Also, (yi) denotes a value at the time of sampling the QRS waveform QRS detected as form abnormality newly at every 500ms, and N denotes the number of sampling points.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to an electrocardiographic waveform processing method, and particularly to an electrocardiographic waveform processing method suitable for automatic morphological classification of electrocardiographic waveforms.

[Conventional technology] As one of the techniques for collecting and analyzing electrocardiogram waveforms, Holter
There is an electrocardiogram method.

This Holter electrocardiography method uses a portable recorder (tape recorder) to record electrocardiogram waveforms in daily life over a long period of 24 hours or more, and automatically analyzes the obtained electrocardiogram waveforms using an analysis system. It's a method.

One of the analysis items of this electrocardiogram waveform is the detection and determination of arrhythmia. In diagnosing this arrhythmia, not only abnormal heart rhythm but also morphological abnormalities in electrocardiographic waveforms are observed.

[Summary of the invention]

In an electrocardiogram waveform processing method, this invention classifies electrocardiogram waveforms based on N-dimensional multidimensional vectors, thereby making it possible to perform morphological classification of electrocardiogram waveforms based on clear criteria, thereby improving the accuracy of classification. It is designed to improve the

[Problem to be solved by the invention]

The above-mentioned morphological abnormality of the electrocardiographic waveform is determined by comparison with a normal electrocardiographic waveform, and is classified into morphology. The electrocardiographic waveforms of the classified forms and the frequency of appearance of the electrocardiographic waveforms for each form, along with the above-mentioned heart rhythm abnormalities, are used as materials for medical diagnosis. As a method for classifying shapes, there is a problem in that there are still no established standards.

By the way, Japanese Unexamined Patent Publication No. 57-96638 proposes a technique for stationary analysis of an electrocardiographic waveform by applying an analysis method called a Lorentz plot to the ejaculation cycle of the electrocardiographic waveform. However, in the above-mentioned conventional techniques, various analyzes are performed mainly focusing on the generation cycle of electrocardiographic waveforms, and the morphological classification of electrocardiographic waveforms is not disclosed.

Therefore, an object of the present invention is to provide a method for processing electrocardiographic waveforms that can perform morphological classification of electrocardiographic waveforms based on clear criteria.

[Failure to solve the problem]

J's invention has a configuration in which electrocardiogram waveforms are classified into classes based on N-dimensional multidimensional vectors M1.

[Effect]

Similarity is calculated between electrocardiographic waveforms obtained in time series, and classification is performed based on the calculated similarity.

Waveform morphology can be classified based on a clear criterion called similarity, which improves the accuracy of waveform classification.

Then, it is possible to determine the forms of electrocardiographic waveforms that are determined to be abnormal and the appearance gradient of each form of electrocardiographic waveforms, thereby providing information useful for medical diagnosis.

〔Example〕

Embodiments of the present invention will be described below with reference to FIGS. 1 to 6.

FIG. 1 shows a block diagram of the main parts of an automatic electrocardiogram waveform analysis device to which the present invention is applied.

When an ultra-small tape cassette 1 in which an electrocardiogram waveform is magnetically recorded as a digital modulation signal is loaded into the data reproducing device 2, the digital modulation signal of the electrocardiogram waveform recorded on the magnetic tape (not shown) is read by the magnetic head. will be played.

The data reproducing device 2 is provided with a DSP, and after demodulating the original digital signal from the reproduced digital modulation signal, the DSP performs filtering on the original digital signal.

Signal processing such as error correction and deinterleaving is performed.

This reproduces the digital data of the electrocardiogram waveform,
This digital data is supplied to the computer 3 for analysis.

The analysis computer 3 performs morphological classification of the electrocardiogram waveform based on the flowchart shown in FIG.

FIG. 2 shows a flowchart regarding morphological classification of electrocardiographic waveforms of arrhythmia.

In step 101, noise removal is performed. This is 1
, since electrocardiogram waveforms are recorded over a long period of time, walking,
This is because noise caused by movements in daily life, such as lying down, is mixed in, making automatic analysis difficult. Next, the process moves to step 102.

In step i02, the QR8 waveform is extracted by peak detection. In FIG. 3, a QRS waveform gR5 consisting of a P wave P, a Q wave Q, an R wave R, and an S wave S, and a one-beat electrocardiogram waveform consisting of a T wave T are omitted.

By peak detection, waveform 1 and QR8 waveform 5 in the figure are extracted. Next, the process moves to step 103.

In step 103, characteristics of the extracted QR3 waveform 5, such as width and height, are extracted. By comparing the characteristics with the normal QR3 waveform, it is determined whether there is an abnormality, that is, whether or not it is an arrhythmia. If it is determined that the arrhythmia is an arrhythmia, the process proceeds to step 104; if it is determined that it is not an arrhythmia, the process ends.6 In step 104, the arrhythmia is
A determination is made as to whether this corresponds to a morphological abnormality, a rhythm abnormality, or only a rhythm abnormality.

For example, when electrocardiographic waveforms as shown in FIG. 4 are obtained in time series, waveforms W1 to W8 have a path-shaped waveform, but 1. Since fluctuations are observed between the cycles TI-T7 of the apexes of the waveforms W1 to W8, it is determined that the rhythm is abnormal.

Further, since the waveform W9 is significantly different in amplitude and shape from the other waveforms W1 to W8, it is determined that the waveform W9 has a morphological abnormality. Note that there is also a type (not shown) in which the rhythm abnormality and the M-type abnormality overlap.

If there is only a rhythm abnormality, the process moves to step 105 and the rhythm abnormality is analyzed. Analysis of this rhythm abnormality is
As shown in waveforms W1 to W8 shown in FIG. The degree of abnormality is determined and the process ends.

Further, in the other two cases, that is, in the case of a shape M abnormality, or in the case that a shape abnormality and a rhythm abnormality overlap, the process moves to step 106. Note that the rhythm abnormality in this case is ignored.

In step 106, similarity is calculated between electrocardiographic waveforms obtained in time series. This similarity calculation is performed based on the following equation, and thereby the similarity Sl is determined.

Σxi yi −1 In the above equation, xi is the QR3 waveform QRS of one of the registered electrocardiographic waveforms determined to be abnormal in shape, sampled every 500 m5 as shown in Fig. 5. N represents the number of sampling points (for example, N=], 60).

In addition, yi is QR3 newly detected as a morphological abnormality.
As shown in FIG. 5, this is the value when waveform 5 is sampled every 500 m9, and represents the number of NU sampling points (for example, N=160).

In addition, in the above-mentioned class 1g calculation, in order to prevent the effects of noise, baseline fluctuation, etc., smoothing differentiation is applied to the electrocardiogram waveform. After determining the degree of similarity, step 1
Move to 07.

In step 107, the magnitude relationship between the similarity Sr and a predetermined reference value (for example, 0.9) is determined.
9≦Sl), the similarity calculation described above determines that the electrocardiogram waveform is the same as one of the registered electrocardiogram waveforms that have already been determined to have a morphological abnormality, and in step 108, one of the registered electrocardiogram waveforms is The electrocardiographic waveforms are classified into similar electrocardiographic waveforms, and the frequency of appearance of the classified electrocardiographic waveforms is counted.

Also, if (0,9>Sr), in step 109,
The QR3 waveform of a new morphological abnormality is newly registered as waveform 5.

For example, if the electrocardiographic waveform Wll shown in FIG. 6 is registered as a morphological abnormality, the obtained electrocardiographic waveform W12 is morphologically significantly different from the already registered electrocardiographic waveform Wll. Therefore, the 0th-order electrocardiographic waveform W13, which is determined to have a degree of similarity SN<0.9 and is newly registered as an abnormal electrocardiographic waveform, is morphologically similar to the electrocardiographic waveform W1.1 described above. , the degree of similarity SI≧0.
9, whereas it is morphologically significantly different from the electrocardiographic waveform W12 described above, and it is determined that the degree of similarity SI<0.9 with respect to the electrocardiographic waveform W12. Therefore, it is determined that it has the same form as the electrocardiographic waveform Wll, and the frequency of appearance of the electrocardiographic waveform Wll is counted up.

The process ends at step 109. - The above-described processing steps are repeated for all electrocardiogram waveforms.

In this embodiment, the morphology of the electrocardiogram waveform is classified based on the degree of similarity Sl, so that the morphology of the waveform can be classified based on clear criteria. Since the appearance frequency and cycle of electrocardiographic waveforms can be determined, the classification accuracy of electrocardiographic waveforms can be improved.

〔Effect of the invention〕

According to this invention, since electrocardiographic waveforms are classified into classes based on N-dimensional multidimensional vectors, it is possible to classify the morphology of electrocardiographic waveforms based on clear criteria, and the accuracy of classification is improved. It has the effect of being able to

According to the embodiment, it is possible to perform morphological classification of electrocardiographic waveforms using a clear standard of similarity, and improve the accuracy of waveform classification, thereby improving the efficiency of diagnosis based on electrocardiographic waveforms. There is an effect called 11-.

[Brief explanation of drawings]

Fig. 1 is a diagram of the main parts of an automatic analysis device that can be used in this invention, and Fig. 2 shows an embodiment of this invention.
-Chart, FIG. 3 is a waveform diagram showing an electrocardiographic waveform in a cycle of one beat, and FIGS. 4 to 6 are diagrams for explaining the degree of similarity p'. Explanation of main symbols in the drawings W1 to W9, Wll to W2B: Waveform 1, S■:
Similarity, R: R@, Q, : Q wave, SO3 wave,
T: 4 T waves, 8 QR8HQR waves, PDP waves.

Claims (1)

[Claims] An electrocardiogram waveform processing method characterized by classifying an electrocardiogram waveform into classes based on an N-dimensional multidimensional vector.