JP6624508B2 - Electrocardiogram analyzer, electrocardiogram analysis method, and electrocardiograph - Google Patents

Description

  The present invention relates to an electrocardiogram analyzer, an electrocardiogram analysis method, and an electrocardiograph, and more particularly to a technique for extracting respiratory information from an electrocardiogram signal.

  Sleep apnea (hypnea) syndrome (Sleep Apnea (Hypopnea) Syndrome: SAS or SAHS) is accompanied by excessive daytime somnolence due to disrupted sleep, and five or more consecutive apnea times of 10 seconds or more during sleep. The disease occurs more than 30 times / 7 hours. Hereinafter, for convenience, SAS and SAHS are simply referred to as SAS.

Diagnosis of SAS includes all-night sleep polygraph (Poly) that measures oral and nasal breathing, snoring, arterial blood oxygen saturation (SpO ₂ ), respiratory effort, electroencephalogram, eye movement, intestinal myocardiogram, electrocardiogram, and body position all night. -Somno-Graphy (PSG) test is required. However, the polysomnography for overnight sleep usually requires staying in a hospital and requires wearing a cannula, a large number of electrodes, and sensors, so that the burden on the subject is relatively large. Therefore, a simpler screening test (simple PSG test) is often performed before performing an overnight polysomnography test. For example, Patent Literature 1 discloses a sleep apnea test apparatus used for a screening test for the presence or absence of apnea and hypopnea, the number of times, SpO ₂ value, snoring, and respiratory effort.

JP 2006-320732 A JP 2010-51387 A

  Although the burden on the subject is reduced by the simple PSG test, it is cumbersome to attach a respiratory sensor near the nose and mouth and go to bed, and a method with less burden is desired.

  On the other hand, a method of detecting apnea or hypopnea from an electrocardiogram has been studied, focusing on the fact that the heart rate changes due to apnea or hypopnea (Patent Document 2). In Patent Literature 2, a cyclic variation of heart rate (CVHR) is detected in which the heart rate repeats bradycardia and tachycardia in synchronization with the periods of apnea and hypopnea. Since the heart rate can be obtained from, for example, the interval (R-R interval) of the R wave of the electrocardiogram (ECG), the CVHR can be detected using a Holter electrocardiogram or the like that has a small measurement load.

  However, since the change in heart rate is not always caused by apnea or hypopnea, it is difficult to accurately measure CVHR for a subject having another factor in which the heart rate is unstable. . For example, since heart rate is unstable in patients with atrial fibrillation, even if they are suffering from SAS, it is not possible to determine from ECG whether heart rate fluctuations are due to atrial fibrillation or SAS. Have difficulty.

  The present invention has been made in view of such problems of the related art, and an electrocardiogram analyzer, an electrocardiogram analysis method, and an electrocardiograph capable of accurately acquiring respiratory information of a subject from an electrocardiogram signal. For the purpose of providing.

  The above-mentioned object is to provide an acquisition unit that acquires a first electrocardiogram signal and a second electrocardiogram signal having different lead directions, an estimation unit that generates an estimation signal of a second electrocardiogram signal from the first electrocardiogram signal, 2. An electrocardiogram, comprising: subtraction means for generating a difference signal between the electrocardiogram signal of No. 2 and the estimation signal generated by the estimation means; and extraction means for extracting a low-frequency component of the difference signal as a signal related to respiratory effort. Achieved by the analyzer.

  With such a configuration, according to the present invention, it is possible to provide an electrocardiogram analyzer, an electrocardiogram analysis method, and an electrocardiograph that can accurately acquire respiratory information of a subject from an electrocardiogram signal.

1 is a block diagram illustrating a functional configuration example of a Holter monitor as an example of an electrocardiogram analyzer according to an embodiment of the present invention. FIG. 4 is a block diagram schematically illustrating an analysis process according to the embodiment of the present invention. FIG. 3 is a diagram illustrating an example of an estimation model in FIG. 2. It is a flowchart regarding the electrocardiogram analysis processing which concerns on embodiment of this invention. FIG. 4 is a waveform chart for explaining the first embodiment of the present invention. FIG. 4 is a waveform chart for explaining the first embodiment of the present invention. FIG. 9 is a waveform chart for explaining Example 2 of the present invention. FIG. 9 is a waveform chart for explaining Example 2 of the present invention.

  Hereinafter, the present invention will be described in detail based on exemplary embodiments with reference to the drawings. FIG. 1 is a block diagram illustrating a functional configuration example of a Holter monitor 100 as an example of an electrocardiogram analyzer according to an embodiment of the present invention. Although the electrocardiogram analyzer of the present embodiment has a configuration for acquiring an electrocardiogram signal (lead signal) from a subject, the configuration for acquiring an electrocardiogram signal from a subject in the present invention is as follows. Not required. The electrocardiogram analysis device according to the present invention may be any device that can acquire electrocardiogram data measured in advance from any storage device or storage medium connected directly or via a network.

  In FIG. 1, a CPU 1 reads out a control program stored in a ROM 2 to a RAM 3 and executes the control program, thereby realizing various functions of the Holter monitor 100 including an electrocardiogram analysis process described later. The ROM 2 is a non-volatile memory that stores parameters required for processing, such as a program executed by the CPU 1, GUI data for displaying a menu screen, user setting data, initial setting data, and the like, and is at least partially rewritable. It may be. The RAM 3 is used as an area for developing a program executed by the CPU 1 and a temporary storage area for variables, data, and the like. The memory card 4 is a storage device that stores an electrocardiogram signal input from a bioelectrode, another bioelectric signal and data in the form of digital data. The memory card 4 is removably attached to the card slot 5.

  The display unit 6 is, for example, a liquid crystal display device. The operation unit 8 includes switches and buttons for turning on / off the power, starting and stopping measurement, inputting various events, and performing various settings. The operation unit 8 may include a touch panel provided on the display unit 6.

The analog-digital converter (A / D converter) 9 converts an analog signal input from the electrocardiogram electrode 12 into a digital signal. The sensor I / F 10 is an interface for acquiring an electrocardiogram signal (lead signal) from the electrocardiogram electrode 12. The sensor I / F 10 may be connectable to an SpO ₂ (arterial blood oxygen saturation) sensor, a blood pressure / pulse wave measurement cuff, or the like, in addition to the electrocardiogram electrode 12.
The communication interface 20 is a communication interface for communicating with an external device 40 such as a host computer or a printer, and has a configuration conforming to a wired and / or wireless communication standard.

  When acquiring and recording an electrocardiogram signal (leading signal) using the Holter monitor 100 having such a configuration, a plurality of electrocardiogram electrodes provided at the distal end of the electrocardiogram electrode 12 at predetermined positions on the body surface of the subject are used. The connector provided at the near end of the electrocardiogram electrode 12 is connected to the connector of the sensor I / F 10.

  In the present invention, in order to detect a respiratory signal from an electrocardiographic signal, electrocardiographic signals in a plurality of lead directions, which are different in the susceptibility to action potential changes in the chest and abdominal muscles due to respiratory effort, are used. Typically, the measurement is performed using a guidance method in which the guidance directions include two directions orthogonal to each other.

  More specifically, two or more different guiding directions including at least the guiding in the X-axis direction (X-leading) among the guiding directions in the left-right (X-axis), up-down (Y-axis), and front-back (Z-axis) directions. It is preferable to measure with. This is because it is considered that many signals caused by the chest and abdominal movements due to respiratory effort are included in the guidance in the X-axis direction.

  For example, the X lead and other leads are selected from the X lead (CM5, mCM5, CC5), the Y lead (NASA, bipolar aVF, CMf), and the Z lead (NASA, mV1, CM2, CB2, bipolar V3). can do. Further, a standard 12 lead or a vector electrocardiogram lead method (for example, a Frank lead method) can be used. When measuring an electrocardiogram signal of two channels, for example, a combination of CC5 and NASA lead can be suitably used as a combination of the lead in the X-axis direction and the lead in the Y-axis direction.

  For example, when the power is turned on by operating the operation unit 8, the CPU 1 can perform an operation mode setting process for acquiring an electrocardiogram signal, such as an initialization process, at a timing before the acquisition of the electrocardiogram signal is started. . Before starting the recording operation of the electrocardiogram signal, the CPU 1 displays an input screen or the like for allowing the user to set the date and time and information (for example, a patient ID) capable of specifying the subject, as necessary. You may.

  Then, the CPU 1 starts the recording operation of the electrocardiogram, for example, in response to the input of the recording start instruction from the operation unit 8 or the lapse of a predetermined time from the power-on. Note that the acquisition of the electrocardiogram signal itself may be performed before the recording start instruction. For example, by displaying a waveform according to the current operation mode on the display unit 6, it is possible for the user to confirm from the waveform display whether or not the operation mode has been correctly set.

  The electrocardiogram signal is sampled at a predetermined sampling rate (for example, 125 Hz to several KHz) by the A / D converter 9 and recorded in the memory card 4 or the external device 40 in the form of a digital signal. Note that the measurement and recording operations of the electrocardiogram signal are not directly related to the present invention and may be the same as those in the related art, so that further description will be omitted.

(Electrocardiogram signal analysis processing)
Next, a method of analyzing two electrocardiogram signals having different lead directions and detecting a signal indicating the movement of the chest and abdomen accompanying the respiratory effort (hereinafter, referred to as a respiration signal) as respiration information will be described. FIG. 2 is a block diagram schematically illustrating an electrocardiogram analysis process according to the present embodiment.

The analysis unit 300 extracts and outputs a respiratory signal from the two-channel electrocardiogram signals E _C1 and E _C2 . Each functional block (described later) of the analysis unit 300 can be realized by software, hardware, or a combination of both. The part realized by software can be implemented, for example, by the CPU 1 executing a program.

  In the analysis unit 300, the low-pass filters 305 and 306 provided as necessary suppress high-frequency components present at discontinuous points (for example, R-waves) included in the electrocardiogram signal, thereby lowering the Improve the estimation accuracy of. The cut-off frequency of the low-pass filters 305 and 306 can be set to, for example, about several Hz.

The estimation signal generation unit 310 uses an estimation model to convert one electrocardiogram signal (here, the electrocardiogram signal E _{C1 of} ch1) to the other electrocardiogram signal (here, the electrocardiogram signal E _{C2 of} ch2) (here, ch2). An estimated signal {E _C2 ) of the electrocardiogram signal E _C2 is generated. In the case where the low-pass filter 305 and 306 are provided, for electrocardiogram signal _{E C1} of ch1 to enter the estimated signal generator 310 is a signal after filtering, estimation signal _{^ E C2} of the electrocardiogram signal _{E C2} of ch2 also it corresponds to electrocardiogram signal E _C2 after the low-pass filter 306 is applied. The subtractor 320 generates a difference signal between the actually measured electrocardiogram signal E _C2 and the estimated signal ΔE _C2 . The respiratory signal extraction filter 330 removes the R-wave component and the bias error from the difference signal, performs a smoothing process (to the extent that the respiratory signal component is not affected), and extracts the respiratory signal. The respiratory signal extraction filter 330 can be realized by, for example, a combination of a median filter for removing an R-wave component and a band-pass filter for passing a respiratory component, but another combination may be used.

Difference signal between the electrocardiogram signal E _C2, which is measured with the respiratory signal is generated by the subtractor 340 is fed back to the estimation signal generating section 310 as an estimation error signal. The estimation signal generation unit 310 sets parameters of the estimation model to, for example, a least square method (Least Mean Square: LMS), a normalized least square method (Normalized LMS: Normalized LMS), and a prediction error method so that the estimation error signal becomes small. Estimate using such as and update.

Here, the estimation signal generation unit 310 will be described. The estimation signal generation unit 310 applies an estimation model to a system that receives one electrocardiogram signal as an input and outputs an estimation signal of the other electrocardiogram signal, identifies a system by estimating parameters of the estimation model, and identifies the estimation signal ＾ Generate E _C2 . In the present embodiment, the estimation signal generation unit 310 uses a polynomial model as an estimation model.

The electrocardiogram signals E _C1 and E _C2 measured through the electrocardiogram electrode 12 are signals transmitted by the action potential of the heart to the body surface. Therefore, the electrocardiogram signal _EC0 near the heart can be considered to be a signal observed through the chest model. Here, an electrocardiogram signal (first electrocardiogram signal) E _{C1 of} the first channel (ch1) is observed through the chest model 1 210, and an electrocardiogram signal (second electrocardiogram signal) E _{C2 of} the second channel (ch2) is obtained. It is assumed that the signal is a signal observed through the model 2 220. Since the guiding directions are different, different chest models are assumed for each channel.

It is also assumed that a signal (respiratory signal) resulting from the movement of the chest and abdomen due to respiratory effort is superimposed on one electrocardiogram signal (here, the electrocardiogram signal E _{C2 of} ch2) by the adder 230. Although the respiration signal can be superimposed on both electrocardiogram signals, the difference in the lead direction causes a difference in the intensity of the respiration signal superimposed on the electrocardiogram signal. Therefore, for convenience, it is assumed that the respiration signal (disturbance due to respiration) is superimposed only on the electrocardiogram signal on which the respiration signal is superimposed more frequently.

Based on such a premise, the measured electrocardiogram signals E _C1 and E _C2 can be expressed as follows.
E _C1 = S1E _C0 (1)
E _C2 = S2E _C0 + rs (2)
Here, rs is a respiration signal. Further, when E _C0 is removed from equation (2),
E _C2 = (S2 / S1) E _C1 + rs (3)
It becomes.
Therefore, to estimate the parameters of the estimation model representing the coefficients S2 / S1, can be obtained estimation signal ^ E _C2 of E _C2 does not contain the respiration signal.

The difference signal between the estimated signal ＾ E _C2 and the actually measured E _C2 is a signal that includes many components other than the electrocardiogram signal included in E _C2 , that is, many components of the respiration signal rs. Therefore, a signal obtained by removing an R-wave component, a bias error, and the like from the difference signal can be extracted as the respiration signal ＾ rs.

In Equations (1) and (2), when E _C1 is estimated from E _C2 , the respiration E _C1 = (S1 / S2) (E _C2 −rs) (3) ′
It becomes. In this case, by estimating the parameters of the estimation model representing the coefficients S1 / S2, the difference signal between the estimated signal ΔE _C1 and the actually measured E _C1 becomes (S1 / S2) rs. In this case, the respiratory signal rs cannot be directly obtained, but a difference signal including the respiratory signal rs can be obtained, so that the respiratory signal can be extracted. Therefore, no matter which of the two measured electrocardiogram signals is input to the estimation signal generator 310, the respiratory waveform can be extracted.

FIG. 3 shows an example of a polynomial model that can be used as an estimation model by the estimation signal generation unit 310, such as an ARX (Auto Regressive with eXogenous) polynomial model, an OE (Output-Error) polynomial model, and a BJ (Box-Jenkins) polynomial model. Is shown. In each model, u (k) indicates an input signal (E _C1 ), y (k) indicates an output signal (ΔE _C2 ), and ω (k) indicates a noise component.

  If the chest models S1 and S2 are each represented by an LCR circuit, the transfer function of the chest model can be represented by a biquad filter. Therefore, the order of each model is basically 2, and can be adjusted according to the high frequency component included in the electrocardiogram signal.

  With reference to the flowchart of FIG. 4, an operation of analyzing an electrocardiogram signal by the Holter monitor 100 of the present embodiment will be described. The analysis process may be performed at the time of recording the electrocardiogram signal, or may be performed on the recorded electrocardiogram signal. Further, as described above, the measurement may be performed by a device having no measurement function, such as the Holter monitor 100, which is different from the measurement device. In the following, it is assumed that an analysis process of an electrocardiogram signal stored in the ROM 2 in advance is executed in the RAM 3 and the CPU 1 executes the analysis process. At least a part of the analysis process is performed by dedicated hardware such as an ASIC. It may be realized.

In S101, the CPU 1 acquires an electrocardiogram signal to be analyzed from the memory card 4 or the external device 40 communicably connected. As described above, the acquired electrocardiogram signal is an electrocardiogram signal representing changes in at least two different electromotive forces in different directions. Here, as a typical example, it is assumed that the Y lead is obtained as the first ECG signal _EC1 and the X lead is obtained as the second ECG signal _EC2 .

  Note that the plurality of electrocardiogram signals acquired in S101 are signals measured at the same time, and even when acquiring a part of the electrocardiogram signal in the measurement period, the electrocardiogram signal for the same measurement period is acquired. The CPU 1 stores the read electrocardiogram signal in the RAM 3. In the case where the analysis interval is long, as in the case of performing SAS screening, an electrocardiogram signal is acquired at predetermined time intervals in S101.

Next, in S103 CPU 1 is to the sample row of the first electrocardiogram signal E _C1 stored in the RAM 3, the estimation model by inputting the estimated signal generator 310 after applying a low pass filtering process if necessary It starts the identification, to enter the sample sequence of the second electrocardiogram signal E _C2 to the subtractor 320. Here, it is assumed that parameter estimation of the estimation model is performed (online identification) each time a new sample is input. The sample of the estimated signal ＾ E _C2 of the second electrocardiogram signal E _C2 is sequentially generated by the estimated signal generator 310 and output to the subtractor 320.

In S105, the subtractor 320 generates a difference signal by subtracting the estimated signal ＾ E _C2 output from the estimated signal generator 310 from the actually measured second electrocardiogram signal E _C2 , and outputs the difference signal to the respiratory signal extraction filter 330. Note that, when the low-pass filter processing is applied before the sample sequence of the first electrocardiogram signal E _C1 is input to the estimation signal generation unit 310, the CPU 1 determines the sample sequence of the second electrocardiogram signal E _C2. Also, before input to the subtractor 320, the same low-pass filter processing as applied to the first ECG signal _EC1 is applied.

  In S107, the respiration signal extraction filter 330 extracts and outputs the respiration signal ＾ rs from the difference signal.

Hereinafter, in the analysis target section, the processing of S101 to S107 is sequentially and repeatedly performed for the samples of the first and second electrocardiogram signals E _C1 and E _C2 . When estimating the parameters of the estimation model without providing a learning period, immediately after the start of the analysis processing, the consistency between the estimated signal ΔE _C2 and the actually measured second electrocardiogram signal E _C2 is not high, but as time elapses, The parameters of the estimation model are optimized, and the consistency is improved.

  When the estimation model is identified off-line by the estimation signal generation unit 310, a process of generating an estimation signal using a predetermined number of consecutive sample groups is performed for each sample group included in a predetermined non-overlapping predetermined period (data window). Are executed sequentially. For example, if an estimated signal of the corresponding section is generated using the sample group included in the first data window, the process of generating the estimated signal of the next section using the sample group included in the next data window is repeatedly performed. . The length of the data window can be, for example, about 60 to 90 seconds.

The CPU 1 stores the respiration signal ＾ rs extracted by the respiration signal extraction filter 330 in the RAM 3 and then records the respiration signal ＾ rs in the memory card 4 or transmits the respiration signal ＾ rs to the external device 40. Alternatively, the CPU 1 may execute the detection of apnea or hypopnea in parallel using the respiration signal ＾ rs. For detection of apnea or hypopnea, another signal (SpO ₂ ) measured in the same analysis target section may be used as necessary.

(Example)
An example of using a previously recorded electrocardiogram signal of the CC5 lead and the NASA lead (sampling rate of 125 Hz) until the respiration signal ＾ rs is obtained by the above-described method will be described with reference to FIGS. 5 and 6 show Example 1 for a healthy person (a subject without a disease in which the heart rate fluctuates), and FIGS. 7 and 8 show Example 2 for a patient with atrial fibrillation.

5 to 8, ecg1 and ecg2 indicate the measured electrocardiogram signals, ecg1 is the NASA lead, and ecg2 is the CC5 lead.

Ecg1 (filtered) and ecg2 (filtered) are electrocardiogram signals obtained by applying a low-pass filter with a cutoff frequency of 4 Hz (corresponding to the low-pass filters 305 and 306 in FIG. 2) to ecg1 and ecg2. ecg2 () is an ecg2 (filtered) signal estimated from ecg1 (filtered), err () is a difference signal obtained by the subtractor 320, and filtered () is a respiration signal extracted by applying the respiration signal extraction filter 330 to the difference signal. ＾ rs.
The respiratory signal extraction filter 330 includes a median filter (80 taps (0.64 seconds) for removing remaining R wave components) and a smoothing filter (150 samples before and after (2. 4 seconds) and a high-pass filter (cutoff frequency: 0.08 Hz) for removing a bias error.

  Resp indicates a respiratory signal measured by a forced respiration sensor attached to the chest to evaluate the method according to the present embodiment. Note that, in order to facilitate comparison with resp, the output of the respiration signal extraction filter 330 multiplied by −1 is shown as Δrs.

ARX, OE, rOE, and BJ in parentheses indicate the types of estimation models used in the estimation signal generation unit 310, respectively. Except for rOE, off-line identification with a data window length of 90 seconds was performed, and on-line identification was performed for rOE. Details regarding the identification are as follows.
(Order)
ARX (Example 1) When Ay = Bu A: 7th order, B: 10th order ARX (Example 2) When Ay = Bu A: 10th order, B: 10th order OE When y = B / Fu + e B: Second order, F: Second order rOE B: Second order, F: Second order BJ When y = B / Fu + C / De B: Second order, F: Third order, C, D: Fifth order Example 1 except for the ARX model , 2 have the same order.
(Parameter identification method)
ARX: Least squares method OE: Prediction error method rOE: Normalized LMS method BJ: Prediction error method Step size: 2.0e ^-4
The size of the step size is an example in which a change in the parameter is set to a value that is not affected by the pulsation.

  It can be seen that also in the respiratory signal filtered () obtained by the method of the present embodiment, a waveform similar to the respiratory signal resp measured by the forced respiration sensor is obtained. For example, an event that seems to be apnea appears in a section around 30 to 40 seconds of the respiratory signal resp in FIGS. 5 to 6 which is an example of a healthy person, and a similar event also appears in the respiratory signal filtered (). ing.

Note that there is a difference in the correlation between the respiratory signal filtered () and resp according to the estimation model. As a result of performing statistical evaluation by a plurality of examples, OE, rOE, BJ, and ARX are arranged in descending order of correlation. there were. The good accuracy of the estimated signal obtained using the OE and rOE models is considered to be due to the good correspondence between the relationship E _C2 = (S2 / S1) E _C1 + rs and the estimation model.

  Further, as shown in FIGS. 7 and 8, even if the relationship between the ECG signal as the estimation source and the estimated ECG signal is reversed, or the ECG signal of the patient with atrial fibrillation, A good respiratory signal was obtained.

  As described above, according to the present embodiment, one of two electrocardiogram signals having different lead directions is used to generate the other estimated signal, and the low-frequency component of the difference signal between the actually measured electrocardiogram signal and the estimated signal is subjected to forced breathing. Extracted as a signal (respiration signal) for Therefore, a respiratory signal can be obtained without depending on the fluctuation of the heart rate, and it becomes possible to screen the SAS from the electrocardiogram signal even for a patient whose heart rate fluctuates, such as a patient with atrial fibrillation.

  In addition, since the electrocardiogram signals having different lead directions can be extracted from the electrocardiogram signals obtained by the lead method which can be used for general electrocardiogram measurement, the results of the electrocardiogram measurement by the Holter electrocardiograph and the like are used for the SAS screening test. It is possible to use it. Therefore, the burden on the subject can be significantly reduced as compared with the conventional simple PSG.

  Note that the electrocardiogram analyzer according to the present invention can also be realized as a program (application software) that causes a generally available general-purpose information processing device such as a personal computer to execute the above-described operation. Therefore, such a program and a storage medium storing the program (an optical recording medium such as a CD-ROM, a DVD-ROM, a magnetic recording medium such as a magnetic disk, a semiconductor memory card, etc.) also constitute the present invention. .

Claims (8)

Acquisition means for acquiring a first electrocardiogram signal and a second electrocardiogram signal having different lead directions;
Estimating means for generating an estimated signal of the second ECG signal from the first ECG signal;
Subtraction means for generating a difference signal between the second electrocardiogram signal and the estimation signal generated by the estimation means;
Extracting means for extracting a low-frequency component of the difference signal as a signal related to respiratory effort,
An electrocardiogram analyzer, comprising:   The electrocardiogram analyzer according to claim 1, wherein one of the first electrocardiogram signal and the second electrocardiogram signal is a lead in the X-axis direction, and the other is a lead in the Y-axis or Z-axis direction. .   3. The electrocardiogram analyzer according to claim 1, wherein one of the first electrocardiogram signal and the second electrocardiogram signal is a CC5 lead, and the other is a NASA lead. 4.   The said estimation means produces | generates the said estimation signal by identifying the system which uses the said 1st electrocardiogram signal as an input signal, and the said estimation signal as an output signal by a polynomial model, The said estimation signal is characterized by the above-mentioned. Item 4. The electrocardiogram analyzer according to any one of items 3.   The electrocardiogram analyzer according to claim 4, wherein the polynomial model is any one of an ARX (Auto Regressive with eXogenous) polynomial model, an OE (Output-Error) polynomial model, and a BJ (Box-Jenkins) polynomial model. . Measuring means for measuring a first electrocardiogram signal and a second electrocardiogram signal having different lead directions using an electrode mounted on the subject;
Storage means for storing the measured first ECG signal and the second ECG signal;
The electrocardiogram analyzer according to any one of claims 1 to 5, which analyzes the first electrocardiogram signal and the second electrocardiogram signal stored in the storage unit.
An electrocardiograph characterized by having: An acquisition step of acquiring a first electrocardiogram signal and a second electrocardiogram signal having different lead directions;
An estimation step of generating an estimation signal of the second electrocardiogram signal from the first electrocardiogram signal;
A subtraction step of generating a difference signal between the second electrocardiogram signal and the estimation signal generated in the estimation step;
Extracting a low frequency component of the difference signal as a signal related to respiratory effort,
An electrocardiogram analysis method, comprising:   A program for causing a computer to function as each unit of the electrocardiogram analyzer according to any one of claims 1 to 5.

Images