JP3738392B2 - Biological signal estimation method - Google Patents

Description

  The present invention relates to a method for estimating a biological signal such as an electrocardiogram and a pulse wave.

  A measurement (measurement) waveform of a biological signal such as an electrocardiogram or a pulse wave usually contains a noise component. Conventionally, as a noise removal method, a method of adding and averaging repeated waveforms of a biological signal (hereinafter referred to as an average method). ) And a method of separating a signal component and a noise component according to a frequency domain.

  Specifically, as an example of the average method, a volume pulse wave is described as an example. A fixed-length average template that is longer than the heartbeat interval of the biological signal to be measured is prepared in advance, and, as shown in FIG. The measurement waveform (a) is cut out by the length of the fixed-length average template, and as shown in FIG. 22, five cut-out waveforms from the first cut-out waveform to the fifth cut-out waveform are shown. An average waveform is obtained, the average waveform corresponding to the average heartbeat interval at this time is determined as the fifth average waveform, and then five cutouts from the second cutout waveform to the sixth cutout waveform are performed. The average waveform for the waveform is obtained, the average waveform for the average heartbeat interval at this time is determined as the sixth average waveform, and thereafter, the average waveform is obtained for each heartbeat by the same calculation. It is way.

  According to the conventional average method, as shown in FIG. 22, distortion due to the influence of the next heartbeat occurs in the latter half part q of the average waveform. Therefore, an accurate average waveform corresponding to one cardiac cycle of the biological signal is generated. This is very difficult, and if the average waveforms for each heartbeat are connected, they become discontinuous, making it difficult to realize a continuous average waveform.

  The main object of the present invention is to solve the above-described problems of the prior art and to provide a biological signal estimation method capable of generating an accurate average waveform for one cardiac cycle of the biological signal. . It is another object of the present invention to provide a biological signal estimation method that can realize a continuous average waveform when the average waveforms for each heartbeat are connected. Furthermore, the present invention provides a biological signal estimation method capable of creating an estimated waveform that can also follow fluctuation components (for example, QRS level and T wave level change for each beat) in each heartbeat that cannot be followed by averaging. With the goal.

The biological signal estimation method according to the present invention creates a cut-out waveform for each heartbeat from the measured waveform of the biosignal, and corrects the cut-out waveform so that all the cut-out waveforms have the same 0 level. A corrected waveform is created, and the corrected waveform is expanded and scaled to match the length of the fixed-length average template according to the characteristics of the biological signal, and the expanded scale waveform is fixed before the update. An updated fixed-length average template waveform averaged by adding to the average template waveform is created, and the updated fixed-length average template waveform is expanded and reduced in accordance with the characteristics of the biological signal and in accordance with the average heartbeat interval. An average waveform for heart rate is created. For example, in the electrocardiogram, as is well known, one heartbeat is composed of a systole and a diastole. However, even if the heartbeat interval changes, the systole does not change much and the diastole mainly changes. If they are different, the ratio between the systole and the diastole is different, and the characteristics of the biological signal as described above refer to such characteristics.

Further, the estimation method of the biological signal according to the present invention is to create a cutting waveform of each heart beat from the measurement waveform of the biological signal, out該切so that all endpoints cutting waveform the created is the same 0 level waveform To generate a corrected waveform, expand and contract the corrected waveform according to the characteristics of the biological signal and match the length of the fixed-length average template, create an expanded scale waveform, and update the expanded scale waveform before the update. An updated fixed-length average template waveform that is averaged by adding to the fixed-length average template waveform is created, and the updated fixed-length average template waveform is scaled according to the characteristics of the biological signal and in accordance with the actual heartbeat interval. Then, an average waveform for one heartbeat is created, and an average waveform for each heartbeat is connected to create a heartbeat-synchronized continuous average waveform.

  In the biological signal estimation method according to the present invention, the heartbeat-synchronized continuous average waveform is subtracted from the measurement waveform of the biological signal to create a differential waveform, and the differential waveform is time-dependently determined for each heartbeat according to the characteristics of the biological signal. In addition, the low frequency and high frequency noises are removed by a digital filter suitable for each division to extract a fluctuation component for each heartbeat, and the fluctuation component is added to the heartbeat-synchronous continuous average waveform to obtain a fluctuation component for each heartbeat. Is also characterized in that an estimated waveform following is created.

  According to the present invention, an updated fixed-length average template waveform is created by expanding and reducing the correction waveform according to the characteristics of the biological signal for each heartbeat and matching the length of the fixed-length average template. An average waveform for one heartbeat of a biological signal is created by expanding and reducing the fixed-length average template waveform to match the average heartbeat interval according to the characteristics of the biological signal and creating an average waveform for one heartbeat. Will be able to.

  Further, according to the present invention, the updated fixed-length average template waveform is expanded and scaled according to the characteristics of the biological signal and matched to the actual heartbeat interval to create an average waveform for one heartbeat, and the average waveform for each heartbeat is generated. Since the continuous average waveform is created by connecting the heartbeats, it becomes possible to obtain a heartbeat-synchronized continuous average waveform with a smooth joint synchronized with the heartbeat.

  In addition, according to the present invention, the fluctuation component is extracted based on the differential waveform, and the fluctuation component is added to the heartbeat-synchronous continuous average waveform, so that an estimated waveform of the biological signal that follows the fluctuation component in each heartbeat is created. Will be able to.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. In addition, the biological signal estimation method according to each embodiment of the present invention is implemented mainly in an electronic circuit such as a microcomputer.

  FIG. 1 shows a flowchart of a biological signal estimation method according to the first embodiment.

  As shown in FIG. 1, the biological signal estimation method according to the first embodiment includes a step 101 for cutting out a waveform from a measurement waveform of a biological signal to create a cut waveform, and correcting the end point of the cut waveform. Step 102 for creating a waveform, Step 103 for updating a fixed-length average template waveform based on the correction waveform, and Step 104 for creating an average waveform for one heartbeat from the updated fixed-length average template waveform. . Hereinafter, taking the case where the biological signal is an electrocardiogram as an example, step 101 to step 104 will be described in order.

(1) Step 101
As shown in FIG. 2 (a), the cut-out waveforms (a1), (a2), and (a3) are created from the measurement waveform (a) of the biological signal accurately in accordance with the actual heartbeat interval for each heartbeat. .

(2) Step 102
As shown in Fig. 2 (b), the cut waveforms (a1), (a2), (a3) are set so that the straight lines connecting the end points of the cut waveforms (a1), (a2), (a3) are at the same level. The corrected waveform (b) ((b1), (b2), (b3)) is created.

(3) Step 103
As shown in FIG. 3, the actual heartbeat interval of the correction waveform (b) ((b1), (b2), (b3)) is divided into a systole and a diastole. As this classification method, there is a method of measuring an actual systole and diastole from a PCG (heart sound diagram) waveform or the like. As an example, a method of estimating from a heartbeat interval is shown. The systole and diastole are estimated by the following formulas (1-a), (1-b), and (2).

(Systole) in = α × (actual heartbeat interval) (1-a)
Or
(Systole) in = α x (average heartbeat interval) (1-b)
(Diastolic phase) in = (actual heartbeat interval)-(systolic phase) in (2)
Where α is a coefficient determined by the actual heartbeat interval or the average heartbeat interval, the unit of (actual heartbeat interval) is seconds, and (systole) in is the contraction of the actual heartbeat estimated from the actual heartbeat interval or the average heartbeat interval. The period (seconds) and (diastolic period) in represent the diastole (seconds) of the actual heart rate obtained by subtracting the systole estimated from the actual heartbeat interval.

  Similarly, a fixed-length average template that is a generator for generating an average waveform is also divided into a systole and a diastole. As this classification method, there is a method of dividing the fixed length average template into the systole and the diastole by averaging the ratio between the actual systole and the diastole using a PCG waveform or the like. Indicates. The systolic and diastolic phases are estimated by the following equations (3-a), (3-b), and (4) (see FIG. 4).

(Systole) tmp = β x (actual heartbeat interval) (3-a)
Or
(Systole) tmp = β x (average heart rate interval) (3-b)
(Diastolic phase) tmp = (fixed length average template length)-(systolic phase) tmp (4)
Where β is a coefficient determined by the actual heartbeat interval or average heartbeat interval, the unit of (fixed length average template length) is seconds, and (systole) tmp is a fixed length estimated from the actual heartbeat interval or average heartbeat interval. Average template systole (seconds), (diastolic phase) tmp represents the fixed-length average template diastole (seconds) obtained by subtracting the systole estimated from the fixed-length average template length.

  Next, the systolic part and the diastolic part of the corrected waveform after the end point correction in step 102 are expanded and contracted according to the length of the fixed period average template and the length of the diastolic period, respectively. The expanded scale waveform is created and added to the fixed-length average template waveform before the update, thereby updating the fixed-length average template waveform after the update.

  As an example, FIG. 5 shows a state where the correction waveform is expanded and reduced in accordance with the length of the fixed-length average template. The systolic and diastolic phases of the corrected waveform (b) ((b1), (b2), (b3)) in FIG. 5 are estimated from the actual heartbeat interval, and the fixed-length average template (e) (( The systolic and diastolic phases of e3)) are estimated from the average heart rate interval.

  The superiority of the biological signal estimation method according to the first embodiment is that the waveform extracted for each heartbeat interval is not simply stretched and scaled according to the fixed-length average template. The difference between the systolic and diastolic fluctuations) is divided into the systolic and diastolic phases according to the difference between the systolic and diastolic phases. This solves the problem of the distortion of the second half q of the average waveform shown in FIG. 22 and shows the effect of taking advantage of the characteristics of the living body with an electrocardiogram as an example in FIG.

  FIG. 6 shows a case where a simple expansion / contraction is performed on the measurement waveform (a) of the biological signal when the constant heartbeat interval is suddenly shortened, and the expansion / contraction is performed according to the characteristics of the biological body. Represents a comparison of cases. In FIG. 6, a fixed-length average template waveform (d) before update is a waveform resulting from learning a waveform with a long heartbeat interval, and an expanded scale waveform (c) is added to the fixed-length average template waveform (d) before update. Is added to obtain an updated fixed-length average template waveform (e). Comparing the updated fixed-length average template waveform (e) when performing a simple expansion scale and performing an expansion scale according to the characteristics of the living body, the advantage of the expansion scale using the characteristics of the living body I understand the sex.

  Another example of the method for updating the fixed-length average template waveform is a weighted average method expressed by the following formula (5). However, the weighted average weight is not constant, and it is preferable to adapt the weight according to the quality of the cut-out waveform and the fluctuation of the heartbeat interval, so that an average waveform that is resistant to noise and follows the fluctuation of the heartbeat can be realized. become.

AveWave [i] = α × AftInput [i] + (1-α) × AveWave [i] (5)
Here, i = 0 to AveSize, α is a weighted average weight for each heartbeat (= 0 to 1), AveSize is the length of the fixed length average template, and AveWave [i] is the i th of the fixed length average template waveform. The value AftInput [i] represents the i-th value of the expanded scale waveform after end point correction.

(4) Step 104
Next, as shown in FIG. 7, the fixed-length average template waveform (e) after the update is expanded and scaled according to the characteristics of the living body and in accordance with the average heartbeat interval to create an average waveform (f) for one heartbeat. . As with the fixed-length average template, the average waveform interval can be divided into a systole and a diastole by averaging the ratio between the actual systole and the diastole. As an example, a method of estimating from a heartbeat interval is shown. The systolic and diastolic phases are estimated by the following equations (6) and (7).

(Systole) ave = γ × (average heartbeat interval) (6)
(Diastolic) ave = (average heart rate interval)-(systolic) ave (7)
Where γ is a coefficient determined by the average heart rate interval, (systole) ave is the systolic phase (seconds) of the average waveform estimated from the average heart rate interval, and (diastolic phase) ave is the systole estimated from the average heart rate interval It becomes the expansion period (second) of the average waveform obtained by subtraction.

  FIG. 8 is a flowchart of the biological signal estimation method according to the second embodiment.

  As shown in FIG. 8, the biological signal estimation method according to the second embodiment includes a step 201 in which a waveform is cut out from a measured waveform of a biological signal to create a cut-out waveform, and an end point correction of the cut-out waveform is performed. Step 202 for creating a waveform, Step 203 for updating the fixed-length average template waveform based on the correction waveform, Step 204 for creating an average waveform for one heartbeat from the updated fixed-length average template waveform, and each heart rate The step 205 generates a heartbeat-synchronized continuous average waveform by connecting the average waveforms. Hereinafter, step 201 to step 205 will be described in order by taking the case where the biological signal is an electrocardiogram as an example.

(1) Step 201 to Step 203
Processing similar to that in steps 101 to 103 in FIG. 1 in the first embodiment described above is performed. Therefore, explanation is omitted.

(2) Step 204
In step 104 of FIG. 1 in the first embodiment described above, the updated fixed-length average template waveform (e) is expanded and scaled to match the average heart rate interval according to the characteristics of the living body, and the average waveform for one heart rate (f However, in step 204 of FIG. 8, the fixed-length average template waveform (e) after the update is expanded and scaled according to the characteristics of the living body and matched to the actual heartbeat interval, and the average for one heartbeat is obtained. Create waveform (f).

(3) Step 205
The average waveform (f) for each heart rate is connected to create a continuous heartbeat-synchronized average waveform (g). As a result, as shown in FIG. 9, the average waveform (f1) synchronized with the actual heartbeat and having a smooth joint is obtained with respect to the measurement waveform (a) including the cut-out waveforms (a1), (a2), and (a3). ), (f2), (f3) to obtain a heartbeat-synchronized continuous average waveform (g). In FIG. 9, (e3) represents a fixed-length average template waveform after updating the average waveform (f3).

  The heartbeat-synchronized continuous average waveform (g) as described above is output by replacing the waveform of each heartbeat with the average waveform (f), but for an indefinite waveform (arrhythmia, abnormal waveform, etc.), the average waveform (f ) Is not assigned, but an indefinite waveform is output as it is.

  Here, as an example of a method for determining an indefinite waveform, a method for determining based on the fluctuation of the heartbeat interval and a method for determining whether or not the waveform is indefinite by comparing with an average waveform are shown.

The former determination method determines that the heartbeat interval is an indefinite waveform when the heartbeat interval fluctuates more than the fluctuation of the heartbeat interval, which is a characteristic of the living body, and a sudden change is seen in the heartbeat interval as shown in FIG. Arrhythmia such as extrasystole can be effectively determined. Specifically, the fluctuation rate of the heartbeat interval is expressed by the following formula (8),
(Variable rate of heartbeat interval) = | (actual heartbeat interval)-(average heartbeat interval) | / (average heartbeat interval) x 100 (8)
For example, when the fluctuation rate of the heartbeat interval is larger than 20, it is determined as an indefinite waveform.

  The latter determination method is used when the heartbeat interval does not fluctuate very much. As shown in FIG. 11, the input waveform (corrected waveform (b)) to be determined for the indefinite waveform (x) is changed to the actual heartbeat interval. The difference waveform (h) is created by subtracting the heartbeat synchronization continuous average waveform (g) developed together, and only the band component of the signal is extracted from this difference waveform (h), the area of the extracted waveform (i), Whether the waveform is indefinite (x) is determined from the height.

  For heartbeats determined to be indefinite waveforms (x), end point correction is performed so that the straight lines connecting the end points of the heartbeats are at the same level, and the corrected waveform (b) after end point correction is converted into a continuous heart rate synchronization average. As shown in FIG. 12, by synthesizing the waveform (g), the heartbeat synchronization continuous average waveform (g) that gradually changes over time while ensuring the continuity of the joint with the previous and subsequent average waveforms is added. It is possible to know the presence of the indefinite waveform (x), and no trouble occurs in the diagnosis of arrhythmia in the electrocardiogram.

  FIG. 13: shows the flowchart of the part performed following the biological signal estimation method which concerns on 2nd Embodiment mentioned above in the biological signal estimation method which concerns on 3rd Embodiment.

  As shown in FIG. 13, this additional part includes step 301 for subtracting the heartbeat synchronization continuous average waveform from the measurement waveform of the biological signal to create a differential waveform, and time-dividing the differential waveform according to the characteristics of the biological signal. A step 302 for creating a fluctuation extraction waveform using a digital filter and a step 303 for adding a fluctuation extraction waveform to a heartbeat synchronization continuous average waveform to create an estimated waveform of a biological signal following the fluctuation component are provided. A process for estimating a waveform following the heartbeat fluctuation is performed by extracting the beat fluctuation that cannot be captured by the continuous average waveform and adding it to the heartbeat synchronization continuous average waveform. Hereinafter, step 301 to step 303 will be described in order.

(1) Step 301
As shown in FIG. 14, a differential waveform (j) is created by subtracting the heartbeat synchronization continuous average waveform (g) from the measurement waveform (a) of the biological signal. The difference waveform (j) includes a fluctuation component (beat fluctuation) and noise (high frequency and low frequency noise) for each heartbeat.

(2) Step 302
When removing the high frequency noise from the differential waveform (j), it is necessary to properly separate the fluctuation component for each heartbeat and the high frequency noise. Therefore, taking advantage of the characteristics of the biological signal (a), for example, in the case of an electrocardiogram, the differential waveform (j) is time-divided into a QRS region that is a relatively high frequency component and other regions, and a digital filter adapted to each division To use. As a result, it is possible to efficiently remove high-frequency noise in a region other than the QRS region while ensuring a fluctuation component in the QRS region.

  Here, there is a moving average method as a method of removing high frequency noise. The digital filter based on the moving average method is expressed by, for example, the following formulas (9) and (10).

X (i) = (1/15) × Σ {x (i + j)} | j = −7 to +7 (9)
y (i) = X (i) x (1-M (i)) + x (i) x M (i) (10)
Where i is the sampling number, x (i) is the i-th value of the difference waveform (j), X (i) is the i-th 15-point moving average value of the difference waveform (j), and M (i) is the difference The i-th synthesis ratio (0 to 1) and y (i) of the waveform (j) represent the i-th value of the waveform obtained by removing high-frequency noise from the differential waveform (j).

  FIG. 15 shows an input waveform (measurement waveform (a) of a biological signal), a difference waveform (j), and a high-frequency noise removal waveform (k) of the digital filter. As shown in FIG. 15, in the QRS region in the high-frequency noise removal waveform (k), the input waveform (a) is an output with an emphasis on the synthesis ratio also in FIG.

  Next, low frequency noise is removed from the high frequency noise removal waveform (k). In the case of an electrocardiogram, the presence of a T wave, which is a low frequency component, becomes a problem when removing low-frequency noise, but only the fluctuation of the T-wave is a problem in the high-frequency noise removal waveform (k). Thus, a low-pass blocking filter can be assembled with a higher cutoff frequency than in the prior art. For example, when a linear phase filter is used as the low-frequency block filter, the fluctuation extraction waveform (l) after low-frequency noise removal is as shown in FIG.

(3) Step 303
The fluctuation extraction waveform (l) is added to the heartbeat-synchronized continuous average waveform (g). As a result, an estimated waveform (m) following the fluctuation component as shown in FIG. 17 is obtained, and a biological signal is estimated from this waveform (m).

  As described above, according to the biological signal estimation method according to the present embodiment, it is possible to create an average waveform for one heartbeat of a biological signal whose heartbeat interval fluctuates, and the temporal change of the entire beat Can be confirmed. In addition, the heartbeat-synchronized continuous average waveform makes it possible to confirm changes over time of the entire beat while grasping each heartbeat interval. In addition, since the beat fluctuation is extracted by a digital filter that matches the characteristics of the biological signal, the effects of distortion on the signal components can be reduced, and low-frequency noise can be removed at a higher cutoff frequency, eliminating noise. Can be performed efficiently.

  FIG. 18 shows an input waveform (measurement waveform (a) of a biological signal) of an electrocardiogram and a heartbeat synchronization continuous average waveform (g) according to the present invention.

  FIG. 19 shows a volume pulse wave input waveform (measurement waveform (a) of biological signal) by impedance measurement and a heartbeat synchronization continuous average waveform (g) according to the present invention.

  Further, FIG. 20 shows an input method (r) of an electrocardiogram and a noise removal waveform (p) obtained by a conventional method of separating a signal component and a noise component according to a frequency domain by an estimation method adapted to heartbeat variability. The estimated waveform (g) from which noise has been removed is shown. It can be seen from the estimated waveform (g) of the biological signal that the QRS level is maintained and noise is efficiently removed.

It is a flowchart of the estimation method of the biomedical signal which concerns on 1st Embodiment of this invention. FIG. 3 is a waveform diagram of an electrocardiogram measurement waveform (a) and its correction waveform (b). It is explanatory drawing of the systole and diastole in correction waveform (b). It is explanatory drawing of the other ratio setting method of a systole and a diastole in a fixed length average template. It is explanatory drawing when developing a correction | amendment waveform (b) to a fixed length average template. FIG. 5 is a comparative explanatory view of a fixed-length average template waveform (e) after update when a simple extension scale is performed and when an extension scale is applied according to the characteristics of a living body. It is explanatory drawing of the fixed-length average template waveform (e) and average waveform (f) after an update. It is a flowchart of the estimation method of the biomedical signal which concerns on 2nd Embodiment of this invention. FIG. 6 is a waveform diagram of an electrocardiogram measurement waveform (a) and a heartbeat synchronization continuous average waveform (g). It is explanatory drawing of an indefinite waveform (x). It is explanatory drawing of the determination method of indefinite waveform (x). FIG. 6 is a waveform diagram of a heartbeat synchronization continuous average waveform (g) including an indefinite waveform (x). It is a flowchart of the principal part of the biological signal estimation method which concerns on 3rd Embodiment of this invention. It is explanatory drawing of the preparation method of a difference waveform (j). FIG. 6 is a waveform diagram of an input waveform (electrocardiogram measurement waveform (a)), a differential waveform (j), and a high-frequency noise removal waveform (k). FIG. 6 is a waveform diagram of a fluctuation extraction waveform (l) after low-frequency noise removal. FIG. 6 is a waveform diagram of an estimated waveform (m) of a biological signal that follows a fluctuation component. FIG. 4 is a waveform diagram of an electrocardiogram (a) and a heartbeat synchronization continuous average waveform (g). It is a waveform diagram of volume pulse wave (a) and heartbeat synchronization continuous average waveform (g) by impedance measurement. FIG. 6 is a waveform diagram of an electrocardiogram (r), an estimated waveform (p) by a conventional method, and an estimated waveform (g) according to the present invention. It is explanatory drawing of the conventional average method. It is explanatory drawing of the problem of the conventional average method.

Explanation of symbols

101, 201 Cutout waveform creation step 102, 202 Endpoint correction step 103, 203 Fixed length average template waveform update step 104, 204 Average waveform creation step 205 Heart rate synchronization continuous average waveform creation step 301 Difference waveform creation step 302 Fluctuation extraction waveform creation step 302 303 Heart Rate Synchronization Continuous Average Waveform Generation Step Including Variation Components

Claims (3)

Create a cut-out waveform for each heartbeat from the measured waveform of the biological signal,
Correcting the extracted waveform so that the end points of all of the generated extracted waveforms are at the same 0 level, creating a corrected waveform;
An expanded scale waveform is created by expanding and contracting the correction waveform according to the characteristics of the biological signal and matching the length of the fixed-length average template,
The expanded scale waveform is added to the fixed-length average template waveform before update and averaged to create a fixed-length average template waveform after update,
A biosignal estimation method comprising: generating an average waveform for one heartbeat by expanding and reducing the updated fixed-length average template waveform according to the characteristics of the biosignal and matching an average heartbeat interval. Create a cut-out waveform for each heartbeat from the measured waveform of the biological signal,
Correcting the extracted waveform so that the end points of all of the generated extracted waveforms are at the same 0 level, creating a corrected waveform;
An expanded scale waveform is created by expanding and contracting the correction waveform according to the characteristics of the biological signal and matching the length of the fixed-length average template,
The expanded scale waveform is added to the fixed-length average template waveform before update and averaged to create a fixed-length average template waveform after update,
The updated fixed-length average template waveform is expanded and reduced in accordance with the characteristics of the biological signal and in accordance with the actual heartbeat interval to create an average waveform for one heartbeat,
A method for estimating a biomedical signal, comprising creating a heartbeat-synchronized continuous average waveform by connecting average waveforms for each heartbeat. Subtract the heart rate synchronized continuous average waveform from the measurement waveform of the biological signal to create a differential waveform,
The differential waveform is time-divided for each heartbeat according to the characteristics of the biological signal, and low-frequency and high-frequency noises are removed by a digital filter suitable for each segment to extract a fluctuation component for each heartbeat,
The biological signal estimation method according to claim 2, wherein the fluctuation component is added to the heartbeat-synchronized continuous average waveform to create an estimated waveform that also follows the fluctuation component in each heartbeat.