JP4196209B2 - Signal processing method and pulse photometer using the same - Google Patents

Description

  The present invention relates to an improvement in an apparatus for measuring biological information using photoelectric pulse waves measured at a plurality of wavelengths, particularly a pulse photometer used for diagnosis of a respiratory / circulatory system.

In medical practice, pulse photometers for measuring the concentration of light-absorbing substances in blood, measuring oxygen saturation SpO2, measuring the concentration of special hemoglobin such as carbon monoxide hemoglobin and Met hemoglobin, and measuring the concentration of injected dye are known. ing.
The measuring device for oxygen saturation SpO2 is called a pulse oximeter.

The principle of the pulse oximeter is that the concentration of the target substance is determined from the pulse wave obtained by transmitting light of multiple wavelengths with different absorption characteristics to the target substance through the living tissue and continuously measuring the amount of the transmitted light. It is what you want.
In the patent No. 32701717 (Patent Document 1), the applicant of the present invention uses the two pulse waves obtained from the transmitted light by irradiating light of two different wavelengths to the living tissue as the vertical axis and the horizontal axis. It has been proposed to draw a graph, find the regression line, and obtain the oxygen saturation or light-absorbing substance concentration in arterial blood based on the slope of the regression line.
According to the present invention, measurement accuracy can be improved and power consumption can be reduced.
However, in order to obtain a regression line or its slope using a large amount of sampling data for pulse waves of each wavelength, a lot of calculation processing is still required.

Japanese Patent No. 3,270,917 (Claims 1, 2, 2, 4)

Next, the principle will be described by taking as an example a pulse oximeter that measures arterial oxygen saturation, which is the premise of the present invention.
This technique is not limited to pulse oximeters, but measures blood absorption substances such as special hemoglobin (carbon monoxide hemoglobin, Met hemoglobin, etc.) and dye injected into blood using the principle of pulse photometry. Applicable to equipment (pulse photometer).

Figure 1 shows a block diagram of the device. A pulse oximeter will be described as an example.
Usually, the measurement is performed by attaching the probe 1 to the part to be measured.
The arithmetic processing / control unit 8 generates the light emission timings of the generating means (LEDs) 2 for generating light of different wavelengths, the red light emitting diode (hereinafter R-LED) and the infrared light emitting diode (hereinafter IR-LED), The light emitting diode drive unit 4 alternately emits R-LEDs and IR-LEDs in the probe.
The light emitted by the LEDs 2 passes through the part to be measured on which the probe mounting part is mounted, and is received by a light receiving part PD (Photo Diode) 3 in the probe.
A signal converted from light to electricity in the light receiving unit PD3 is converted into a voltage in the input unit 5.
In this received light signal, a component that reflects the optical characteristics of pulsation at the measured portion appears as an AC component.
The signal obtained from the light receiving unit PD3 is separated and demodulated by the demodulation circuit 6 into the wavelengths of infrared light (IR) and red light (R), and the digitized signals are respectively processed by the AD conversion circuit 7. For example, the oxygen saturation SpO2 is calculated by giving to the control unit 8 and performing arithmetic processing.

Signals corresponding to infrared light (IR) and red light (R) digitized by the AD conversion circuit 7 form respective measured pulse wave data.
The absorbance A of the transmitted light of the tissue is expressed by the following equation, where Ab is the light attenuation of blood and At is the light attenuation of other tissues.

Iout: transmitted light intensity Iin: incident light intensity E: blood extinction coefficient C: blood hemoglobin concentration D: blood thickness When the blood thickness becomes D + ΔD due to the blood artery, equation (1) is as follows: become.

(2) Formula-(1) Formula

For each of the wavelengths λ1 and λ2, equation (3) is as follows.

Subscripts 1 and 2 indicate that they are related to wavelengths λ1 and λ2, respectively. (4) ÷ (5) = Φ _S

Then, Δln (Iout1) and Δln (Iout2) are measured to obtain Φ _S and the oxygen saturation SpO2 is calculated.

As required.

In addition to the component that reflects pulsation in Δln (Iout1) and Δln (Iout2), if the influence of body motion is superimposed, the absorbance at each wavelength is

En: Noise absorption coefficient of the noise source Cn: Concentration of the noise source Dn: A waveform (pulse wave) that is expressed by a change in the thickness of the noise source and reflects the pulsation component cannot be obtained, and a signal superimposed with noise is observed.
In this case, the absorbance ratio Φ ′ is

And does not match the arterial oxygen saturation.
FIG. 2A shows a waveform obtained by taking a logarithm of transmitted light intensity data for each predetermined time at each wavelength and performing a process of making the average value zero or a low-pass removing filter process.
Using the waveform data measured at approximately the same time in FIG. 2A (which is sufficiently shorter than the frequency component of the pulse wave), the horizontal axis represents the infrared light amplitude and the vertical axis represents the red light amplitude. The graph is shown in FIG.
If the observation data reflects only the pulse wave component, this graph is almost on a straight line, and the slope of this straight line represents the absorbance ratio Φs.
However, when noise is superimposed as shown in FIG. 3 (a), the noise is different from the pulse wave component, resulting in a combined result of the noise absorbance ratio and the arterial absorbance ratio (FIG. 3 (b)).

As described above, if noise is included in the measured pulse wave data, it is not possible to accurately measure the absorbance ratio, so it is necessary to remove the noise.
As a method for removing this noise, a frequency analysis method and an independent component analysis are conventionally known.

However, the frequency analysis method has a problem that it is difficult to extract the signal component when the fundamental wave of the signal component of the measured pulse wave data or its harmonic and the frequency of the noise overlap.
In addition, the independent component analysis method has a problem that separation is difficult when a plurality of noise sources are present because the measurement signal cannot be decomposed into independent components higher than the measured pulse wave data.

An object (object) of the present invention is to easily remove noise from an observation signal on which noise is superimposed, extract pulse wave data, and obtain an absorbance ratio.
Another object of the present invention is to provide a pulse photometer using pulse wave data from which noise is removed.

In order to solve the above-mentioned problem, the measured pulse wave data of each wavelength obtained by irradiating a living tissue with light of two different wavelengths and converting the light of each wavelength transmitted or reflected into an electrical signal is previously transmitted. Processing steps,
Dividing the measured pulse wave data of each wavelength into a plurality of frequency bands, and obtaining an absorbance ratio for each frequency band; and
The signal processing of the measured pulse wave data is executed in the step of determining the presence or absence of noise in the measured pulse wave data based on the presence or absence of the absorbance ratio for each frequency band. (Claim 1)

  In addition, the determination of the presence or absence of the same absorbance ratio for each frequency band is performed when the absorbance ratio is within a predetermined range in all the divided frequency bands, or in a frequency band in which the power after the division is dominant. When the absorbance ratio is within a predetermined range, signal processing of the measured pulse wave data is performed. (Claim 2)

（ここでSは信号成分であり、Nはノイズ成分である。s１は第１の電気信号、s２は第２の電気信号、φ＝tan ⁻¹ Φ。Φは吸光度比、θは−φからπ/２−φの範囲で変化）により実行し、θは信号成分Ｓのノルムを最小とする。（請求項３）
また、請求項２に記載の方法で求めたノイズの混入が無いと判定された吸光度比を、前記既知の吸光度比として用いる。（請求項４） Further, it is a method of removing noise from measured pulse wave data of each wavelength obtained by irradiating a living tissue with light of two different wavelengths and transmitting or reflecting light of each wavelength converted into an electrical signal,
For the measured pulse wave data of each wavelength, the measured pulse wave data is whitened by affine transformation using a known or previously obtained absorbance ratio.

(Where S is a signal component, N is a noise component, s1 is a first electrical signal, s2 is a second electrical signal, φ = tan ⁻¹ Φ, Φ is an absorbance ratio, and θ is from −φ. (change in the range of π / 2−φ), and θ minimizes the norm of the signal component S. (Claim 3)
In addition, an absorbance ratio determined by the method according to claim 2 to determine that there is no noise mixing is used as the known absorbance ratio. (Claim 4)

A step of pre-processing the measurement pulse wave data of each wavelength obtained by irradiating a living tissue with light of two different wavelengths and converting the light of each wavelength transmitted or reflected into an electrical signal; Dividing the measured pulse wave data of each wavelength into a plurality of frequency bands and whitening each frequency band, and separating the pulse wave signal and the noise signal. (Reference example)

Further, independent component analysis is used for the whitening treatment. (Reference example)
Further, measurement of each wavelength obtained by irradiating a living tissue with light of two different wavelengths and converting the light of each wavelength transmitted or reflected by the method according to any one of claims 3 and 4 into an electrical signal. The S / N ratio for each frequency component is obtained by frequency analysis of the pulse wave signal and the noise signal separated by processing the pulse wave data. (Claim 5 )

Further, measurement of each wavelength obtained by irradiating a living tissue with light of two different wavelengths and converting the light of each wavelength transmitted or reflected by the method according to any one of claims 3 to 5 into an electrical signal. A pulse photometer that processes pulse wave data and separates noise and signal components. (Claim 6 )
A pulse photometer that obtains a pulse waveform from the separated signal components. (Claim 7 )
A pulse photometer that calculates a pulse rate from pulse detection or frequency analysis from the separated signal components. (Claim 8)

The inventions according to claims 1 to 5 are inventions of a method in which a procedure (step) is limited. However, it is also possible to make an apparatus invention as a means for achieving the function of each procedure (step). .

In the present invention, by using a known absorbance ratio in advance, it is possible to separate a pulse wave signal and a noise signal and obtain a pulse wave signal with reduced noise.
In addition, the affine transformation is used to use the known absorbance ratio, the observation signal is whitened, the pulse wave signal and the noise signal are separated, and the pulse wave signal with reduced noise can be obtained. is there.
In addition, it is possible to determine whether or not noise is mixed in the observation signal by separating the pulse wave signal frequency band into a plurality of ranges, obtaining an absorbance ratio principal component for each frequency band, and using the identity. it can.
In addition, affine transformation using a known absorbance ratio can determine the first rotation using a known absorbance ratio and a coefficient suitable for removing noise, and obtain a pulse wave signal with reduced noise. It is possible.
Further, the pulse rate can be determined from pulse detection or frequency analysis from the pulse wave signal with reduced noise.
Also, by dividing the pulse wave signal frequency band into multiple ranges, the number of signal sources is reduced and whitening is performed for each frequency range, so that the pulse wave signal absorbance ratio and the noise signal absorbance ratio are Can be requested.
Further, since the pulse wave signal and the noise signal are obtained separately, an S / N ratio for each frequency is obtained and can be used as a signal evaluation index.

A method for separating a signal component and a noise component using a known absorbance ratio First, light received by a light receiving part PD (Photo Diode) 3 in the probe having the configuration shown in FIG. The
In this received light signal, a component that reflects the optical characteristics of pulsation at the measured portion appears as an AC component.
The signal obtained from the light receiving unit PD3 is separated and demodulated into wavelengths of infrared light (IR) and red light (R) by the demodulation circuit 6, and digitized by the AD conversion circuit 7, respectively.

This infrared light observation signal data
IR = {IRi; i = 0,1,2, ...} (11)
Red light observation signal data
R = {Ri; i = 0,1,2, ...} (12)
The pulse wave component of the observed signal is S = {Si; i = 0,1,2, ...} (13)
N = {Ni; i = 0,1,2, ...} (14)
And
As shown in FIG. 4, when the slope of the pulse wave component is φ and the slope of the noise component is ψ, the obtained observation signals (IRi, Ri)

Is expressed as a composite vector Xi of signal component: Si and noise component: Ni.
Therefore,

If the inverse matrix A of W is applied to the observation data, the signal component: S and the noise component: N can be separated.
This indicates that the signal (S) component is converted to the horizontal axis and the noise (N) component is converted to the vertical axis, and when a conversion matrix A for realizing whitening is obtained,

 It becomes. Therefore,

Get.
The absorbance ratio Φ is the slope of the signal component.
The relationship between the slope φ and the absorbance ratio Φ is Φ = tanφ.
If it is within a time sufficiently shorter than the pulse rate, the previous absorbance (eg, 10 msec before) or the past absorbance determined to be free of noise can be treated as a known value.
It is also possible to use φ candidates obtained by a method as described later.
Next, θ is
θ = ψ-φ (20)
Since ψ is unknown, the value of θ is changed in the range of −φ to π / 2−φ, and the norm in the horizontal (S) axis direction is determined as the minimum angle.
When the observation signal is processed with the obtained transformation matrix A using φ and θ, a signal component is obtained on the horizontal axis and a noise component is obtained on the vertical axis.
Specifically, the matrix transformation expressed by the equation (19) is performed from the first coordinate system in which two types of observation signals IR and R shown in FIG. 3B are selected as coordinate axes to FIG. 3C. This means coordinate conversion to the second coordinate system in which the signal component and noise component shown are selected as coordinate axes.
FIG. 4 shows observed values (IRi, Ri) at a certain time point t = ti, and a vector Xi representing the observed value is a combined vector of the signal component vector Si and the noise component vector Ni. In executing the above coordinate conversion, first, an operation of matching the signal component vector Si with the horizontal axis in the coordinate system after conversion is performed. Specifically, an operation of rotating the vector Si in FIG. 4 clockwise by an angle φ is performed. The value of φ is obtained from the above equation using the known value of Φ.
In equation (19), θ is a coefficient corresponding to an operation for causing the noise component vector Ni to coincide with the vertical axis in the coordinate system after conversion. The waveform shown in FIG. 3 (b) is a set of the observed values described above, and naturally the value of θ differs for each observed value if the value of φ is fixed. In order to determine θ as a conversion coefficient as a single value, θ is changed in the range of −φ to π / 2−φ, and a change in the variation degree (norm) of the value in the horizontal axis (signal component) direction is examined. The final value is determined as the value with the smallest norm.
FIG. 3C shows a graph obtained by processing the transformation matrix A in FIG.
Further, FIG. 3D shows a waveform obtained by separating noise by this processing.
FIG. 5 shows spectra of infrared light and red light obtained by frequency analysis before processing, and FIG. 6 shows spectra obtained by frequency analysis of corresponding signal components and noise components after processing.

・ Pulse rate calculation method By calculating the peak-peak interval Ts (Fig. 3 (d)) from the signal waveform extracted by the above conversion processing, the pulse rate = 1 / Ts x 60 [beats / min] is easily obtained. be able to.
Alternatively, the spectrum (FIG. 6) obtained by frequency analysis from the signal after processing compared to the spectrum obtained by frequency analysis before processing (FIG. 5) makes the fundamental frequency fs of the pulse wave clear. The number can be calculated as fs × 60 [beats / min].

-Method for judging whether noise is mixed in the observation signal The observation signal is separated for each predetermined frequency range.
Separation may be performed for each continuous frequency range or for each synthesis of discrete ranges.
For each frequency range, the horizontal axis (or vertical axis) takes the infrared light component, and the axis orthogonal to it takes the red light component.
FIG. 7 is a graph in which the horizontal axis represents the infrared light component and the vertical axis represents the red light component.
A slope corresponding to the principal component is obtained for the entire pulse wave frequency band and for each frequency range after separation, and it is determined whether a single signal or a plurality of signals are combined based on the identity of the values.
Specifically, as a method of determining a single signal, the entire pulse wave frequency band and the separated frequency range are all within a predetermined range, or the value of the dominant power range and the pulse wave frequency band The overall value may be within a predetermined range.
As a specific example, FIG. 7 and FIG. 8 show the case where the entire pulse wave frequency band and the range from 0.5 Hz to 6 Hz are divided every 0.5 Hz, and the inclination of the main component is shown in the title part.
FIG. 7 shows different values in the range showing a large power of 0.5-1 Hz: 0.54 3.5-4.0 Hz: 0.84 4.0-4.5: 0.84, and it can be determined that the synthesis of multiple signals was observed.
FIG. 8 shows the same 0.50 value in the range of 1-1.5 Hz, 2-2.5 Hz, and 2.5-3 Hz indicating a large power, and is determined as a single signal.

-Oxygen saturation candidate (absorbance ratio) calculation The pulse wave signal frequency band is separated into a plurality of ranges in the same manner as the determination of the noise contamination. Independent component analysis is performed for each separated frequency range.
If the observed signal vertical vector is X and the transpose is T, from Equations (11) and (12),

  The variance-covariance matrix H of the observed signal vector X is

Let the eigenvalue diagonal matrix be Λ and the eigenvector matrix be Γ.
The independent component Y can be represented by U as the transformation matrix.

And is decomposed into independent components.
If a signal component exists in the separated frequency range, it can be converted into a signal component-noise component plane as shown in FIG.
When the eigenvector matrix is transformed by the inverse matrix U ⁻¹ of the transformation matrix U, the vector is projected onto the IR-R plane, and the slope of the vector representing the signal component is obtained as Φs, and the slope of the vector representing the noise component is obtained as Φn.
FIG. 10 shows the result of obtaining the slope for each frequency range in FIG. 8 with respect to the observation signal in FIG. 2, and it can be seen that the absorbance ratio 1 is close to each frequency (0.5 ± 0.02).
The result of processing FIG. 7 is FIG. 9, and the result showing Φs in the range including fs in FIG. 6 is obtained.
Oxygen saturation SpO2 is a function of Φs
SpO2 = f (Φs)
As obtained.

Next, the operation of the processing unit of the apparatus block shown in FIG.
FIG. 11 is a flowchart showing a processing flow.
S1: When measurement starts, preprocess digital signals observed with red and infrared light. Here, logarithmic calculation based on Lambert-Beer's law, pulse wave band filter processing, and the like are performed.
S2: Each signal is divided into a predetermined frequency range (for example, in increments of 0.5 Hz) using a bandpass filter, and an absorbance ratio (inclination of the main component on the IR-R plane) for each frequency range is obtained.
S3: The identity of the absorbance ratio obtained for each frequency range is determined, and a pulse wave absorbance ratio candidate is obtained. There may be a plurality of candidates.
S4: Determine whether the signal source is single or plural from the result of S3.
S5: In the case of a single signal, the oxygen saturation is obtained from the absorbance ratio, and the basic period is obtained from the processing result of S1, and the pulse rate is calculated.
S6: Absorbance ratio candidate obtained in S3 or the result of the absorbance ratio corresponding to the previous SpO2 value (sufficiently early compared to the pulse rate: for example 10 msec before) or past absorbance determined to be free of noise Affine transformation is performed using the ratio. The transformation matrix is rotated by φ (value corresponding to the absorbance ratio) using equation (19), and θ is obtained as an angle at which the norm value in the φ axis direction is minimized.
S7: The basic period is obtained by the period or frequency analysis of the signal obtained in the φ direction of S6, and the pulse rate is obtained. In addition, the S / N for each frequency is obtained and used as one index for evaluating the reliability of the absorbance ratio.
S8: Independent component analysis is performed using the time domain data of each frequency range divided in S2.
S9: Obtain the slope of the eigenvector obtained in S8 on the IR-R plane, determine the identity between each frequency range, and obtain the absorbance ratio candidate. There may be a plurality of candidates.
S10: The output of S3, S5, S7, and S9 is compared with the past absorbance ratio / bandwidth, and the current oxygen saturation is determined from the common terms, continuity, and reliability (S7 output). After averaging as necessary, the oxygen saturation and pulse rate are output.
S11: When continuing the measurement, return to S1 to repeat the process, and when not continuing, end the measurement.

It is a block diagram which shows the structure of the pulse oximeter of this invention. (a) is the waveform which performed the process which makes the average value in each measurement wavelength zero, or the low-pass removal filter process. (b) is a graph with the horizontal axis representing infrared light amplitude and the vertical axis representing red light amplitude. (a) is an observation (measurement) waveform on which noise is superimposed. (b) is a graph with the horizontal axis representing infrared light amplitude and the vertical axis representing red light amplitude. (c) is a graph obtained by processing (b) by the processing of the transformation matrix A. (d) is a waveform obtained by separating noise by this processing. It is a figure which shows the relationship between a signal component and a noise component. It is a spectrum figure of infrared light and red light obtained by frequency analysis of the waveform of Drawing 3 (a) before processing. It is the spectrum figure obtained by the frequency analysis of the waveform of FIG.3 (d). It is the graph which took the component of the infrared light on the horizontal axis, and took the component of the red light on the vertical axis. (Noise superimposition) It is the graph which took the component of the infrared light on the horizontal axis, and took the component of the red light on the vertical axis. (Single signal) It is the result of calculating | requiring the inclination for every frequency range of FIG. 7 by the independent component analysis. It is the result of having calculated | required the inclination for every frequency range of FIG. 8 by independent component analysis. It is a flowchart which shows the processing flow performed with the pulse oximeter shown in FIG.

Explanation of symbols

1 Probe 2 Light emitting part (LEDs)
3 Light receiver (PD)
4 Light-emitting diode drive unit 5 Input unit 6 Demodulation circuit 7 AD conversion 8 Arithmetic processing / control unit

Claims (8)

Pre-processing the measured pulse wave data of each wavelength obtained by irradiating a living tissue with light of two different wavelengths and converting the light of each wavelength transmitted or reflected into an electrical signal;
Dividing the measured pulse wave data of each wavelength into a plurality of frequency bands, and obtaining an absorbance ratio for each frequency band; and
Determining the presence or absence of noise in the measured pulse wave data according to the presence or absence of the absorbance ratio for each frequency band; and
A signal processing method of measured pulse wave data. The determination of the presence or absence of the same absorbance ratio for each frequency band is performed when the absorbance ratio is within a predetermined range in all the divided frequency bands, or in the frequency band where the power after the division is dominant. 2. The signal processing method of measured pulse wave data according to claim 1, wherein it is determined that noise is not mixed in the measured pulse wave data when the signal is within a predetermined range. A method of removing noise from measured pulse wave data of each wavelength obtained by irradiating a living tissue with light of two different wavelengths and converting the light of each wavelength transmitted or reflected into an electrical signal,
For the measured pulse wave data of each wavelength, the measured pulse wave data is whitened by affine transformation using a known or previously obtained absorbance ratio.

(Where S is a signal component, N is a noise component, s1 is a first electrical signal, s2 is a second electrical signal, φ = tan ⁻¹ Φ, Φ is an absorbance ratio, and θ is from −φ. The signal processing method is characterized in that the signal processing is executed in accordance with a change in the range of π / 2−φ, and θ is a signal processing that minimizes the norm of the signal component S and is separated into a pulse wave signal and a noise signal.
The signal processing method according to claim 3, wherein an absorbance ratio determined by the method according to claim 2 to determine that there is no noise mixing is used as the known absorbance ratio .
5. A measurement pulse wave of each wavelength obtained by irradiating a living tissue with light of two different wavelengths and converting the light of each wavelength transmitted or reflected into an electrical signal by the method of claim 3 or 4 A signal processing method characterized in that an S / N ratio for each frequency component is obtained by frequency analysis of a pulse wave signal and a noise signal separated by processing data. 6. A measurement pulse wave of each wavelength obtained by irradiating a living tissue with light of two different wavelengths and converting the light of each wavelength transmitted or reflected by the method according to any one of claims 3 to 5 into an electrical signal A pulse photometer that processes a data string to separate a noise component and a signal component. The pulse photometer according to claim 6 , wherein a pulse waveform is obtained from the separated signal component. The pulse photometer according to claim 6 , wherein the pulse rate is calculated from the separated signal component by pulse detection or frequency analysis.

Images