JP2008049091A - Body component concentration measuring method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To prevent estimation precision from lowering even if the tendencies of spectrum shapes caused by individual difference, skin conditions and environmental factors are different. <P>SOLUTION: Data of difference table are previously created using spectra for creating a difference table. A spectrum data set obtained by adding the difference table data to criterion spectrum data is multivariately analyzed to find a calibration function, and absorbance spectrum measured by irradiating the living body with near-infrared light is analyzed using the calibration function to measure the concentration of the body component. When obtaining the concentration, this method corrects it with a ratio of the characteristic amount of the difference table creation spectrum to the characteristic amount of the criterion spectrum. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention irradiates near-infrared light on the skin tissue surface of a living body, receives diffuse reflection light from the skin tissue, performs spectrum measurement, and performs qualitative and quantitative analysis of biological components by a near-infrared spectroscopic analysis method. The present invention relates to a method for measuring the concentration of a biological component.

  Near-infrared spectroscopy, which performs qualitative and quantitative analysis of biological tissue from spectral signals obtained by receiving diffusely reflected light inside the biological tissue when irradiated with near-infrared light, Information is attracting attention in many medical fields because information can be obtained immediately on the spot, non-invasively and without reagents.

  Especially regarding blood glucose measurement, not only is there a demand for blood glucose management in diabetics, but in recent years there has been a report that the mortality rate of patients will be significantly reduced by managing blood glucose within an appropriate range in the intensive care unit. Non-invasive continuous blood glucose monitoring has been attracting attention.

  In near-infrared spectroscopy, which is one of the technical means for non-invasively measuring blood glucose level, the absorbance changes at the glucose-specific absorption wavelength in the near-infrared spectrum due to the increase or decrease in glucose concentration. Japanese Patent Application Laid-Open No. 11-70101, Japanese Patent Application Laid-Open No. 2002-165645, and Japanese Patent Application Laid-Open No. 2006-87913 measure the spectrum from the signal and detect the glucose concentration in the living body by multivariate analysis of the spectrum signal. It is shown in the gazette.

  FIG. 1 shows a non-invasive optical blood glucose level measuring system disclosed in Japanese Patent Application Laid-Open No. 2006-87913. Near infrared light emitted from a halogen lamp 1 is a heat shielding plate 2 and a pinhole 3. Then, the light enters the living tissue 6 through the lens 4 and the optical fiber bundle 5. In the optical fiber bundle 5, one end of the measurement optical fiber 7 and one end of the reference optical fiber 8 are connected, the other end of the measurement optical fiber 7 is connected to the measurement probe 9, and the other end of the reference optical fiber 8. Is connected to the reference probe 10, and the measurement probe 9 and the reference probe 10 are connected to the measurement-side emitting body 11 and the reference-side emitting body 12 via optical fibers.

  When the near-infrared spectrum measurement is performed by bringing the tip surface of the measurement probe 9 into contact with the surface of the living tissue 6 such as the forearm of the human body at a predetermined pressure, the near-infrared light incident on the optical fiber bundle 5 from the light source 1 is The light is transmitted through the measurement optical fiber 7 and irradiated from the tip of the measurement probe 9 as shown in FIG. 1B to the surface of the living tissue 6 from 12 light emitting fibers 20 arranged concentrically. The measurement light applied to the living tissue 6 is diffusely reflected in the living tissue, and then a part of the diffuse reflected light is received by the light receiving fiber 19 disposed at the center of the tip of the measurement probe 9. The received light is emitted from the measurement-side emitting body 11 through the light-receiving side optical fiber 19, is incident on the diffraction grating 14 through the lens 13, and is dispersed and then detected by the light-receiving element 15. The optical signal detected by the light receiving element 15 is AD converted by the A / D converter 16 and then manually operated by an arithmetic unit 17 such as a personal computer. The blood glucose level is calculated by analyzing the spectrum data.

  In the reference measurement, light reflected by a reference plate 18 such as a ceramic plate is measured and used as reference light. The reference probe 10 has the same configuration as that of the assumed probe 9, and the optical fiber bundle from the light source 1 is used. 5 is irradiated from the tip of the reference probe 10 to the surface of the reference plate 18 through the reference optical fiber 8. The reflected light from the reference plate 18 is emitted from the reference-side emitter 12 via the light receiving optical fiber 19 disposed at the tip of the reference probe 10.

  A shutter 22 is disposed between the measurement-side emitter 11 and the lens 13 and between the reference-side emitter 12 and the lens 13, and the light from the measurement-side emitter 11 is changed by opening and closing the shutter 22. One of the light from the reference-side emitter 12 is selectively passed and guided to the light receiving element 15.

  The end faces of the measurement probe 9 and the reference probe 10 are composed of 12 light emitting fibers 20 arranged on a circle and one light receiving fiber 19 arranged in the center as described above. The distance L between the centers of the light emitting fiber 20 and the light receiving fiber 19 is 650 μm. Further, the end faces of the measurement-side emitter 11 and the reference-side emitter 12 are arranged with the emission fiber 21 (the other end of the light receiving fiber 19) as the center as shown in the right figure of FIG.

  When performing quantitative and qualitative analysis by near infrared spectroscopy, usually perform multivariate analysis on the data set obtained by collecting data in advance, and create a calibration function when performing quantitative analysis, It is common to use a method for calculating the target component concentration. At this time, in order to obtain a robust calibration function, it is said that it is good to collect the widest possible and wide range of data and create the above data set. However, the time, labor, and cost associated with the collection of a wide range of data are considered. The burden is a big problem.

  On the other hand, for quantitative analysis of biological components, particularly blood glucose level measurement, the absorbance of the target component that is a signal is very small. For example, when the glucose signal change is measured by the apparatus shown in the above-mentioned conventional example, the change in absorbance of the glucose signal with respect to the concentration change of 10 mg / dl required for clinically meaningful measurement is about 100 μAU, and the biological change And the noise associated with measurement reproducibility is extremely small (poor signal-to-noise ratio), so the calibration function creation method used in normal near-infrared spectroscopy can be used directly for qualitative and quantitative analysis of biological components. It is difficult to obtain a useful calibration function. Also, the calibration function is often created for each individual in order to eliminate noise caused by individual differences among subjects and improve adaptability.

  However, in a medical facility such as an intensive care unit (ICU), it is practically impossible to prepare a calibration function by performing spectrum measurement on the patient prior to monitoring in treatment.

  On the other hand, what is disclosed in the above Japanese Patent Application Laid-Open No. 2006-87913 is a method of creating a calibration function without collecting spectrum data in advance, and here, a plurality of obtained from a living body or a simulation or the like. A plurality of differential absorbance spectra, which are differences between the near-infrared absorbance spectrum and a reference absorbance spectrum selected from these, are obtained, and a plurality of reference absorbance spectra of a subject measured in advance are combined with each of the difference absorbance spectra. A calibration function is created by obtaining a synthetic absorbance spectrum of the sample and performing multivariate analysis using the obtained plurality of synthesized absorbance spectra.

However, by applying only the calibration function creation method shown here, it is difficult to deal with all the biological spectra with various differences in solids. FIG. 2 shows an absorption spectrum B (water peak height) of skin measured from a subject different from an example of the reference spectrum A (created by computer simulation and having a water peak height of 1.10 described later) used in the present invention. Is about 1.4), and the absorption spectrum C (water peak height is about 1.1) is shown, but there is a spectrum with such a large tendency, and thus the skin spectrum When blood glucose levels of subjects having different tendencies are measured using the same difference table, there is a case in which the measurement accuracy for subjects having greatly different tendencies from the reference spectrum deteriorates.
JP 2006-87913 A

  The present invention has been invented in view of the above-described conventional problems, and the accuracy of prediction is reduced when blood glucose levels are measured when the trends in spectrum shapes due to individual differences, skin conditions, environmental factors, and the like are different. It is an object of the present invention to provide a biological component concentration measurement method that solves the above-mentioned drawback.

  In order to solve the above problems, the present invention provides a multivariate analysis of a spectrum data set obtained by previously creating a difference table data using a spectrum for creating a difference table and adding the difference table data to the reference spectrum data. In the biological component concentration measurement method, the calibration function is obtained, and the concentration of the biological component is measured by analyzing the absorbance spectrum measured by irradiating the living body with near infrared light using the calibration function. It is characterized in that the blood sugar level is calculated by performing correction based on the ratio between the characteristic amount of the table creation spectrum and the characteristic amount of the reference spectrum. Even when the tendency of the spectrum shape due to individual differences, skin conditions, environmental factors, and the like is different, it can be dealt with by the above correction, and the biological component concentration, particularly blood glucose level can be measured with high accuracy.

  In this correction, the measurement spectrum may be corrected using a ratio between the feature amount of the spectrum for creating the difference table and the feature amount of the pre-primary reference spectrum.

  Further, a plurality of difference tables may be prepared as the difference table, and a difference table that approximates an absorbance spectrum in which the feature amount is measured among these difference tables may be used.

  As the feature amount, the absorbance height of the water component of the biological component can be suitably used.

  Also, by creating the difference table data in advance using the spectrum for creating the difference table, the calibration function is obtained by performing multivariate analysis on the spectrum data set obtained by adding the difference table data to the reference spectrum data, In the biological component concentration measurement method for measuring the concentration of a biological component by analyzing the absorbance spectrum measured by irradiating the living body with near infrared light using the calibration function, the feature amount of the reference spectrum is considered. A calibration function may be created by correcting the data in the difference table.

  The correction of the difference table at this time may be performed by changing any of the scattering coefficient, moisture concentration, temperature, protein concentration, lipid concentration, and glucose concentration of the living body.

  In the present invention, because the blood glucose level is calculated using a correction means based on the ratio between the characteristic amount of the spectrum for creating the difference table and the characteristic amount of the reference spectrum, it is caused by individual differences, skin conditions, environmental factors, etc. Even when the spectrum shapes tend to be different, the biological component concentration, particularly the blood glucose level, can be accurately measured.

  In addition, in the case of creating a calibration function by correcting the difference table data in consideration of the characteristic amount of the reference spectrum, even in the case of subjects with different spectral shape trends due to individual differences, skin conditions, environmental factors, etc. The blood sugar level can be measured without reducing the prediction accuracy. In addition, since it is a simple correction method, it is not necessary to perform a long-time calculation with a high-performance computer such as Monte Carlo simulation when creating a calibration function for each individual, and the storage means represented by memory is also large. It does not require a capacity and can be used with a normal personal computer.

  Hereinafter, the present invention will be described with reference to the embodiments shown in the accompanying drawings. In the present invention, spectrum measurement relating to a living tissue covers a dermis layer. The skin tissue of a living body is mainly composed of three layers of the epidermis, dermis, and subcutaneous tissue. The epidermal tissue is a tissue including the stratum corneum, and the capillaries are not so developed in the tissue. The subcutaneous tissue is mainly composed of adipose tissue. Therefore, it is considered that the correlation between the water-soluble biological component concentrations contained in the two tissues, particularly, the glucose concentration and the blood glucose concentration (blood glucose level) is low. On the other hand, the dermal tissue has developed capillary blood vessels and the concentration of biological components with high water solubility, especially glucose has high permeability in the tissue. As is the case with fluid (ISF: Interstitial Fluid), it is considered to change following blood glucose level. For this reason, if spectrum measurement targeting the dermal tissue is performed, the concentration of biological signals, especially the spectral signal correlated with blood glucose level fluctuation, Measurement is possible.

The measuring probe composed of an optical fiber having a center-to-center distance L of 0.65 mm shown in FIG. 1 has been developed to target this dermis layer. In the present invention, this measuring probe is preferably used. Can be used. The skin spectrum during the verification experiment is automatically controlled so that the measurement probe placed on the forearm of the subject is moved up and down using a load sensor and a stepping motor, and contacted at a constant pressure (470 gf / cm ² ) at intervals of 5 minutes. went. The time required for one spectrum measurement (probe contact time) is about 2 minutes.

  In addition to glucose (blood glucose level), biological component concentrations expected to correlate with blood concentration include uric acid levels, cholesterol levels, triglyceride levels, albumin levels, globulin levels, oxygen saturation levels, hemoglobin levels, etc. .

  In the present invention, the measuring apparatus having the configuration shown in FIG. 1 was used. And here, the correction based on the ratio between the feature quantity of the spectrum for creating the difference table and the feature quantity of the reference spectrum is performed. As the feature quantity of the spectrum, the water peak height of the near infrared absorption spectrum is used. The method of multiplying the difference between the water peak height of the reference spectrum used to create the difference table and the water peak height of the reference spectrum used to create the calibration function in the experiment by the difference from the calibrated blood glucose level measured at the start of the verification experiment Was used.

  The near-infrared light wavelength used for the measurement was 1400 to 1850 nm, and the water peak height was a numerical value obtained by subtracting the absorbance of the 1650 nm base line from the absorbance of the water absorption peak of 1450 nm, which was used to create the difference table of this example. The feature amount of the spectrum was 1.10, and the feature amount of the reference spectrum was 1.43.

  After the warm-up of the blood glucose monitor was completed, the blood glucose monitor was calibrated using the blood glucose level actually measured by blood collection. A virtual data set is synthesized by adding the difference table obtained by the technique described later to the absorption spectrum measured simultaneously with blood collection, and a calibration function is created by performing multivariate analysis (PLS regression analysis in this example) from this data set. To do. The near-infrared wavelength range used for creating the calibration function is 1520 nm to 1700 nm, and the PLS parameter is 8. Thereafter, the blood glucose level measured by the near-infrared spectrum was obtained as a relative change value from the actually calibrated blood glucose level obtained by blood collection that simultaneously performed this absorption spectrum. The verification of this example was performed by comparing the blood glucose level measured with near-infrared light and the blood glucose level actually measured by blood sampling. The measurement was carried out for 3 to 4 hours after the start of measurement.

  The accuracy verification result is shown in FIG. The correction used to calculate the measured blood glucose level is the difference between the calculated blood glucose level calculated by substituting the measured absorbance spectrum into the calibration function and the calibration blood glucose level. This was done by adding the calibrated blood glucose level to the value multiplied by the ratio (ie 1.10 / 1.43). Comparing the accuracy comparison between the case where correction is not performed (conventional example) and the present embodiment where correction is performed with the predicted standard error (SEP), the conventional example is 29.5 mg / dl, while the implementation side is 22.9 mg / dl. It can be seen that the error is reduced.

  The reason why the accuracy is improved by this correction can be understood as correction of the average optical path length of the measured absorption spectrum with respect to the average spectrum of the difference table. That is, the main factor of the change in the shape of the absorption spectrum is a change in the average optical path length due to the change in the optical characteristics of the biological components, and a feature amount typified by the water peak height can be used as a substitute characteristic for the change in the average optical path length. Therefore, the rationality of using the feature ratio for correction can be understood.

  In addition, the difference table creation in the present embodiment was performed by adopting the Monte Carlo method as the light propagation simulation. The Monte Carlo method is a statistical method that can accurately reproduce a target event using a function based on the probability distribution of the target event with respect to uniform random numbers in the range of 0 to 1 generated by a computer. When reproducing light propagation, the light incident on the medium is regarded as a collection of photons, and the behavior of each photon in the medium is determined based on the optical characteristic values of the medium (absorption coefficient, scattering coefficient, scattering phase function, refractive index). ) To track. As a result, light propagation can be statistically reproduced from the behavior of all photons.

  When the light propagation simulation is performed based on the Monte Carlo method, optical characteristic values such as the absorption coefficient, scattering coefficient, refractive index, and anisotropic scattering parameter of the epidermis layer, the dermis layer, and the subcutaneous tissue layer are required. As a variation factor used in the simulation, glucose, protein, lipid, moisture, and temperature are changed in 5 parameters in the body and scattering coefficient within the preset range so that the daily variation of the in vivo parameter is covered. The absorbance spectrum was prepared by repeating 64 combinations (maximum of 2 6) combining the maximum and minimum values of each parameter.

The difference table is obtained by subtracting the average spectrum of the simulation spectra from the plurality of absorbance spectra obtained by simulating in this way, thereby obtaining a plurality of differential absorbance spectra. FIG. 8 shows an example of a simulation spectrum used for creating the difference table. The merits and demerits of creating a calibration function from the absorption vector data set obtained by such numerical simulation are considered as follows.
(1) No preliminary experiment is required to create a calibration function.
(2) A disturbance factor and its change amount can be set arbitrarily. For example, if the amount of change is set at regular intervals, the possibility of accidental correlation is very low.
(3) The objective variable (glucose concentration) is clear and there are no uncertainties such as the time delay of the ISF with respect to the blood glucose level.
(4) The change of the absorption spectrum cannot be completely expressed by several kinds of disturbance factors to be set.
(5) The set disturbance factor When such a disturbance occurs, the created calibration function cannot handle it.
(6) Optical simulation technology in a scattering system such as a living body is still in the development stage.

  Although the method using numerical simulation was shown as a method for creating the difference table, it is not limited to this, and a difference table obtained from a data set consisting of absorbance spectra obtained by actually measuring human skin, Instead of numerical simulation, a difference table obtained by measuring a simulated living body represented by intra-ribbed may be used.

  The correction of the measured blood glucose level in the present embodiment is substantially the same as that of the first embodiment except that the measurement spectrum is corrected using the ratio between the feature amount of the spectrum for creating the difference table and the feature amount of the reference spectrum. Is the same. Therefore, the overlapping description is omitted. In addition, an apparatus having the same configuration as that of the first embodiment can be used in this embodiment to perform quantitative analysis.

  The correction here uses the water peak height of the near-infrared absorption spectrum as the feature amount of the spectrum, and the difference between the absorbance at each wavelength of the measurement spectrum and the absorbance at the base line of 1650 nm is used to create the difference table. The absorption spectrum was converted by adding the absorbance of the base line to the value obtained by multiplying the ratio of the average spectrum of the spectrum to the reference spectrum (that is, 1.10 / 1.43).

  The accuracy verification result is shown in FIG. Further, when the accuracy comparison of the case where correction is not performed (conventional example) and the present embodiment where correction is performed is compared with the predicted standard error (SEP), the conventional example is 29.5 mg / dl, whereas the embodiment is 22.8 mg. / Dl, which shows that the error is reduced.

  In the correction of the measured blood glucose level of the present embodiment, a technique for converting the characteristic amount of the measurement spectrum is used, but the spectrum used for creating each component value of the difference table used in the embodiment without performing the conversion of the spectrum. The same effect can be obtained even if correction is performed by adding the constituent blood glucose level to the value obtained by multiplying the ratio of the reference spectrum (ie, 1.10 / 1.43).

  The correction of the measured blood sugar level in the present embodiment is substantially the same as in the first and second embodiments except that one of a plurality of difference tables prepared in advance as a difference table is used. As the difference table, the water peak height of the average spectrum of the simulation used in Examples 1 and 2 is 1.10 (difference table A), and the water peak height of the average spectrum of the simulation is 1.40 ( Two types of difference tables B) were prepared. The determination of which difference table to use is made with a standard spectrum for creating a calibration function. Here, blood glucose level measurement was performed using the difference table A for water peak heights of less than 1.3 and the difference table B for water peak heights of 1.3 or more.

  Since the measured water peak height of the standard spectrum is 1.41, the accuracy verification result when the difference table B is used is shown in FIG. Moreover, the water peak height of the near-infrared absorption spectrum was used as the spectrum feature amount, and the method of correcting the measurement spectrum was used in the same manner as in Example 2. However, the correction was performed by the method shown in the other examples. It doesn't matter. Comparing the accuracy comparison between the case where correction is not performed (conventional example) and the present embodiment where correction is performed with the predicted standard error (SEP), the conventional example is 25.0 mg / dl, whereas the example is 10.7 mg / dl. It can be seen that the error is reduced.

  In the present embodiment, a method of providing a reference value for a spectrum feature amount (water peak height in this embodiment) is exemplified as a method of selecting a difference table to be used from a plurality of difference tables, but the present invention is not limited to this. . For example, a discriminant analysis such as principal component analysis may be used to compare the shape of the standard spectrum and the average spectrum used for creating the difference table, and a method using the closer difference table may be used. In addition to discriminant analysis, compare the centroid distance of the standard spectrum and the average spectrum used to create the difference table, use the closer difference table, or compare the Mahalanobis generalized distance and A method of using a difference table may be used.

  The correction of the measured blood glucose level in the present embodiment is performed by creating a difference table in accordance with the shape of the reference spectrum measured when creating the calibration function. The apparatus and protocol used for verification are the same as those in the first to third embodiments.

  Conventionally, the average spectrum of the simulation spectrum that created the difference table obtained by actual measurement or simulation is a fixed value, whereas the calibration function was calculated by adding the value of this difference table to the measured reference spectrum. In this embodiment, the difference table is calculated after the average spectrum of the simulation spectrum used to create the conventional difference table is simply approximated to the actually measured reference spectrum. For this reason, a difference table closer to the change in the measured spectrum is created, and a better biological component concentration, particularly a blood glucose level, can be estimated without performing complicated simulation calculations such as Monte Carlo simulation.

The big flow is as follows.
(1) Measure the measured reference spectrum.
(2) The spectrum is corrected and calculated so that the average spectrum of the simulation spectrum that created the difference table 1 is similar to the actually measured reference spectrum.
(3) The difference table 2 is created using the spectrum calculated in 2 as the standard spectrum of a new difference table.
(4) The difference table 2 is added to the actually measured reference spectrum, a spectrum group for creating a calibration function is created, and a calibration function is created.
(5) Thereafter, the concentration of the biological component, particularly the blood glucose level, is calculated from the measured spectrum using the calibration function.

  Examples will be described. First, the reference spectrum of the actually measured spectrum was measured, and then the spectrum was corrected so that the average spectrum of the simulation spectrum that created the difference table was similar to the actually measured reference spectrum.

  In this example, the water peak height is used as a feature amount, and the reference spectrum of the difference table is corrected so that the actually measured reference spectrum matches the water peak height. In this correction, biological components (six components of glucose, water, fat, and protein) existing in the epidermis, dermis, and subcutaneous tissue, temperature, and scattering coefficient in advance are added as biological component elements. The difference spectrum for the fluctuation of the six components is set, and the difference spectrum for the fluctuation of the six components is added to and subtracted from the average spectrum of the simulation spectrum for which the difference table is created. Correction and calculation were performed so as to approximate the reference spectrum.

  In this example, since the water peak height of the actually measured reference spectrum used as the feature amount is 1.40, and the water peak height conventionally used is 1.10, the difference table 1 was created. Correction calculation was performed so that the water peak height of the standard spectrum of the difference table 2 newly created by adding the difference spectrum of water to the average spectrum of the simulation spectrum becomes 1.40.

  Next, the difference table 2 is created. To each newly calculated standard spectrum, glucose is ± 100 mg / dl, temperature is ± 6 ° C., other biological component elements are changed by ± 8%, and a difference spectrum of each component is added to obtain a plurality of spectra. Created. The plurality of spectra created this time are a total of 64 spectra obtained by varying each of the six biological component elements (glucose, water, fat, protein, temperature, scattering coefficient) within the above ranges. A new difference table 2 was created by subtracting the average spectra of the created 64 spectra from the plurality of simulation spectra.

  Next, a spectrum group used to create a calibration function for calculating a blood glucose level by adding the difference table 2 to an actually measured reference spectrum was created, and a calibration function was calculated from the spectrum group. Then, the blood glucose level was calculated by multiplying this calibration function by the spectrum actually measured from the next.

  The accuracy verification result is shown in FIG. Comparing the accuracy comparison of the present example with no correction (value estimated using the difference table used in the past) and the corrected example with the predicted standard error (SEP), the conventional example is 37.9 mg / dl. On the other hand, in the example, it is 14.3 mg / dl, which shows that the error is reduced.

  Further, as described below, the standard spectrum of the difference table and the spectrum for the difference table may be calculated by dividing into three layers of the epidermis, dermis, and subcutaneous tissue constituting the skin. By calculating each optical characteristic value in three layers, a standard spectrum and a difference table can be created with higher accuracy.

  More specifically, first, an extinction coefficient and a scattering coefficient corresponding to the ratio of the biological component elements existing in the skin for creating the difference table 2 in accordance with the shape of the reference spectrum were created. That is, the structure of the skin, which is the measurement site, is defined as being composed of three layers of the epidermis, dermis, and subcutaneous tissue, and biological components present in the skin (this time, four components of glucose, water, fat, and protein) ), Temperature, and scattering coefficient are defined as biological component elements, and the optical property values (absorption coefficient, scattering coefficient) of each layer are determined according to the proportion of the six biological component elements present in each of the three layers. Calculated by simulation. That is, a total of six optical characteristic values of the absorption coefficient and scattering coefficient of the epidermis, the absorption coefficient and scattering coefficient of the dermis, and the absorption coefficient and scattering coefficient of the subcutaneous tissue were determined according to the existence ratio of the six biological component elements. Then, the pre-calculated relationship between the existence ratio of such biological component elements and the six optical characteristic values was stored in a memory as a table.

  Next, in order to be able to calculate the absorbance spectrum from the six optical characteristic values defined above, multiple regressions with a total of six variables of the absorption coefficient and scattering coefficient value of the skin structure three layers as explanatory variables and the absorbance spectrum as the objective variable An equation was created and this multiple regression equation was also stored in memory.

  In this case, if the ratio of the six biological component elements such as glucose and water is arbitrarily set in the skin, the absorption coefficient and scattering coefficient in the skin tissue are determined from the table held in the memory. An absorbance spectrum is uniquely calculated by substituting it into the multiple regression equation as an explanatory variable.

  The above is an explanation of calculating the absorbance spectrum separately for each skin structure. If a standard spectrum for a new difference table is created using this technique and a difference table is created, six biological component elements are contained in the skin. Can be set arbitrarily, so that various absorbance spectra can be calculated from the absorption coefficient and scattering coefficient divided into epidermis, dermis, and subcutaneous tissue, and the standard spectrum can be approximated to the reference spectrum more accurately. .

  In addition, this time, the standard spectrum of the difference table 2 is calculated so as to be similar to the actually measured reference spectrum with the water peak height as a special amount in particular, but the present invention is not limited to this, and other components such as fat peaks are calculated. The difference table may be created by further approximating the actually measured reference spectrum and the standard spectrum of the difference table 2 using the height as a feature amount.

  Further, the difference table may be created by adjusting the existence ratio of the six biological component elements without limiting the feature amount to one so that the actually measured reference spectrum is similar to the standard spectrum of the difference table 2. .

  Furthermore, in this embodiment, in the average spectrum of the difference table, 1.40 which is the water peak height of the actually measured reference spectrum based on the difference table where the water peak height used as the extra quantity is 1.10. Although a standard spectrum was created so as to approximate to, the present invention is not limited to this. The difference table may be used based on a difference table created with a water peak height of 1.40 in the average spectrum, or based on a difference table created with a completely different water peak height, for example, 1.30. As an alternative, a standard spectrum may be approximated.

  Further, as in the third embodiment, two difference tables may be provided. That is, two average spectra may be provided, an average spectrum having a shape close to the reference spectrum may be determined, and the average spectrum may be approximated to the reference spectrum as described above.

  The correction of the measured blood glucose level of the present embodiment is substantially the same as that of the second embodiment except that the fat peak height is used as the ratio between the feature amount of the spectrum for creating the difference table and the feature amount of the reference spectrum. This is the same as the embodiment. Therefore, the overlapping description is omitted. In addition, an apparatus having the same configuration as that of the first embodiment can be used to perform quantitative analysis in this embodiment.

  The fat peak height as the trace amount is a value obtained by subtracting the absorbance at the 1650 nm base line from the absorbance at the fat absorption peak at 1725 nm, and the fat peak height of the average spectrum used for preparation in the difference table of this example is 0. 135, the reference spectrum is 0.147. Then, the correction is based on the difference between the absorbance at each wavelength of the measured spectrum and the absorbance at the baseline of 1650 nm, and the ratio of the average spectrum of each spectrum used for creating the difference table to the fat peak height of the reference spectrum (ie, 0.13). /0.147) was added by adding the baseline absorbance to convert the measured spectrum.

  The accuracy verification result is shown in FIG. In addition, when comparing the accuracy comparison of the case where correction is not performed (conventional example) and the present embodiment where correction is performed with the predicted standard error (SEP), the conventional example is 39.6 mg / dl, whereas the embodiment is 32.5 mg. / Dl, which shows that the error is reduced.

  In this embodiment, the feature amount of the spectrum for creating the difference table and the fat peak height are used as the feature amount of the reference spectrum. However, the feature amount is not limited to this, for example, as shown in FIG. A stable portion near 1800 nm as seen in the absorption spectrum may be used as the feature amount. In the examples, water and fat peaks are defined as differences from the baseline, but this is not limited, and the definition may be made using absorbance at other wavelengths. Further, an integral value (area) of the absorption spectrum or barycentric coordinates may be used as the feature amount.

The measuring apparatus used in this invention is shown, (a) is the schematic of the whole structure, (b) is the schematic of a probe structure. It is a graph which shows an example of the average spectrum of the simulation spectrum used for preparation of the difference table used by this invention, and a measurement spectrum. It is a graph which shows the result of the Example of this invention. It is a graph which shows the result of another Example. It is a graph which shows the result of other examples. It is a graph which shows the result of another Example. It is a graph which shows the result of another Example. It is explanatory drawing of an example of the simulation spectrum used for difference table preparation.

Claims (6)

  Create the difference table data in advance using the spectrum for the difference table creation, add the difference table data to the reference spectrum data, and perform multivariate analysis to obtain the calibration function. In the biological component concentration measurement method for measuring the concentration of the biological component by analyzing the absorbance spectrum measured by irradiating the biological body with near infrared light using a function, the feature amount of the spectrum for creating the difference table and the reference A biological concentration measurement method comprising calculating a blood glucose level by performing correction based on a ratio to a spectral feature amount.   The biological component concentration measurement method according to claim 1, wherein the measurement spectrum is corrected using a ratio between a feature quantity of the spectrum for creating the difference table and a feature quantity of the reference spectrum.   The biological component concentration measurement method according to claim 1 or 2, wherein a plurality of difference tables are prepared as the difference table, and a difference table that approximates an absorbance spectrum in which a feature amount of the difference tables is measured is used. .   The biological component concentration measurement method according to any one of claims 1 to 3, wherein the absorbance height of the water component of the biological component is used as the feature amount.   Create a difference table data in advance using the spectrum for creating the difference table, add the difference table data to the reference spectrum data, and perform multivariate analysis to obtain a calibration function. In a biological component concentration measurement method for measuring the concentration of a biological component by analyzing an absorbance spectrum measured by irradiating a living body with near infrared light using a function, the difference is calculated in consideration of a feature amount of the reference spectrum. A biological component concentration measuring method, wherein a calibration function is created by correcting data in a table.   6. The biological component concentration measurement according to claim 5, wherein the correction of the difference table is performed by changing any one of a scattering coefficient, moisture concentration, temperature, protein concentration, lipid concentration and glucose concentration of the living body. Method.

Images