Glucose concentration estimation device

Abstract

To estimate glucose concentration at low cost and with high accuracy by using near-infrared light of multiple types of wavelengths having relatively large differences in absorbance between glucose and water.SOLUTION: A glucose concentration estimation device 1 includes a first light source 3-1 that irradiates a human body 150 with first light including light of a wavelength of 1384 nm, a second light source 3-2 that irradiates the human body 150 with second light including light of a wavelength of 1582 nm, a third light source 3-3 that irradiates the human body 150 with third light including light of a wavelength of 1847 nm, a fourth light source 3-4 that irradiates the human body with fourth light including light of a wavelength of 2216 nm, a light receiving module 5 having a light receiving element that receives light reflected from the human body 150 after irradiation with the first to fourth lights, and a control circuit 6 that estimates a glucose concentration on the basis of an output of the light receiving module 5.SELECTED DRAWING: Figure 4

Description

[Technical Field]

[0001] The present invention relates to a glucose concentration estimation device that estimates a glucose concentration using an optical detection means. [Background technology]

[0002] In the past, medical devices and healthcare products typically required blood sampling to test blood components such as glucose levels. Recently, non-invasive detection methods using optical detection means have been attracting attention in order to avoid the physical burden on patients and concerns about infection.

[0003] Patent Document 1 discloses a technique in which a 9.26 μm mid-infrared laser beam is generated by an optical parametric oscillator (OPO) using a 1.06 μm near-infrared excitation light generated by a light source, and the generated laser beam is locally irradiated onto the epithelium of a subject's body, and the diffusely reflected light is detected by a photodetector. Furthermore, the technique discloses a technique for calculating the glucose concentration in interstitial fluid using a normalized light intensity calculated from the signal ratio between the monitor photodetector and the photodetector. Glucose has high absorption sensitivity in the 9.26 μm mid-infrared wavelength range, which is expected to enable highly accurate glucose concentration detection.

[0004] Patent document 2 discloses a blood glucose monitoring device that includes a reference blood glucose measurement means for invasively measuring a reference blood glucose level, a blood glucose estimation means for non-invasively estimating a blood glucose level using near-infrared light, and a calibration means for automatically calibrating the estimated blood glucose level estimated by the blood glucose estimation means using the reference blood glucose level.

[0005] Patent Document 3 discloses a technique for quantifying glucose concentrations in biological tissues or fluids using the absorption of light in the near-infrared wavelength range (1300 nm to 1900 nm). At least one wavelength or wavelength region is selected from each of four wavelength ranges: 1530 nm to 1560 nm, 1580 nm to 1640 nm, 1640 nm to 1720 nm, and 1720 nm to 1750 nm. The technique quantifies glucose based on the absorption signal obtained using these four wavelengths or wavelength regions. In this technique, at least one wavelength or wavelength region is selected from the 1580 nm to 1640 nm wavelength range as a specific absorption wavelength for glucose. Furthermore, the absorption signals from the other three wavelength ranges are used to remove disturbances caused by biological components other than glucose that are superimposed on the glucose absorption spectrum of the selected wavelength or wavelength region. [Prior art documents] [Patent documents]

[0006] [Patent Document 1] Patent No. 6415606 [Patent Document 2] Japanese Patent Application Laid-Open No. 2011-62335 [Patent Document 3] Japanese Patent Application Laid-Open No. 2000-131322 Summary of the Invention [Problem to be solved by the invention]

[0007] However, in the technology of Patent Document 1, near-infrared excitation light is generated by a YAG light source, but YAG light sources are expensive. Furthermore, the optical system compatible with the 9.26 μm wavelength is made of compound materials, which are also expensive. This results in the problem of high manufacturing costs for the device.

[0008] Furthermore, while the technology in Patent Document 2 can reduce costs by using near-infrared light, the wavelength of near-infrared light is also strongly absorbed by water in the human body. Therefore, with this method, it is impossible to distinguish between glucose concentration and water concentration based on the detection results using near-infrared light alone, and it is necessary to calibrate the blood glucose level estimation results using near-infrared light using the results of invasive blood glucose level detection by blood sampling. Furthermore, the technology of Patent Document 3 is designed to remove superposition of components other than the glucose concentration, but within the wavelength range specified in Patent Document 3, it is impossible to determine the concentration of each disturbing component with high accuracy. The present invention has been made in consideration of the above-mentioned problems, and aims to provide a glucose concentration estimation device that can estimate glucose concentrations at low cost and with high accuracy using near-infrared light of multiple wavelengths where the difference in absorbance between glucose and water is relatively large. [Means for solving the problem]

[0009] The present invention provides a device for estimating a glucose concentration based on an output from the light receiving element, the device including at least three light sources, including a first light source for irradiating a living organism with first light, the first light being light having one of the wavelengths in the wavelength range of 1375 nm to 1395 nm, a second light source for irradiating the living organism with second light, the second light being light having one of the wavelengths in the wavelength range of 1575 nm to 1595 nm, and a third light source for irradiating the living organism with third light, the third light being light having one of the wavelengths in the wavelength range of 1835 nm to 1855 nm; a light receiving element for irradiating the living organism with light of a plurality of wavelengths including at least the first to third light and receiving light returning from the light receiving element; and an estimation unit for estimating a glucose concentration based on an output from the light receiving element. The glucose concentration estimation device includes:

[0010] In the above configuration, at least four light sources may be provided, further including a fourth light source for irradiating the living body with fourth light, which is light including any one wavelength in the wavelength range of 2175 nm to 2255 nm, and the light receiving element may be configured to be able to irradiate the living body with light of multiple wavelengths including at least the first to fourth light and receive the light returned.

[0011] In the above configuration, the device may further include a memory unit that stores estimation information including a trained model obtained by previously learning the relationship between information on reflected light when light of the multiple wavelengths is incident on the living body and glucose concentration, and the estimation unit may estimate the glucose concentration corresponding to the output of the light receiving element using the estimation information stored in the memory unit. In the above configuration, the trained model may be a model trained on a combination of different concentrations of glucose, water, and a third component corresponding to another component in the dermis layer set for the artificial skin.

[0012] In the above configuration, the device further includes an attribute information acquisition unit that acquires attribute information regarding the attributes of the subject, and the trained model is a model that has learned the relationship between the attribute information, the reflected light information, and the glucose concentration based on the subject's attribute information, information on the reflected light reflected from the subject's skin, and the actual measured value of the subject's glucose concentration, and the estimation unit may use the information for estimation to estimate the glucose concentration of a single subject from the attribute information of the single subject acquired by the attribute information acquisition unit and the output of the light receiving element corresponding to the single subject.

[0013] In the above configuration, the device may further include a memory unit that stores estimation information including information indicating the correspondence between information on reflected light and glucose concentration, the information being generated using a trained model obtained by previously learning the relationship between information on reflected light when light of the multiple wavelengths is incident on the living body and glucose concentration, and the estimation unit may estimate the glucose concentration corresponding to the output of the light receiving element based on the estimation information stored in the memory unit. In the above configuration, the trained model may be a model trained on a combination of different concentrations of glucose, water, and a third component corresponding to another component in the dermis layer set for the artificial skin.

[0014] In the above configuration, the device may further include an attribute information acquisition unit that acquires attribute information regarding the attributes of the subject, wherein the trained model is a model that has learned the relationship between the reflected light information and the glucose concentration for each subject based on information about the reflected light reflected from the skin of the subject and the actual measured value of the glucose concentration of the subject, and the information indicating the correspondence is information indicating the correspondence between the subject's attribute information, the reflected light information of the subject corresponding to the attribute information, and the glucose concentration of the subject corresponding to the attribute information, and the estimation unit may use the information for estimation to estimate the glucose concentration of one subject from the attribute information of the subject acquired by the attribute information acquisition unit and the output of the light receiving element corresponding to the one subject. [Effects of the Invention]

[0015] According to the present invention, it is possible to estimate the glucose concentration at low cost and with high accuracy by using near-infrared light of a plurality of wavelengths in which the difference in absorbance between glucose and water is relatively large. [Brief explanation of the drawings]

[0016] [Figure 1] 1A and 1B are diagrams showing an image of the mixture of glucose and water in the dermis layer of a human body, and an image of near-infrared light irradiation and reflection. [Figure 2] 1 is a graph showing the relationship between the detection sensitivity and changes in glucose, moisture, and epidermal layer thickness in response to near-infrared light with a wavelength of 1550 nm. [Figure 3] 3 is a diagram showing the light absorption spectra of glucose units and water alone, in which the horizontal axis represents wavelength (nm) and the vertical axis represents absorbance. [Figure 4] 1 is a diagram showing an example of a schematic configuration of a glucose concentration estimation device 1 according to a first embodiment. [Figure 5] (a) is an external view of the trapezoidal prism 2 and the first to fourth triangular prisms 4-1 to 4-4 in the first embodiment as seen from the top side, and (b) is a diagram showing the positional relationship between the trapezoidal prism 2 and the first to fourth triangular prisms 4-1 to 4-4 and the first to fourth light sources 3-1 to 3-4. [Figure 6A] 10 is a diagram showing the incidence states of first and third light beams on first and third triangular prisms 4-1 and 4-3. FIG. [Figure 6B] 10 is a diagram showing the incidence states of second and fourth light beams on second and fourth triangular prisms 4-2 and 4-4. FIG. [Figure 7] This is a graph showing the relationship between wavelength and absorbance for combinations of concentrations of multiple types of glucose, water, and nanocellulose when they are mixed using artificial skin. [Figure 8A] 10 is a table showing the correspondence between the detected values ​​of reflected light and the measured values ​​of glucose concentration for subjects A and B. [Figure 8B] FIG. 8B is a graphical representation of the table of FIG. 8A. [Figure 9] 4 is a flowchart showing a glucose concentration estimation process according to the first embodiment. [Figure 10] FIG. 10 is a diagram showing the relationship between true values ​​and estimated values ​​when using detected values ​​of reflected light for light of four wavelengths. [Figure 11] FIG. 10 is a diagram showing a part of the schematic configuration of a glucose concentration estimation device 1A according to a second embodiment. [Figure 12] 10 is a flowchart showing a glucose concentration estimation process according to a second embodiment. [Figure 13] FIG. 10 is a diagram showing the relationship between true values ​​and estimated values ​​when detection values ​​of three wavelengths are used. DETAILED DESCRIPTION OF THE INVENTION

[0017] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. The embodiment described below is an example of a means for realizing the present invention, and should be appropriately modified or changed depending on the configuration of the device to which the present invention is applied and various conditions, and the present invention is not limited to the embodiment described below.

[0018] In addition, in the following description of the drawings, the same or similar parts are designated by the same or similar reference numerals. However, it should be noted that the drawings are schematic, and the vertical and horizontal dimensions and scales of the components or parts may differ from those of the actual parts. Therefore, the specific dimensions and scales should be determined by taking into consideration the following explanation. Furthermore, it goes without saying that the dimensional relationships and ratios may differ between the drawings. [First embodiment] First, a first embodiment of the present invention will be described, with reference to Figures 1 to 6 showing the first embodiment. The glucose concentration estimation device 1 according to the first embodiment irradiates the epidermis of a human body with multiple near-infrared light beams of specific wavelengths, and estimates the concentration of glucose in the blood based on the intensity of each light beam returning from the epidermis of the living body in response to the irradiation of light of each wavelength. [Reflection of incident light on the dermis layer] FIG. 1 shows an image of the mixture of glucose and water in the dermis layer of the human body, as well as images of irradiation and reflection of near-infrared light.

[0019] The light irradiated here is, for example, near-infrared light in the band of 1300 nm to 2400 nm. The outermost layer of human skin is covered by the epidermis layer. The dermis layer is located below the epidermis layer. In the case of the human forearm, the thicknesses of the epidermis layer and dermis layer are approximately 0.2 mm and 2 mm, respectively.

[0020] As shown in Fig. 1, the dermis layer contains a mixture of glucose 100 and water 101. Therefore, when light 200 is irradiated onto a living body, part of the light 200 is reflected from the surface of the epidermis layer, but the remaining light penetrates the epidermis layer. Furthermore, part of the light 200 that penetrates the epidermis layer is reflected within the epidermis layer and returns to the outside of the body.

[0021] The remaining light 200 that penetrates the epidermis penetrates the dermis below the epidermis. A portion of the light 200 that penetrates the dermis is reflected within the dermis and returns to the outside of the body, as shown by reflected light 201, 202, 203, and 204 in Figure 1. At this time, the reflected light 201-204 becomes light that has been absorbed by water 101, glucose 100, and the like within the dermis. Therefore, the absorbance of the reflected light 201-204 can be determined from the intensity of the incident light 200 and the intensity of the reflected light 201-204. As will be described later, glucose and water have different absorbances at different wavelengths of light, so the glucose concentration can be estimated from the absorbance of the reflected light 201-204. Figure 1 illustrates the principle of estimating glucose concentration in a situation where glucose and water are mixed. In reality, the dermis contains components other than glucose and water that absorb light. For example, other components such as tolazamide and triglyceride are said to be present in the dermis. Therefore, in order to improve the accuracy of estimating the glucose concentration, it is necessary to take other components into consideration. [Detection sensitivity for biological changes] Figure 2 is a graph showing the relationship between the detection sensitivity and changes in glucose, moisture, and epidermal thickness in response to near-infrared light with a wavelength of 1550 nm. In Figure 2, the horizontal axis represents changes in glucose, moisture, and epidermal thickness, and the vertical axis represents detection sensitivity. Here, the detection sensitivity is defined as the sensitivity where the difference between the absorbance before a meal and the absorbance after a meal is used as the reference value. Specifically, it is said that the fluctuation in glucose levels before and after a meal in humans generally results in an increase of about 70 mg / dl of glucose. After measuring the absorbance of the artificial skin, the absorbance when 70 mg / dl is added to the artificial skin is measured and the difference is calculated. This difference in absorbance is used as the reference value for detection sensitivity.

[0022] As shown in Figure 2, the detection sensitivity for glucose fluctuations before and after a meal was 1. Compared to this reference sensitivity, the detection sensitivity for daily fluctuations in human body water (±10%) was 263. This result was obtained by adjusting the water content of the artificial skin, calculating the difference in absorbance, and then calculating the ratio to the reference value of detection sensitivity. The detection sensitivity for variations in epidermal thickness (±10%) was 22. This result was obtained by adjusting the thickness of the artificial skin, calculating the difference in absorbance, and then calculating the ratio of detection sensitivity. That is, compared to a change in absorbance before and after a meal of 1, the change in absorbance for thickness variation was 22, and the change in absorbance for water variation was 263. The reason for the particularly large change in absorbance for water variation is that the glucose concentration in the dermis is 0.1% to 1.0%, while the water concentration is approximately 50%. Because water absorbs light, water dominates the change in absorbance in the dermis. [About optical absorption spectrum] Figure 3 shows the optical absorption spectra of glucose units and water alone. In Figure 3, the horizontal axis represents wavelength (nm) and the vertical axis represents absorbance. In Figure 3, the solid line represents the optical absorption spectrum of glucose alone, and the dashed line represents the optical absorption spectrum of water alone. Figure 3 shows the results of measuring the optical absorption spectra of glucose and water using a device that uses FTIR (Fourier Transform Infrared Spectroscopy) to measure the infrared absorption spectrum specific to the object being measured.

[0023] As shown in Figure 3, the near-infrared wavelength range below 1350 nm has low glucose absorbance, making it difficult to use as a sensor. Furthermore, at wavelengths where the optical absorption spectrum of glucose intersects with that of water, it is impossible to determine whether the fluctuation in absorbance is due to glucose or water. At near-infrared wavelengths of 1550 nm, where glucose absorbance is relatively high, the difference in absorbance between glucose and water is large. However, because water absorbs light, the influence of water on absorbance is unavoidable. Therefore, when using near-infrared light, the absorption of water must also be taken into consideration. Furthermore, the other components mentioned above also exhibit different absorbances at different wavelengths than glucose and water, so these must also be taken into consideration. 〔composition〕

[0024] 4 is a diagram showing an example of the schematic configuration of a glucose concentration estimation device 1 according to the first embodiment. The glucose concentration estimation device 1 includes a trapezoidal prism 2, a first light source 3-1, a second light source 3-2, a third light source 3-3, and a fourth light source 3-4. The glucose concentration estimation device 1 further includes a first triangular prism 4-1, a second triangular prism 4-2, a third triangular prism 4-3 (not shown in FIG. 4), a fourth triangular prism 4-4, a light-receiving module 5, a control circuit 6, and a display device 7.

[0025] The first light source 3-1 emits a first light including far-infrared light of a predetermined wavelength. The first light is near-infrared light including light of any one wavelength in the wavelength range from 1375 nm to 1395 nm. For example, a laser diode that emits the first light can be used as the first light source 3-1.

[0026] The second light source 3-2 emits second light containing far-infrared light of a predetermined wavelength. The second light is near-infrared light containing light of any one wavelength in the wavelength range from 1575 nm to 1595 nm. For example, a laser diode that emits the second light can be used as the second light source 3-2.

[0027] The third light source 3-3 emits third light including far-infrared light of a predetermined wavelength. The third light is near-infrared light including light of any one wavelength in the wavelength range from 1835 nm to 1855 nm. For example, a laser diode that emits the third light can be used as the third light source 3-3.

[0028] The fourth light source 3-4 emits fourth light including far-infrared light of a predetermined wavelength. The fourth light is near-infrared light including light of any one wavelength in the wavelength range from 2175 nm to 2255 nm. For example, a laser diode that emits the fourth light can be used as the fourth light source 3-4. Here, the aforementioned "near-infrared light including any one wavelength" may be near-infrared light including light of any one wavelength included in the wavelength ranges indicated by the respective numerical ranges above, and may be, for example, light of only this one wavelength, or light having light of this one wavelength as its peak wavelength and including an unwanted spectrum that is an unwanted wavelength component other than this wavelength. Furthermore, if the light emitted by the light source includes an unwanted spectrum, a configuration may be provided in which a filter is provided to remove this unwanted spectrum. The first to fourth light sources 3-1 to 3-4 each include a collimating lens inside so that they can emit collimated light. Alternatively, the collimating lens may be provided outside the first to fourth light sources 3-1 to 3-4.

[0029] The trapezoidal prism 2 has two parallel faces of different sizes and four side walls that connect the two parallel faces and are inclined relative to the two parallel faces. The trapezoidal prism is an optical element filled with a medium that transmits predetermined light. Here, the predetermined light is the light used for measurement, i.e., the first to fourth light beams. The medium that transmits the predetermined light is a light-transmitting material such as glass or transparent plastic. Hereinafter, the surface with the larger area of ​​the two parallel surfaces of the trapezoidal prism 2 will be referred to as the bottom surface, and the surface with the smaller area of ​​the two parallel surfaces of the trapezoidal prism 2 will be referred to as the top surface. When estimating the glucose concentration, the human body 150 is pressed against the bottom surface of the trapezoidal prism 2 . In the example shown in the figure, the human body 150 includes fingers, wrists, and arms. The parts of the human body 150 that are pressed against the bottom surface of the trapezoidal prism 2 are not limited to fingers, wrists, and arms.

[0030] The first to fourth triangular prisms 4-1 to 4-4 are optical elements each having a triangular prism shape and filled with a medium that transmits the light used in the measurement. The refractive index of the first to fourth triangular prisms 4-1 to 4-4 is equal to the refractive index of the trapezoidal prism 2. The medium that transmits the predetermined light is a light-transmitting material such as glass or transparent plastic.

[0031] The first to fourth light beams emitted by the first to fourth light sources 3-1 to 3-4 pass through the first to fourth triangular prisms 4-1 to 4-4 and enter the trapezoidal prism 2 from the sidewalls of the trapezoidal prism 2. The first to fourth light beams that entered the trapezoidal prism 2 reach the bottom surface of the trapezoidal prism 2 and are incident on the human body 150 that is pressed against the bottom surface. Meanwhile, the reflected light of the first to fourth light beams that returned from the human body 150 enters the bottom surface of the trapezoidal prism 2 and is focused on the top surface of the trapezoidal prism 2 after multiple reflections within the trapezoidal prism 2 or directly without multiple reflections.

[0032] A light-receiving module 5 is disposed on the top surface of the trapezoidal prism 2. Any module can be used as the light-receiving module 5 as long as it can detect the first to fourth light beams. For example, a module having a light-receiving element using GaAs can be used as the light-receiving module 5.

[0033] Although not shown, the light-receiving module 5 has a structure in which a light-receiving element and a circuit are sealed with resin. For example, the light-receiving element is a photodiode, and the circuit is an IC in which a drive circuit for the light-receiving element and an arithmetic circuit are configured on a single chip. The light-receiving module 5 is integrally molded with resin with a transparent plate such as a glass plate provided on the light-receiving surface side of the light-receiving element, and the surface of the glass plate is exposed. When the first to fourth light beams enter the light-receiving module 5 through the top surface of the trapezoidal prism 2, the light-receiving module 5 outputs an electrical signal according to the intensity of the incident light. The electrical signal output from the light-receiving module 5 is input to the control circuit 6.

[0034] The control circuit 6 drives the first light source 3-1, the second light source 3-2, the third light source 3-3, the fourth light source 3-4 and the light-receiving module 5, and estimates the concentration of glucose in the blood of the living body based on the intensities of the first to fourth light rays detected by the light-receiving module 5. To achieve this, the control circuit 6 includes a processor 11 and a memory 12. The memory 12 pre-stores estimation information 13 for estimating the glucose concentration in blood from the output signal of the light-receiving module 5. Also, various control programs required for estimating the glucose concentration are pre-stored.

[0035] The estimation information 13 is composed of, for example, a trained model that has learned the relationship between information on the reflected light of the first to fourth light beams incident on the human body 150 and glucose concentration, and an information table that is generated using this trained model and indicates the correspondence between information on the reflected light of the first to fourth light beams and glucose concentration. The trained model is, for example, a model that receives a detection value, which is the voltage value of the output signal from the light-receiving module 5 that receives the reflected light of the first to fourth light beams, as input and outputs a glucose concentration. In other words, the detection value becomes information on the reflected light. Note that the trained model may be configured to output another value that can be converted to a glucose concentration rather than the glucose concentration itself. The estimation information 13 may include both a trained model and an information table, or may include only one of them. As a learning method for the learning model, there is a method of performing learning using information on the reflected light of the first to fourth light beams on the human bodies 150 of a plurality of subjects and actual measured values ​​of glucose concentration. In this case, for example, blood is collected from the subjects to obtain actual measured values ​​of glucose concentration.

[0036] Another training method is to use artificial skin, also known as a phantom. In this method, the concentrations of glucose, water, and a third component are set in the dermis layer of the artificial skin, and training is performed on a large number of different combinations of these concentrations. Nanocellulose, for example, can be used as the third component. This third component can be a substitute for other components such as tolazamide and triglyceride mentioned above. The training method using artificial skin generates a trained model for an unspecified number of people.

[0037] In addition, learning algorithms that can be used include, for example, linear regression analysis, MARS (Multivariate Adaptive Regression Splines), support vector regression (SVR), regression trees, model trees, genetic programs, binary classification, logistic regression, k-nearest neighbors, support vector machines, decision trees, random forests, and neural networks. In the first embodiment, as an example, a trained model is generated using deep learning, which is one of the learning methods using a neural network. Furthermore, when learning is performed using attribute information of the subject, for example, the estimation information 13 may include information on the reflected light of the first to fourth light beams and an information table showing the correspondence between the attribute information of the subject and the glucose concentration. The attribute information includes, for example, information such as the subject's age, sex, height, and weight. The processor 11 causes the first light source 3-1, the second light source 3-2, the third light source 3-3, and the fourth light source 3-4 to emit light simultaneously or at different timings. In the case of simultaneous light emission, the processor 11 further acquires a signal output from the light receiving module 5 that indicates the intensity of the reflected light of the mixed light of the first to fourth light beams.

[0038] On the other hand, in the case of different light emission timings, the processor 11 acquires a signal indicating the intensity of the reflected light of the first light, a signal indicating the intensity of the reflected light of the second light, a signal indicating the intensity of the reflected light of the third light, and a signal indicating the intensity of the reflected light of the fourth light, which are output from the light receiving module 5.

[0039] The processor 11 further acquires attribute information of the subject to be estimated when the estimation information 13 includes an information table associated with attribute information of the subject. Then, the processor 11 estimates the glucose concentration in the blood of the subject based on the acquired signals, the attribute information, and the estimation information 13. On the other hand, when the estimation information 13 includes a trained model using artificial skin or an information table generated using this trained model, the processor 11 estimates the glucose concentration in the blood of a subject based on each acquired signal and the estimation information 13.

[0040] The processor 11 may be configured, for example, by a CPU (Central Processing Unit), or may be realized by hardware (circuitry) using an FPGA (Field-Programmable Gate Array) or an ASIC (Application Specific Integrated Circuit), etc. Furthermore, the control circuit 6 may include an amplifier circuit that amplifies the signal from the light receiving module 5, an analog-to-digital conversion circuit that converts the signal into a digital value, and the like, as appropriate. The control circuit 6 sends the estimated blood glucose concentration to the display device 7 as an estimation result.

[0041] The display device 7 is, for example, an LCD (Liquid Crystal Display) or an OLED (Organic Electroluminescent Display). The display device 7 displays the blood glucose concentration sent from the control circuit 6 in a manner that is visible to the user. The display device 7 may display the concentration as numerical information, or may display it as a graph, bar, or the like. The glucose concentration estimation device 1 does not necessarily have to include the display device 7. The glucose concentration estimation device 1 may be provided with a speaker instead of the display device 7, and the estimated glucose concentration in the blood may be output as audio information by the speaker. In addition, the glucose concentration estimation device 1 may be provided with a wired or wireless interface that can be connected to an external device such as a personal computer, and the control circuit 6 may output the estimated glucose concentration in the blood to the external device via the interface. [Configuration of trapezoidal prism 2 and arrangement of components around trapezoidal prism 2]

[0042] Fig. 5(a) is an external view of the trapezoidal prism 2 and first to fourth triangular prisms 4-1 to 4-4 according to the first embodiment, as seen from the top side. Fig. 5(b) is a diagram showing the relative positions of the trapezoidal prism 2 and first to fourth triangular prisms 4-1 to 4-4 and the first to fourth light sources 3-1 to 3-4. Fig. 6A is a diagram showing the incidence of first and third light beams on the first and third triangular prisms 4-1 and 4-3. Fig. 6B is a diagram showing the incidence of second and fourth light beams on the second and fourth triangular prisms 4-2 and 4-4.

[0043] The cross-sectional view of the trapezoidal prism 2 and the first and third triangular prisms 4-1 and 4-3 in Fig. 6A corresponds to the cross-sectional view taken along line A-A' in Fig. 5(a), and the cross-sectional view of the trapezoidal prism 2 and the second and fourth triangular prisms 4-2 and 4-4 in Fig. 6B corresponds to the cross-sectional view taken along line B-B' in Fig. 5(a).

[0044] 6A, reference numeral 1000-1 indicates the optical path of the first light emitted from the first light source 3-1, reference numeral 1000-3 indicates the optical path of the third light emitted from the third light source 3-3, and reference numeral 1000-2 indicates the optical path of the second light emitted from the second light source 3-2, and reference numeral 1000-4 indicates the optical path of the fourth light emitted from the fourth light source 3-4.

[0045] 5(a) and (b) and 6A and 6B, surface 2a is the top surface on which light-receiving module 5 is disposed, surfaces 2b, 2c, 2d, and 2e are side walls, and surface 2f is the bottom surface that is pressed against human body 150. Hereinafter, surface 2a will be referred to as top surface 2a, and surface 2f will be referred to as bottom surface 2f. Furthermore, surfaces 2b, 2c, 2d, and 2e will be referred to as side walls 2b, 2c, 2d, and 2e. A first light is incident on side wall 2b, a second light is incident on side wall 2c, a third light is incident on side wall 2d, and a fourth light is incident on side wall 2e.

[0046] Here, the intensity of light reflected back within the dermis layer is strongest at the light irradiation position and decreases with increasing distance from the light irradiation position. Therefore, the positions and orientations of the first to fourth light sources 3-1 to 3-4 are determined so that the first to fourth light rays are emitted toward the human body 150 from position 401, which is the center of the area of ​​the bottom surface 2f in front of the top surface 2a. When the bottom surface 2f is pressed against the human body 150, the light emitted from position 401 to the human body 150 returns strongly from position 401. The light-receiving module 5 is located directly in front of position 401, where the light returns strongly, and therefore can receive a large amount of light returning from the human body 150.

[0047] A first triangular prism 4-1 is arranged on the side wall 2b where the first light is incident to suppress loss due to reflection of the first light, and a third triangular prism 4-3 is arranged on the side wall 2d where the third light is incident to suppress loss due to reflection of the third light. Similarly, a second triangular prism 4-2 is arranged on the side wall 2c where the second light is incident to suppress loss due to reflection of the second light, and a fourth triangular prism 4-4 is arranged on the side wall 2e where the fourth light is incident to suppress loss due to reflection of the fourth light.

[0048] Specifically, the first light beam is incident perpendicularly on the wall surface 4-1a of the first triangular prism 4-1, the second light beam is incident perpendicularly on the wall surface 4-2a of the second triangular prism 4-2, the third light beam is incident perpendicularly on the wall surface 4-3a of the third triangular prism 4-3, and the fourth light beam is incident perpendicularly on the wall surface 4-4a of the fourth triangular prism 4-4. The refractive indices of the first to fourth triangular prisms 4-1 to 4-4 are equal to the refractive index of the trapezoidal prism 2. [Wavelengths used for the first to fourth lights] Figure 7 is a graph showing the relationship between wavelength and absorbance for combinations of concentrations of glucose, water, and nanocellulose when they are mixed using artificial skin. In Figure 7, the horizontal axis represents wavelength and the vertical axis represents absorbance. As shown in FIG. 7, the waveform group 300 is composed of a plurality of types of waveforms corresponding to different combinations of concentrations of glucose, water, and the third component in the artificial skin.

[0049] Waveform group 300 shows two absorbance peaks: a first peak 310 and a second peak 320. In the wavelength range near first peak 310 and second peak 320, the fluctuations in absorbance are so large that using wavelengths in this range results in large estimation errors. On the other hand, in the wavelength range at the base of first peak 310 and second peak 320, where the fluctuations are relatively small, the fluctuations in absorbance are small and stable.

[0050] Therefore, in the first embodiment, a total of four wavelengths of light are adopted as the first to fourth lights, one selected from each of the foot portions on both the shorter and longer wavelength sides of the first peak 310, and one selected from each of the foot portions on both the shorter and longer wavelength sides of the second peak 320. In the example shown in Figure 7, as indicated by the dashed lines in the figure, "1384 nm" is selected as the wavelength of the first light, "1582 nm" as the wavelength of the second light, "1847 nm" as the wavelength of the third light, and "2216 nm" as the wavelength of the fourth light.

[0051] Although an example of selecting these four wavelengths has been shown, the wavelength of the first light can be selected from the wavelength range of 1375 nm to 1395 nm, the wavelength of the second light can be selected from the wavelength range of 1575 nm to 1595 nm, the wavelength of the third light can be selected from the wavelength range of 1835 nm to 1855 nm, and the wavelength of the fourth light can be selected from the wavelength range of 2175 nm to 2255 nm. [Example of learning method] Fig. 8A is a table showing the correspondence between the detected values ​​of reflected light and the measured values ​​of glucose concentration for subjects A and B. Fig. 8B is a graph showing the table of Fig. 8A. 8A and 8B, the learning method involves repeatedly learning the relationship between the detected value of reflected light when the first to fourth light beams are simultaneously irradiated onto the skin of each subject and the glucose concentration measured by sampling blood. In other words, a trained model is generated for each individual. Here, the detected value is a voltage value of an output signal indicating the intensity of the reflected light from the light receiving module 5, and is expressed in mV. Figure 8A shows an example of data from two subjects, A and B, but if the target of estimation is not a specific individual, it is desirable to conduct learning on a large number of subjects with different attributes such as age, sex, weight, and height. As shown in FIG. 8B, even if the detected values ​​are close to each other, blood glucose levels vary depending on the individual. Instead of learning one detection value when the first to fourth lights are irradiated simultaneously, the relationship between four detection values ​​when the first to fourth lights are irradiated in order at different timings and the actual measurement value of the glucose concentration may be learned.

[0052] On the other hand, although not shown, a learning method can be used in which the water concentration, glucose concentration, and third component concentration of the artificial skin are changed and learning is performed repeatedly. In this case, a trained model for an unspecified number of people will be generated. In this case, too, the relationship with one detection value obtained by simultaneous irradiation may be learned, or the relationship with four detection values ​​corresponding to the first to fourth light may be learned. [Glucose concentration estimation process] Next, the glucose concentration estimation process executed by the control circuit 6 will be described. FIG. 9 is a flowchart showing the glucose concentration estimation process according to the first embodiment. The processor 11 of the control circuit 6 starts a control program stored in a predetermined area of ​​the memory, and executes the glucose concentration estimation process shown in the flowchart of FIG. 9 in accordance with the program. When the glucose concentration estimation process is executed by the processor 11, as shown in FIG. 9, the process first proceeds to step S100.

[0053] In step S100, the first to fourth light sources 3-1 to 3-4 are caused to emit light, the first to fourth light beams are irradiated onto the human body 150 of one subject to be estimated, and an output signal is obtained from the light-receiving module 5 that receives the first to fourth light beams reflected from the human body 150. Then, the process proceeds to step S102.

[0054] Here, the light emission process of the first to fourth light sources 3-1 to 3-4 is performed by using estimation information 13 that includes a trained model corresponding to the detection values ​​when the first to fourth light sources 3-1 to 3-4 are simultaneously emitted, or an information table generated from this trained model. In this case, the first to fourth light sources 3-1 to 3-4 are caused to emit light simultaneously. Then, an output signal of the light receiving module 5 corresponding to the reflected light when the light sources 3-1 to 3-4 are simultaneously emitted is obtained.

[0055] On the other hand, the estimation information includes a trained model corresponding to the detection values ​​when the first to fourth light sources 3-1 to 3-4 are individually emitted at different times, or an information table generated from this trained model. In this case, the output signals of the light-receiving module 5 are acquired in the order in which they were emitted. That is, four output signals corresponding to the reflected light of the first to fourth light sources, respectively, are acquired. In step S102, the glucose concentration of one subject corresponding to the detection value indicated by the signal acquired in response to the emitted light of the four wavelengths is estimated using the estimation information 13 stored in the memory 12. Thereafter, the process proceeds to step S104.

[0056] Here, the estimation information 13 includes a plurality of information tables showing the correspondence between the subject's attribute information, the detected values, and the glucose concentrations, which are generated from a trained model that has learned the relationship between the detected values ​​of the reflected light in response to irradiation with the first to fourth light beams and the actual measured glucose concentrations for each subject. In this case, the attribute information of one subject is acquired, and an information table corresponding to attribute information that is the same as or most similar to the acquired attribute information of the one subject is selected from the plurality of information tables. Then, the glucose concentration corresponding to the detected value that is the same as or most similar to the detected value of the one subject is acquired from the selected information table, thereby estimating the glucose concentration of the one subject.

[0057] If the estimation information 13 includes only a trained model and does not include an information table, a trained model corresponding to attribute information that is the same as or most similar to the acquired attribute information of a subject is selected. That is, the attribute information and the trained model are associated in advance. Then, a detection value corresponding to the output signal acquired in step S100 is input to the selected trained model, and the output of the trained model in response to this input is used as the estimated value of the glucose concentration. Furthermore, if the output value of the trained model does not directly indicate the glucose concentration, a value converted into the glucose concentration is used as the estimated value.

[0058] On the other hand, the estimation information 13 includes an information table showing the correspondence between the detection value and the glucose concentration, which is generated from a trained model that has learned the relationship between the detection value of the reflected light when the first to fourth lights are irradiated onto the artificial skin and the glucose concentration. In this case, the glucose concentration of a subject is estimated by obtaining from the information table the glucose concentration corresponding to the detection value that is the same as or closest to the detection value of the subject. In addition, if the estimation information 13 includes only a trained model and does not include an information table, a detection value corresponding to the acquired output signal is input to the trained model, and the output of the trained model in response to this input is used as an estimated value of the glucose concentration. In step S104, the glucose concentration in the blood estimated in step S102 is output and displayed on the display device 7. After that, the series of processes ends. [Estimation accuracy when using four wavelengths]

[0059] FIG. 10 is a diagram showing the relationship between true and estimated values ​​when using the detected values ​​of reflected light for four wavelengths of light. In FIG. 10, the horizontal axis represents the true glucose concentration value, and the vertical axis represents the estimated glucose concentration value. 2 is the coefficient of determination, and the closer it is to 1, the more accurate the estimation. The dashed line in Figure 10 is an approximate straight line.

[0060] As shown in Figure 10, learning was performed based on the detected values ​​when the four wavelengths were "1384 nm," "1582 nm," "1847 nm," and "2216 nm." As a result, it can be seen that the estimated values ​​and true values ​​are almost identical. In this case, the coefficient of determination is "0.9992," indicating that highly accurate estimation is possible. The estimation results shown in Figure 10 are the estimation results when learning is performed using artificial skin. As described above, learning was performed by varying the concentrations of the three components contained in the artificial skin: glucose, water, and a third component. In the second embodiment described below, it is explained that using a light source with three wavelengths can improve estimation accuracy. The inventors speculated on the reason why the estimation accuracy in the first embodiment was significantly higher than that in the second embodiment. When using a light source with three wavelengths, the coefficient of determination was 0.7996 because there was a measurement error in the sensor, such as a deviation when attaching the sensor to the skin, when measuring the true value. On the other hand, in this embodiment, even if there was a measurement error in this sensor, the coefficient of determination was 0.9992 because four wavelengths were used. Thus, taking into account the measurement error, it is speculated that the estimation accuracy was significantly improved when four wavelengths were used. It is speculated that such a significant improvement in estimation accuracy was achieved by using a light source with four wavelengths or a wavelength range nearby, as disclosed in this embodiment. [Effects of the first embodiment]

[0061] As described above, the glucose concentration estimation device 1 according to the first embodiment includes a first light source 3-1 for irradiating the human body 150 with first light including light of any one wavelength in the wavelength range of 1375 nm to 1395 nm, a second light source 3-2 for irradiating the human body 150 with second light including light of any one wavelength in the wavelength range of 1575 nm to 1595 nm, a third light source 3-3 for irradiating the human body 150 with third light including light of any one wavelength in the wavelength range of 1835 nm to 1855 nm, a fourth light source 3-4 for irradiating the human body 150 with fourth light including light of any one wavelength in the wavelength range of 2175 nm to 2255 nm, a light-receiving module 5 having a light-receiving element that receives light returning from the irradiating of the first to fourth light onto the human body 150, and a control circuit 6 that estimates the glucose concentration based on the output of the light-receiving module 5.

[0062] With this configuration, the glucose concentration can be estimated based on information about the reflected light from irradiation with light in four wavelength ranges where the difference in absorbance between glucose and water contained in the dermis layer is relatively large and the absorbance is stable. This allows the glucose concentration to be estimated according to the change in absorbance due to water, thereby enabling the glucose concentration to be estimated with high accuracy. Furthermore, relatively inexpensive laser diodes can be used as the first to fourth light sources 3-1 to 3-4, and the first to fourth light rays fall within wavelength ranges that do not require special optical materials, allowing products to be manufactured at lower cost than conventional methods.

[0063] In addition, the glucose concentration estimation device 1 according to the first embodiment further includes a memory 12 that stores estimation information 13 including a learned model obtained by previously learning the relationship between information on reflected light when the first to fourth light beams are incident on the human body 150 and the glucose concentration, and the control circuit 6 uses the estimation information 13 stored in the memory 12 to estimate the glucose concentration corresponding to the output of the light receiving module 5.

[0064] With this configuration, the glucose concentration can be estimated based on the estimation information 13 including a trained model that has learned the relationship between the glucose concentration and information on the reflected light in response to the irradiation of the first to fourth light beams onto the human body 150. This makes it possible to more accurately estimate the glucose concentration in response to changes in absorbance due to moisture. In addition, the glucose concentration estimation device 1 of the first embodiment has a trained model that is trained on combinations of multiple different concentrations of glucose, water, and a third component corresponding to another component in the dermis layer set on the artificial skin. This configuration eliminates the need to measure many subjects and allows for easy learning by setting desired concentration combinations, making it easy to generate a trained model that corresponds to an unspecified number of people.

[0065] In addition, in the glucose concentration estimation device 1 according to the first embodiment, the estimation information 13 includes an information table showing the correspondence between the reflected light information and the glucose concentration, which is generated using a trained model obtained by previously learning the relationship between the reflected light information and the glucose concentration when the first to fourth light beams are incident on the human body 150, and the control circuit 6 estimates the glucose concentration corresponding to the output of the light receiving module 5 based on the estimation information 13 stored in the memory 12. With this configuration, the glucose concentration can be estimated by selecting the glucose concentration corresponding to the output of the light receiving module 5 from the information table, so estimation can be performed with simpler processing than calculation processing using a trained model.

[0066] In addition, in the glucose concentration estimation device 1 of the first embodiment, the control circuit 6 acquires attribute information regarding the attributes of the subject, the trained model is a model that learns the relationship between the information on the reflected light in response to the irradiation of the first to fourth lights onto the subject's skin and the glucose concentration based on the information on the reflected light in response to the irradiation of the first to fourth lights onto the subject's skin and the actual measured value of the subject's glucose concentration, and the estimation information 13 is an information table generated using the trained model that shows the correspondence between the subject's attribute information, the information on the reflected light of the subject corresponding to the attribute information, and the glucose concentration of the subject corresponding to the attribute information, and the control circuit 6 uses the information table to estimate the glucose concentration of a single subject from the acquired attribute information of the single subject and the output of the light receiving module 5 corresponding to the single subject.

[0067] With this configuration, the glucose concentration of the subject to be estimated can be estimated using an information table generated from trained models for other subjects with the same or similar attributes as the subject to be estimated, thereby enabling glucose concentrations to be estimated easily and with high accuracy for subjects with the same or similar attributes. [Correspondence in the first embodiment] In the first embodiment, the control circuit 6 and step S102 correspond to an estimation unit and an attribute information acquisition unit, the human body 150 corresponds to a living body, and the memory 12 corresponds to a storage unit. Second Embodiment Next, a second embodiment of the present invention will be described below. Figures 11 to 13 are diagrams showing the second embodiment. 〔composition〕

[0068] Fig. 11 is a diagram showing a part of the schematic configuration of a glucose concentration estimation device 1A according to the second embodiment. Fig. 12 is a flowchart showing the glucose concentration estimation process according to the second embodiment. Fig. 13 is a diagram showing the relationship between the true value and the estimated value when detection values ​​of three wavelengths are used. The second embodiment differs from the first embodiment in that light of three wavelengths, the first to third light, in the first embodiment is incident on the human body 150, and the glucose concentration is estimated based on the detected values ​​of the reflected light of the first to third light. As shown in FIG. 11, a glucose concentration estimation device 1A according to the second embodiment has a configuration in which the fourth light source 3-4 and the fourth triangular prism 4-4 are removed from the glucose concentration estimation device 1 of the first embodiment. A first light beam enters side wall 2b via a first triangular prism 4-1, a second light beam enters side wall 2c via a second triangular prism 4-2, and a third light beam enters side wall 2d via a third triangular prism 4-3. As in the first embodiment, the wavelength of the first light emitted by the first light source 3-1 is "1384 nm," the wavelength of the second light emitted by the second light source 3-2 is "1582 nm," and the wavelength of the third light emitted by the third light source 3-3 is "1847 nm."

[0069] The estimation information 13 according to the second embodiment is composed of, for example, a trained model that has learned the relationship between the detection values ​​of the reflected light of the first to third light beams incident on the human body 150 and the glucose concentration, and an information table that is generated using this trained model and indicates the correspondence relationship between the reflected light of the first to third light beams and the glucose concentration. The trained model is a model that receives, for example, the detection values, which are the voltage values ​​of the output signal of the light-receiving module 5 that receives the reflected light of the first to third light beams, as input, and outputs the glucose concentration. Similarly to the first embodiment, the estimation information 13 according to the second embodiment may include both a trained model and an information table, or may include only one of them. The learning method for the learning model is the same as that of the first embodiment, except that the detection values ​​of the reflected light from the first to third lights are used. [Glucose concentration estimation process] Next, the glucose concentration estimation process executed by the control circuit 6 according to the second embodiment will be described. FIG. 12 is a flowchart showing the glucose concentration estimation process according to the second embodiment. The processor 11 of the control circuit 6 starts a control program stored in a predetermined area of ​​the memory, and executes the glucose concentration estimation process shown in the flowchart of FIG. 12 in accordance with the program. When the glucose concentration estimation process is executed by the processor 11, as shown in FIG. 12, the process first proceeds to step S200.

[0070] In step S200, the first to third light sources 3-1 to 3-3 are caused to emit light, the first to third light beams are irradiated onto the human body 150 of one subject to be estimated, and an output signal is obtained from the light-receiving module 5 that receives the first to third light beams reflected from the human body 150. Then, the process proceeds to step S202. Here, similarly to the first embodiment, the first to third light sources 3-1 to 3-3 are caused to emit light simultaneously or individually at different timings according to the content of the estimation information 13. Then, an output signal from the light receiving module 5 for each reflected light is obtained. In step S202, the glucose concentration of one subject corresponding to the detection value indicated by the signal acquired in response to the light emission at three wavelengths is estimated using the estimation information 13 stored in the memory 12. Thereafter, the process proceeds to step S204.

[0071] Here, the second embodiment differs from the first embodiment in that the trained model is a trained model of the relationship between the detected value of the reflected light for each subject in response to irradiation with the first to third lights and the actual measured value of the glucose concentration, and that the trained model is used to generate the information table. The remaining details of the estimation process are the same as those of the first embodiment. In step S204, the glucose concentration in the blood estimated in step S202 is output and displayed on the display device 7. After that, the series of processes ends. [Estimation accuracy when using three wavelengths]

[0072] Fig. 13 is a diagram showing the relationship between true and estimated values ​​when using the detected values ​​of reflected light for light of three wavelengths. In Fig. 13, the horizontal axis represents the true value of glucose concentration, and the vertical axis represents the estimated value of glucose concentration. 2is the coefficient of determination, and the closer it is to 1, the more accurate the estimation. The dashed line in FIG. 13 is an approximate straight line.

[0073] 13, learning was performed based on the detected values ​​when the three wavelengths were "1384 nm," "1582 nm," and "1847 nm." As a result, it can be seen that the number of cases where the estimated value does not match the true value increases compared to the case of four wavelengths in the first embodiment. However, even in this case, the coefficient of determination is "0.7996," indicating that estimation can be performed with a relatively high degree of accuracy. Furthermore, since the estimation accuracy is higher with four wavelengths than with three wavelengths, it can be predicted that the estimation accuracy can be improved by increasing the number of wavelengths. The estimation results shown in FIG. 13 are estimation results obtained when learning is performed using artificial skin. [Effects of the second embodiment]

[0074] As described above, the glucose concentration estimation device 1A according to the second embodiment includes a first light source 3-1 for irradiating the human body 150 with first light including light of any one wavelength in the wavelength range of 1375 nm to 1395 nm, a second light source 3-2 for irradiating the human body 150 with second light including light of any one wavelength in the wavelength range of 1575 nm to 1595 nm, a third light source 3-3 for irradiating the human body 150 with third light including light of any one wavelength in the wavelength range of 1835 nm to 1855 nm, a light-receiving module 5 having a light-receiving element that receives light returned from the irradiating the human body 150 with the first to third light, and a control circuit 6 that estimates the glucose concentration based on the output of the light-receiving module 5.

[0075] With this configuration, the glucose concentration can be estimated based on information about the reflected light from irradiation with light in three wavelength ranges where the difference in absorbance between glucose and water contained in the dermis layer is relatively large and the absorbance is stable. This allows the glucose concentration to be estimated according to the change in absorbance due to water, thereby enabling highly accurate estimation of the glucose concentration. Furthermore, relatively inexpensive laser diodes can be used as the first to third light sources 3-1 to 3-3, and the first to third light beams fall within wavelength ranges that do not require special optical materials, allowing products to be manufactured at lower cost than conventional methods. Furthermore, compared to the case of using four wavelengths of light in the first embodiment, the number of light sources and triangular prisms can be reduced, allowing products to be manufactured at even lower cost.

[0076] In addition, the glucose concentration estimation device 1A according to the second embodiment further includes a memory 12 that stores estimation information 13 including a learned model obtained by previously learning the relationship between information on reflected light when the first to third light beams are incident on the human body 150 and the glucose concentration, and the control circuit 6 uses the estimation information 13 stored in the memory 12 to estimate the glucose concentration corresponding to the output of the light receiving module 5.

[0077] With this configuration, the glucose concentration can be estimated based on the estimation information 13 including a trained model that has learned the relationship between the glucose concentration and information on the reflected light in response to the irradiation of the first to third light beams onto the human body 150. This makes it possible to more accurately estimate the glucose concentration in response to changes in absorbance due to moisture. In addition, the glucose concentration estimation device 1A of the second embodiment has a trained model that is trained on a combination of multiple different concentrations of glucose, water, and a third component corresponding to another component in the dermis layer set on the artificial skin. This configuration eliminates the need to measure many subjects and allows for easy learning by setting desired concentration combinations, making it easy to generate a trained model that corresponds to an unspecified number of people.

[0078] In addition, in the glucose concentration estimation device 1A of the second embodiment, the estimation information 13 includes an information table showing the correspondence between reflected light information and glucose concentration, which is generated using a trained model obtained by previously learning the relationship between reflected light information and glucose concentration when the first to third light beams are incident on the human body 150, and the control circuit 6 estimates the glucose concentration corresponding to the output of the light receiving module 5 based on the estimation information 13 stored in the memory 12. With this configuration, the glucose concentration can be estimated by selecting the glucose concentration corresponding to the output of the light receiving module 5 from the information table, so estimation can be performed with simpler processing than calculation processing using a trained model.

[0079] In addition, in the glucose concentration estimation device 1A of the second embodiment, the control circuit 6 acquires attribute information regarding the attributes of the subject, the trained model is a model that learns the relationship between the information on the reflected light in response to the irradiation of the first to third lights onto the subject's skin and the glucose concentration based on the information on the reflected light in response to the irradiation of the first to third lights onto the subject's skin and the actual measured value of the subject's glucose concentration, and the estimation information 13 is an information table generated using the trained model that shows the correspondence between the subject's attribute information, the information on the reflected light of the subject corresponding to the attribute information, and the glucose concentration of the subject corresponding to the attribute information, and the control circuit 6 uses the information table to estimate the glucose concentration of a single subject from the acquired attribute information of the single subject and the output of the light receiving module 5 corresponding to the single subject.

[0080] With this configuration, the glucose concentration of the subject to be estimated can be estimated using an information table generated from trained models for other subjects with the same or similar attributes as the subject to be estimated, thereby enabling glucose concentrations to be estimated easily and with high accuracy for subjects with the same or similar attributes. [Corresponding relationship in the second embodiment] In the second embodiment, the control circuit 6 and step S202 correspond to an estimation unit and an attribute information acquisition unit, the human body 150 corresponds to a living body, and the memory 12 corresponds to a storage unit. [Modification]

[0081] In the first embodiment, four wavelengths are used: one wavelength in the wavelength range of 1375 nm to 1395 nm, one wavelength in the wavelength range of 1575 nm to 1595 nm, one wavelength in the wavelength range of 1835 nm to 1855 nm, and one wavelength in the wavelength range of 2175 nm to 2255 nm. However, the present invention is not limited to this configuration, and five or more wavelengths may be used from these wavelength ranges, or wavelengths from other wavelength ranges may be added.

[0082] In the second embodiment, three wavelengths are used: one wavelength in the wavelength range of 1375 nm to 1395 nm, one wavelength in the wavelength range of 1575 nm to 1595 nm, and one wavelength in the wavelength range of 1835 nm to 1855 nm. However, the present invention is not limited to this configuration, and four or more wavelengths may be used from these wavelength ranges, or wavelengths from other wavelength ranges may be added.

[0083] Furthermore, in the above embodiment and its modified examples, a configuration has been described in which an individual trained model is generated for each subject and attribute information is associated with an individual information table generated using the trained model, but this configuration is not limited to this. A trained model may be generated that learns the relationship between the subject's attribute information, information about the reflected light from the subject, and the subject's actual measured glucose concentration. With this configuration, the relationship between attributes can also be learned, so that by inputting information about the reflected light of a single subject and the attribute information of the single subject into the trained model, it is possible to estimate the glucose concentration corresponding to the attribute information of the single subject. [Explanation of symbols]

[0084] 1, 1A... glucose concentration estimation device, 2... trapezoidal prism, 3-1 to 3-4... first to fourth light sources, 4-1 to 4-4... first to fourth triangular prisms, 5... light receiving module, 6... control circuit, 7... display device, 11... processor, 12... memory, 13... estimation information, 150... human body

Claims

1. a first light source for irradiating a living body with first light, which is light including any one wavelength in a wavelength range from 1375 nm to 1395 nm; a second light source for irradiating the living body with second light, which is light including any one wavelength in a wavelength range of 1575 nm to 1595 nm; At least three light sources including a third light source for irradiating the living body with third light, the third light being light including any one wavelength in a wavelength range of 1835 nm to 1855 nm; a light receiving element that receives light returning from irradiation of the living body with light of a plurality of wavelengths including at least the first to third light; an estimation unit that estimates a glucose concentration based on an output of the light receiving element; A glucose concentration estimation device comprising:

2. In claim 1, At least four light sources are provided, further including a fourth light source for irradiating the living body with fourth light, which is light having any one wavelength in a wavelength range of 2175 nm to 2255 nm; The light receiving element is configured to be able to receive light reflected from the living body when the living body is irradiated with light of a plurality of wavelengths including at least the first to fourth light beams.

3. In claim 1 or 2, a storage unit configured to store estimation information including a trained model obtained by previously training a relationship between information on reflected light when light of the plurality of wavelengths is incident on the living body and a glucose concentration; The estimation unit is a glucose concentration estimation device that estimates a glucose concentration corresponding to an output of the light receiving element using the information for estimation stored in the storage unit.

4. In claim 3, The trained model is a glucose concentration estimation device that has been trained on a combination of multiple different concentrations of glucose, water, and a third component corresponding to another component in the dermis layer set for artificial skin.

5. In claim 3, further comprising an attribute information acquisition unit for acquiring attribute information relating to attributes of the subject; the trained model is a model that has trained a relationship between attribute information of a subject, information on the reflected light reflected from the skin of the subject, and the glucose concentration based on the attribute information, information on the reflected light reflected from the skin of the subject, and an actual measurement value of the glucose concentration of the subject; The estimation unit uses the estimation information to estimate the glucose concentration of a single subject from the attribute information of the single subject acquired by the attribute information acquisition unit and the output of the light receiving element corresponding to the single subject.

6. In claim 1 or 2, a storage unit configured to store estimation information including information indicating a correspondence relationship between information on reflected light and a glucose concentration, the information being generated using a trained model obtained by previously training the relationship between information on reflected light and a glucose concentration when light of the plurality of wavelengths is incident on the living body; The estimation unit is a glucose concentration estimation device that estimates a glucose concentration corresponding to an output from the light receiving element based on the information for estimation stored in the storage unit.

7. In claim 6, The trained model is a glucose concentration estimation device that has been trained on a combination of multiple different concentrations of glucose, water, and a third component corresponding to another component in the dermis layer set for artificial skin.

8. In claim 6, further comprising an attribute information acquisition unit for acquiring attribute information relating to attributes of the subject; the trained model is a model that has trained, for each subject, a relationship between information on the reflected light reflected from the skin of the subject and the glucose concentration based on the information on the reflected light and an actual measurement value of the glucose concentration of the subject; the information indicating the correspondence relationship is information indicating a correspondence relationship between attribute information of a subject, information on the reflected light of the subject corresponding to the attribute information, and a glucose concentration of the subject corresponding to the attribute information, The estimation unit uses the estimation information to estimate the glucose concentration of a single subject from the attribute information of the single subject acquired by the attribute information acquisition unit and the output of the light receiving element corresponding to the single subject.

Images