Temperature estimation system and temperature estimation method
The system uses multiple temperature and heart rate sensors with machine learning to adaptively estimate internal body temperature, addressing transient errors and environmental variations, ensuring accurate core body temperature measurement.
Patent Information
- Application Number
- JP2023528810
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- Filing Date
- 2021-06-15
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2041-06-15
AI Technical Summary
Conventional methods for non-invasive temperature estimation of a living body suffer from transient errors due to changing thermal resistance and blood flow, requiring constant calibration of proportionality coefficients, and are inaccurate during dynamic movements or varying environmental conditions.
A temperature estimation system using multiple temperature sensors and a heart rate sensor, combined with machine learning, to estimate the internal temperature of a living body by learning the proportionality coefficient based on surface and internal temperatures, and heart rate data, without the need for constant recalibration.
Accurately estimates internal body temperature with reduced estimation errors, even during dynamic movements and varying conditions, by adaptively adjusting to the subject's state and environment, thus providing precise core body temperature readings.
Smart Images

Figure 0007593494000002 
Figure 0007593494000003 
Figure 0007593494000004
Abstract
Description
[Technical field]
[0001] The present invention relates to a temperature estimation system and a temperature estimation method for non-invasively and accurately estimating the internal temperature of a subject such as a living body. [Background technology]
[0002] Recent research in chronobiology has revealed that the human circadian rhythm, or so-called biological clock, is closely related to various aspects of our bodies, including not only the quality of sleep, exercise, and work, but also the effectiveness of medication and the onset of diseases. Although the circadian rhythm is almost constant, it is known to vary greatly depending on the light we are exposed to in our daily lives, exercise, diet, as well as age and gender.
[0003] Deep body temperature is known as an index for measuring circadian rhythms, but the most common methods for measuring deep body temperature are to insert a thermometer into the rectum or to measure the temperature of the eardrum with the ear sealed, which are very stressful methods for measuring deep body temperature during daily activities or while sleeping.
[0004] On the other hand, there is a technique for non-invasively measuring the deep body temperature of a living body, which estimates the deep body temperature of a living body by replacing the heat flow with a one-dimensional equivalent circuit model (see Patent Document 1 and Non-Patent Document 1).
[0005] The method disclosed in Patent Document 1 and Non-Patent Document 1 uses a thermal equivalent circuit model of a living body 100 and a sensor 101 as shown in FIG. cbt The deep body temperature T of the living body 100 is estimated. cbt is the thermal resistance R sensor When the sensor 101 having the above-mentioned temperature is placed on the living body 100, the temperature T skin and the temperature T of the top surface of the sensor 101 opposite to the surface in contact with the living body 100. top Therefore, it can be estimated using equation (1). T cbt =T skin +α×Hskin (1)
[0006] Here, H skin is the heat flux on the skin surface of the living body 100, and is expressed by equation (2). H skin =(T skin -T top ) / R skin (2) In addition, α is the thermal resistance R of the living body 100 body The proportionality factor related to skin is the thermal resistance of the sensor 101.
[0007] However, in the conventional methods disclosed in Patent Document 1 and Non-Patent Document 1, the flow of heat transported from the living body 100 through the sensor 101 to the outside air is assumed to be steady. Therefore, when the living body 100 is exposed to wind, when the living body 100 runs, or when the living body 100 moves suddenly from a warm place to a cold place, the deep body temperature T cbt However, there was a problem that transient errors occurred in the estimation of the
[0008] In addition, in the conventional method, the thermal resistance R body is assumed to be constant regardless of time, and the proportionality coefficient α is also assumed to be constant. However, the state of blood flow near the skin of the living body 100 changes depending on the posture and movement of the living body 100. For this reason, the thermal resistance R body is not constant but changes from time to time.
[0009] When a person goes from indoors to outdoors or lies down in daily life, the core body temperature T estimated by the conventional method cbt and true deep body temperature (tympanic membrane temperature) T measured by a tympanic membrane thermometer. ref This is shown in FIG.
[0010] True core temperature T ref and the estimated core body temperature T cbt The difference between the temperature T top and the temperature of the human skin surface T skinThis is due to the difference in the time it takes for the temperature to settle into a steady state, and the fact that blood flow changes depending on a person's posture. In addition, in a state where the wind is constantly changing, it is not possible to expect the temperature to settle into a steady state, even if a person is at rest and there is no change in blood flow. Furthermore, in the conventional method, even when the heat flow can be assumed to be steady, it is necessary to calibrate the proportionality coefficient α each time a measurement is made using a separate sensor such as a tympanic thermometer.
[0011] The temperature distribution within an object can generally be described by the following heat conduction equation:
[0012]
number
[0013] In equation (3), T is temperature, k is thermal conductivity, c is heat capacity, ρ is density, and Q is internal heat generation. Q is a term that occurs during motion, etc. Δ is the second derivative operator with respect to space, and in a three-dimensional orthogonal coordinate system, it is expressed as ∂ 2 / ∂x 2 +∂ 2 / ∂y 2 +∂ 2 / ∂z 2 The thermal property values k, ρ, and c change from moment to moment due to various factors, such as the moisture content of human skin, the expansion and contraction of capillaries due to human activity, sweating, and the expansion and contraction of blood vessels due to blood pressure that changes with a person's posture, etc. Since α in formula (1) is a proportionality coefficient equivalent to k / (ρc), the proportionality coefficient α also changes from moment to moment.
[0014] The conventional methods disclosed in Patent Document 1 and Non-Patent Document 1 use a one-dimensional heat equivalent circuit model, but as mentioned above, when parameters such as heat capacity c and density ρ change, the one-dimensional heat equivalent circuit model no longer holds. In other words, the internal temperature can be estimated for an object in which the flow of heat is sufficiently stable, but the internal temperature cannot be estimated for an unsteady, dynamic object such as a living body. For this reason, it is necessary to somehow take into account equation (3) and calculate the internal temperature (core body temperature T cbt) or the proportionality coefficient α. [Prior art documents] [Patent documents]
[0015] [Patent Document 1] JP 2020-003291 A [Non-patent literature]
[0016] [Non-Patent Document 1] K.Kitamura et al., “Development of a new method for the noninvasive measurement of deep body temperature without a heater”,Medical Engineering & Physics,vol.32,No.1,pp.1-6,2010 Summary of the Invention [Problem to be solved by the invention]
[0017] The present invention has been made to solve the above-mentioned problems, and aims to provide a temperature estimation system and a temperature estimation method that can reduce estimation errors in the internal temperature of a subject such as a living organism without having to calibrate the proportionality coefficient each time a measurement is performed. [Means for solving the problem]
[0018] The temperature estimation system of the present invention is characterized in that it comprises an insulating material, a first temperature sensor provided on the surface of the insulating material facing a test object and configured to measure the temperature of the surface of the test object, a second temperature sensor configured to measure the temperature inside the insulating material directly above the first temperature sensor, a third temperature sensor configured to measure the temperature of the surface of the test object at a position away from the first temperature sensor, a learning device configured to estimate a proportionality coefficient related to the thermal resistance of the test object based on the measurement results of the first, second, and third temperature sensors, and a temperature calculation unit configured to calculate the internal temperature of the test object based on the measurement results of the first and second temperature sensors and the proportionality coefficient. Furthermore, one configuration example of the temperature estimation system of the present invention further includes a heart rate measuring unit configured to measure the heart rate of the subject, and the learning device estimates the proportionality coefficient based on the measurement results of the first, second, and third temperature sensors and the measurement result of the heart rate measuring unit.
[0019] The temperature estimation system of the present invention is characterized in that it comprises an insulating material, a first temperature sensor provided on the surface of the insulating material facing a test subject and configured to measure the temperature of the surface of the test subject, a second temperature sensor configured to measure the temperature inside the insulating material directly above the first temperature sensor, a third temperature sensor configured to measure the temperature of the surface of the test subject at a position away from the first temperature sensor, and a learning device configured to estimate the internal temperature of the test subject based on the measurement results of the first, second, and third temperature sensors. Furthermore, one configuration example of the temperature estimation system of the present invention further includes a heart rate measuring unit configured to measure the heart rate of the subject, and the learning device estimates the internal temperature of the subject based on the measurement results of the first, second, and third temperature sensors and the measurement result of the heart rate measuring unit.
[0020] Furthermore, one configuration example of the temperature estimation system of the present invention is characterized in that it further comprises a selection unit configured to include a plurality of learning devices prepared in advance to correspond to the state of the subject or the environment surrounding the subject, and to select a learning device from the plurality of learning devices that corresponds to the state of the subject or the environment surrounding the subject for the estimation based on at least one of the measurement results of the first temperature sensor and the measurement results of the second temperature sensor. Furthermore, one configuration example of the temperature estimation system of the present invention is characterized in that it further comprises a selection unit configured to include a plurality of learning devices prepared in advance to correspond to the state of the subject or the environment surrounding the subject, and to select a learning device from the plurality of learning devices that corresponds to the state of the subject or the environment surrounding the subject for the estimation based on at least one of the measurement result of the first temperature sensor, the measurement result of the second temperature sensor, and the measurement result of the heart rate measurement unit.
[0021] In addition, the temperature estimation method of the present invention is characterized by including a first step of measuring the surface temperature of the specimen using a first temperature sensor provided on the surface of the insulation facing the specimen, a second step of measuring the temperature inside the insulation directly above the first temperature sensor using a second temperature sensor, a third step of measuring the surface temperature of the specimen at a position away from the first temperature sensor using a third temperature sensor, a fourth step of estimating a proportionality coefficient related to the thermal resistance of the specimen based on the measurement results of the first, second and third steps using a trained learning device, and a fifth step of calculating the internal temperature of the specimen based on the measurement results of the first and second steps and the proportionality coefficient. Furthermore, one configuration example of the temperature estimation method of the present invention further includes a sixth step of measuring the heart rate of the subject, and the fourth step includes a step of estimating the proportionality coefficient based on the measurement results of the first, second, and third steps and the measurement result of the sixth step. Effect of the Invention
[0022] According to the present invention, by estimating a proportionality coefficient based on the measurement results of the first, second, and third temperature sensors, or based on the measurement results of the first, second, and third temperature sensors and the measurement results of the heart rate measuring unit, and calculating the internal temperature of the subject, it is possible to estimate the internal temperature of the subject with high accuracy without having to calibrate the proportionality coefficient each time measurement is performed.
[0023] In addition, in the present invention, by estimating the internal temperature of the subject based on the measurement results of the first, second, and third temperature sensors, or based on the measurement results of the first, second, and third temperature sensors and the measurement results of the heart rate measuring unit, the internal temperature of the subject can be estimated with high accuracy without having to calibrate the proportionality coefficient each time measurement is performed. [Brief description of the drawings]
[0024] [Figure 1] FIG. 1 is a schematic diagram showing a sensor placed on the surface of a living body. [Diagram 2] FIG. 2 is a block diagram showing a configuration of a temperature estimation system according to a first embodiment of the present invention. [Diagram 3] FIG. 3 is a plan view showing the surface of the housing of the temperature measuring unit in FIG. [Figure 4] FIG. 4 is a block diagram showing another example of the temperature measuring unit according to the first embodiment of the present invention. [Diagram 5] FIG. 5 is a plan view showing the surface of the housing of the temperature measuring unit in FIG. [Figure 6] FIG. 6 is a block diagram showing another example of the temperature measuring unit according to the first embodiment of the present invention. [Figure 7] FIG. 7 is a block diagram showing another example of the temperature measuring unit according to the first embodiment of the present invention. [Figure 8] FIG. 8 is a block diagram showing another example of the temperature measuring unit according to the first embodiment of the present invention. [Figure 9] FIG. 9 is a plan view showing the surface of the housing of the temperature measuring unit in FIG. [Figure 10] FIG. 10 is a block diagram showing another example of the temperature measuring unit according to the first embodiment of the present invention. [Figure 11] FIG. 11 is a block diagram showing another example of the temperature measuring unit according to the first embodiment of the present invention. [Figure 12] FIG. 12 is a plan view showing the surface of the housing of the temperature measuring unit in FIG. [Figure 13] FIG. 13 is a plan view showing another example of the arrangement of the temperature sensors according to the first embodiment of the present invention. [Figure 14] FIG. 14 is a plan view showing another example of the arrangement of the temperature sensors according to the first embodiment of the present invention. [Figure 15] FIG. 15 is a block diagram showing another example of the temperature measuring unit according to the first embodiment of the present invention. [Figure 16] FIG. 16 is a flowchart illustrating the operation of the temperature estimation system according to the first embodiment of the present invention. [Figure 17] FIG. 17 is a diagram for explaining a method for generating a tensor in the first embodiment of the present invention. [Figure 18] FIG. 18 is a flowchart illustrating another operation of the temperature estimation system according to the first embodiment of the present invention. [Figure 19A] FIG. 19A is a diagram showing an example of deep body temperature estimated by the method according to the first embodiment of the present invention. [Figure 19B] FIG. 19B is a diagram showing an example of true core body temperature measured by a tympanic thermometer. [Figure 20] FIG. 20 is a block diagram showing a configuration of a server device of a temperature estimation system according to the second embodiment of the present invention. [Figure 21] FIG. 21 is a flowchart illustrating the operation of the temperature estimation system according to the second embodiment of the present invention. [Figure 22] FIG. 22 is a flowchart illustrating another operation of the temperature estimation system according to the first embodiment of the present invention. [Diagram 23] FIG. 23 is a flowchart illustrating another operation of the temperature estimation system according to the first embodiment of the present invention. [Figure 24] FIG. 24 is a flowchart illustrating another operation of the temperature estimation system according to the second embodiment of the present invention. [Diagram 25] FIG. 25 is a block diagram showing an example of the configuration of a computer that realizes the temperature estimation system according to the first and second embodiments of the present invention. [Figure 26] FIG. 26 is a diagram showing a thermal equivalent circuit model of a living body and a sensor. [Figure 27] FIG. 27 is a diagram showing an example of deep body temperature estimated by a conventional method and a true deep body temperature measured by a tympanic thermometer. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0025] [Principle of the invention] If values corresponding to ∂T / ∂t, ΔT, and Q in equation (3) can be measured, k / (cρ) in equation (3) can be estimated. A schematic diagram of a sensor 101a placed on the surface of a living body 100 is shown in FIG. 1. The heat conduction equation shown in equation (3) also holds true in the living body 100 near the sensor 101a.
[0026] The thermal property values k, ρ, and c are parameters that change from moment to moment, but the spatial change in the properties of k, ρ, and c is relatively small. In other words, although the values of k, ρ, and c change from moment to moment, the assumption that they are almost uniform spatially is valid. The heat conduction equation of equation (3) in the vicinity of the sensor 101a is expressed as the deep temperature T cbt In addition, the proportionality coefficient α related to the thermal resistance of the living body 100 corresponds to k / (cρ), and is therefore related to the heat conduction equation of formula (3) in the vicinity of the sensor 101a.
[0027] Therefore, based on the relationship of equation (3) in the vicinity of the sensor 101a, the deep body temperature T cbt It is also possible to estimate the deep body temperature T from equation (1) using the proportionality coefficient α estimated based on equation (3). cbt can be estimated.
[0028] Considering the relationship of formula (3) near the sensor 101a, in order to estimate ∂T / ∂t,ΔT, it is necessary to measure the time series data of temperature, the temperature distribution on the surface of the living body 100, and the heat flowing from the surface of the living body 100 to the sensor 101a. For example, in the conventional temperature T skin ,T top In addition, T skin The temperature of the skin surface at a distance T side If you measure T skin and T side and T skin and T top A value corresponding to ΔT can be obtained from the difference between Q and the intensity of the exercise of the living body 100. On the other hand, Q is closely related to the intensity of the exercise of the living body 100, and therefore can be inferred from the heart rate N and the amount of activity A of the living body 100.
[0029] That is, temperature T skin ,T top ,T side and heart rate N and core body temperature T cbt The function f that shows the relationship between cbt can be estimated. T cbt =f(T skin ,T top ,T side ,N) (4)
[0030] Also, the temperature T skin ,T top ,T side The proportionality coefficient α can be estimated using a function g that indicates the relationship between the heart rate N and the proportionality coefficient α. α = g(T skin ,T top ,T side ,N) (5)
[0031] In general, equation (3) and T skin ,T top ,T side , N using elementary functions, it is possible to associate them using dimensionless numbers or convolutional neural networks.
[0032] [First Example] Hereinafter, embodiments of the present invention will be described with reference to the drawings. Fig. 2 is a block diagram showing the configuration of a temperature estimation system according to a first embodiment of the present invention. The temperature estimation system includes a temperature measurement unit 1 (temperature measurement device), a heartbeat measurement unit 2 (heartbeat measurement device), a terminal 3 such as a PC (Personal Computer) or a smartphone, and a server device 4.
[0033] The temperature measuring unit 1 measures the temperature T skin The temperature sensor 10 measures the temperature T top The temperature sensor 11 measures the temperature T of the skin surface of the living body 100 at a position away from the temperature sensor 10. side A temperature sensor 13 for measuring the temperature T skin ,T top ,T side The data is provided to the terminal 3 through a communication unit 15, and a control unit 16 that controls reading and writing of data to the storage unit 14 and communication.
[0034] The temperature measuring unit 1 is arranged so that the surface of the housing 17 made of resin and the insulating material 12 exposed on this surface come into contact with the skin of the living body 100. The temperature sensor 10 is provided on the surface of the insulating material 12 facing the living body. The temperature sensor 11 is provided inside the insulating material 12 directly above the temperature sensor 10. The insulating material 12 holds the temperature sensors 10 and 11, and acts as a resistor against heat flowing into the temperature sensor 11. An example of the material for the insulating material 12 is PET resin. The temperature sensor 13 is arranged at a position away from the temperature sensor 10 so as to come into contact with the skin of the living body 100.
[0035] Fig. 3 is a plan view showing the surface of the housing 17 of the temperature measuring unit 1 that comes into contact with the living body 100. As shown in Fig. 3, the temperature sensor 13 is disposed at one location around the temperature sensor 10. As described later, a plurality of temperature sensors 13 may be disposed around the temperature sensor 10.
[0036] The heart rate measuring unit 2 measures the heart rate N of the living body 100 by, for example, photoplethysmography. An example of the heart rate measuring unit 2 is a wristwatch-type heart rate measuring device.
[0037] The server device 4 includes a communication unit 40 for transmitting and receiving data to and from the terminal 3, a storage unit 41 for storing data, and a temperature T skin ,T side ,T top and the heart rate N into a tensor. skin ,T side ,T top Proportionality coefficient α to temperature T skin ,T side ,T top A trained learning device 43 for estimating a proportionality coefficient α for the heart rate N, a machine learning unit 44 for performing machine learning on the learning device 43, and a deep body temperature T cbt The temperature calculator 45 calculates the internal temperature.
[0038] Fig. 4 is a diagram showing another example of the temperature measurement unit. The temperature measurement unit 1a in Fig. 4 is equipped with a plurality of temperature sensors 13. Fig. 5 is a plan view showing the surface of the housing 17 of the temperature measurement unit 1a that comes into contact with the living body 100. As shown in Fig. 5, a plurality of temperature sensors 13 are arranged around the temperature sensor 10 on the surface of the housing 17 of the temperature measurement unit 1a.
[0039] Fig. 6 is a diagram showing another example of the temperature measurement unit. The temperature measurement unit 1b in Fig. 6 is configured by separating a housing 17 that houses the temperature sensors 10, 13 and the heat insulating material 12 from a housing 18 that houses the memory unit 14, the communication unit 15, and the control unit 16. The temperature sensors 10, 13 are connected to the device on the housing 18 side by wiring 19. The arrangement of the temperature sensors 10, 13 as viewed from the living body 100 side is the same as that in Fig. 5.
[0040] Fig. 7 is a diagram showing another example of a temperature measurement unit. The temperature measurement unit 1c of Fig. 7 is obtained by arranging a plurality of insulating materials 20 around the insulating material 12 in the temperature measurement unit 1b of Fig. 6. The temperature sensor 10 is provided on the surface of the insulating material 12 facing the living body. The temperature sensor 13 is provided on the surface of the insulating material 20 facing the living body. The temperature sensor 11 is provided inside the insulating materials 12, 20 directly above the temperature sensors 10, 13. The arrangement of the temperature sensors 10, 13 as viewed from the living body 100 side is the same as in Fig. 5.
[0041] Fig. 8 is a diagram showing another example of the temperature measurement unit. The temperature measurement unit 1d in Fig. 8 is obtained by arranging temperature sensors 13 in two layers around the temperature sensor 10 in the temperature measurement unit 1b in Fig. 6. Fig. 9 is a plan view showing the surface of the housing 17 of the temperature measurement unit 1d that comes into contact with the living body 100.
[0042] FIG. 10 is a diagram showing another example of the temperature measuring unit. The temperature measuring unit 1e of FIG. 10 is the temperature measuring unit 1a of FIG. 4, in which an internal structure 21 for suppressing the influence of outside air and wind is provided inside the housing 17. The temperature sensors 10, 11, and 13 are isolated from the outside air by the internal structure 21. The internal structure 21 is made of a material with good thermal conductivity, such as aluminum. The internal structure 21 is, for example, in a truncated cone or dome shape, and covers the temperature sensors 10, 11, and 13 and the insulating material 12. The arrangement of the temperature sensors 10 and 13 as viewed from the living body 100 side is the same as that of FIG. 5.
[0043] Fig. 11 is a diagram showing another example of a temperature measuring unit. The temperature measuring unit 1f of Fig. 11 has a configuration similar to that of the temperature measuring unit 1e of Fig. 10. However, the temperature sensor 13 is arranged on the outside of the internal structure 21, not on the inside. Fig. 12 is a plan view showing the surface of the housing 17 of the temperature measuring unit 1f that comes into contact with the living body 100.
[0044] In addition, in the temperature measurement units 1a, 1b, 1c, 1e, and 1f, the temperature sensors 13 may be arranged as shown in Fig. 13. In addition, in the temperature measurement units 1a, 1b, 1c, 1e, and 1f, the temperature sensors 13 may be arranged as shown in Fig. 14.
[0045] 15 is a diagram showing another example of a temperature measurement unit. A temperature measurement unit 1g in FIG. 15 incorporates a heart rate measurement unit 2 therein. The arrangement of temperature sensors 10, 13 as viewed from the living body 100 side is the same as in FIG.
[0046] Fig. 16 is a flowchart for explaining the operation of the temperature estimation system of this embodiment. Note that Fig. 16 explains an example in which the heart rate measurement unit 2 is not used.
[0047] The temperature sensor 10 of the temperature measuring unit 1, 1a to 1g measures the temperature T skin The temperature sensor 13 measures the temperature T of the skin surface of the living body 100 at a position distant from the temperature sensor 10. side The temperature sensor 11 measures the temperature T top (Step S100 in FIG. 16). The measurement data of the temperature sensors 10, 13, and 11 are temporarily stored in the memory unit 14. The temperature sensors 10, 13, and 11 each measure a temperature T skin ,T side ,T top is measured at regular intervals, for example, every second.
[0048] The communication unit 15 of the temperature measurement unit 1, 1a to 1g detects the temperature T skin ,T side ,T top The data is transmitted to a terminal 3 such as a PC or a smartphone (FIG. 16, step S101). Terminal 3 receives the temperature T skin ,T side ,T top The communication unit 40 of the server device 4 transmits the received temperature T skin ,T side ,T top The data is stored in the storage unit 41.
[0049] Next, after the time τ (for example, 1024 seconds) required for acquiring time-series data required for the data set used in deep learning has elapsed (YES in step S103 of FIG. 16), the tensor generation unit 42 of the server device 4 generates the temperature Tskin ,T side ,T top Each time series data is converted into a tensor (Step S104 in FIG. 16).
[0050] For example, the temperature T measured every second as shown in Fig. 17 skin ,T side ,T top The tensor generator 42 generates time series data for temperature T skin ,T side ,T top When converting each time series data of 1024 seconds into a tensor, the temperature T skin ,T side ,T top Then, 32°C is subtracted from each of the data, and the subtracted data is divided by 4°C to obtain the temperature T skin ,T side ,T top Then, as shown in the example of tensor Te in Figure 17, the normalized temperature T skin ,T side ,T top All you have to do is arrange the data for each time period as pixels in a tensor (image).
[0051] The temperature T top When measuring the temperature T top Normalize the normalized temperature T top The data for each time can be arranged as pixels of a tensor. Similarly, the temperature T side When measuring the temperature T side Normalize the normalized temperature T side All we need to do is arrange the data for each time period as pixels in a tensor.
[0052] Next, the learning device 43 of the server device 4 calculates the temperature T skin ,T side ,T top and the proportionality coefficient α, or temperature T skin ,T side ,T topIt is a model constructed by software that models the relationship between the heart rate N and the proportionality coefficient α. The learning device 43 outputs an estimation result of the proportionality coefficient α when a tensor is input from the tensor generating unit 42 (step S105 in FIG. 16). An example of the learning device 43 is a CNN (Convolutional Neural Network) learning device.
[0053] The learning device 43 needs to be trained in advance. Specifically, in a preliminary test, the temperature T skin ,T side ,T top At the same time, the true deep body temperature T ref Obtain the time series data of the temperature T skin ,T top ,T ref The proportionality coefficient α is calculated for each time period using the data in accordance with equation (6), thereby obtaining time-series data for the proportionality coefficient α. α = (T ref -T skin ) / (T skin -T top ) ···(6)
[0054] And the temperature T skin ,T side ,T top The time series data of temperature T skin ,T side ,T top Each of the time series data is used as an input variable of the learning device 43, and the proportionality coefficient α is used as an output variable of the learning device 43, and machine learning is performed in the learning device 43 so as to obtain a target output variable. In this way, the learning device 43 can be trained before the proportionality coefficient α is estimated in step S105.
[0055] Next, the temperature calculation unit 45 of the server device 4 calculates the temperature T skin ,T top Based on the heat flux H on the skin surface of the living body 100skin is calculated using equation (2), and the temperature T skin and heat flux H skin Based on the proportionality coefficient α estimated by the learning device 43, the deep body temperature T cbt is calculated by the formula (1) (step S106 in FIG. 16). skin is stored in advance in the storage unit 41. The temperature calculation unit 45 calculates the heat flux H skin Without using cbt may be calculated. T cbt =T skin +α×(T skin -T top ) ···(7)
[0056] Core body temperature T cbt The temperature T used to calculate skin ,T top For time τ, the temperature T skin ,T top The deep body temperature T cbt or the temperature T for time τ skin ,T top Alternatively, a representative value (for example, an average value) of each of the temperatures T top When measuring the temperature T top Using the deep body temperature T cbt Alternatively, the temperature T top A representative value (for example, an average value) of the above may be used.
[0057] In this embodiment, the deep body temperature T is calculated to about one decimal place in the range of about 30°C to 42.0°C with a temperature accuracy of ±0.1°C. cbt Therefore, for example, when using a CNN learning device, it is sufficient to prepare 120 output layers for the learning device 43.
[0058] In addition, if the learning of the learning device 43 is insufficient or if overlearned data is included, the deep body temperature T cbtIn such cases, the estimated core temperature T cbt The probability of abnormal values of is small. Therefore, the core temperature T is calculated by statistical signal processing such as particle filters and a first-order delay filter used in classical control. cbt This allows for sufficient removal of outliers.
[0059] The communication unit 40 of the server device 4 receives the deep body temperature T cbt The data is transmitted to the terminal 3 (step S107 in FIG. 16). The terminal 3 receives the deep body temperature T cbt The value of is displayed (step S108 in FIG. 16).
[0060] In this way, the deep body temperature T cbt In this embodiment, the deep body temperature T cbt This can reduce the estimation error.
[0061] When the exercise intensity of the living body 100 is high, it is preferable to also obtain the heart rate N of the living body 100. The operation of the temperature estimation system in this case will be described with reference to Fig. 18. The processes in steps S100 and S101 are as described above.
[0062] The heart rate measurement unit 2 measures the heart rate N (instantaneous heart rate) of the living body 100 at regular intervals, for example, every second (FIG. 18, step S109). The heart rate measurement unit 2 transmits data on the heart rate N to the terminal 3 (FIG. 18, step S110). Note that, as shown in FIG. 15, if the heart rate measurement unit 2 is provided inside the temperature measurement unit 1g, the data on the heart rate N is transmitted from the communication unit 15 to the terminal 3.
[0063] Terminal 3 receives the temperature T skin ,T side ,T top The data of the temperature T and the data of the heart rate N are transmitted to the server device 4 (step S102a in FIG. 18). skin ,T side ,T topThe data of the number of heart beats and the data of the number of heart beats N are stored in the memory unit 41.
[0064] After the time τ has elapsed (YES in step S103 of FIG. 18), the tensor generating unit 42 of the server device 4 calculates the temperature T skin ,T side ,T top The time series data of the temperature T skin ,T side ,T top Similarly, we can arrange the heart rate data N as pixels of a tensor.
[0065] The learning device 43 of the server device 4 outputs an estimation result of the proportionality coefficient α when the tensor is input from the tensor generating unit 42 (step S105a in FIG. 18). When using data on the heart rate N, the temperature T skin ,T side ,T top At the same time, the true core body temperature T ref Obtain the time series data of the temperature T skin ,T top ,T ref The proportionality coefficient α is calculated for each time period using the data in accordance with equation (6), thereby obtaining time-series data for the proportionality coefficient α.
[0066] And the temperature T skin ,T side ,T top The time series data of the temperature T skin ,T side ,T top and the time series data of the heart rate N are used as input variables of the learning device 43, and the proportionality coefficient α is used as an output variable of the learning device 43, and machine learning is performed in the learning device 43 so as to obtain a target output variable. In this way, the learning device 43 can be trained before estimating the proportionality coefficient α in step S105a.
[0067] The processes in steps S106 to S108 are as described above. In this way, by using the data on the heart rate N, the deep body temperature T cbt This can reduce the estimation error.
[0068] When a person moves between indoors and outdoors in daily life or lies down, the deep body temperature T estimated by the method of this embodiment is cbt The true deep body temperature (tympanic membrane temperature) T ref The core body temperature T estimated by the conventional method shown in FIG. cbt In comparison, according to the present embodiment, noise is suppressed and errors are corrected, and an estimation result close to the eardrum temperature is obtained. According to the present embodiment, it is not necessary to calibrate the proportionality coefficient α every time a measurement is performed, and the deep body temperature T is highly accurately estimated without being affected by the blood flow of the living body, the outside temperature, or the wind. cbt can be estimated.
[0069] [Second Example] Next, a second embodiment of the present invention will be described. Fig. 20 is a block diagram showing the configuration of a server device of a temperature estimation system according to the second embodiment of the present invention. The server device 4a of this embodiment includes a communication unit 40, a storage unit 41, a tensor generation unit 42, a plurality of learning devices 43-1 to 43-3 prepared in advance to correspond to the state of the living body 100 or the environment around the living body 100, a machine learning unit 44, a temperature calculation unit 45, and a temperature T skin ,T top , and a selection unit 46 that selects a learning device that corresponds to the state of the living body 100 or the surrounding environment from among the plurality of learning devices 43-1 to 43-3, based on at least one of the heart rate N. The temperature measuring units 1, 1a to 1g are configured as described in the first embodiment.
[0070] The accuracy of the proportionality coefficient α can be determined based on the required temperature accuracy. In this embodiment, for a temperature accuracy of ±0.1°C, the proportionality coefficient α, which changes depending on the state of blood flow, etc., is sufficient if it is determined to about 2.0 to 12.0 to about one decimal place. Therefore, for example, when using a CNN learning device, it is possible to deal with the change and accuracy of the proportionality coefficient α by preparing 100 output layers of the learning device. When the range of α can be limited to a certain extent, there is an advantage in that the output layer can be made smaller than when the deep body temperature is estimated directly.
[0071] 21 is a flow chart for explaining the operation of the temperature estimation system of the present embodiment. As in the first embodiment, the temperature sensors 10, 13, and 11 of the temperature measurement units 1, 1a to 1g measure the temperature T skin ,T side ,T top is measured, for example, every second (FIG. 21, step S200). The communication unit 15 of the temperature measurement unit 1, 1a to 1g detects the temperature T skin ,T side ,T top The data is transmitted to the terminal 3 (FIG. 21, step S201).
[0072] The heart rate measurement unit 2 measures the heart rate N (instantaneous heart rate) of the living body 100, for example, every second (FIG. 21, step S202). The heart rate measurement unit 2 transmits data on the heart rate N to the terminal 3 (FIG. 21, step S203). As shown in FIG. 15, when the heart rate measurement unit 2 is provided inside the temperature measurement unit 1g, the data on the heart rate N is transmitted from the communication unit 15 to the terminal 3.
[0073] Terminal 3 receives the temperature T skin ,T side ,T top The data of the number of heartbeats and the data of the number of heartbeats N are transmitted to the server device 4a (Step S204 in FIG. 21). After the time τ has elapsed (YES in step S205 of FIG. 21), the tensor generating unit 42 of the server device 4a calculates the temperature T skin ,T side ,T top and the time series data of the heart rate N for time τ are converted into tensors (FIG. 21, step S206).
[0074] Next, the selection unit 46 of the server device 4a selects a learning device to be used for estimating the proportionality coefficient α from among the learning devices 43-1 to 43-3 prepared in advance. By switching the range of the output layer of the learning device depending on the environment or usage scene, more accurate estimation can be performed. For example, when a healthy person is generally sleeping, the healthy person is indoors and the outside temperature is stable. For this reason, the proportionality coefficient α is about 2 to 6, and the output layer of the learning device may be about 40. On the other hand, when a person exercises outdoors, the change in deep body temperature is also large in a hot or cold environment, so it is desirable to widen the range of the output layer of the learning device to be prepared.
[0075] Therefore, in order to improve the accuracy of the estimation of the deep body temperature, it is desirable to prepare a plurality of learning devices and switch between them for each situation. For example, a learning device 43-1 corresponding to an exercise state or a hot / cold environment, a learning device 43-2 corresponding to daily life, and a learning device 43-3 corresponding to a sleeping state are prepared in advance. The selection unit 46 selects the temperature T skin , the temperature at the top of the sensor T top , or the heart rate N as an index to select the learning machine to use.
[0076] Specifically, the selection unit 46 selects the temperature T top is the threshold T top _ th _ high If higher, temperature T top is the threshold T top _ th _ low If lower, temperature T skin is the threshold T skin _ th If heart rate N is higher than threshold N th If any of the above cases holds (YES in step S207 in FIG. 21), it is determined that learning device 43-1 is to be used (step S208 in FIG. 21).
[0077] T top _ th _ high is the temperature threshold for determining whether the outside air is in a hot environment.top _ th _ low is the temperature threshold for determining whether the outside air is cold or not. skin _ th is a temperature value for determining whether the living body 100 is in a feverish state due to a cold or the like, or whether the living body 100 is in a feverish state due to exercise. th is a heart rate threshold for determining whether or not the living body 100 is in an exercising state.
[0078] On the other hand, if the determination in step S207 is not true and the heart rate N is equal to or lower than the threshold N th _ low If it is lower (YES in step S209 in FIG. 21), it is determined that the learning device 43-3 is to be used (step S210 in FIG. 21). th _ low is a heart rate threshold for determining whether or not the living body 100 is in a sleeping state or in a resting state.
[0079] When neither the determination in steps S207 nor S209 is satisfied (NO in step S209), the selection unit 46 determines to use the learning device 43-2 (step S211 in FIG. 21).
[0080] In addition, the temperature T skin ,T top The determinations in steps S207 and S209 may be performed using the latest values of the time series data of the temperature T skin ,T top , and the heart rate N may be determined using a representative value (for example, an average value) of each of the temperatures T top When measuring the temperature T top The determination in step S207 may be performed using the temperature T top The judgment may be made using a representative value (for example, an average value) of the above.
[0081] The learning device selected by the selection unit 46 from among the learning devices 43-1 to 43-3 of the server device 4a outputs the estimation result of the proportionality coefficient α for the tensor input from the tensor generation unit 42 (Step S212 in FIG. 21).
[0082] The learning units 43-1 to 43-3 need to be trained in advance according to the corresponding environment and usage scene. That is, the learning unit 43-1 learns the temperature T skin ,T side ,T top ,T ref The learning unit 43-2 acquires data on the temperature T skin ,T side ,T top ,T ref The learning unit 43-3 acquires data on the temperature T skin ,T side ,T top ,T ref The data on heart rate N is acquired and learning is performed.
[0083] The processing in steps S213 to S215 in FIG. 21 is the same as the processing in steps S106 to S108 in the first embodiment. In this way, the deep body temperature T cbt When the learning devices 43-1 to 43-3 are provided on the cloud server side, the learning of the learning devices 43-1 to 43-3 can be sequentially advanced, and updating and improvement of accuracy can be promoted while being used.
[0084] In order to protect the temperature sensors 10 and 13, the temperature sensors 10 and 13 may be prevented from directly contacting the skin of the living body 100. For example, a thin sheet-like member made of a material with a small heat capacity such as PET resin is provided on the surface of the housing 17 on the living body side, and the temperature sensors 10 and 13 detect the temperature T of the skin surface of the living body 100 through this member. skin ,T side It is also possible to measure the following. In the first and second embodiments, the proportionality coefficient α is estimated by the learning devices 43, 43-1 to 43-3, and then the deep body temperature T cbt The deep body temperature T cbt may be estimated.
[0085] The deep body temperature T cbt When estimating the temperature T skin ,T side ,T top ,T ref The learning units 43, 43-1 to 43-3 perform machine learning using the tensors converted from each of the time series data. Specifically, the machine learning unit 44 performs machine learning on the temperature T skin ,T side ,T top The time series data of each of the above is used as input variables for the learning units 43, 43-1 to 43-3, and the deep body temperature T ref The learning devices 43, 43-1 to 43-3 perform machine learning using the above as output variables so as to obtain target output variables.
[0086] When using data on the heart rate N, the machine learning unit 44 calculates the temperature T skin ,T side ,T top ,T ref The learning units 43, 43-1 to 43-3 perform machine learning using tensors converted from the time series data of the temperature T skin ,T side ,T top The time series data of each of the above and the time series data of the heart rate N are used as input variables of the learning units 43, 43-1 to 43-3, and the deep body temperature T ref The learning devices 43, 43-1 to 43-3 perform machine learning using the above as output variables so as to obtain target output variables.
[0087] The deep body temperature T cbtIn the case where the deep body temperature T is estimated, the temperature calculation unit 45 is not required. cbt The operation for estimating the deep body temperature T cbt The operation for estimating is shown in Fig. 24. Fig. 22 shows an example in which the heartbeat measuring unit 2 is not used, and Fig. 23 shows an example in which the heartbeat measuring unit 2 is used.
[0088] In the examples of FIGS. 22 and 23, the learning unit 43 calculates the deep body temperature T cbt The estimation result is output (step S106a). In the example of FIG. 24, the learning device selected by the selection unit 46 from among the learning devices 43-1 to 43-3 is used to calculate the deep body temperature T cbt The estimation result is output (step S213a).
[0089] In addition, the proportionality coefficient α corresponds to the deep body temperature T cbt Since the fluctuations are continuous and gradual, it is usually more accurate to estimate the proportionality coefficient α.
[0090] The storage unit 14, communication unit 15, and control unit 16 of the temperature measuring units 1, 1a to 1g described in the first and second embodiments can be realized by a computer equipped with a CPU (Central Processing Unit), a storage device, and an interface, and a program that controls these hardware resources. An example of the configuration of this computer is shown in FIG.
[0091] The computer includes a CPU 200, a storage device 201, and an interface device (I / F) 202. Hardware such as temperature sensors 10, 11, and 13 and a communication unit 15 are connected to the I / F 202. In such a computer, a temperature estimation program for implementing the temperature estimation method of the present invention is stored in the storage device 201. The CPU 200 executes the processes described in the first and second embodiments in accordance with the program stored in the storage device 201. Each of the terminal 3 and the server devices 4 and 4a can also be realized by a computer. [Industrial Applicability]
[0092] The present invention can be applied to a technique for estimating the internal temperature of a subject such as a living body. [Explanation of symbols]
[0093] 1, 1a to 1g... temperature measurement unit, 2... heart rate measurement unit, 3... terminal, 4, 4a... server device, 10, 11, 13... temperature sensor, 12, 20... insulation material, 14, 41... memory unit, 15, 40... communication unit, 16... control unit, 17, 18... housing, 19... wiring, 21... internal structure, 42... tensor generation unit, 43, 43-1 to 43-3... learning device, 44... machine learning unit, 45... temperature calculation unit, 46... selection unit.
Claims
1. Insulation material; a first temperature sensor provided on a surface of the thermal insulation material facing an object to be inspected and configured to measure a temperature of a surface of the object to be inspected; a second temperature sensor configured to measure a temperature inside the insulation directly above the first temperature sensor; a third temperature sensor configured to measure a temperature of the surface of the object at a position remote from the first temperature sensor; a learning device configured to estimate a proportionality coefficient related to a thermal resistance of the object based on the measurement results of the first, second, and third temperature sensors; a temperature calculation unit configured to calculate an internal temperature of the subject based on the measurement results of the first and second temperature sensors and the proportionality coefficient.
2. 2. The temperature estimation system according to claim 1, a heart rate measuring unit configured to measure a heart rate of the subject; A temperature estimation system characterized in that the learning device estimates the proportionality coefficient based on the measurement results of the first, second, and third temperature sensors and the measurement result of the heart rate measurement unit.
3. Insulation material; a first temperature sensor provided on a surface of the thermal insulation material facing an object to be inspected and configured to measure a temperature of a surface of the object to be inspected; a second temperature sensor configured to measure a temperature inside the insulation directly above the first temperature sensor; a third temperature sensor configured to measure a temperature of the surface of the object at a position remote from the first temperature sensor; a learning device configured to estimate an internal temperature of the subject based on the measurement results of the first, second, and third temperature sensors.
4. 4. The temperature estimation system according to claim 3, a heart rate measuring unit configured to measure a heart rate of the subject; A temperature estimation system characterized in that the learning device estimates the internal temperature of the subject based on the measurement results of the first, second, and third temperature sensors and the measurement result of the heart rate measurement unit.
5. 4. The temperature estimation system according to claim 1, a plurality of the learning units prepared in advance to correspond to a state of the subject or an environment surrounding the subject; A temperature estimation system further comprising a selection unit configured to select a learning machine from the plurality of learning machines that corresponds to a state of the subject or an environment surrounding the subject for the estimation based on at least one of the measurement results of the first temperature sensor and the measurement results of the second temperature sensor.
6. 5. The temperature estimation system according to claim 2, a plurality of the learning units prepared in advance to correspond to a state of the subject or an environment surrounding the subject; A temperature estimation system further comprising a selection unit configured to select a learning machine from the plurality of learning machines that corresponds to the state of the subject or the environment surrounding the subject for the estimation based on at least one of the measurement results of the first temperature sensor, the measurement results of the second temperature sensor, and the measurement results of the heart rate measurement unit.
7. a first step of measuring a temperature of a surface of the object by a first temperature sensor provided on a surface of a thermal insulating material facing the object; a second step of measuring a temperature inside the insulation material directly above the first temperature sensor with a second temperature sensor; a third step of measuring a temperature of the surface of the object at a position away from the first temperature sensor by a third temperature sensor; a fourth step of estimating a proportionality coefficient related to the thermal resistance of the test object based on the measurement results of the first, second, and third steps by a trained learning device; and a fifth step of calculating an internal temperature of the object based on the measurement results of the first and second steps and the proportionality coefficient.
8. 8. The temperature estimation method according to claim 7, A sixth step of measuring a heart rate of the subject, A temperature estimation method characterized in that the fourth step includes a step of estimating the proportionality coefficient based on the measurement results of the first, second and third steps and the measurement result of the sixth step.
Citation Information
Patent Citations
Deep body temperature measuring apparatus, deep body temperature measuring system, and deep body temperature measuring method
JP2018183564A
In-vivo temperature measurement instrument and in-vivo temperature measurement method
JP2019207124A
Device and method for measuring temperature in living body
JP2020003291A
Learning device, estimation device, and environment adjusting system
JP2021010721A
Wrist temperature rhythm acquisition apparatus and method, core temperature rhythm acquisition apparatus and method, and wearable device
US20180092545A1
