Core body temperature estimating method, core body temperature estimating device, and program

Abstract

This core body temperature estimating method includes: a first step for acquiring the time-varying behavior of an observed core body temperature of a living body on the basis of a sensor output, acquiring the time-varying behavior of a predicted core body temperature of the living body by means of a model indicating the time-varying behavior of the core body temperature of the living body, and estimating the time-varying behavior of the actual core body temperature of the living body as the time-varying behavior of an estimated core body temperature on the basis of a plurality of observed core body temperatures and the time-varying behavior of the predicted core body temperature; and a second step for updating a parameter constituting the model so as to bring the time-varying behavior indicated by the model closer to the time-varying behavior of the estimated core body temperature. Furthermore, the method involves repeatedly performing, a plurality of times, the second step followed by the first step for acquiring the time-varying behavior of the predicted core body temperature, using, as the model, the model of which the parameters have been updated in the second step. In this way, the time-varying behavior of the core body temperature is accurately estimated.

Description

Deep body temperature estimation method, deep body temperature estimation device, and program

[0001] The present invention relates to a method for estimating a core body temperature, a device for estimating a core body temperature, and a program.

[0002] A method has been developed for estimating the deep body temperature (core temperature) of a living body using a predetermined estimation model (Patent Document 1).

[0003] Japanese Patent Application Laid-Open No. 2020-003291

[0004] In conventional methods for estimating deep body temperature, such as the method described in Patent Document 1, when a living body runs, the estimation model for deep body temperature changes transiently, resulting in systematic estimation errors, and it may not be possible to accurately estimate the time-varying pattern of deep body temperature.

[0005] The present invention has been made to accurately estimate the time-varying pattern of core body temperature.

[0006] In order to solve the above problem, the deep body temperature estimation method of the present invention comprises a first step of acquiring the time-varying pattern of an observed deep body temperature of a living organism based on sensor output, acquiring the time-varying pattern of a predicted deep body temperature of the living organism using a model that indicates the time-varying pattern of the deep body temperature of the living organism, and estimating the time-varying pattern of the actual deep body temperature of the living organism as the time-varying pattern of an estimated deep body temperature based on the multiple observed deep body temperatures and the time-varying pattern of the predicted deep body temperature, and a second step of updating parameters that constitute the model so that the time-varying pattern indicated by the model approaches the time-varying pattern of the estimated deep body temperature, and repeating the second step and newly performing the first step of acquiring the time-varying pattern of the predicted deep body temperature multiple times using the model whose parameters have been updated in the second step.

[0007] The deep body temperature estimation device of the present invention has a deep body temperature estimation unit that acquires the time-varying pattern of an observed deep body temperature of a living organism based on sensor output, acquires the time-varying pattern of a predicted deep body temperature of the living organism using a model that indicates the time-varying pattern of the deep body temperature of the living organism, and estimates the time-varying pattern of the actual deep body temperature of the living organism as the time-varying pattern of an estimated deep body temperature based on the multiple observed deep body temperatures and the time-varying pattern of the predicted deep body temperature, and a parameter update unit that updates the parameters that constitute the model so as to bring the time-varying pattern indicated by the model closer to the time-varying pattern of the estimated deep body temperature, and when the parameters of the model are updated by the parameter update unit, the deep body temperature estimation unit acquires the time-varying pattern of the predicted deep body temperature using the updated model with the updated parameters.

[0008] A program according to the present invention causes a computer to function as the core body temperature estimation device.

[0009] According to the present invention, the manner in which core body temperature changes over time can be estimated with high accuracy.

[0010] Fig. 1 is a configuration diagram of a core body temperature estimation device and a heat flux sensor according to an embodiment of the present invention. Fig. 2 is a block diagram of the core body temperature estimation device. Fig. 3 is a flowchart of the estimation process. Fig. 4 is a flowchart of the update process. Fig. 5 is a graph of the estimated core body temperature. Fig. 6 is a graph of the estimated core body temperature. Fig. 7 is a configuration diagram of a heat flux sensor according to a modified example.

[0011] As shown in FIG. 1, a core body temperature estimation system 10 according to this embodiment includes a heat flux sensor 20 and a core body temperature estimation device 30 (hereinafter also referred to as device 30).

[0012] The heat flux sensor 20 is attached to the epidermis U1 of a given location (chest, abdomen, forehead, etc.) of a user (living body) U (depicted as a cross section with dots on the cross section in FIG. 1 ) by any means such as a fixing band or adhesive, and is configured to detect the heat flux (see arrow HF) emitted by the user U. In the following description, the direction toward the epidermis U1 when the sensor 20 is attached to the epidermis U1 is referred to as down, and the opposite direction as up. This up and down is for convenience and does not necessarily coincide with the actual up and down directions.

[0013] The heat flux sensor 20 includes a housing 21 , a measuring unit 22 , a control unit 23 , and a communication unit 24 .

[0014] The housing 21 houses and supports the other members 22 to 24. The housing 21 houses and supports the measuring unit 22 in a state where it is exposed downward.

[0015] The measuring unit 22 includes a thermal resistor 22A made of a heat insulating material such as urethane resin, and sensor elements 22B and 22C such as thermistors supported by the thermal resistor 22A. The sensor element 22B is provided at the lower end of the thermal resistor 22A and is configured to come into contact with the skin U1 when the sensor 20 is attached to the skin U1. The sensor element 22C is provided above the thermal resistor 22A and faces the sensor element 22B in the vertical direction across the thermal resistor 22A.

[0016] The control unit 23 includes a control circuit such as a microcomputer. The control unit 23 is configured to control the operation of the heat flux sensor 20. The control unit 23 measures the temperature Tskin of the epidermis U1 using the sensor element 22B, and measures the temperature Tu at a position distant from the epidermis U1 using the sensor element 22C. For example, the control unit 23 measures the temperature Tskin by detecting the resistance value of the sensor element 22B and calculating the temperature Tskin from the resistance value. The control unit 23 measures the temperature Tu by detecting the resistance value of the sensor element 22C and calculating the temperature Tu from the resistance value. The temperatures Tskin and Tu represent heat fluxes.

[0017] The control unit 23 wirelessly communicates with the device 30 via a communication unit 24 formed of a wireless communication module or the like. The communication unit 24 may be connected to the device 30 via a wired connection. The control unit 23 measures the temperatures Tskin and Tu based on a measurement instruction transmitted from the device 30 via the communication unit 24, and transmits the measured temperatures Tskin and Tu to the device 30 via the communication unit 24.

[0018] The device 30 is configured to estimate the core body temperature, which is the body temperature of the deep part (also called the core) U2 of the user U, based on the temperatures Tskin and Tu (heat flux) measured by the heat flux sensor 20.

[0019] The device 30 is, for example, a computer such as a smartwatch. The device 30 includes a storage device 31 that stores a program and various data (including the models described below), and a processor 32 that executes the program and performs the processes described below using the various data. The processor 32 includes a CPU (Central Processing Unit) and the like. The device 30 further includes a main memory 33 that provides a working area for the processor 32, a communication unit 34 that establishes communication between the processor 32 and the heat flux sensor 20, a user interface 35 that accepts user operations, and a display device 36 that displays various images.

[0020] The processor 32 executes the programs stored in the storage device 31 to operate as a core body temperature estimation unit 32A, a parameter update unit 32B, and a display control unit 32C shown in FIG.

[0021] The core body temperature estimation unit 32A acquires the core body temperature of the user U as the observed core body temperature using an observation model that shows the relationship between the temperatures Tskin and Tu and the core body temperature based on the output of the heat flux sensor 20, i.e., the temperatures Tskin and Tu (heat flux) measured by the heat flux sensor 20. Furthermore, the core body temperature estimation unit 32A acquires the predicted core body temperature of the user U at the time of measuring the temperatures Tskin and Tu using a model that shows the temporal change in the core body temperature of the user U (here, a time-evolving model in which the observation model is evolved over time, but the details of the model are arbitrary). The observation model and each data (parameters, etc.) that indicates the time-evolving model are stored in the storage device 31. The core body temperature estimation unit 32A estimates the actual core body temperature of the user U as the estimated core body temperature based on the acquired observed core body temperature and predicted core body temperature. A set of the observed core body temperature and the predicted core body temperature at the same time is periodically acquired by the core body temperature estimation unit 32A and used to estimate the estimated core body temperature. For this reason, the core body temperature estimator 32A acquires or estimates multiple observed core body temperatures, multiple predicted core body temperatures, and multiple estimated core body temperatures. These are also time-varying patterns. Therefore, the core body temperature estimator 32A acquires the time-varying patterns of the observed core body temperatures and the predicted core body temperatures, and estimates the time-varying pattern of the estimated core body temperature based on the acquired time-varying patterns. The core body temperature estimator 32A stores the time-varying patterns (time-series data) of the estimated core body temperature for a certain period, such as one day, 18 hours, or 12 hours, in the storage device 31. The certain period may be any period during which the rhythm (period, phase, amplitude) of the core body temperature appears, and may be a day to obtain a circadian rhythm, or a predetermined period during sleep.

[0022] The parameter update unit 32B updates the parameters constituting the time evolution model so that the time evolution pattern indicated by the time evolution model approaches the time evolution pattern of the estimated deep body temperature for a certain period of time estimated by the deep body temperature estimation unit 32A (here, the time evolution pattern stored in the memory device 31).

[0023] After the parameter update unit 32B updates the time evolution model, the core body temperature estimation unit 32A performs the above operation again. At this time, the core body temperature estimation unit 32A obtains a predicted core body temperature using the time evolution model updated by the parameter update unit 32B.

[0024] The operations of the core body temperature estimation unit 32A and the parameter update unit 32B are repeatedly executed. This repeatedly updates the parameters of the time evolution model based on the time change pattern (time series data) of the actually estimated estimated core body temperature. This improves the accuracy of the time evolution model, and the time change pattern of the core body temperature is accurately estimated. In the repeated operations, the above-mentioned certain period of time should preferably be the same time period for each day or multiple days. This further improves the accuracy of the time evolution model.

[0025] The display control unit 32C reads out the time-varying pattern of the estimated core body temperature from the storage device 31 and displays it on the display device 36 in response to a display request from the user U via the user interface 35. At this time, the display control unit 32C may detect the rhythm (period, phase, shape, amplitude, etc.) of the time-varying pattern of the estimated core body temperature and display it together with the time-varying pattern. The rhythm (particularly the period) can be detected using an autocorrelation method (correlogram), a power spectrum method, a cosinor method, a periodogram method, etc. The cosinor method or the periodogram method is particularly commonly used. The display control unit 32C may also superimpose the time-varying pattern for each of multiple fixed periods on the display device 36 to display a discrepancy in the rhythm of the change in core body temperature.

[0026] As an example of this embodiment, the following equation (1) is used as the observation model for observing the core body temperature (the observed core body temperature described below) y based on the above observation model, i.e., Tskin and Tu measured by the heat flux sensor 20. Furthermore, the following equation (2) is used as a time-evolved model (a function showing the change in core body temperature over time) obtained by time-evolving this observation model. In the following equation (2), Tcbt is the predicted core body temperature predicted by the time-evolved model, t is time, A is the amplitude, ω is the angular frequency, φ is the phase, and M is the average core body temperature. A, ω, φ, and M are collectively referred to as biological rhythm parameters B. It is assumed that an appropriate value is set as the initial value of biological rhythm parameter B. y = Tskin + α(Tskin - Tu) (1) Tcbt(t) = A cos(ωt + φ) + M (2)

[0027] As can be seen from equation (2), the time evolution model is assumed to be a cosine wave. In addition to the cosine wave, other equations that describe vibration, such as a spring-mass-damper model or the van der Pol equation, may be used as the time evolution model.

[0028] In the time evolution model, the change ΔT in the core body temperature Tcbt after time Δt is expressed by the following equation (3) based on the above equation (2). Also, the predicted core body temperature Tcbt(t+Δt) after Δt is expressed by the following equation (4). ΔT = ωA sin(ωt+φ) * Δt (3) Tcbt(t+Δt) = Tcbt(t) + ΔT (4)

[0029] 3, the core body temperature estimation unit 32A estimates the core body temperature every Δt and stores the estimation results in the memory unit as the estimated core body temperature Tcbt_a in chronological order. This allows the time-varying pattern of the estimated core body temperature Tcbt_a to be obtained. Here, a Kalman filter is used to estimate the estimated core body temperature Tcbt_a as a post-estimate value, using the observed core body temperature y as the observed value and the predicted core body temperature Tcbt(t) as the pre-estimate value.

[0030] Before the estimation process begins, appropriate values ​​are input and set as initial values ​​for the time t, variance P, constant C, system noise Q, and observation noise w. These initial values ​​are set to optimal values ​​depending on the sampling rate (here, Δt), etc. The constant C may be A in the above equation (2). In addition, in order to determine the proportionality coefficient α in equation (1), a reference temperature Tcbt_ref is also input and set. This reference temperature Tcbt_ref is measured by any method using an axillary thermometer, sublingual thermometer, tympanic thermometer, etc., and input by the user U through the user interface 35.

[0031] In the estimation process of FIG. 3, the core body temperature estimation unit 32A first sets the reference temperature Tcbt_ref as the estimated core body temperature Tcbt_a (step S11).

[0032] The core body temperature estimator 32A instructs the heat flux sensor 20 to measure the temperatures Tskin and Tu, and acquires the measured Tskin and Tu from the heat flux sensor 20 (step S12).

[0033] Based on the values ​​obtained in steps S11 and S12, the core body temperature estimating unit 32A calculates a proportionality coefficient α using the following equation (5) (step S13): α=(Tcbt_ref−Tskin) / (Tskin−Tu) (5)

[0034] The core body temperature estimation unit 32A then waits until Δt has passed since step S12 (step S14), and after Δt has passed, instructs the heat flux sensor 20 to measure Tskin and Tu, and obtains the measured Tskin and Tu from the heat flux sensor 20 (step S15).

[0035] The core body temperature estimation unit 32A substitutes α, Tskin, and Tu obtained in steps S13 and S15 into the above equation (1) to derive the observed core body temperature y (step S16).

[0036] The core body temperature estimation unit 32A updates the predicted core body temperature Tcbt_f using the following equation (6) (step S17). ΔT is derived from the above equation (3). Using equation (6), the change in core body temperature after Δt has elapsed, obtained based on the time evolution model, is added to the estimated core body temperature Tcbt_a before Δt has elapsed (i.e., at time t). This gives the predicted core body temperature Tcbt_f after Δt has elapsed (i.e., the present). Tcbt_f=Tcbt_a+ΔT (6)

[0037] The core body temperature estimation unit 32A updates the variance P using the following equation (7) (step S18). ΔT is the ΔT derived in step S17 (i.e., the change from the time Δt before to the present time). P = P + ΔT ２ *Q...(7)

[0038] The core body temperature estimation unit 32A corrects the observed core body temperature y using the following equation (8) based on the observation noise w and the variance P, and calculates the Kalman gain K for obtaining the estimated core body temperature Tcbt_a (step S19). Note that the observation noise w is calculated based on the observed core body temperature y and the predicted core body temperature Tcbt_f as follows: w=|y ２ -Tcbt_f ２ |. The observation noise w may also be calculated using the Mahalanobis distance. In this way, by setting a value corresponding to the difference between the observed value and the predicted value as the observation noise w, it is possible to effectively remove the transient response described below. K=P / (P+C*w) (8)

[0039] The core body temperature estimation unit 32A calculates the difference between the observed core body temperature y and the predicted core body temperature Tcbt_f and derives it as an estimation error e (see equation (9) below) (step S20). Here, the estimation error e is derived using the observed core body temperature y at a certain point in time and the predicted core body temperature Tcbt_f from before. e=|y−Tcbt_f| (9)

[0040] The core body temperature estimating unit 32A updates the estimated core body temperature Tcbt_a using the following equation (10), sets t=t+Δt, and stores the updated estimated core body temperature Tcbt_a in the storage device 31 in association with the current time (step S21). Tcbt_a=Tcbt_f+K*e (10)

[0041] The core body temperature estimating unit 32A updates the variance P using the following equation (11) (step S22): P=(1−K)P (11)

[0042] Thereafter, the core body temperature estimation unit 32A sets t=t+Δt, and then waits until Δt has elapsed from step S15 (step S23), after which the process returns to step S15. The processes of steps S15 to S23 are repeated for a fixed period (sampling rate=Δt), after which execution is suspended, and the process is repeated again for a fixed period on the next day.

[0043] By the above-described processing, time-series data indicating the time variation of the estimated core body temperature for a certain period such as one day, 12 hours, or 18 hours is registered in the storage device 31.

[0044] 4 and updates the biological rhythm parameter B of the time evolution model of the above equation (2), for example, when the time change pattern of the estimated core body temperature for a certain period of time is registered in the storage device 31. The parameter update unit 32B updates the biological rhythm parameter B so that the time change pattern of the core body temperature indicated by the time evolution model approaches the time change pattern of the estimated core body temperature. Here, an observed biological rhythm parameter B_y derived from the time change pattern of the estimated core body temperature (core body temperature graph) is used as the observed value using a Kalman filter, and a predicted biological rhythm parameter B_f of the current time evolution model is used as a pre-estimated value to estimate an estimated biological rhythm parameter B_a as a post-estimated value.

[0045] Before the estimation process begins, appropriate values ​​are input and set as initial values ​​for the variance Pb, constant Cb, system noise Qb, and observation noise wb. These initial values ​​are set to optimal values ​​depending on the sampling rate (here, Δt), etc. These initial values ​​may be the same as the initial values ​​for the variance P, constant C, system noise Q, and observation noise w. The update process in Figure 4 is performed for each of the multiple constants (such as period) that make up the biological rhythm parameter B (for example, the following process is performed with period B, and then the following process is performed with phase B).

[0046] 4, the parameter update unit 32B first observes an observed biorhythm parameter B_y from the time-varying pattern of the estimated core body temperature (step S51). This parameter B_y is derived by fitting using the autocorrelation method (correlogram), the power spectrum method, the cosinor method, the periodogram method, or the like.

[0047] The parameter update unit 32B sets the current predicted biorhythm parameter B_f as the estimated biorhythm parameter B_a (step S52).

[0048] The parameter update unit 32B updates the variance Pb using the following equation (12) (step S53): Pb=Pb+Qb (12)

[0049] The parameter update unit 32B corrects the observed biological rhythm parameter B_y using the following equation (13) based on the observed noise wb and the variance Pb, and calculates the Kalman gain Kb for obtaining the estimated biological rhythm parameter B_a (step S54): Kb=Pb / (Pb+Ab*wb) (13)

[0050] The parameter update unit 32B calculates the difference between the observed biorhythm parameter B_y and the predicted biorhythm parameter B_f, and derives it as an estimation error eb (see the following equation (14)) (step S55). eb=|B_y−B_f| (14)

[0051] The parameter update unit 32B updates the estimated biological rhythm parameter B_a using the following equation (15) (step S56), and updates the parameter B of the time evolution model with the updated estimated biological rhythm parameter B_a (step S57). This update updates the parameters constituting the time evolution model so that the time change pattern indicated by the time evolution model approaches the time change pattern of the estimated core body temperature for a certain period estimated by the core body temperature estimation unit 32A. B_a = B_f + Kb * eb (15)

[0052] The parameter update unit 32B updates the variance Pb using the following equation (16) (step S58): Pb=(1−Kb)Pb (16)

[0053] Thereafter, the parameter update unit 32B waits until the time-varying pattern of the new estimated deep body temperature is stored in the memory device 31 (step S59), and when the time-varying pattern of the new estimated deep body temperature is stored, it performs processing from step S51.

[0054] The above process realizes the process of updating the model used in statistical filtering to remove transient responses of tens of minutes using the biological rhythm parameters (period, amplitude, phase, and average) of core body temperature, which changes from moment to moment. In other words, the time evolution model in "statistical filtering that models transient responses over tens of minutes" is estimated by "statistical filtering that models circadian rhythms that change over several days to several weeks."

[0055] The effects of the above process will now be explained. Changes in core body temperature, a type of biological activity, have a periodicity (rhythm). To determine whether this rhythm is normal, it is important to understand the parameters that characterize the rhythm (at least one of period, amplitude, phase, and average). However, conventional estimations of core body temperature using sensors attached to a living organism (e.g., user U), such as the heat flux sensor 20, assume a steady flow of heat transported from the living organism through the sensor to the outside air. Therefore, when the living organism is exposed to wind, runs, or suddenly moves from a warm place to a cold place, the core body temperature estimation model transiently changes, resulting in systematic estimation errors. This makes it difficult to easily determine the above parameters using conventional cosinor and visual methods. Furthermore, because changes in core body temperature are not necessarily cosine or sine waves, rhythms cannot be determined using conventional cosinor and visual methods. Furthermore, it is conceivable to use statistical filtering that models the convection-induced error to remove the transient systematic error. However, this method cannot completely remove the error before sleep or immediately before waking up, when the rhythmic changes in core body temperature are large. Attempting to forcibly remove the error at these times can also result in problems such as the waveform of the time-varying core body temperature being distorted. However, as in this embodiment, by repeatedly updating the parameters of the time-varying model based on the actual time-varying core body temperature (time-series data), the accuracy of the time-varying model used to estimate the core body temperature is improved, and the time-varying core body temperature can be accurately estimated. Therefore, it is easy to grasp the rhythm of the change in core body temperature based on the time-varying core body temperature.

[0056] Figures 5 and 6 show the time variation of deep body temperature during sleep that was actually estimated. In the figures, G1 is a graph (time variation) of the observed deep body temperature observed by the observation model. G2 is a graph of the estimated deep body temperature when the parameters of the time evolution model are not updated. G3 is a graph of the estimated deep body temperature obtained by this embodiment. According to this embodiment, the time delay of the response due to signal processing is eliminated (Figure 5), and errors due to transient response are eliminated (Figure 6).

[0057] The core body temperature estimation method, core body temperature device, and program including the above-described embodiment will be described below.

[0058] A method for estimating deep body temperature, comprising: a first step of acquiring the time-varying pattern of an observed deep body temperature of a living organism based on sensor output; acquiring the time-varying pattern of a predicted deep body temperature of the living organism using a model that indicates the time-varying pattern of the deep body temperature of the living organism; and estimating the time-varying pattern of the actual deep body temperature of the living organism as the time-varying pattern of an estimated deep body temperature based on the multiple observed deep body temperatures and the time-varying pattern of the predicted deep body temperature; and a second step of updating parameters that constitute the model so that the time-varying pattern indicated by the model approaches the time-varying pattern of the estimated deep body temperature, wherein the second step is performed and the first step of acquiring the time-varying pattern of the predicted deep body temperature is newly performed multiple times using the model whose parameters have been updated in the second step.

[0059] The sensor may be any sensor capable of measuring temperature that can be used to monitor core body temperature. For example, a rectal thermometer or tympanic thermometer that directly measures core body temperature may be used. Other examples of heat flux sensors include a heat flux sensor with a heater, or the heat flux sensor 120 shown in FIG. 7. The heat flux sensor 120 includes a thermal resistor 122A and sensor elements 122B-122E, such as thermistors, supported by the thermal resistor 122A. The example in FIG. 7 has four sensor elements compared to the example in FIG. 1. Sensor elements 122B and 122C face each other vertically, and sensor elements 122D and 122E face each other vertically. The distance between sensor elements 122B and 122C is different from the distance between sensor elements 122D and 122E. The observation model is expressed by a different equation depending on the type of sensor. An observation model when such a heat flux sensor 120 is employed is expressed by, for example, the following equation (17): Here, the temperatures measured by the sensor elements 122B and 122D, respectively, are defined as Tskin1 and Tskin2, and the temperatures measured by the sensor elements 122C and 122E, respectively, are defined as Tu1 and Tu2.

[0060] In the first step, the estimated deep body temperature is estimated one by one by performing estimation based on a pair of the observed deep body temperature and the predicted deep body temperature at the same timing from the time change pattern of the observed deep body temperature and the time change pattern of the predicted deep body temperature, thereby estimating the time change pattern of the estimated deep body temperature.

[0061] In the above-mentioned method, a statistical filtering technique such as a Kalman filter is effective for estimating the core body temperature. Instead of the above-mentioned method, a part or all of the time-varying pattern of the estimated core body temperature may be estimated by batch processing a part or all of the time-varying pattern of the observed core body temperature and a part or all of the time-varying pattern of the predicted core body temperature. In such a case, a part or all of the time-varying pattern of the estimated core body temperature is estimated by fitting the above-mentioned trigonometric function or polynomial.

[0062] A deep body temperature estimation device comprising: a deep body temperature estimation unit that acquires the time-varying pattern of an observed deep body temperature of a living organism based on sensor output, acquires the time-varying pattern of a predicted deep body temperature of the living organism using a model that indicates the time-varying pattern of the deep body temperature of the living organism, and estimates the time-varying pattern of an actual deep body temperature of the living organism as the time-varying pattern of an estimated deep body temperature based on the multiple observed deep body temperatures and the time-varying pattern of the predicted deep body temperature; and a parameter update unit that updates parameters that constitute the model so as to bring the time-varying pattern indicated by the model closer to the time-varying pattern of the estimated deep body temperature, wherein when the parameters of the model are updated by the parameter update unit, the deep body temperature estimation unit acquires the time-varying pattern of the predicted deep body temperature using the updated model with the updated parameters.

[0063] At least a part of the core body temperature estimation unit 32A and the parameter update unit 32B may be configured by an ASIC or FPGA.

[0064] A program that causes a computer to function as the core body temperature estimation device, the program being stored in a computer-readable non-transitory storage medium.

[0065] The present invention is not limited to the above-described embodiments and modifications. For example, the present invention includes various modifications to the above-described embodiments and modifications that can be understood by a person skilled in the art within the scope of the technical concept of the present invention. The configurations listed in the above-described embodiments and modifications can be combined as appropriate within a range that does not cause contradictions. In addition, any of the above-described configurations can be deleted.

[0066] 10...core body temperature estimation system, 20...heat flux sensor, 20...sensor, 21...housing, 22...component, 22...measuring unit, 22A...thermal resistor, 22B...sensor element, 22C...sensor element, 23...component, 23...control unit, 24...component, 24...communication unit, 30...device, 30...core body temperature estimation device, 31...storage device, 32...processor, 32A...core body temperature estimation unit, 32B...parameter update unit, 32C...display control unit, 33...main memory, 34...communication unit, 35...user interface, 36...display device, 120...heat flux sensor Sa, 122A...thermal resistor, 122B...sensor element, 122C...sensor element, 122D...sensor element, 122E...sensor element, B...biorhythm parameter, B...parameter, B_a...estimated biorhythm parameter, B_f...predicted biorhythm parameter, B_y...observed biorhythm parameter, B_y...parameter, C...constant, Cb...constant, e...estimated error, eb...estimated error, HF...arrow, K...Kalman gain, Kb...Kalman gain, P...variance, Pb...variance, Q...system noise, U1...epidermis, U2...deep layer.

Claims

1. A method for estimating deep body temperature, comprising: a first step of acquiring the time-varying pattern of an observed deep body temperature of a living organism based on sensor output; acquiring the time-varying pattern of a predicted deep body temperature of the living organism using a model that indicates the time-varying pattern of the deep body temperature of the living organism; and estimating the time-varying pattern of the actual deep body temperature of the living organism as the time-varying pattern of an estimated deep body temperature based on the time-varying pattern of the observed deep body temperature and the time-varying pattern of the predicted deep body temperature; and a second step of updating parameters that constitute the model so that the time-varying pattern indicated by the model approaches the time-varying pattern of the estimated deep body temperature; wherein the second step is performed, and the first step of acquiring the time-varying pattern of the predicted deep body temperature is newly performed multiple times using the model whose parameters have been updated in the second step.

2. A method for estimating deep body temperature as described in claim 1, wherein in the first step, the estimated deep body temperature is estimated one by one by performing a pair-by-pair estimation based on the observed deep body temperature and the predicted deep body temperature at the same timing, out of the time-varying pattern of the observed deep body temperature and the time-varying pattern of the predicted deep body temperature, thereby estimating the time-varying pattern of the estimated deep body temperature.

3. A deep body temperature estimation device comprising: a deep body temperature estimation unit that acquires the time-varying pattern of an observed deep body temperature of a living organism based on sensor output, acquires the time-varying pattern of a predicted deep body temperature of the living organism using a model that indicates the time-varying pattern of the deep body temperature of the living organism, and estimates the time-varying pattern of an actual deep body temperature of the living organism as the time-varying pattern of an estimated deep body temperature based on the time-varying pattern of the observed deep body temperature and the time-varying pattern of the predicted deep body temperature; and a parameter update unit that updates parameters that constitute the model so as to bring the time-varying pattern indicated by the model closer to the time-varying pattern of the estimated deep body temperature, wherein when the parameters of the model are updated by the parameter update unit, the deep body temperature estimation unit acquires the time-varying pattern of the predicted deep body temperature using the updated model with the updated parameters.

4. A program that causes a computer to function as the deep body temperature estimation device according to claim 3.

Images