Noninvasive sensor and measurement method

Through the reverse modeling algorithm and corresponding table optimization irradiation parameters, the problem of accuracy and energy consumption of non-invasive sensors in a layered environment is solved, and a high-precision and low-energy-consuming sensor system is realized.

CN118891005APending Publication Date: 2024-11-01ECLYPIA

Patent Information

Application Number
CN202380028163.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Priority Date
2022-03-18
Filing Date
2023-03-17
Publication Date
2024-11-01

AI Technical Summary

Technical Problem

Existing non-invasive sensors based on photoacoustic detection or photothermal detection are difficult to achieve high-precision measurement in a layered environment, and have high energy consumption, especially in living organisms such as human skin. Skin structure changes cause sensors to require frequent calibration. Increased energy consumption and accuracy requirements.

Method used

A sensor system including light source, detection unit, memory and adaptation module is adopted. Through reverse modeling algorithms and corresponding tables, the irradiation parameters are dynamically adjusted to optimize measurements, reducing unnecessary energy consumption and calibration times, and improving accuracy.

Benefits of technology

It realizes that the measurement accuracy of the non-invasive sensor is improved, adapts to changes in the target environment, reduces the energy consumption and calibration frequency of the sensor, and improves the measurement accuracy of the parameters of interest.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118891005A_ABST
    Figure CN118891005A_ABST
Patent Text Reader

Abstract

The invention relates to a method for measuring a parameter of interest in a target environment (2) by means of a non-invasive sensor (1) based on photoacoustic or photothermal detection, comprising: a) providing a sensor comprising an adaptation module (14) comprising a processor implementing a reverse modeling algorithm; b) the adaptation module selects an initial model irradiation configuration; c) determining an optimal irradiation case in the corresponding table; d) the light source irradiates the target environment according to the optimal irradiation case; e) the detection unit detects the signal; f) the processor of the adaptation module returns the current model configuration (CMmes) and the estimated value (Pest) of the parameter of interest; g) the processor of the adaptation module evaluates the selected model illumination configuration and only if the illumination model configuration differs from the current model configuration; g1) receiving a current model configuration (CMmes), returning a new model configuration, and then repeatedly executing c), d), e) and f); h) the value (Pmes) of the parameter of interest measured by the sensor is the last estimated value (Pest) of the parameter of interest.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to a process measurement and a non-invasive sensor which enables the measurement of one or more parameters of interest in a target environment.

[0002] More specifically, the present invention relates to a non-invasive sensor based on photothermal effect or photoacoustic detection, which is particularly configured to measure parameters in a target environment (such as a layered environment and / or an evolving environment). For example, the measured parameter can be blood sugar in the skin. Background Art

[0003] In the field of sensors for living organisms, it is known to realize non-invasive sensors based on photoacoustic detection or photothermal detection.

[0004] The region of interest of the environment to be analyzed, called the target, is illuminated by means of a laser beam of selected wavelength and modulation frequency depending on the parameter of interest to be measured. The laser beam is absorbed by the target at a certain depth, which depends on the structure of the target. The energy-absorbing rays cause local heating of the target. As a reaction to this heating, a heat wave with a frequency equal to the laser modulation frequency is generated in the target. This heat wave propagates in the target and in particular to the outer surface of the target.

[0005] This heat wave can be directly detected and analyzed. Next, let's talk about photothermal effects. Photoacoustic detection uses the correlation between the heat wave and the pressure wave frequency that is the same as the modulation frequency.

[0006] In the case of indirect photoacoustic detection, when a heat wave generated in the target reaches the interface between the target and the fluid external environment after propagation, a pressure wave generated in the fluid external environment is detected.

[0007] This photoacoustic effect has been the subject of many theoretical studies. Allan Rosencwaig and Allen Gersho have developed a theoretical model of photoacoustic signals in particular. The model involves the physicochemical properties of the sample to be analyzed, including the optical diffusion length, thickness and thermal diffusion length of the sample. (Rosencwaig, A. and Gersho, A. (1976), Theory of the photoacoustic effect with solids, Journal of Applied Physics, 47, 64).

[0008] Hu et al. developed a generalized theory of the photoacoustic effect in a multilayer material. (Hu, H., Wang, X., & Xu, X. (1999). Generalized theory of the photoacoustic effect in a multilayer material. Journal of Applied Physics, 86, 3953-3958).

[0009] Compared to other detection techniques, photoacoustic detection presents many advantages, among which the aspect of positive exchange energy can be cited: an optical signal at the input of the environment to be analyzed is converted into an acoustic signal that is specific to the phenomenon to be observed and allows the use of inexpensive miniaturized sensors.

[0010] The difficulties of photoacoustic or photothermal testing are particularly:

[0011] - the number of parameters usually affects the detected signal; and

[0012] - For certain analytes of interest present at lower concentrations in the environment to be analyzed, the ratio of the detected signal for each of these parameters of interest is smaller.

[0013] In a layered material, in order to be able to deduce the concentration of the analyte of interest in a given layer from a photoacoustic or photothermal signal, all other parameters that influence this signal must be known. In particular, the structure of the material must be known, i.e. the thickness of the different layers that make up the material, their physicochemical composition (in addition to the parameter of interest to be measured) and possibly also their thermal conductivity or even the thermal resistance associated with each interface between two consecutive layers.

[0014] It is only possible to calibrate a sensor based on photoacoustic or photothermal detection once and for all (or not calibrate it at all during a considerable period of use) only when the parameter of interest changes in the target. On the other hand, when the characteristics of the stratified environment to be analyzed change, the sensor must be calibrated regularly to obtain measurement results with acceptable accuracy. This problem arises in particular in sensors for living organisms. For example, in the case of noninvasive sensors for interstitial blood glucose monitoring, calibration of noninvasive sensors based on photoacoustic or photothermal detection can only be achieved with limited accuracy, not only because the skin composition of different patients is different, but also because the skin composition of a given patient also changes over time.

[0015] Document EP2460470 describes a conventional calibration procedure for a non-invasive blood glucose sensor including a near-infrared laser source. It is worth noting that this sensor uses a detection method other than photoacoustic detection or photothermal detection, since it is a part of the incident light wave transmitted or diffused by the substance being detected. Therefore, the technical constraints and limitations of this sensor (especially in terms of energy consumption and measurement accuracy) are different from those of sensors based on photoacoustic detection or photothermal detection.

[0016] In the calibration process of EP2460470, a plurality of calibration models, which may be obtained by numerical simulation, or a plurality of data sets for creating these models are stored in the sensor memory.

[0017] During calibration, the spectrum of the tissue to be analyzed (called the reference spectrum) is measured by illuminating the tissue in a predetermined wavelength range, on the basis of which a calibration model is selected for subsequent measurements.

[0018] On the subsequent analysis date, the spectrum of the tissue to be analyzed is measured again, and the difference between the measured spectrum and the reference spectrum is calculated. The quality of the previously selected calibration model is evaluated based on the difference, and if the evaluation result is not good, a new calibration model is selected to determine the measured blood glucose value. Then, by substituting the absorbance measured at each wavelength of the spectrum in the calibration model, the blood glucose value is determined by the sensor.

[0019] Compared to a blood glucose sensor that is calibrated only once when the sensor is fitted to a patient, gradually adjusting the calibration model over time according to the process of EP2460470 can achieve a higher accuracy measurement of blood glucose. The disadvantage of the process of EP2460470 is that it requires the spectrum of the tissue to be analyzed to be acquired over the full wavelength range every time recalibration is performed. This acquisition consumes a lot of energy.

[0020] Therefore, if the method of EP2460470 is simply modified with a continuous sensor based on indirect photoacoustic detection, the energy cost of the process is not sustainable, or an electrical unit is required, but the weight of which is unacceptable to the patient who will be equipped with such a sensor.

[0021] Furthermore, photoacoustic detection requires not only the selection of the wavelength, but also the selection of the laser modulation frequency. In fact, the depth to which the laser penetrates the target depends on the modulation frequency of the laser. If the structure of the target (e.g., skin) changes over time, the modulation frequency used to measure the parameter of interest (e.g., interstitial blood glucose) also changes over time. Therefore, in order to implement the calibration process of EP2460470 for a sensor based on photoacoustic detection or photothermal detection, it is necessary not only to change the wavelength of the laser as before so that the difference between the reference spectrum and the calibrated spectrum can be calculated, but also to change the modulation frequency of the laser. In other words, a series of spectra of the target at different modulation frequencies must be obtained before each measurement (or each series of measurements). Therefore, if the method of EP2460470 is simply changed to a non-invasive sensor based on indirect photoacoustic detection, the energy consumption problem will be further exacerbated.

[0022] Finally, EP2460470 proposes an a priori skin model consisting of three layers: a 0.1 mm thick surface layer, a 0.9 mm thick inner layer and a 2.0 mm thick subcutaneous layer. Although EP2460470 (in Article 79) mentions the possibility of using the skin (total) thickness as a variable in the simulation, it neither proves the feasibility of such an implementation nor indicates whether it is possible to consider a number of layers other than three. Therefore, the calibration process of EP2460470 does not seem to be able to take into account all the inherent variability of the layered environment to be studied. Therefore, the present invention aims to improve the accuracy of non-invasive sensors (especially layered and / or scalable non-invasive sensors based on photoacoustic detection or photothermal detection) in the target environment, while controlling the energy consumption of the sensor or reducing the energy consumption of the sensor while controlling its accuracy. Summary of the invention

[0023] Therefore, the present invention relates to a method for measuring a parameter of interest in a target environment by means of a non-invasive sensor based on photoacoustic detection or photothermal detection, comprising:

[0024] (a) Providing sensors, including:

[0025] -light source;

[0026] - means for controlling parameters of the illumination from the light source;

[0027] - a detection unit configured to detect an acoustic signal or a thermal signal;

[0028] - a memory, wherein a correspondence table is stored, the correspondence table comprising model configurations and optimal irradiation cases, each model configuration representing a given state of a target environment, each optimal irradiation case comprising a set of irradiation parameters, and each model configuration being associated with one optimal irradiation case;

[0029] - and an adaptation module exchanging information with the detection unit and the means for controlling the illumination parameters of the light source, the adaptation module comprising a processor adapted to implement an inverse modeling algorithm receiving as input an illumination case comprising a set of illumination parameters and the acoustic or thermal signal and providing as output a model configuration and the value of the parameter of interest;

[0030] (b) The adaptation module selects the initial irradiation model configuration;

[0031] (c) the adaptation module determines in the correspondence table the best irradiation case for the selected irradiation model configuration (i.e., the irradiation case that allows measuring the parameter of interest with a predetermined accuracy and / or the amount of measurement data that achieves the lowest energy consumption);

[0032] (d) the light source illuminates the target environment according to the illumination parameter set of the best illumination case;

[0033] (e) a detection unit detects an acoustic signal or a thermal signal generated in response to the irradiation;

[0034] (f) a processor of the adaptation module implements an inverse modeling algorithm, receives as input the acoustic signal or thermal signal detected by the detection unit and the best irradiation case for irradiation, and returns as output the current model configuration and estimates of the parameters of interest;

[0035] (g) The processor of the adaptation module evaluates the selected irradiation model configuration by comparing it with the current model configuration, and only if the evaluation result is not good: (g1) The adaptation module (14) receives the current model configuration (CMmes) as input and returns a new model configuration for irradiation (CMirrad) as output, and then repeats (c), (d), (e) and (f);

[0036] (h) The value of the parameter of interest measured by the sensor (Pmes) is the final estimate of the parameter of interest (Pest).

[0037] Due to these regulations, blood glucose measurements can be obtained in the following ways:

[0038] - Single irradiation (if the evaluation result is good, sub-step g1 is not performed). Then, the irradiation is performed according to an irradiation case selected by default but corresponding to controlled energy consumption (for example, a limited number of irradiation parameters are selected). The step (g) of evaluating the irradiation model configuration can achieve the certainty that the measurement accuracy is indeed the highest accuracy in the current case given the irradiation cases available in the corresponding table, which is advantageous compared to a process that does not provide step (g) and does not have this certainty;

[0039] - Two irradiations (in case of a bad evaluation, sub-step g1 is performed in step (g), and thus comprises repetition of (c), (d), (e) and (f)). Then, the second irradiation is performed according to the irradiation case, so that the highest accuracy can be obtained for this specific case in view of the additional information acquired due to the first irradiation and the irradiation cases available in the correspondence table. Therefore, the accuracy of the process is known and improved compared to a process without step (g).

[0040] Thus, in the first case, the method allows to verify the chosen irradiation parameters taking into account the additional information acquired during the irradiation, so as to reach the certainty that the measurement accuracy is optimal for this sensor configuration and the state of the target on the measurement date; and in the second case, the method allows to adjust the irradiation parameters in a second step to obtain the best measurement accuracy taking into account the additional information acquired during the first irradiation; all this in a controlled energy consumption (in particular, without the need to acquire the complete spectrum at the irradiation frequency or the modulation frequency). It is worth noting that the correspondence table is essential to this process. It is precisely this correspondence table that allows to associate each current model configuration or one selected for irradiation with the most appropriate irradiation case, thus limiting the energy consumption of each irradiation.

[0041] According to different aspects, one and / or another of the following features may be predicted separately or simultaneously.

[0042] According to one embodiment, (g1) comprises: after (f), repeating (g). In this case, the irradiation is repeated as long as the irradiation parameters used are not optimal irradiation parameters for the state of the target on the measurement date, and optionally as long as the number of repetitions is below a predetermined threshold.

[0043] In particular, if the irradiation is repeated at a higher frequency at a characteristic frequency of the target structure change, this embodiment allows a gradual convergence to the irradiation parameters, so that the best accuracy can be obtained. Furthermore, since the number of irradiation parameters for each irradiation case is limited, the energy consumption of each irradiation is controlled, even if convergence can only be achieved after three, four, five or even ten consecutive irradiations, the total energy consumption can be controlled and in particular less than the energy consumption required to collect a complete spectrum at the irradiation frequency and / or modulation frequency, while the measurement accuracy can be controlled or even improved.

[0044] According to one embodiment of the process, a correspondence table is pre-generated using a processor and a model configuration database, which includes bytes (model configuration, irradiation cases, and parameters of interest) and acoustic or thermal signals associated with each byte detected by a detection unit, and the correspondence table is stored in a memory of the non-invasive sensor.

[0045] Due to this arrangement, the correspondence table can be obtained in an automatic manner and can be updated, for example, when the model configuration database is enriched or has to be adapted to a given patient or patient type.

[0046] According to one embodiment, the measuring method comprises:

[0047] The processor learns at least one inverse modeling algorithm from the model configuration database and stores the at least one inverse modeling algorithm in a memory of the non-invasive sensor.

[0048] Due to this arrangement, one or more inverse modeling algorithms can be learned in an automatic manner, for example each inverse modeling algorithm being suitable for one model configuration and one given irradiation case or for a set of model configurations and given irradiation cases.According to one embodiment of the method, at least a part of the acoustic or thermal signals detected by the detection units associated with the bytes stored in the model configuration database are simulated, i.e. they are generated using computer simulation means.

[0049] Thanks to this arrangement, a large number of bytes can be generated, for example, in order to correspond to various states of the target environment and various irradiation cases, much faster and at a much lower cost than if these bytes were obtained experimentally in a real situation. In the case where the target environment is a biological tissue (such as skin), bytes can be generated corresponding to rare physiological situations that are extreme or simply more unusual than the physiological situations to which a group of test patients are susceptible. Therefore, the correspondence tables generated from such a database can cover a larger range of situations with a much finer granularity than without simulation, so that ultimately the best irradiation case can be generated that is better adapted to each model configuration in terms of energy consumption and / or accuracy.

[0050] The present invention further relates to a non-invasive sensor based on photoacoustic detection or photothermal detection, configured to measure a parameter of interest in a target environment, comprising:

[0051] -light source;

[0052] - means for controlling parameters of the illumination from the light source;

[0053] - a detection unit configured to detect an acoustic signal or a thermal signal; and

[0054] - a memory, wherein a correspondence table is stored, the correspondence table comprising model configurations and optimal irradiation cases, each model configuration representing a given state of the target environment, each optimal irradiation case comprising a set of irradiation parameters, and each model configuration is associated with one optimal irradiation case;

[0055] The non-invasive sensor further comprises an adaptation module adapted to exchange information with the detection unit and the device for controlling the illumination parameters of the light source, the adaptation module comprising a processor adapted to implement an inverse modeling algorithm, the inverse modeling algorithm receiving an illumination case comprising a set of illumination parameters and an acoustic signal or a thermal signal as input and providing a model configuration and a value of a parameter of interest as output, the adaptation module further configured to:

[0056] i-Select the initial irradiation model configuration;

[0057] ii-determine the best irradiation case corresponding to the irradiation model configuration in the correspondence table;

[0058] iii-transmitting the best illumination case from the light source to the device for controlling the illumination parameters;

[0059] iv- receiving a signal detected by a detection unit;

[0060] v-determine the current model configuration and the estimation of the parameters of interest based on the received detected photoacoustic signal or photothermal signal and the best irradiation case;

[0061] vi-evaluate the irradiation model configuration by comparing it with the current model configuration;

[0062] vii-only in the case where the illumination model configuration of the selected target layered environment is different from the model configuration of the current target layered environment being compared, determining a new illumination model configuration when the current model configuration is received, determining a new best illumination case corresponding to the new illumination model configuration from the correspondence table, transmitting the new best illumination case to the illumination parameter control device, so that the light source illuminates the target layered environment according to the illumination parameter set of the new best illumination case, and the detection unit detects a new thermal signal or acoustic signal generated in response to the illumination, and transmits it to the illumination parameter control device. The target layered environment according to the illumination parameter set of the new best illumination case, and the detection unit detects a new thermal signal or acoustic signal generated in response to the illumination, and transmits it to the processor of the adaptation module (14) configured to repeatedly execute iv, v, vi and vii;

[0063] viii- Based on the last estimate of the parameter of interest, determine a measured value of the parameter of interest.

[0064] Due to this arrangement, the sensor is able to measure parameters of interest in a target environment, in particular biological tissue, with controlled accuracy and / or energy consumption. Finally, the invention relates to a computer program comprising instructions that drive a non-invasive sensor following the previous embodiments to perform the process steps according to any of the above embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] The embodiments of the present invention will be described below in conjunction with the accompanying drawings, and the accompanying drawings are briefly described below:

[0066] Figure 1 A typical modeling of a skin-like target layered environment is shown.

[0067] Figure 2 The steps of optimizing the measurement process in a specific embodiment of the present invention are shown in case it is desired to measure the interstitial blood glucose of a patient.

[0068] Figure 3 The main elements of a sensor 1 according to the invention are shown, the sensor 1 being in this case positioned in contact with a target environment 2 .

[0069] Figure 4 The steps implemented by the simulation module 15 are shown to simulate the photoacoustic signal to be detected by the photoacoustic unit triggered in response to the illumination of a target layered environment according to the illumination parameters of the illumination instance Ij, the target layered environment being modeled using the parameters of the model configuration CMk.

[0070] Figure 5 shows how a model works in reverse.

[0071] Figure 6a , 6b and 6c show the results of an impact analysis of SHAP (“SHapley Additive exPlanations”) type variables, performed on three inverse models, trained on:

[0072] - Figure 6a : The first “thick layer” group, corresponding to the model configuration of a double-layer skin, in which the thickness of the upper layer of the simulated stratum corneum is greater than 18 μm;

[0073] - Figure 6b : The second “thin layer” group, corresponding to the model configuration of a double-layer skin, in which the thickness of the upper layer of the simulated stratum corneum is less than 18 μm;

[0074] - Figure 6c : The entire first zone and the second zone;

[0075] Each of these models was trained on the signals detected after illumination at six different modulation frequencies.

[0076] exist Figure 6a , 6bIn 6c and 6d, the horizontal axis represents the SHAP value. Each point in the figure corresponds to a Shapley value of a variable and an instance. The position on the vertical axis is determined by the variable, and the position on the horizontal axis is determined by the Shapley value. The intensity (grey shading) represents the value of the variable.

[0077] Figure 7a Shown are the blood glucose values ​​predicted by a first inverse model trained on the basis of a first "thick slice" group from signals detected after irradiation at six different modulation frequencies.

[0078] Figure 7b shows the blood glucose value predicted by the second inverse model, which is based on Figure 6a The two variables identified in the paper that have the greatest impact on the first “thick layer” group and the signals detected after irradiation are trained based on the first “thick layer” group.

[0079] Figure 8a The blood glucose values ​​predicted by the third inverse model, which was trained on the basis of the second "thin layer" group according to the signals detected after irradiation at six different modulation frequencies, are shown.

[0080] Figure 8b shows the blood glucose value predicted by the fourth inverse model, which is based on Figure 6b The two variables identified in the paper that have the greatest impact on the second "thin layer" group and the signals detected after irradiation are trained based on the second "thin layer" group.

[0081] Figure 9a The blood glucose values ​​predicted by the fifth inverse model are shown, which is trained based on both groups (ie, the second "thin layer" group and the first "thick layer" group) according to the signals detected after irradiation at six different modulation frequencies.

[0082] Figure 9b shows the blood glucose value predicted by the sixth inverse model, which is based on Figure 6c The two most important variables identified above that have an impact on both groups (i.e., the second "thin layer" group and the first "thick layer" group) and the signals detected after irradiation are trained based on both groups (i.e., the second "thin layer" group and the first "thick layer" group).

[0083] In the drawings, the same reference numerals refer to the same or similar objects. DETAILED DESCRIPTION

[0084] The invention relates to a non-invasive sensor 1 for measuring one or more parameters of a target environment 2, in particular a target layered environment 2 whose structure may evolve over time. For example, the target layered environment 2 may be a surface layer (such as the skin) of a human or animal organism.

[0085] The parameter to be measured is referred to hereinafter as the "parameter of interest".

[0086] In case the layered environment is tissue from a human or an animal, the parameters of interest may in particular be physiological parameters.

[0087] For example, the physiological parameter to be measured is blood glucose (especially interstitial blood glucose). This may also involve measuring the moisture content in a specific layer of the skin or even the lactate concentration in a specific layer. These examples are non-limiting.

[0088] The sensor 1 may be wearable and may allow continuous monitoring of one or more parameters of interest.

[0089] The non-invasive sensor 1 may be based on photoacoustic detection or photothermal detection.

[0090] The measurement method is particularly suitable for improving the accuracy of non-invasive sensors based on indirect photoacoustic detection, for which the sensor detects the sound waves generated by irradiation in the fluid environment (especially gas) around the target environment 2, and controls the energy consumption of the sensor or reduces its energy consumption while controlling its accuracy. However, it is entirely possible to implement the process on non-invasive sensors based on photothermal energy or direct photoacoustic detection to obtain one of the two technical effects mentioned above. To simplify understanding, the exemplary indirect photoacoustic detection will be described in more detail below, but it can be easily analogized to sensors based on indirect photoacoustic or photothermal energy.

[0091] Figure 3 A non-invasive sensor 1 is schematically shown in FIG. The non-invasive sensor comprises:

[0092] - an illumination device 11, comprising a light source 11a, an intensity modulation device of the light source 11b and a control device 11c having at least one modulation frequency, the intensity modulation device 11b modulating the intensity of the light emitted by the light source according to the modulation frequency;

[0093] at least one photoacoustic detection unit 12, detecting a signal generated in response to irradiating the target environment 2 with emitted light (e.g. directly or indirectly detecting a heat wave propagating in the target environment 2);

[0094] - a signal processing module 13 configured to receive and process data from at least one detection unit 12;

[0095] - An adaptation module 14 of the illumination parameters and the calibration model;

[0096] - An analog module 15 , remote from the other elements of the non-invasive sensor 1 or on board together with them.

[0097] In a specific embodiment, the light source 11a emits a laser beam to the target environment 2 whose intensity is modulated at at least one specific wavelength.

[0098] The light source 11a can be in particular a light emitting diode (LED) or a laser chip. Additionally or alternatively, the light source 11a can be understood as a quantum cascade laser (QCL) (MIR-QCL) emitting in the mid-infrared region, an ICL laser ("interband cavity laser"), an internal or external cavity laser or a GaSb laser. These examples are non-limiting. The light source 11a can be selected according to the target environment 2 and / or the parameters of interest.

[0099] The non-invasive sensor 1 may include several light sources 11 a .

[0100] The non-invasive sensor 1 further comprises a circuit associated with one or more light sources 11a and at least one control device 11c, the control device being configured to control the illumination parameters of the light sources 11a, specifically:

[0101] - for at least one intensity modulation means 11 b to modulate the intensity of at least one light source 11 a to modulate the frequency of the intensity of light emitted by the light source 11 a at an adjustable modulation frequency.

[0102] - Optionally, the wave number (or equivalent wavelength) of the light emitted by the light source 11a

[0103] - and / or optionally, the luminous power of the light source 11a at a given wavelength.

[0104] The frequency used to modulate the intensity of a given light source 11a at a given wavelength is referred to in the remainder of this description as fmod(λ).

[0105] Several wavelengths may be associated with the same modulation frequency, and several modulation frequencies may be associated with the same wavelength. The light emitted by the light source 11a at a given wavelength λ may be characterized by the wavelength and the corresponding modulation frequency and optionally the corresponding optical power and / or other parameters such as the integration time or duty cycle characterizing the laser pulses.

[0106] The intensity of the light source 11a may be modulated by any known electrical or mechanical means.

[0107] The light may be emitted by the light source 11a continuously or in a pulsed manner.

[0108] The light incident on the target environment 2 emitted by the intensity modulated one or more light sources 11a propagates to the target environment 2 and then passes through the target environment 2 (at Figure 3The incident light is then gradually absorbed by different components of the target 2 at a depth zmax, which depends on the structure of the target environment 2 and its physicochemical composition. The absorption of light energy causes the target environment 2 to be heated locally. As a result, a thermal wave with a frequency equal to the modulation frequency of the light source propagates in the target environment 2 (at Figure 3 The heat wave generates a pressure wave of the same frequency in the gas environment outside the target, and the pressure wave propagates in the gas environment around the target environment 2 and in particular in the photoacoustic detection unit 12 (in the Figure 3 The phenomenon is indicated by the alternating dashed arrows in the figure).

[0109] In the case of photoacoustic detection, the detection unit 12 comprises a gas (e.g. air) filled chamber in which the sound waves propagate and in which one or more sensors are suitably placed, e.g. facing the target environment 2. For example, these are one or more electroacoustic sensors (such as microphones or piezoelectric transducers) configured to convert the pressure of the sound waves into electrical signals.

[0110] Each electroacoustic sensor is functionally connected to a signal processing module 13 .

[0111] The signal processing module 13 may include an analog-to-digital converter configured to convert the analog electrical signal from the electroacoustic sensor into a digital signal.

[0112] The signal processing module 13 may include a synchronous detection device adapted to demodulate and extract the signal of interest from the detected signal.

[0113] The signal processing module 13 may include an operational amplifier operatively connected to the analog-to-digital converter and configured to amplify the electrical signal transmitted by the electroacoustic sensor originating from the acoustic response of the target 2 .

[0114] In an exemplary embodiment, the analog-to-digital converter is functionally connected to a digital signal processor for digital signal processing.

[0115] The non-invasive sensor 1 according to the invention further comprises a parameter adaptation module 14 for the illumination and calibration model and a simulation module 15. These two elements will be described in the following sections.

[0116] Adaptation module 14

[0117] The adaptation module 14 is a computerized device including at least one processor

[0118] - it can exchange information with the simulation module 15 described below;

[0119] - it may receive information from the detection unit 12 and / or the signal processing module 13 (if applicable);

[0120] -and it can transmit information on at least one illumination parameter of the light source 11a (for example, the modulation frequency of the intensity modulation device 11b modulating the light intensity emitted by the light source 11a at a given wavelength, the wave number (or equivalent wavelength) of the light emitted by the light source 11a, the optical power of the light source 11a at a given wavelength and optionally a given modulation frequency, etc.) to one or more control devices 11c.

[0121] In the following specific cases, Figure 2 The steps performed by the non-invasive sensor 1 for measuring a parameter of interest comprising an adaptation module 14 are schematically shown:

[0122] - Target environment 2 is the skin (so in this case, target environment 2 is the layered environment)

[0123] -The parameter of interest is interstitial glucose.

[0124] The non-invasive sensor 1 further comprises a memory for a model configuration database, a correspondence table and one or more inverse models as described below.

[0125] The memory may be distributed in the adaptation module (14) and / or the simulation module (15) and / or shared by them.

[0126] The adaptation module 14 performs the steps of the process which allow the selection in the model configuration database of the model configuration of the target environment 2 which best suits the target environment 2 for the measurement date, the selection of the best illumination parameters for the measurement based on the correspondence table and the determination of the inverse model which best suits (i.e., is most accurate) for calculating the parameters measured from the signals detected for this particular measurement. These steps will be described after they have been described which allow the generation of the model configuration database and the correspondence table.

[0127] The model configuration database and the correspondence table are generated by a simulation module 15 embedded in the sensor 1 or remotely. In the case of a remote simulation module 15, the non-invasive sensor 1 comprises communication means so that the simulation module 15 and the adaptation module 14 can exchange data.

[0128] Simulation Module 15

[0129] like Figure 4As shown, the simulation module 15 is a computerized device configured to generate a set of configurations of a model (or even modeling or description) each corresponding to a specific state of the target environment 2 from an analytical model of the target environment 2 and a photoacoustic detection unit (or applicable photothermal detection unit) 12, a set of illumination cases of the target environment 2, and a photoacoustic signal (or applicable photothermal signal) theoretically detected in response to each illumination case for each model configuration CMk of the target environment 2.

[0130] The simulation module 15 is particularly useful in cases where the target environment 2 evolves over time and / or is layered. In this case, the target environment 2 adopts a real configuration that changes over time (because one or more concentrations change in one or more layers of the target environment 2, so one or several dimensions (such as, for example, the thickness of one of the layers of the target environment 2 and / or the number of layers in the target environment 2) change), each of which can be modeled by a specific model configuration.

[0131] (a) Multi-physics field analysis model of target environment 2

[0132] like Figure 1 As shown, the target environment 2 to be analyzed is modeled. The target environment 2 separates the external environment A from the internal environment B and is composed of N continuous layers, the interfaces of which are assumed to be local planes, for example. Each layer is described by one or more parameters of interest and a plurality of explicit parameters applicable to the target environment 2 of interest (referred to as level 1 parameters because their values ​​will be provided as input to the simulation module 15 for each simulation) For example, if the target environment 2 is the skin and one seeks to measure the glucose concentration in layer j of this target environment 2, the target environment 2 will be described by its number of layers N and the concentration of interest [Glc]j, and for modeling purposes, each layer i can be described by its thickness ei and its concentration in water [H2O]i. For i≠j, the glucose concentration [Glc]i can be considered. In this case, concentration and concentration The value of is a level 1 parameter.

[0133] If more accurate modeling is desired, the list of level 1 parameters can be enriched. In particular, the concentrations of other components of the skin (such as fat, lactic acid, oxygen, etc.) can be included in the list of level 1 parameters describing a layer of the skin.

[0134] Still for the example of skin, skin color, patient age, or any other anthropometric parameter may be considered in order to expand or restrict the space of potential models.

[0135] The analytical model may be used to calculate non-explicit parameters (referred to as Level 2 parameters, since no values ​​are provided as input to the simulation module 15) of the model target environment 2. For example, the thermal conductivity, heat capacity, density, or absorption coefficient of each layer of the target environment 2 at each wavelength may be derived from the Level 1 parameters and known equations.

[0136] One or more parameters of interest that play a specific role, which are not included in the list of settings at level 1. Depending on the stage of the process, this parameter of interest will be known or unknown: in order to perform a simulation using the simulation module 15, its value is known; but in the case of actual measurements using the non-invasive sensor 1, its value is of course unknown.

[0137] The number of layers N of the target environment 2 can also be a variable in the model. Still in the example of the skin, N can therefore be greater than or equal to one or two, depending on the physiological situation. Thus, for certain physiological situations, the skin can be appropriately described by two layers; the first layer corresponds to the stratum corneum, whose glucose concentration can be, for example, lower; and the second layer corresponds to the rest of the skin, the glucose concentration of the second layer being equivalent to the interstitial glucose concentration to be measured.

[0138] For other cases, a model with three or more layers may be more appropriate. In the latter case, the moisture concentration of a layer may, for example, increase with the depth of the layer.

[0139] Therefore, the number of layers N of the target environment 2 may not be a constant.

[0140] In the case of the skin, the external environment A is typically the atmosphere surrounding the patient, which is also filled with the photoacoustic detection unit.

[0141] The simulation module 15 implements a multiphysics analysis model of the layered environment 2, which is for example based on physical and / or chemical equations such as, as non-limiting examples, the Beer-Lambert equations for light absorption and the thermodynamic equations for heat (Fourier's laws and conservation laws).

[0142] The model configuration CMk (k positive integers) of the target environment 2 corresponds to (or even models) a specific state of a given target environment 2. It is assumed that this specific state is properly represented by data of the number of layers N and the values ​​of the level 1 parameters of each layer.

[0143] For each model configuration CMk, if the parameters of interest are better known, the multiphysics field analysis model allows simulating the generation of thermal waves at the interface between layer 1 and the external environment A (interface 1 / A) in response to the illumination of the light source 11a, and the illumination parameters of the light source 11a are known (i.e., for example, the wavelength λ, the modulation frequency fmod(λ) of this wavelength, and the surface power density of this wavelength).

[0144] As a variant, the multiphysics analysis model allows the simulation of pressure waves generated in the external environment A.

[0145] In both cases, the signal obtained at the output of the processor implementing the multiphysics analysis model is called the “simulated response wave”.

[0146] The simulated response wave may be provided as input to a processor implementing the detection unit model.

[0147] (b) Detection unit model

[0148] The non-invasive sensor 1 based on photoacoustic detection or photothermal detection includes a photoacoustic detection unit (or photothermal detection unit) 12, which is configured to detect and analyze the wave pressure (or thermal wave) generated in the external environment A when the thermal wave generated in the target environment 2 in response to irradiation reaches the interface 1 / A.

[0149] The entire detection unit 12 can be analytically modeled. The detection unit model 12 allows the output signal of the detection unit 12 to be predicted based on the response waves simulated by the multi-physics analytical model (theoretically received at the input of the detection unit 12).

[0150] Different models can be considered.

[0151] Therefore, in the case of indirect photoacoustic detection, the parameters of the detection unit model 12 (hereinafter referred to as unit parameters) may include these dimensions (e.g., the size of the through-port, the height of the unit, etc.) and thermodynamic state parameters (temperature, air pressure, relative or absolute humidity, etc.). In particular, the photoacoustic detection unit may be modeled using an equivalent RLC circuit. For example, the model described in "Tea Infineon Silicon MEMS Microphone" (2013) by Dehe and Alfons et al. may be suitable.

[0152] If necessary, step model signal processing performed by the signal processing module 13 can be included in the photoacoustic detection unit model to generate a corresponding signal theoretically obtained at the output end of the photoacoustic detection unit from each simulated response wave generated by the processor implementing the multi-physics field analysis model (as well as a corresponding case applicable after the signal processing module 13 processes the signal).

[0153] The processor of the simulation module 15 may be configured to implement a photoacoustic detection unit model.

[0154] Due to the multi-physics field analysis model of the target environment 2 and the detection unit model 12, a global analysis model can be provided, which allows predicting the expected signal at the output of the detection unit 12 or (where applicable) the signal processing module 13 based on the model configuration CMk of the target environment 2 and the data of the irradiation case Ij, provided the parameters of interest (since it is a simulation, the parameters of interest can be effectively selected). Figure 4 This is shown in .

[0155] In summary, the simulation module 15 therefore receives as input the parameters of the model configuration CMk of the target environment 2 (i.e. the number of layers N of the target environment 2 and the level 1 parameters of each layer as well as the parameters of interest and the illumination parameters of the illumination case Ij (i.e. the illumination case comprising one or more wavelengths of light emitted by one or more lasers, one or more corresponding modulation frequencies of the intensity of said one or more lasers and optionally the corresponding illumination power of said one or more lasers). At output, for the selected model configuration CMk of the target environment 2, the simulation module 15 provides the signal theoretically expected at the output of the detection unit 12 or, if necessary, the simulated signal theoretically expected at the output of the signal processing module 13, called the simulated output signal.

[0156] The analog output signal can be stored in memory in the form of a Fourier spectrum.

[0157] The bytes {model configuration CMk of target environment 2, irradiation case Ij, parameters of interest, amplitudes and phases of components of the simulated output signal} can be stored in the model configuration database.

[0158] Model Configuration Database

[0159] Thus, a large number of model configurations CMk can be generated in an automatic and / or random manner, each model configuration CMk corresponding to multiple layers N and a set of level 1 parameters, and optionally corresponding to a value or value range of a parameter of interest (describing a specific situation in the target environment 2 of interest).

[0160] For each model configuration CMk, a large number of illumination cases Ij may be generated in an automatic and / or random manner, each illumination case corresponding to a set of illumination parameters (describing the parameters of one or more light sources 11a used for illumination).

[0161] Thus, an irradiation case Ij can include one or more intensity modulation frequencies of one or more lasers, the wavelength of each of these lasers, and optionally the irradiation power of each laser.

[0162] The amplitude and phase of each component of the analog output signal obtained at the processor output of the simulation module 15 are calculated using the simulation module 15, which implements a global analysis model, which includes, where applicable, a multi-physics field analysis model and a sensing unit model for each model configuration CMk for each irradiation case Ij, and also provides the values ​​of the parameters of interest.

[0163] The illumination case Ij may be the same for several different model configurations CMk and possibly for several values ​​of the parameter of interest, or the illumination case Ij may be different for different model configurations CMk and / or different values ​​of the parameter of interest. Once the simulation has been performed, we may store all these model configurations CMk, illumination cases, parameter of interest values, and the associated photoacoustic signal (or photothermal signal where applicable) simulated as bytes {template configuration CMk, illumination case Ij, parameter of interest, amplitude and phase of components of the simulated output signal} in a model configuration database.

[0164] The generation of model configurations CMk and / or irradiation cases Ij may not be completely random.

[0165] The generation of the model configuration CMk can be based on physiological considerations to limit the space of possibilities to physiologically realistic model configurations. For example, the possible thickness of the first layer of the skin can be limited to the range actually observed by experiments [8mm, 40mm], and the water concentration of this layer can be limited to a limited range for each thickness, the water concentration of the stratum corneum being related to its thickness.

[0166] The generation of illumination cases Ij may in particular take into account the limitations of the light source 11a of a given wavelength and / or power (or even a modulation frequency range related to the type of target environment 2 to be analyzed or a wavelength related to the parameter of interest) available for the non-invasive sensor 1.

[0167] As a variant, the model configuration database may comprise only the bytes {model configuration CMk, illumination case Ij, parameters of interest, amplitudes and phases of the components of the measured signal} obtained by experiments in real situations, or simultaneously these bytes obtained in real situations and those obtained by simulation.

[0168] Correspondence table

[0169] According to the model configuration database, the artificial intelligence model can be trained in the simulation module 15.

[0170] After learning, the AI ​​model is able to solve the inverse problem of finding one or more parameters of interest and the model configuration CMk of the target environment 2 given the simulated photoacoustic signal and the illumination parameters of the illumination case Ij, thereby finding the number of layers N and the level 1 parameters, such as Figure 5 shown.

[0171] The learned model (hereinafter referred to as the inverse model) may be transmitted to the adaptation module 14 and stored in the memory of the module.

[0172] Several different inverse models can be learned using learning games and / or different learning rules.

[0173] To the extent that several irradiation cases Ij can be associated with the same model configuration CMk, statistical analysis techniques and / or artificial intelligence can also be employed to identify irradiation cases Ij such that the parameters of interest can be measured with the required accuracy and / or data measurement quality with minimal energy consumption.

[0174] In particular, the influence of variables (“feature importance”) can be analyzed in the trained model. At the end of the analysis, each model configuration CMk can be associated with an optimized illumination case I opt.k (including a limited number of illumination wavelengths and a limited number of modulation frequencies fmod for each illumination wavelength).

[0175] It is possible to work model configuration by model configuration or to classify the model configurations and associate each class of model configurations with an optimal irradiation case I opt.cat.k.

[0176] according to Figures 6a to 9b The meaning of the corresponding table can be understood. The background of these figures is the measurement of interstitial blood glucose in the skin. The inverse model is trained based on a model configuration database, which in this case includes bytes (configuration model (CMk), irradiation case (Ij) and parameters of interest) and an acoustic signal or thermal signal associated with each byte detected by the detection unit.

[0177] The data in the database is divided into two groups:

[0178] - The first "thick layer" group corresponds to a two-layer skin model configuration, in which the thickness of the upper layer simulating the stratum corneum is greater than 18 mm;

[0179] - The second "Thin Layer" group corresponds to a two-layer skin model configuration, where the thickness of the upper layer simulating the stratum corneum is less than 18 mm.

[0180] Based on the signals detected for six different modulation frequencies, an inverse model is trained for each group. Then, the influence of the variables is analyzed for each of the two inverse models obtained, as Figure 6a and Figure 6bThe analysis of the impact of variables leads to the following conclusions:

[0181] - For the model configuration of the "thick layer" group, the illumination parameters {wavelength 1034 cm-1; two different modulation frequencies: 50 Hz and 200 Hz} are sufficient to obtain the required accurate blood glucose measurements, and the analysis of the modulus (equivalent to the amplitude) of the signal detected at each of these frequencies is sufficient;

[0182] - For the model configuration of the "Thin Layer" group, the analysis of the illumination parameters {wavelength 1034 cm-1; two different modulation frequencies: 50 Hz and 400 Hz} and the modulus (equivalent to the amplitude) of the signal detected at each of these frequencies is sufficient to obtain blood glucose measurements with the required accuracy.

[0183] Therefore, it should be understood that since the shot cases match the detected model configuration, the process can reduce the energy consumption of each shot while controlling the measurement accuracy.

[0184] This can be seen by observing that with full illumination (i.e., in this case following the six modulation frequencies envisaged ( Figure 7a For the "thick layer" group; Figure 8a For the "thin layer" group)) or with irradiation limited to the best irradiation cases ( Figure 7b For the "thick layer" group; Figure 8b This is confirmed by measurements obtained for the "thin layer" group).

[0185] For the "thick layer" group, when only two modulation frequencies are selected from the six potential modulation frequencies (50 Hz and 200 Hz in this case) and only the amplitude (or equivalently the modulus) of the signal component detected at each of these frequencies is analyzed, such that:

[0186] - The mean square error has changed by 7% relative;

[0187] -Energy consumption has been reduced by at least one third.

[0188] The mean square error varied from 10.5 mg / dL to 11.2 mg / dL, thus remaining well below the target threshold of 20 mg / dL (considered the maximum acceptable RMSE) when selecting the variables of interest for the best irradiation case.

[0189] For the "thin layer" group, when only two modulation frequencies are selected from the six potential modulation frequencies (50 Hz and 400 Hz in this case) and only the amplitude (or equivalently the modulus) of the signal component detected at each of these frequencies is analyzed, such that:

[0190] - The mean square error has changed by 20% relative;

[0191] -Energy consumption has been reduced by at least one third.

[0192] The mean square error changes from 4.9 mg / dL to 5.8 mg / dL, thus remaining well below the target threshold of 20 mg / dL (considered the maximum acceptable RMSE) when selecting the variable of interest for the best irradiation case. Even though the relative change in the mean square error is not negligible in this case, the absolute value of the mean square error is thus kept under control during the selection of the best irradiation case.

[0193] On the other hand, in the case where the model configuration for a specific irradiation case is not associated with a specific inverse model (i.e., in the case where a single inverse model is trained for all model configurations of the “thin layer” group and the “thick layer” group and in the case where irradiation is performed according to irradiation cases of all these model configurations applicable to the “thin layer” group and the “thick layer” group), it can be found that no technical effect is obtained.

[0194] exist Figure 6c It can be observed that the best irradiation case is selected for both groups, for which the irradiation parameters to be selected are {wavelength 1034 cm-1; two different modulation frequencies: 200 Hz and 400 Hz}, and an analysis of the modulus (equivalent to the amplitude) of the signal detected at each of these frequencies must be completed.

[0195] from Figure 9a and Figure 9b It can be observed in Figure 1 that the mean square deviation increases threefold when switching from irradiation according to six different modulation frequencies (RMSE = 13.0 mg / dL) to irradiation according to the two most influential identified modulation frequencies, namely 200 Hz and 400 Hz (RMSE = 26.9 mg / dL), analyzing only the amplitude of the signal detected at each of these frequencies. The threshold of the RMSE (fixed at 20 mg / dL in this example) is also exceeded, indicating that other influencing variables must be considered in order to take this threshold into account.

[0196] The correspondence table may include the best irradiation cases for all groups, especially taking into account the initial irradiation. In the above example, it should be understood that at least one additional modulation frequency (i.e., 50 Hz) must be included to meet the criterion for the RMSE value of the initial irradiation. Then, the energy consumption may be reduced, but only by a factor of two.

[0197] Therefore, in Figures 7a to 9b It can be seen that the technical effect of reducing energy consumption during precision-controlled measurement is achieved, in particular with the help of the correspondence table, so that a set of potential states of the target environment 2 can be divided into several categories and the best illumination case can be associated with each of these categories.

[0198] In this case, according to usage in the field of blood glucose sensors, the number of illumination parameters (i.e., the two quantities that carry the most information) is selected to ensure that the RMSE remains less than 20 mg / dL even under illuminations that do not include all possibilities and the accuracy is better than this threshold in hypoglycemia. However, three, four, or more illumination parameters can be selected among all potential illumination parameters to ensure that the RMSE value is below another threshold (e.g., 15 mg / dL, 10 mg / dL, or even 5 mg / dL).

[0199] The number of shot parameters chosen is a compromise between available power per shot and an acceptable RMSE.

[0200] The {CMk; Iopt.k} pairs or {model configuration category k, Iopt.cat.k} pairs are stored in a corresponding table in the memory of the non-invasive sensor 1 (eg, the memory of the simulation module 15 and / or the adaptation module 14 ).

[0201] The inverse model corresponding to each model configuration CMk or each model configuration class is also stored in the memory of the non-invasive sensor 1 (eg, the memory of the simulation module 15 and / or the adaptation module 14 ).

[0202] At this stage, the necessary elements for carrying out the measuring method according to the invention are already in place.

[0203] It has been learned that, thanks to the correspondence table, it is not necessary to establish a complete spectrum of the target for the state of the measured target during the measurement process: the best illumination case corresponding to the current model configuration (i.e. the value of the level 1 parameter that best describes the target environment 2 currently being measured) can be automatically selected from the correspondence table.

[0204] Therefore, compared with the prior art process, the measuring method according to the invention can reduce the energy consumption of the sensor 1 with the same or even higher accuracy.

[0205] Methods for measuring parameters of interest

[0206] Figure 2 A method for measuring a parameter of interest by means of a non-invasive sensor 1 based on indirect photoacoustic detection is shown in FIG. The method comprises the following steps:

[0207] (a) Initialization:

[0208] An initial model configuration of the target environment 2 is selected for the first irradiation based on predetermined criteria. Figure 2 , the initial model configuration CMirrad is the model configuration CMk.

[0209] Different options can be used for this initialization.

[0210] In particular, in the case of interstitial blood glucose measurements based on a model configuration database and / or a database of experimental measurements, a set of level 1 parameters and blood glucose averages can be determined for a patient population or a given patient. This set of average parameters corresponds to an initial model configuration CMk that is most likely to be the best fit for the next measurement in the absence of any other information (especially in the absence of historical measurements).

[0211] (b) Based on the correspondence table, the processor of the adaptation module 14 determines the illumination case Iopt.k associated with the initial model configuration (eg, CMk).

[0212] (c) First illumination: The target environment 2 is illuminated for the first time using one or more light sources 11a according to the illumination parameters of the best illumination case Iopt.k for the model configuration CMk.

[0213] (d) PA detection:

[0214] The real photoacoustic signal generated in response to the illumination is detected by means of the photoacoustic detection unit 12. The real photoacoustic signal is received and processed by the signal processing module 13 where applicable and transmitted to the adaptation module 14 after processing.

[0215] (e) Inverse problem solving:

[0216] The processor of the adaptation module 14 implementing the learned one or more inverse models (and in particular the inverse model corresponding to at least the model configuration CMk) receives as input the real photoacoustic signal and the illumination case Iopt.k and determines the current model configuration of the target environment 2 (denoted as CMmes) and the parameters of interest (such as Figure 5 shown).

[0217] (f) Model configuration verification:

[0218] The adaptation module 14 compares the current model configuration CMmes of the target environment 2 with the model configuration CMirrad used for illumination (the model configuration CMk from the initialization step for the first measurement, a possibly different model configuration CM1 for further measurements). There are two potential cases:

[0219] (Case 1) If the model configuration CMmes of the measured target environment 2 is the same as the model configuration CMirrad used for irradiation, then the selected irradiation parameters are optimal for the current physiological situation (which, it should be noted, is not known a priori and evolves over time), and similarly, the inverse model used to determine the parameters of interest is also optimal. Therefore, the parameters of interest determined by the adaptation module 14 in step f are the measurement results.

[0220] (Case 2) If the model configuration CMmes of the measured target environment 2 is different from the model configuration CMirrad, the processor of the adaptation module uses the correspondence table to determine a new model configuration CMl (l≠k) that is more suitable for the model configuration CMmes estimated by the adaptation module 14. In particular, CMl=CMmes can be selected. However, other options are also possible, especially if it is desired to take into account the measurement history over a longer period of time (considering at least two previously selected model configurations). The processor of the adaptation module 14 then searches the correspondence table for the best irradiation case Iopt.l for the model configuration CMl of the target environment 2 and transmits the corresponding parameters to the irradiation device 11.

[0221] Then, another irradiation (subsequent irradiation) is performed on the target environment 2 according to the irradiation parameters of the irradiation case Iopt.1. Then, the detection and (if applicable) photoacoustic signal processing of step (d) and the inverse problem solving of step (e) are repeated.

[0222] The method may include the model configuration verification of step (f) at most once.

[0223] Alternatively, the method may further include: repeatedly performing the model configuration verification of step (f).

[0224] The model configuration CMmes for target environment 2 can be different from the model configuration CMirrad. In fact, CMirrad is selected based on the following conditions:

[0225] - initialization criteria (especially in the absence of patient measurement history, CMirrad is the "generic" model configuration. This is the case for first irradiations); or

[0226] - Previous measurement (CMirrad is the most appropriate phantom configuration for the patient, knowing the measurement results after the previous irradiation).

[0227] Thus, a verification of CMirrad is performed thanks to the additional information acquired by the current shot (i.e. the signal detected by the photoacoustic detection unit 12). If by chance CMmes=CMirrad, there is no need to perform another shot, and although the value of the parameter of interest is the most accurate value that can be obtained, the model configuration verification step provides additional information (i.e. confirmation that the measurement accuracy is the highest for this case).

[0228] In this case, CMmes differs from CMirrad in that the matching of the irradiation parameters and the configuration model improves the measurement accuracy at the expense of at least one additional irradiation, but the energy consumption is always kept under control and, thanks to the second validation step, it is confirmed that the measurement accuracy is the highest possible.

[0229] In cases where repeated verification steps are allowed (such as Figure 2 As shown), additional information with the highest measurement accuracy is obtained at the output of the process.

[0230] Typically, in this case, the measurements of the parameters of interest are obtained after the first irradiation or after two irradiations. However, in order to ensure process convergence and / or limit energy consumption, a limitation on the number of model configuration verification steps can be foreseen.

[0231] In all cases, it can be seen that by selecting the illumination case Ij based on the correspondence table, the number of modulation frequencies and wavelengths used for illumination can be limited by retaining only values ​​that carry non-redundant information about the model configuration of the current physiological target environment 2 and are sufficient to obtain the required and / or optimal measurement accuracy to limit the energy consumption of the sensor to a predetermined value.

[0232] For example, if the physiological situation corresponds to 3 layers, each characterized by 2 concentrations (e.g., water and blood glucose), then the solution of the inverse problem is a problem with 2*3=6 unknowns. Knowing the amplitude and phase of each of the three components of the photoacoustic signal corresponding to the modulation frequency and the selected wavelength, it should be possible to solve the inverse problem with the required accuracy. The difficulty lies in the optimal choice of these modulation frequencies and the best wavelength, which can be solved using a correspondence table.

[0233] Therefore, the energy consumption of the sensor 1 is limited or controlled.

[0234] Even in cases where two or three irradiations are required, each irradiation step requires a limited power, which is lower than the power required to obtain a complete absorption spectrum for each potential frequency modulation.

[0235] The model configuration validation step enables:

[0236] - automatic recalibration of the sensor 1 , which ensures that the inverse model from which the value of the parameter of interest of the detected photoacoustic signal can be derived is the one most relevant to the current physiological situation, without requiring a priori knowledge of this situation;

[0237] - Adjustment of the irradiation case, which ensures that the parameters used for subsequent irradiations are optimal with respect to the knowledge of the target environment and in particular provided by the last detected photoacoustic signal.

[0238] Several embodiments can be considered to further optimize the process measurement according to the invention.

[0239] In particular, corresponding to the actual situation of a given patient or a given group of patients, the artificial intelligence model of the simulation device 15 can be pre-trained based on the simulation model configuration database and then re-trained based on the experimental model configuration database to ensure that the inverse model correctly predicts the actual situation without requiring a set of detailed experimental data. In this case, the inverse learning model can include a transfer learning phase ("transfer learning").

[0240] Reference numerals list

[0241] 1 Non-invasive sensor based on indirect photoacoustic detection

[0242] 11 Irradiation device

[0243] 11a Light Source

[0244] 11b means for modulating the intensity of the light source 11a

[0245] 11c Modulation frequency control means fmod for intensity of light source 11a

[0246] 12 Detection Unit

[0247] 13 Signal Processing Module

[0248] 14 Adapter Module

[0249] 15 Analog Module

Claims

1. A method for measuring a parameter of interest in a target environment (2) by means of a non-invasive sensor (1) based on photoacoustic detection or photothermal detection, include: a) providing a sensor, the sensor comprising: - Light source (11a) - means for controlling the illumination parameters of said light source (11a); - a detection unit (12) configured to detect an acoustic signal or a thermal signal; - a memory, wherein a correspondence table is stored, the correspondence table comprising model configurations (CMk) and optimal irradiation cases (Iopt.k), each model configuration representing a given state of the target environment (2), each optimal irradiation case comprising a set of irradiation parameters, and each model configuration (CMk) is associated with one optimal irradiation case; - an adaptation module (14) for exchanging information with the detection unit (12) and the illumination parameter control means of the light source, the adaptation module (14) comprising a processor suitable for implementing an inverse modeling algorithm, the inverse modeling algorithm receiving as input an illumination case (Ij) comprising a set of illumination parameters and an acoustic signal or a thermal signal, and providing as output a model configuration (CMk) and the value of the parameter of interest; b) the adaptation module (14) selects an initial irradiation model configuration (CMirrad); c) the adaptation module (14) determines from the correspondence table the best irradiation case for the selected irradiation model configuration (CMirrad), i.e. the one that allows measuring the parameter of interest with a predetermined accuracy and / or with a minimum amount of measurement data that consumes the least energy; d) the light source (11a) illuminates the target environment (2) according to the illumination parameter set of the best illumination case; e) the detection unit (12) detects an acoustic signal or a thermal signal generated in response to the irradiation; f) the processor of the adaptation module (14) implements the inverse modeling algorithm, receives as input the acoustic signal or thermal signal detected by the detection unit (12) and the best irradiation case for irradiation, and returns as output the current model configuration (CMmes) and the estimation of the parameter of interest (Pest); g) the processor of the adaptation module (14) evaluates the selected irradiation model configuration (CMirrad) by comparison with the current model configuration (CMmes), and only if the selected irradiation model configuration (CMirrad) is different from the current model configuration (CMmes): g1) the adaptation module (14) receives the current model configuration (CMmes) as input and returns a new model configuration (CMirrad) for irradiation as output, and then repeats c), d), e) and f); h) The value of the parameter of interest measured by the sensor (Pmes) is the final estimate of the parameter of interest (Pest).

2. The measuring method according to claim 1, g1) further comprising: include: After f), repeat g).

3. The measuring method according to any one of claims 1 and 2, include: I—Generate a correspondence table with the aid of a processor and a model configuration database, wherein the model configuration database includes bytes (model configuration (CMk), irradiation case (Ij) and parameters of interest) and an acoustic signal or a thermal signal associated with each byte detected by the detection unit (12), and the correspondence table is stored in the memory of the non-invasive sensor (1).

4. The measuring method according to claim 3, include: II—The processor learns at least one inverse modeling algorithm from the model configuration database, and the at least one inverse modeling algorithm is stored in the memory of the non-invasive sensor (1).

5. A measurement method according to any one of claims 3 and 4, wherein at least a part of the acoustic signal or thermal signal detected by the detection unit (12) and associated with the bytes stored in the model configuration database is simulated, that is, at least a part of the acoustic signal or thermal signal is generated by means of a computerized simulation device.

6. A non-invasive sensor (1) based on photoacoustic detection or photothermal detection, configured to measure a parameter of interest in a target environment (2), include: - Light source (11a) - means for controlling the illumination parameters of said light source (11a); - a detection unit (12) configured to detect an acoustic signal or a thermal signal; - a memory, wherein a correspondence table is stored, the correspondence table comprising model configurations (CMk) and optimal irradiation cases (Iopt.k), each model configuration representing a given state of the target environment (2), each optimal irradiation case comprising a set of irradiation parameters, and each model configuration (CMk) is associated with one optimal irradiation case (Iopt.k); The non-invasive sensor further comprises an adaptation module (14), which is suitable for exchanging information with the detection unit (12) and the light source to control the illumination parameters of the light source (11a), and comprises a processor suitable for implementing an inverse modeling algorithm, which receives an illumination case (Ij) including a set of illumination parameters and an acoustic signal or a thermal signal as input, and provides a model configuration (CMk) and the value of the parameter of interest as output; The adaptation module (14) is further configured as follows: i—Select the initial model irradiation configuration (CMirrad); ii—determining the best irradiation case corresponding to the irradiation model configuration from the correspondence table; iii—transmitting the best irradiation case to the light source irradiation parameter control device; iv—receiving a signal detected by the detection unit (12); v—determine the current model configuration (CMmes) and the estimation of the parameter of interest (Pest) based on the received detected photoacoustic signal or photothermal signal and the best irradiation case; vi—Evaluate the model irradiation configuration (CMirrad) by comparing it with the current model configuration (CMmes); vii—only when the model irradiation configuration (CMirrad) of the selected target layered environment is different from the model configuration (CMmes) of the compared current target layered environment, a new model irradiation configuration (CMirrad) is determined when the current model configuration (CMmes) is received, a new best irradiation case corresponding to the new irradiation model configuration (Cmirrad) is determined from the correspondence table, the new best irradiation case is transmitted to the irradiation parameter control device, so that the light source irradiates the target layered environment according to the irradiation parameter set of the new best irradiation case, and the detection unit detects a new thermal signal or acoustic signal generated in response to the irradiation and transmits it to a processor of the adaptation module (14) configured to repeatedly execute iv, v, vi and vii; - Based on the last estimate of said parameter of interest (Pest), determining a measured value of the parameter of interest (Pmes).

7. A computer program comprising instructions for directing the non-invasive sensor (1) according to claim 6 to perform the steps of the method according to any one of claims 1 to 5.

Citation Information

Patent Citations

  • Device for estimating blood sugar level

    EP2460470A1

Cited By

  • Differential near-infrared probe sensor for precise non-invasive intracranial pressure monitoring and monitoring method

    CN120130981A