Method and device for multi-frequency bioelectric impedance analysis

Abstract

FIELD: medicine.SUBSTANCE: group of inventions relates to a method for multi-frequency bioelectrical impedance analysis and a device for implementing it. Alternating electric current signals are passed through the human body. Each signal has one of three different frequencies. At each frequency, the amplitude of the voltage of the signals passing through the human body is measured. An initial phase angle is selected for each specific impedance module. The phase angle characterizes the phase shift of the current of signals entering the human body relative to the phase of the voltage of signals passing through the human body. At each frequency, the phase angle is estimated using an equivalent electrical model of the human body, the initial phase angle and the impedance modulus. A body model is created for a specific person. The ratio of extracellular water content to total body water content in a human is determined by inputting extracted parameters of the human body model into a trained machine learning model. The device comprises a source of alternating electric current signals, electrodes for conducting and reading signals, a measuring device for measuring signal parameters, a processor and memory.EFFECT: device is compact and has small-area electrodes for integration into wearable and / or portable devices while maintaining measurement accuracy.14 cl, 6 dwg

Description

BACKGROUND OF THE INVENTIONThe field of technology to which the invention relates

[0001] The present invention relates generally to the evaluation of body composition of a living being using multi-frequency bioelectrical impedance analysis, and in particular to a method and apparatus for multi-frequency bioelectrical impedance analysis. Description of the Prior Art

[0002] Bioelectrical impedance analysis (BIA) is widely used to determine body composition parameters such as body fat, muscle, and water content. This analysis is performed using a device that measures the body's electrical impedance, which contains multiple electrodes placed on the skin's surface.

[0003] There are single-frequency and multi-frequency bioelectrical impedance analysis methods. Single-frequency bioelectrical impedance analysis is performed using only one signal with a single frequency to measure impedance, while multi-frequency bioelectrical impedance analysis is performed using multiple signals with different frequencies to measure impedance. Single-frequency bioelectrical impedance analysis can determine body fat content, muscle tissue content, and total body water (TBW) in the body. Multi-frequency bioelectrical impedance analysis can determine a wider range of parameters. In addition to body fat content, muscle tissue content, and TBW, multi-frequency bioelectrical impedance analysis can determine intracellular water (ICW) and extracellular water (ECW) in the body.

[0004] The ECW / TBW ratio is an important indicator of the body's water balance. Monitoring the ECW / TBW ratio is critical for certain conditions associated with excess fluid accumulation in the body, such as heart failure or kidney failure.

[0005] The prior art includes technical solutions for determining body composition using bioelectrical impedance analysis.

[0006] In the US patent application US2011301489 A1, published on 08.12.2011 and entitled "FLUID INDICATOR", a device is proposed for use in performing impedance measurements in a subject, wherein the device includes a processing system for determining, at each of three frequencies, first and second parameter values ​​for first and second impedance parameters related to the impedance of at least one segment of the subject's body, solving a system of equations representing a circle defined with respect to the first and second impedance parameters to thereby determine the parameter values ​​of the circle, wherein the equations are solved using the first and second parameter values ​​at each of the three frequencies, using the parameter values ​​of the circle to determine a third impedance parameter value at the corresponding frequency, and using the third impedance parameter value to determine an indicator indicating relative fluid levels in the segment of the subject's body.The proposed technical solution requires measuring two signal parameters: amplitude and phase. The proposed device does not have a small form factor suitable for use in wearable and / or portable devices such as smartwatches, smartphones, fitness watches, etc.

[0007] Korean Patent Application KR20250010622 A, published on January 21, 2025, and entitled "MEDICAL DEVICE FOR MONITORING BIOLOGICAL PARAMETERS," proposes a medical device for monitoring a user's biological parameters, comprising: a plate for measuring the user's weight, at least first and second electrodes for measuring impedance, and a device for measuring blood pressure and / or heart rate. The proposed technical solution requires measuring two signal parameters at each frequency, namely, amplitude and phase. The proposed device cannot accurately determine the ECW / TBW ratio based on the impedance of the upper body. The proposed device does not have a small form factor that allows it to be used in wearable and / or portable devices such as smart watches, smartphones, fitness watches, etc.

[0008] Korean Patent KR102397753 B1, issued on May 16, 2022, and entitled "AN ELECTRONIC DEVICE," proposes an electronic device for biometric user authentication based on bioelectrical impedance analysis (BIA) measurements. The basic idea is to use the relationships between BIA values ​​measured under various conditions to generate user-specific biometric characteristics. The proposed technical solution requires measuring two signal parameters: amplitude and phase. Although the proposed device performs BIA measurements, it cannot determine the ECW / TBW ratio and is intended for biometric user authentication, not for determining body composition parameters such as body fat, muscle mass, and water content.

[0009] The paper “Santiago F. Scagliusi, Luis Giménez-Miranda, Pablo Pérez-García, Daniel Martín Fernández, Francisco J. Medrano, Gloria Huertas, Alberto Yúfera, “Bioimpedance Spectroscopy-Based Edema Supervision Wearable System for Noninvasive Monitoring of Heart Failure”, Browse Journals & Magazines, IEEE Transactions on Instrumentation and Measurement, vol. 72, published May 17, 2023, https: / / ieeexplore.ieee.org / document / 10128797” proposes a miniature, portable, and wireless device that provides accurate and reliable measurement of the bioelectrical impedance analysis frequency spectrum in healthy individuals and patients with acute heart failure. The proposed technical solution requires measuring two signal parameters at each frequency: amplitude and phase. Multi-frequency measurement during bioelectrical impedance analysis is performed only on a small area of ​​one body part. The electrodes have an area of ​​129 mm. 2, which sets a form factor that does not allow the use of this device in wearable and / or portable devices, such as smart watches, smartphones, fitness watches, etc. ESSENCE OF THE INVENTION

[0010] To address the above-mentioned drawbacks, a method and device for multi-frequency bioelectrical impedance analysis are proposed. The device for multi-frequency bioelectrical impedance analysis is compact, has small-area electrodes, and can be integrated into wearable and / or portable devices, such as smart watches, smartphones, fitness watches, etc. However, devices in which the multi-frequency bioelectrical impedance analysis device can be integrated are not limited to the aforementioned wearable and / or portable devices. The method and device for multi-frequency bioelectrical impedance analysis are intended for measuring the impedance of a section of one body part or half of the body, depending on the location of the electrodes on the wearable and / or portable device, as well as the entire body.A method and device for multi-frequency bioelectrical impedance analysis can determine body fat content, body muscle content, TBW, ICW, and ECW, as well as the ECW / TBW ratio. The method and device for multi-frequency bioelectrical impedance analysis can determine a user's biological parameters based solely on measuring the amplitude of the signal passing through the user's body.

[0011] One aspect of the present invention provides a method for multi-frequency bioelectrical impedance analysis, wherein said method comprises the steps of: conducting (S101) alternating electric current signals with a given current amplitude and a current phase through a human body, wherein each signal has one of at least three frequencies different from each other; at each of said frequencies, measuring (S103) the voltage amplitude of said alternating electric current signals passed through the human body, and calculating the voltage drop value of said alternating electric current signals passed through the human body; at each of said frequencies, determining (S105) the modulus of the impedance of the human body using the calculated voltage drop value of said alternating electric current signals passed through the human body and the given current amplitude of said alternating electric current signals entering the human body;selecting (S107) an initial phase angle for each determined impedance module, wherein the phase angle characterizes the phase shift of the current of said alternating electric current signals entering the human body relative to the phase of the voltage of said alternating electric current signals that have passed through the human body; at each of said frequencies, estimating (S109) the phase angle using an equivalent electrical model of the human body, the initial phase angle, and the corresponding determined impedance module; generating (S111) an equivalent electrical model of the human body for a specific person based on the determined impedance modules and the estimated phase angles; extracting (S113) at least two parameters of the equivalent electrical model of the human body from the generated equivalent electrical model of the human body;and determining (S115) the ratio of extracellular water (ECW) to total body water (TBW) in a human body by inputting at least two extracted parameters of the equivalent electrical model of the human body into a trained machine learning model, wherein the machine learning model is trained on at least predetermined at least two parameters of the equivalent electrical models of the human body and the corresponding ECW / TBW ratios of a plurality of people.

[0012] In a further aspect, said alternating electric current signals are conducted between two different parts of a human body.

[0013] In another further aspect, said alternating electric current signals are conducted over a section of one part of the human body.

[0014] In yet another additional aspect, an initial phase angle is selected for each determined impedance module from a database containing pre-measured impedance modules and phase angles at each frequency.

[0015] In another additional aspect, at each of the mentioned frequencies, the phase angle is estimated by an error minimization method using an equivalent electrical model of the human body, an initial phase angle, and a corresponding determined impedance modulus.

[0016] In yet another additional aspect, the method further comprises the steps of: measuring the phase of the voltage of said alternating electric current signals passed through the human body at each of said frequencies; determining a phase angle at each of said frequencies that characterizes the phase shift of the current of said alternating electric current signals entering the human body relative to the phase of the voltage of said alternating electric current signals passed through the human body, wherein for each specific impedance modulus, a corresponding specific phase angle is selected as the initial phase angle.

[0017] In yet another additional aspect, alternating electric current signals with a given current amplitude and current phase are conducted through a human body, wherein each signal has one of at least four frequencies that are different from each other; wherein said method, after step (S111), further comprises the steps of: obtaining impedances from the determined impedance moduli and the corresponding estimated phase angles; calculating an impedance at each of said frequencies using the generated equivalent electrical model of the human body; determining errors in the obtained impedances, which are the differences between the obtained impedances and the corresponding calculated impedances; and excluding at least one determined impedance moduli for the impedance obtained with the greatest error from the set of determined impedance moduli, wherein at least three determined impedance moduli are left in the set of determined impedance moduli.

[0018] In yet another additional aspect, after step (S111), the method further comprises the steps of: obtaining impedances from the determined impedance modules and the corresponding estimated phase angles; calculating an impedance at each of said frequencies using the generated equivalent electrical model of the human body; determining errors in the obtained impedances, which are differences between the obtained impedances and the corresponding calculated impedances; if the determined errors in the obtained impedances are less than a predetermined threshold value at three or more of said frequencies, said method proceeds to step (S113); and if the determined errors in the obtained impedances are less than a predetermined threshold value at less than three of said frequencies, said method proceeds to step (S101).

[0019] In yet another additional aspect, after step (S113), the method further comprises the steps of: checking whether each of the extracted parameters of the equivalent electrical model of the human body is within a predetermined range; if all of the extracted parameters of the equivalent electrical model of the human body are within the predetermined range, said method proceeds to step (S115); and if at least one extracted parameter of the equivalent electrical model of the human body is outside the predetermined range, said method proceeds to step (S101).

[0020] In yet another additional aspect, said alternating electric current signals are conducted between two different parts of a human body, wherein determining the ECW / TBW ratio in the human body by inputting at least two extracted parameters of the equivalent electrical model of the human body into a trained machine learning model is performed for the entire human body.

[0021] In yet another additional aspect, said alternating electric current signals are conducted between two different parts of a human body, wherein determining the ECW / TBW ratio in the human body by inputting at least two extracted parameters of the equivalent electrical model of the human body into a trained machine learning model is performed for a section of the body between the other two different parts of the human body.

[0022] In yet another additional aspect, when determining the ECW / TBW ratio in a human body by inputting at least two extracted parameters of the equivalent electrical model of the human body into a trained machine learning model, anthropometric indicators, age and gender of the user are further input into the trained machine learning model, wherein the machine learning model is trained on at least predetermined at least two parameters of the equivalent electrical models of the human body, anthropometric indicators, age, gender and corresponding ECW / TBW ratios of a plurality of people.

[0023] In yet another additional aspect, when determining the ECW / TBW ratio in a human body by inputting at least two extracted parameters of the equivalent electrical model of the human body, anthropometric indicators, age and gender of the user into a trained machine learning model, the body composition indicators of the user are further input into the trained machine learning model, wherein the machine learning model is trained on predetermined at least two parameters of the equivalent electrical models of the human body, anthropometric indicators, age, gender, body composition indicators and the corresponding ECW / TBW ratios of a plurality of people.

[0024] Another aspect of the present invention provides a device for multi-frequency bioelectrical impedance analysis, wherein said device comprises: a signal source configured to generate alternating electric current signals with a given current amplitude and current phase, wherein each signal has one of at least three frequencies different from each other; two first electrodes configured to conduct said alternating electric current signals through the human body; two second electrodes configured to read said alternating electric current signals passed through the human body; a measuring device for measuring parameters of said alternating electric current signals passed through the human body; at least one processor;and a memory that stores a trained machine learning model and instructions that, when executed by at least one processor, cause the at least one processor to perform the method of any of the embodiments, wherein the machine learning model is trained on at least predetermined at least two parameters of the equivalent electrical models of the human body and the corresponding ECW / TBW ratios of a plurality of people. BRIEF DESCRIPTION OF THE DRAWINGS;

[0025] Fig. 1 shows the Nyquist diagram.

[0026] Fig. 2 shows the difference between the impedance values ​​calculated from the equivalent electrical model of the human body and the impedances obtained from the measurement.

[0027] Fig. 3 is a flow chart of a method 100 for multi-frequency bioelectrical impedance analysis.

[0028] Fig. 4 is a block diagram of a device 200 for multi-frequency bioelectrical impedance analysis.

[0029] Fig. 5 shows the arrangement of electrodes 202 and 203 on the wearable device.

[0030] Fig. 6 shows the arrangement of electrodes 202 and 203 on the wearable device. DETAILED DESCRIPTION OF EMBODIMENTS OF THE PRESENT INVENTION

[0031] The following description with reference to the accompanying drawings is provided to facilitate a thorough understanding of the various embodiments of the present invention defined by the claims and its equivalents. The description includes various specific details to facilitate such understanding, but these details are to be considered only exemplary. Accordingly, those skilled in the art will find that various changes and modifications can be made to the various embodiments described in this application without departing from the scope of the present invention. Furthermore, descriptions of well-known functions and constructions may be omitted for clarity and conciseness.

[0032] The terms and phrases used in the following description and claims are not limited by bibliographic meanings, but are simply used by the inventor to ensure a clear and consistent understanding of the present invention. Accordingly, those skilled in the art will understand that the following description of various embodiments of the present invention is provided for illustrative purposes only.

[0033] It should be understood that the designation of elements in the singular includes a plurality of elements unless the context clearly indicates otherwise.

[0034] It should be understood that while the terms "first," "second," etc. may be used herein in relation to elements of the present disclosure, such elements should not be construed as limited by these terms. The terms are used only to distinguish one element from other elements.

[0035] It should be further understood that the terms “comprises,” “comprising,” “includes,” and / or “including,” when used in this application, mean the presence of the stated features, values, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, values, operations, elements, components, and / or groups thereof.

[0036] In various embodiments of the present disclosure, a "module" or "block" may perform at least one function or operation and may be implemented using hardware, software, or a combination of both. A "multiple modules" or "multiple blocks" may be implemented with at least one processor by integrating it with at least one module other than the "module" or "block" that must be implemented using dedicated hardware.

[0037] At least one of the plurality of modules may be implemented by an artificial intelligence (AI) model (machine learning model). The AI-related function may be performed by non-volatile memory, volatile memory, and a processor.

[0038] The processor may include one or more processors. The one or more processors may be a general-purpose processor such as a central processing unit (CPU), an application processor (AP), or the like, a graphics processor such as a graphics processing unit (GPU), a visual processing unit (VPU), and / or a specialized artificial intelligence (AI) processor such as a neural processing unit (NPU). The present invention is not limited to the above processors, and any processor capable of performing the required task may be used.

[0039] One or more processors control the processing of input data according to a predetermined operating rule or AI model stored in non-volatile memory or volatile memory. The predetermined operating rule or AI model is achieved through training.

[0040] In this case, provisioning through training means that by applying a learning algorithm to a set of training data, a predetermined operating rule or AI model with the desired characteristic is created. Training may be performed within the device itself that contains the AI ​​according to the embodiment and / or may be implemented through a separate server / system.

[0041] An AI model can consist of multiple neural network layers. Each layer has multiple weight values ​​and performs the layer operation using the calculations of the previous layer and the multiple weights. Examples of neural networks include, but are not limited to, convolutional neural networks (CNNs), deep neural networks (DNNs), recurrent neural networks (RNNs), restricted Boltzmann machines (RBMs), deep belief networks (DBNs), bidirectional recurrent deep neural networks (BRDNNs), generative adversarial networks (GANs), and deep Q-networks.

[0042] A learning algorithm is a method for training a predetermined target device (e.g., an automated device) using a set of training data to force, enable, or control the target device to perform a determination or prediction. Examples of learning algorithms include, but are not limited to, supervised learning, unsupervised learning, semi-supervised learning, and reinforcement learning.

[0043] Hereinafter, various embodiments of the present invention are described in more detail with reference to the accompanying drawings.

[0044] A method 100 and a device 200 for multi-frequency bioelectrical impedance analysis are proposed. The present invention provides a device for multi-frequency bioelectrical impedance analysis that is compact, has electrodes 202 and 203 with a small area, and can be built into wearable and / or portable devices, such as a smartwatch, smartphone, fitness watch, etc.

[0045] The multi-frequency bioelectrical impedance analysis device 200 comprises electrodes 202 for supplying alternating electric current signals to the human body and electrodes 203 for measuring the alternating electric current signals passed through the human body. Since the present multi-frequency bioelectrical impedance analysis device 200 is designed to be embedded in wearable and / or portable devices, the area of ​​each of the electrodes 202 and 203 is less than 30 mm 2Electrodes with such a small area provide significant impedance measurement error. This error is caused by the difficulty of accurately measuring the phase of AC signals passing through the human body. Any movement or changes in pressure of fingers or body parts on the electrodes can cause distortions in measurements at one or more frequencies. Furthermore, the lower the frequency of the AC signals, the higher the contact resistance between the electrodes and the human body, which also reduces the accuracy of AC signal phase measurements at low frequencies.

[0046] However, it is possible to accurately measure the voltage amplitude of alternating electric current signals passed through the human body using electrodes 202 and 203. The present invention is based on measuring the voltage amplitudes of alternating electric current signals of different frequencies passed through the human body, based on which the impedance modules for each frequency are calculated and a phase angle is selected for each impedance module in order to calculate impedances at different frequencies.

[0047] Fig. 3 shows a flow chart of the method 100 for multi-frequency bioelectrical impedance analysis.

[0048] At step S101, alternating current signals with specified current amplitude and current phase are passed through the human body. Each signal has one of at least three frequencies that differ from each other. The frequencies may range from 1 kHz to 1000 kHz. The use of at least three frequencies is due to the fact that the phase angles are selected using a Nyquist diagram, which is a graph of the dependence of reactance on active resistance at different frequencies in the form of a circle. The Nyquist diagram is shown in Fig. 1. To plot the circle, at least three points are required. The specified current amplitudes at different signal frequencies may differ from each other or may be the same. The specified current phases at different signal frequencies may also differ from each other or may be the same. The number of signals with different frequencies may be arbitrary and is not limited to three signals with different frequencies.

[0049] In step S103, the voltage amplitude of the AC signals passing through the human body is measured at each of the mentioned frequencies, and the voltage drop value of the AC signals passing through the human body is calculated. The voltage drop value of the AC signals passing through the human body is the difference between the voltage amplitude of the AC signals passing through the human body and the voltage amplitude of the AC signals entering the human body. The voltage amplitudes of the AC signals entering the human body are known from the signal generation conditions of signal source 201.

[0050] In step S105, at each of the mentioned frequencies, the absolute value of the human body impedance is determined using the calculated voltage drop value of the mentioned AC electric current signals passed through the human body and the specified current amplitude of the mentioned AC electric current signals input to the human body. The voltage drop value divided by the specified current amplitude represents the absolute value of the impedance.

[0051] In step S107, an initial phase angle is selected for each specific impedance modulus. The phase angle characterizes the phase shift of the current of the said alternating current signals entering the human body relative to the phase of the voltage of the said alternating current signals passing through the human body. At each frequency, the initial phase angle can be selected from a database containing impedance modulus values ​​and their corresponding phase angles, which are pre-obtained at different frequencies for a plurality of people. The initial phase angle can also be selected at each frequency from a set of phase angle constants, i.e., for each frequency, the initial phase angle is specified as a constant. The initial phase angle can also be selected at each frequency from arbitrary phase angle values.

[0052] In step S109, at each of the mentioned frequencies, the phase angle is estimated using the equivalent electrical model of the human body, the initial phase angle, and the corresponding determined impedance modulus.

[0053] The equivalent electrical model of the human body is characterized by a Nyquist diagram, which is a graph of the reactance versus active resistance at different frequencies in the form of a circle, and an equivalent electrical circuit of the human body. The equivalent electrical circuit of the human body can be any equivalent electrical circuit of two resistors and one capacitor, the Cole-Cole model, and any circuits consisting of five electrical elements. For these equivalent electrical circuits of the human body, the Nyquist diagram is a graph in the form of a circle. The equivalent electrical model of the human body contains the following parameters: the nominal values ​​of the electrical elements of the equivalent electrical circuit of the human body and the parameters of the Nyquist diagram. The parameters of the Nyquist diagram are shown in Fig.1 and contain the values ​​of the impedance spectrum, R0, R∝, r, D, β, where the values ​​of the impedance spectrum are the values ​​of the impedance Z. n кГц at the frequencies at which the measurements were made, n is the frequency value, R0 is the resistance at point A on the Nyquist diagram, point A is the point where the circle intersects the R-axis at the lowest frequency, R ∞ is the resistance at point B on the Nyquist diagram, point B is the point where the circle intersects the R-axis at the highest frequency, r is the radius of the Nyquist diagram circle, D is the center of the Nyquist diagram circle, β is the value of the angle between the radius to point A and the radius to point B, divided by π.

[0054] Estimation of phase angles ϕ at the frequencies at which measurements were made using an equivalent electrical model of the human body, initial phase angles ϕ and the corresponding determined impedance moduli |Z n кГц| is accomplished by completing steps 1-6 below.

[0055] Step 1: Calculating ImpedancesZ n кГц at the frequencies at which the measurements were carried out, using the initial phase angles ϕ and the corresponding specific impedance moduli |Z n кГц | according to the formulas:R n кГц = |Z n кГц | ×cos(ϕ)(1),X n кГц = |Z n кГц | ×sin(ϕ)(2),Z n кГц =R n кГц ×iX n кГц (3).

[0056] Step 2: Plotting the Calculated ImpedancesZ n кГц on the graph of the dependence of reactive resistance on active resistance and performing an approximation based on the calculated impedancesZ n кГц to construct a Nyquist diagram and select an equivalent electrical circuit of the human body based on the constructed Nyquist diagram.

[0057] Step 3: Calculating Impedances at the frequencies at which the measurements were taken, based on the nominal values ​​of the electrical elements of the selected equivalent electrical circuit of the human body. For example, the following formulas are used for the Cole-Cole model: , (4) , (5)ω = 2 ×π× f, (6)where f are the frequencies at which the measurements were taken, ω are the angular frequencies corresponding to f, is the capacitance of the CPE element β constant phase, and is a time constant that determines the characteristic frequency of the model. The formulas are given for the Cole-Cole model, but other formulas can be used for other equivalent electrical models of the human body. Such formulas for calculating impedance based on the nominal values ​​of electrical elements of other equivalent electrical circuits of the human body are known in the art. For example, for equivalent electrical circuits of the human body containing a capacitor, formulas (4), (5), (6) can be applied and in formula (5) replace on the capacitance of the capacitor.

[0058] Step 4: Calculate the differences between the calculated impedancesZ n кГц and the corresponding calculated impedances at all frequencies at which measurements were taken.

[0059] Step 5: Change the phase angles ϕ taking into account the calculated differences between the calculated impedances Z n кГц and the corresponding calculated impedances using a selection method or an error minimization method, such as a gradient method, a gradient-free method, or a global optimization method. Gradient methods may include gradient descent, stochastic gradient descent (SGD), Momentum, Nesterov accelerated gradient (NAG), AdaGrad, RMSprop, Adam, Newton, Broyden, Fletcher, Goldfarb, Shanno quasi-Newton (BFGS / L-BFGS), and conjugate gradient methods. Gradient-free methods may include evolutionary algorithms, direct search methods, and particle swarm methods. Global optimization methods may include simulated annealing, differential evolution, and Bayesian optimization. The error minimization method is not limited to the listed methods. The error minimization method may be any known error minimization method.

[0060] Step 6: Repeat steps 1-5 until the difference between the calculated impedancesZ n кГц and the corresponding calculated impedances do not reach a threshold value or cease decreasing over a specified number of iterations. The threshold value can be set to a small value, such as 0.0001, and the number of iterations can range from units to hundreds. In this case, phase angles ϕ, modified in step 5, are used in step 1.

[0061] In step S111, an equivalent electrical model of the human body is generated for a specific person based on the determined impedance moduli and estimated phase angles. The generated equivalent electrical model of the human body for a specific person is the Nyquist diagram and the equivalent electrical circuit of the human body obtained in step 2 in an iteration, in which, in step 6, the differences between the calculated impedances Z n кГцand the corresponding calculated impedances reached a predetermined threshold value.

[0062] In step S113, at least two parameters of the equivalent electrical model of the human body are extracted from the generated equivalent electrical model of the human body. Different parameters of the equivalent electrical model of the human body and different numbers of parameters of the equivalent electrical model of the human body are used for different areas and parts of the human body to obtain the best result for determining the ECW / TBW ratio. For example, using only R0 and R ∞ For determining the ECW / TBW ratio for hands, this method yields the best results compared to using other parameters of the equivalent electrical model of the human body. Parameters of the equivalent electrical model of the human body for obtaining the best results in determining the ECW / TBW ratio for different areas and parts of the human body are known in the art.

[0063] In step S115, the ECW / TBW ratio in the human body is determined by inputting at least two extracted parameters of the equivalent electrical model of the human body into a trained machine learning model. The machine learning model may be any model from a linear regression model with feature interaction, linear models, tree models, ensemble methods, neural networks, including fully connected, convolutional, recurrent, and transformer architectures. The machine learning model is not limited to the listed models. The machine learning model may be any known machine learning model. The choice of a specific machine learning model depends on the characteristics of the dataset, the required prediction accuracy, computing resources, and training / execution time.The machine learning model is trained on at least two pre-defined parameters of the equivalent electrical models of the human body and the corresponding ECW / TBW ratios of a set of people.

[0064] Alternating electric current signals can be conducted between two different parts of the human body when one first electrode 202 for conducting alternating electric current signals through the human body and one second electrode 203 for measuring alternating electric current signals passed through the human body are located on one surface of a wearable and / or portable device, and one other first electrode 202 and one other second electrode 203 are located on one other surface of a wearable and / or portable device.For example, when one first electrode 202 and one second electrode 203 are located on the back surface of a wearable and / or portable device, and one other first electrode 202 and one other second electrode 203 are located on the side surface of a wearable and / or portable device, the palm of a person in the case of a smartphone or the wrist of a person in the case of a smart watch and fitness watch touches one first electrode 202 and one second electrode 203, and the fingers touch one other first electrode 202 and one other second electrode 203. In this case, alternating electric current signals pass between the two hands of a person.

[0065] Alternating current signals can be transmitted across a single part of the human body when all electrodes 202 and 203 are located on the same surface of the wearable and / or portable device. For example, if electrodes 202 and 203 are located on the back surface of the wearable and / or portable device, the person's palm touches all electrodes 202 and 203. In this case, the alternating current signals are transmitted across the person's palm.

[0066] The initial phase angle may not be selected from a database. In this case, method 100 may further comprise the steps of: measuring the voltage phase of said alternating current signals passed through the human body at each of said frequencies; determining the phase angle at each of said frequencies based on the measured voltage phase and the current phase specified when generating the alternating current signals. For each determined impedance modulus, selecting the corresponding determined phase angle as the initial phase angle.

[0067] To form a more accurate equivalent electrical model of the human body, measurements can be taken at more than three frequencies and three impedances obtained with the lowest error can be determined. Their impedance moduli can then be used to form an equivalent electrical model of the human body for a specific individual. In this case, alternating current signals with a given current amplitude and phase are passed through the human body, with each signal having one of at least four distinct frequencies.The method 100, after step (S111), further comprises the steps of: obtaining impedances from the determined impedance modules and the corresponding estimated phase angles; calculating the impedance at each of the said frequencies using the formed equivalent electrical model of the human body; determining the errors of the obtained impedances, which are the differences between the obtained impedances and the corresponding calculated impedances; and excluding at least one determined impedance module for the impedance obtained with the greatest error from the set of determined impedance modules, while leaving at least three determined impedance modules in the set of determined impedance modules. The obtained impedances will have different errors. Those impedance modules are left that relate to the obtained impedances, the errors of which are less than those of the excluded impedances. Preferably, the error is no more than 6 ohms.If the error is more than 6 Ohm, the parameters of the equivalent electrical model of the human body may have significant errors, which will worsen the result of determining the ECW / TBW ratio.

[0068] Determination of impedance errorsZ n кГц , obtained from the determined impedance moduli and the corresponding estimated phase angles, are determined as follows. The impedances Z are obtained n кГц from the determined impedance moduli and the corresponding estimated phase angles according to formulas (1), (2) and (3). The impedances are then calculated at the frequencies at which the measurements were taken, based on the nominal values ​​of the electrical elements of the generated equivalent electrical circuit of the human body. Since the Nyquist diagram is constructed based on impedances, n кГц , obtained from certain impedance moduli and the corresponding estimated phase angles, by means of approximation, then not all the resulting impedancesZ n кГцmay coincide with the calculated impedances . Therefore, the obtained impedances Z n кГц , which do not match the calculated impedances , have an error. The difference between the calculated impedance and the resulting impedance Z n кГц is an error. The error in the obtained impedances n кГц shown in Fig. 2, where the white circles indicate the calculated impedances , and the resulting impedances are indicated by crossesZ n кГц . And indicate the error in active resistance and reactive resistance, respectively.

[0069] To obtain a fully formed equivalent electrical model of the human body, at least three correctly determined impedance moduli at different frequencies are required. Otherwise, the user will have to repeat the measurement. To avoid significantly inaccurate ECW / TBW ratio determination, the difference between the obtained impedance and the calculated impedance can be estimated and method 100 can be continued. If the difference is acceptable for at least three obtained impedances, method 100 should be restarted.In this case, after step (S111), the method 100 may further comprise the steps of: obtaining impedances from the determined impedance modules and the corresponding estimated phase angles; calculating the impedance at each of the mentioned frequencies using the formed equivalent electrical model of the human body; determining the errors in the obtained impedances, which are the differences between the obtained impedances and the corresponding calculated impedances; if the determined errors in the obtained impedances are less than a given threshold value at three or more of the mentioned frequencies, the said method 100 proceeds to step (S113); and if the determined errors in the obtained impedances are less than a given threshold value at less than three of the mentioned frequencies, the said method 100 proceeds to step (S101). The threshold value may be, for example, 6 ohms or less.If the difference between the obtained impedance and the calculated impedance is more than 6 Ohms, the parameters of the equivalent electrical model of the human body may have significant errors, which will worsen the result of determining the ECW / TBW ratio.

[0070] The parameters of the equivalent electrical model of the human body must be within a predetermined range to determine the ECW / TBW ratio with the smallest error. In this case, the method 100, after step (S113), may further comprise the steps of: checking whether each of the extracted parameters of the equivalent electrical model of the human body is within the predetermined range; if all the extracted parameters of the equivalent electrical model of the human body are within the predetermined range, said method 100 proceeds to step (S115); and if at least one extracted parameter of the equivalent electrical model of the human body is outside the predetermined range, said method 100 proceeds to step (S101). A separate range is set for each parameter. The ranges can be any within the following limits. For example, 100 ohms to 1500 ohms is the limit for R0, 50 ohms to 1200 ohms is the limit for R ∞ , 0F – 1E-4F is the limit for CPE β, 50 Ohm – 500 Ohm is the limit for r, 0.3 – 0.9 is the limit for β, 0.94 – 2.83 rad is the limit for D.

[0071] A machine learning model can be trained to determine the ECW / TBW ratio for the entire human body. In this case, the aforementioned alternating electric current signals are conducted between two different parts of the human body, and the ECW / TBW ratio in the human body is determined for the entire human body by inputting at least two extracted parameters of the equivalent electrical model of the human body into the trained machine learning model. The machine learning model is trained on at least two predetermined parameters of the equivalent electrical models of the human body and the corresponding ECW / TBW ratios for the entire human body for a plurality of people.

[0072] The trained machine learning model can be trained to determine the ECW / TBW ratio for a body region between two different parts of the human body, different from the body region between two different parts of the human body through which alternating electric current signals are passed. In this case, said alternating electric current signals are passed between two different parts of the human body, and the ECW / TBW ratio in the human body is determined for the body region between the other two different parts of the human body by inputting at least two extracted parameters of the equivalent electrical model of the human body into the trained machine learning model. The machine learning model is trained on at least predetermined at least two parameters of the equivalent electrical models of the human body and the corresponding ECW / TBW ratios between the required two different parts of the human body for a plurality of people.

[0073] To improve the accuracy of ECW / TBW ratio determination, in addition to at least two extracted parameters of the equivalent electrical model of the human body, the user's anthropometric indicators, age, and gender can be input into the trained machine learning model. In this case, when training the machine learning model, in addition to the predetermined at least two parameters of the equivalent electrical model of the human body and the corresponding ECW / TBW ratios of a plurality of people, the predetermined anthropometric indicators, age, and gender of a plurality of people are used as training data. The anthropometric indicators include at least the weight and height of a person.

[0074] To further improve the accuracy of ECW / TBW ratio determination, in addition to at least two extracted parameters of the human equivalent electrical body model, anthropometric indicators, age, and gender of the user, the user's body composition indicators can be additionally input into the trained machine learning model. In this case, when training the machine learning model, in addition to the predetermined at least two parameters of the human equivalent electrical body model, anthropometric indicators, age, gender, and the corresponding ECW / TBW ratios of multiple individuals, the predetermined body composition indicators of multiple individuals are used as training data. The body composition indicators include body fat content, body muscle content, and total body water content.

[0075] Fig. 4 shows a block diagram of a device 200 for multi-frequency bioelectrical impedance analysis. The device 200 comprises a signal source 201, two first electrodes 202, two second electrodes 203, a measuring device 204, at least one processor 205 and memory 206.

[0076] Signal source 201 is designed to generate alternating current signals with a specified current amplitude and phase. Each generated signal has one of at least three distinct frequencies.

[0077] The first two electrodes 202 are designed to conduct the said alternating electric current signals through the human body. The second two electrodes 203 are designed to read the said alternating electric current signals passed through the human body. The area of ​​each of the first two electrodes 202 and the second two electrodes 203 is 30 mm 2 or less.

[0078] Fig. 5 shows a possible arrangement of electrodes 202 and 203 on a smart watch. A smart watch is chosen only as an example of a wearable and / or portable device in which device 200 is built. The wearable and / or portable device is not limited to a smart watch and can be any wearable and / or portable device, such as a smart watch, a fitness watch, a smartphone, etc. One first electrode 202 and one second electrode 203 can be located on one surface of the wearable and / or portable device, and one other first electrode 202 and one other second electrode 203 can be located on one other surface of the wearable and / or portable device.One first electrode 202 and one second electrode 203 may be located on the back surface of the wearable and / or portable device, and one other first electrode 202 and one other second electrode 203 may be located on the side surface of the wearable and / or portable device. As shown in Fig. 5, one first electrode 202 and one second electrode 203 may be located on the back surface of the smart watch, and one other first electrode 202 and one other second electrode 203 may be located on the buttons of the smart watch.

[0079] Fig. 6 shows another possible arrangement of electrodes 202 and 203 on a smartwatch. Electrodes 202 and 203 can be located on the back surface of a wearable and / or portable device. As shown in Fig. 6, electrodes 202 and 203 can be located on the back surface of a smartwatch.

[0080] Measuring device 204 is designed to measure the parameters of the aforementioned alternating electric current signals passed through the human body. Measuring device 204 comprises at least a voltmeter for measuring the amplitude of the voltage of the alternating electric current signals passed through the human body. Measuring device 204 may further comprise a phase detector for measuring the phase of the voltage of the alternating electric current signals passed through the human body. Measuring device 204 should have dimensions that allow for embedding device 200 in a wearable and / or portable device.

[0081] At least one processor 205 may be any processor that ensures the execution of the operations of the method 100. The processor may be a processor of a wearable and / or portable device.

[0082] Memory 206 may be any memory capable of storing a trained machine learning model and instructions that, when executed by at least one processor, cause at least one processor to perform method 100 according to any of the embodiments. The memory may be the memory of a wearable and / or portable device.

[0083] The above descriptions of embodiments of the invention are illustrative, and modifications of configuration and implementation do not depart from the scope of the present description. For example, although embodiments of the invention are described generally in connection with Figures 1-6, the provided descriptions are exemplary. Although the subject matter of the invention is described in language specific to structural features or methodological operations, it is understood that the subject matter of the invention is not necessarily limited to the specific features or operations described above. Moreover, the specific features and operations described above are disclosed as exemplary forms of implementing the claims.

[0084] Accordingly, it is intended that the scope of the embodiments of the invention be limited only by the following claims.

Claims

1. A method for multi-frequency bioelectrical impedance analysis, wherein said method comprises the steps of: conduct (S101) alternating electric current signals with a given current amplitude and current phase through the human body, wherein each signal has one of at least three frequencies that differ from each other; at each of the mentioned frequencies, the amplitude of the voltage of the mentioned signals of alternating electric current that have passed through the human body is measured (S103), and the value of the voltage drop of the mentioned signals of alternating electric current that have passed through the human body is calculated; at each of the mentioned frequencies, the modulus of the impedance of the human body is determined (S105) using the calculated value of the voltage drop of the mentioned alternating electric current signals passed through the human body and the given current amplitude of the mentioned alternating electric current signals entering the human body; select (S107) an initial phase angle for each specific impedance module, wherein the phase angle characterizes the phase shift of the current of the said alternating electric current signals entering the human body relative to the phase of the voltage of the said alternating electric current signals passing through the human body; at each of the mentioned frequencies, the phase angle is estimated (S109) using an equivalent electrical model of the human body, the initial phase angle and the corresponding determined impedance module; form (S111) an equivalent electrical model of the human body for a specific person based on the determined impedance moduli and estimated phase angles; extracting (S113) at least two parameters of the equivalent electrical model of the human body from the generated equivalent electrical model of the human body; and determine (S115) the ratio of extracellular water (ECW) to total body water (TBW) in a human body by inputting at least two extracted parameters of the equivalent electrical model of the human body into a trained machine learning model, wherein the machine learning model is trained on at least predetermined at least two parameters of equivalent electrical models of the human body and the corresponding ECW / TBW ratios of a plurality of people.

2. The method according to paragraph 1, in which the said alternating electric current signals are conducted between two different parts of the human body.

3. The method according to paragraph 1, in which the said alternating electric current signals are conducted along a section of one part of the human body.

4. The method according to any one of paragraphs 1-3, in which the initial phase angle is selected for each determined impedance module from a database containing previously measured impedance modules and phase angles at each frequency.

5. The method according to any one of paragraphs 1-4, in which at each of the mentioned frequencies the phase angle is estimated by the error minimization method using an equivalent electrical model of the human body, an initial phase angle and a corresponding determined impedance modulus.

6. The method according to any one of paragraphs 1-3, further comprising the steps of: at each of the mentioned frequencies, the phase of the voltage of the mentioned alternating electric current signals passing through the human body is measured; at each of the mentioned frequencies, a phase angle is determined that characterizes the phase shift of the current of the mentioned alternating electric current signals entering the human body relative to the phase of the voltage of the mentioned alternating electric current signals passing through the human body, in this case, for each specific impedance module, the corresponding specific phase angle is selected as the initial phase angle.

7. The method according to any one of paragraphs 1-6, in which conduct alternating electric current signals with a given current amplitude and current phase through the human body, wherein each signal has one of at least four frequencies that differ from each other; wherein the said method after step (S111) further comprises the steps of: impedances are obtained from the determined impedance moduli and the corresponding estimated phase angles; at each of the mentioned frequencies, the impedance is calculated using the formed equivalent electrical model of the human body; determine the errors in the obtained impedances, which are the differences between the obtained impedances and the corresponding calculated impedances; and excluding at least one specific impedance module for the impedance obtained with the greatest error from the set of specific impedance modules, while leaving at least three specific impedance modules in the set of specific impedance modules.

8. The method according to any one of paragraphs 1-7, after step (S111) further comprising the steps of: impedances are obtained from the determined impedance moduli and the corresponding estimated phase angles; at each of the mentioned frequencies, the impedance is calculated using the formed equivalent electrical model of the human body; determine the errors in the obtained impedances, which are the differences between the obtained impedances and the corresponding calculated impedances; if the determined errors of the obtained impedances are less than a given threshold value at three or more of the said frequencies, the said method proceeds to step (S113); and if the determined errors of the obtained impedances are less than a given threshold value at less than three of the mentioned frequencies, the said method goes to step (S101).

9. The method according to any one of paragraphs 1-8, after step (S113) further comprising the steps of: check whether each of the extracted parameters of the equivalent electrical model of the human body is within a given range; if all the extracted parameters of the equivalent electrical model of the human body are within a given range, the said method proceeds to step (S115); and if at least one extracted parameter of the equivalent electrical model of the human body is outside a predetermined range, the said method proceeds to step (S101).

10. The method according to any one of paragraphs 2, 4-9, in which the said alternating electric current signals are conducted between two different parts of the human body, wherein the determination of the ECW / TBW ratio in the human body by inputting at least two extracted parameters of the equivalent electrical model of the human body into the trained machine learning model is performed for the entire human body.

11. The method according to any one of paragraphs 2, 4-9, in which the said alternating electric current signals are conducted between two different parts of the human body, wherein the determination of the ECW / TBW ratio in the human body by inputting at least two extracted parameters of the equivalent electrical model of the human body into the trained machine learning model is performed for a section of the body between two other different parts of the human body.

12. The method according to any one of paragraphs 1-11, in which, when determining the ECW / TBW ratio in the human body by inputting at least two extracted parameters of the equivalent electrical model of the human body into the trained machine learning model, anthropometric indicators, age and gender of the user are additionally input into the trained machine learning model, wherein the machine learning model is trained on at least predetermined at least two parameters of equivalent electrical models of the human body, anthropometric indicators, age, gender and corresponding ECW / TBW ratios of a plurality of people.

13. The method according to claim 12, wherein when determining the ECW / TBW ratio in the human body by inputting at least two extracted parameters of the equivalent electrical model of the human body, anthropometric indicators, age and gender of the user into the trained machine learning model, the body composition indicators of the user are additionally input into the trained machine learning model, wherein the machine learning model is trained on predetermined at least two parameters of equivalent electrical models of the human body, anthropometric indicators, age, gender, body composition indicators and corresponding ECW / TBW ratios of a plurality of people.

14. A device for multi-frequency bioelectrical impedance analysis, wherein said device comprises: a signal source configured to generate alternating electric current signals with a given current amplitude and current phase, wherein each signal has one of at least three frequencies that differ from each other; two first electrodes configured to conduct said alternating electric current signals through the human body; two second electrodes configured to read the said alternating electric current signals passed through the human body; a measuring device for measuring the parameters of the said alternating electric current signals passed through the human body; at least one processor; and a memory that stores a trained machine learning model and instructions that, when executed by at least one processor, cause at least one processor to perform the method according to any one of paragraphs 1-13, wherein the machine learning model is trained on at least predetermined at least two parameters of equivalent electrical models of the human body and the corresponding ECW / TBW ratios of a plurality of people.

Images