A real-time detection method for multiple body fluids based on bioelectrical impedance

Abstract

This invention discloses a real-time detection method for multiple body fluids in the human body based on bioelectrical impedance analysis. The method employs a lightweight detection system composed of a four-electrode array, an AC excitation module, a signal acquisition module, a power supply module, a main control module, and a communication module. It achieves impedance signal acquisition from multiple sites through flexible conductive patch electrodes. The system adjusts the electrode placement and excitation signal parameters according to different detection scenarios, extracting impedance amplitude, phase, rate of change, and frequency domain features, and constructing a scenario-based dataset by combining this with the individual information of the subject. For scenarios with abnormal body fluids such as pulmonary edema and cerebral edema, a differentiated prediction model is established to achieve real-time assessment of the corresponding body fluid status. Simultaneously, the method establishes a long-term detection risk warning model based on individual baseline impedance, continuous impedance change trends, weight percentage changes, and individual characteristics, generating individualized body fluid load warning thresholds, and outputting graded risk alerts when impedance changes reach the thresholds.

Description

Technical Field

[0001] This invention relates to the field of biomedicine, specifically to a method for real-time detection of multiple bodily fluids in the human body based on bioelectrical impedance. Background Technology

[0002] Fluid balance is crucial for maintaining normal physiological functions. In hemodialysis patients, conditions such as volume overload, pulmonary edema, intestinal edema, and cerebral edema are all closely related to fluid metabolism disorders. Currently, commonly used clinical methods for fluid analysis include ultrasound examination, blood biochemistry analysis, chest X-ray / CT / MRI, and invasive intracranial pressure monitoring.

[0003] Bioelectrical impedance analysis (BIA) technology uses microampere-level weak alternating current for detection, featuring non-invasiveness, no ionizing radiation, and high safety. It is applicable to a wide range of people and supports repeated measurements. This method is fast, easy to operate, and uses a variety of portable devices that are easy to deploy. It requires no special computer room environment, has a low learning curve for operators, and has low purchase and maintenance costs with no disposable consumables, making it convenient for grassroots institutions, health check centers, and homes to adopt. The detection process can quickly output multiple indicators such as body fat percentage, skeletal muscle mass, body fluid distribution, and phase angle. Multi-frequency BIA can also distinguish between intracellular and extracellular water, used to assess nutritional status, edema, and fluid imbalance. Under standardized measurement conditions, it exhibits good repeatability, making it ideal for long-term dynamic trend monitoring. It can be applied to many scenarios such as weight management, sarcopenia screening, fluid management for dialysis patients, and lymphedema assessment.

[0004] However, the measures currently being taken have the following problems: 1) Commonly used clinical methods for detecting body fluids have many limitations: ultrasound, chest X-ray, CT, and MRI rely on large-scale professional equipment and certified operators, making it impossible to achieve real-time continuous bedside monitoring. Furthermore, CT and MRI involve radiation or have contraindications, making them unsuitable for dynamic monitoring of critically ill patients. Blood biochemistry analysis and invasive intracranial pressure monitoring are invasive tests that can cause patient discomfort and pose a risk of infection. The test results are also delayed, making it difficult to meet the needs for early warning and dynamic monitoring of body fluid imbalances. 2) Bioelectrical impedance analysis (BIA) devices for detecting body fluids still have shortcomings: First, the biological tissue equivalent model is simplified and does not fully consider the differences in electrical properties between intracellular and extracellular fluids and cell membranes, making it unable to adapt to the electrical properties of different tissues, resulting in low accuracy of impedance data analysis; Second, the detection function is limited, mostly targeting a single site or a single body fluid indicator, failing to cover the multi-scenario detection needs of body fluid imbalance in multiple sites throughout the body, especially lacking specific body fluid monitoring solutions for abdominal intestines and intracranial brain tissue; Third, the electrode layout and signal processing technology are not designed differently for different detection sites, making them susceptible to interference from physiological movement and tissue stratification, resulting in poor monitoring stability and specificity. Summary of the Invention

[0005] To address the aforementioned problems, the objective of this invention is to propose a real-time detection method for multiple human body fluids based on bioelectrical impedance analysis. This method constructs a mathematical model describing the relationship between the electrical impedance of biological tissues and body fluids based on the principle of bioelectrical impedance analysis. Using this established mathematical model, a method for assessing changes in body fluids by detecting transdermal tissue electrical impedance is derived. A lightweight impedance detection system is designed using a microcontroller and automatic control principles, adjusting its detection position according to the specific detection target to achieve multifunctional detection of changes in human body fluids.

[0006] This was achieved through the following technical solutions: A method for real-time detection of multiple bodily fluids in the human body based on bioelectrical impedance, comprising the following steps: S1. Test preparation and parameter configuration: Complete electrode placement and personalized system parameter settings; S2, Signal Excitation, Acquisition and Impedance Calculation: Based on the main control and communication modules, AC excitation, ADC sampling and DFT demodulation are completed, and the impedance amplitude, real part and imaginary part of impedance are calculated. S3. Data preprocessing and feature construction: The host computer system uses a moving average of M sampling points to calculate the continuous impedance sequence of each data channel of the impedance real part, imaginary part, amplitude and phase output in step S2, and performs low-pass filtering. S4. Contextualized Extremely Random Tree Prediction: Integrates impedance features, weight, and age information, and calls the extreme randomly tree model for prediction; S5. Results Output and Long-Term Risk Warning: Output body fluid status and graded prompts to construct individualized impedance thresholds.

[0007] Optionally, in step S1, the parameter configuration includes: basic information of the test subject, excitation parameters, and warning parameters; the basic information of the test subject includes age, gender, height, weight, BMI, and clinical diagnostic information; the excitation signal parameters include: excitation frequency, excitation current amplitude, sampling parameters, and DFT points; the warning parameters include: individual baseline setting rules, primary fluid load warning threshold, secondary fluid load warning threshold, weight change percentage reference threshold, continuous abnormality judgment duration, and baseline update cycle.

[0008] Optionally, in step S2, the main control and communication module writes the excitation frequency, excitation current amplitude, ADC sampling and DFT parameters to the register through the serial peripheral interface; the internal waveform generator generates a sinusoidal excitation signal and applies it to the two excitation electrodes; the two measurement electrodes acquire the differential response voltage.

[0009] Optionally, in step S2, the voltage and current response signals are respectively input to a buffer, a programmable gain amplifier, and an anti-aliasing filter for conditioning; the analog signal is then converted into a discrete digital sequence by a 16-bit ADC, and the real and imaginary parts of the excitation frequency components corresponding to the voltage and current channels are extracted by the DFT module.

[0010] The complex impedance is calculated by the main control and communication modules based on the real and imaginary parts of the voltage and current output from the voltage and current channels, using the following formula: Z = V / I = (V Re +jV Im ) / (I Re +jI Im ) = R + jX, Among them, the real and imaginary parts of the voltage and current outputs of the voltage channel and current channel are V, respectively. Re V Im I Re and I Im R is the real part of the impedance related to tissue conductivity loss, X is the imaginary part of the impedance related to the energy storage characteristics corresponding to cell membrane capacitance, j is the imaginary unit, j=√-1, and the impedance amplitude is |Z|=√(R) 2 +X 2 The phase is θ = arctan(X,R).

[0011] Optionally, in step S3, the host computer system uses a moving average of M sampling points for the continuous impedance sequence of each channel and performs low-pass filtering; within an analysis window containing N sampling points, it extracts the mean, variance, maximum value, minimum value, peak-to-peak value and rate of change of the real part, imaginary part and amplitude of the impedance, and simultaneously extracts the mean phase and rate of change of the phase.

[0012] Optionally, in step S4, extreme random tree models are established for the four detection scenarios. The four detection scenarios include: hemodialysis volume load monitoring, pulmonary fluid monitoring, abdominal fluid monitoring, and intracranial fluid monitoring. The construction process of the extreme random tree model is as follows: collect the scenario-based feature vector x and clinical reference label y at the same time point; remove or impute missing values, and divide the training set and validation set according to a preset ratio; randomly select a subset of input features at each node of each tree and randomly generate candidate split thresholds, and then complete the node split based on classification purity or regression error; construct multiple independent decision trees, with majority voting for classification tasks and the average value of the outputs of each tree for regression tasks.

[0013] Optionally, in step S5, the host computer system displays R(t), X(t), |Z(t)|, θ(t) and the relative change curve of impedance, and displays the fluid change trend and risk level output by the scenario-based extreme random tree model.

[0014] The present invention also provides a real-time detection system for multiple human body fluids based on bioelectrical impedance, comprising: a four-electrode array, an AC excitation and signal acquisition module, a power supply module, a main control and communication module, and a host computer system; A four-electrode array is used to achieve high-precision human tissue impedance detection, including two current-excited electrodes and two voltage-measuring electrodes; The AC excitation and signal acquisition module is used for sinusoidal excitation signal generation, signal pickup, and digital demodulation. The AC excitation and signal acquisition module uses registers to configure the operating parameters of each functional unit. Each functional unit includes: digital waveform generator, high-speed digital-to-analog converter, constant current drive path, input buffer, programmable gain amplifier, anti-aliasing filter, 16-bit analog-to-digital converter, and DFT module. The power supply module is used to supply power to the system; The main control and communication modules are used for system measurement and control, data acquisition, and data interaction; The host computer system is used to receive the timestamps, detection scenarios, real part R, imaginary part X, amplitude |Z|, and phase θ from the main control and communication modules, and to complete data storage, 50-point moving average, low-pass filtering, feature extraction, feature vector construction, extreme random tree model calling, threshold comparison, and result display.

[0015] Optionally, in the main control and communication modules, the main control uses the serial peripheral interface SPI to configure registers and measurement sequences, control measurement startup, channel switching, data reading, abnormal state detection and data packaging, and transmit the obtained complex impedance data to the host computer system.

[0016] The beneficial effects of this invention compared to the prior art are: 1) This invention constructs an AC equivalent circuit model to describe the electrical impedance of biological tissues, and realizes the assessment and analysis of changes in human body fluids by detecting the impedance changes of biological tissues; it also builds a dynamic and lightweight bioelectrical impedance acquisition system, which adopts an integrated design of flexible electrode patches and microcontrollers, and can realize continuous and real-time detection of changes in human body fluids.

[0017] 2) This invention achieves multifunctional body fluid detection by adjusting the system configuration. Based on the detection target, the corresponding electrode patch placement position and AC excitation signal are configured to complete the detection of various body fluid indicators such as changes in body fluid in dialysis patients and pleural fluid load in patients with pulmonary edema. For four major detection scenarios, namely hemodialysis volume load, pulmonary fluid in pulmonary edema, abdominal fluid in intestinal edema, and intracranial fluid in cerebral edema, a customized electrode layout, excitation signal configuration, and model input dataset are matched to construct a differentiated detection solution for the entire process. This ensures the detection accuracy and stability under different sites and different diseases, and adapts to the physiological characteristics and clinical monitoring needs of different populations. Attached Figure Description

[0018] Figure 1 This is a flowchart of the real-time detection method for body fluids in multiple scenarios based on bioelectrical impedance according to the present invention; Figure 2 This is a system block diagram of the bioimpedance detection device of the present invention; Figure 3 This is a graph showing the relative change of patient impedance over time during dialysis in an embodiment of the present invention. Detailed Implementation

[0019] The technical solutions in the embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0020] A method for multifunctional detection of changes in body fluids throughout or in specific areas of the human body using bioelectrical impedance analysis, such as... Figure 1 The diagram shown is a flowchart of the real-time detection method for body fluids in multiple scenarios based on bioelectrical impedance of the present invention, which includes the following steps: S1. Test Preparation and Parameter Configuration: Complete electrode placement and personalized system parameter settings. Select the hemodialysis volume load, pulmonary fluid, abdominal fluid, or intracranial fluid testing scenario based on the clinical testing objectives, complete electrode placement, and configure parameters in the host computer system. Parameter configuration includes: basic information of the test subject, excitation parameters, and warning parameters.

[0021] Basic information of the test subject includes age, gender, height, weight, BMI, and clinical diagnosis. For hemodialysis volume load monitoring, all four electrodes are arranged longitudinally along the upper arm, with two current-excitation electrodes on the outer side and two voltage-measuring electrodes on the inner side. For pulmonary fluid monitoring, the excitation electrodes are placed in the mid-axillary region on both sides of the chest, and the measuring electrodes are placed symmetrically on both sides of the sternum. For abdominal fluid monitoring, the excitation electrodes are placed in the mid-axillary region on both sides of the abdomen at the level of the umbilicus, and the measuring electrodes are placed in the midline region above and below the umbilicus, avoiding surgical scars and areas of significant gas. For intracranial fluid monitoring, the excitation electrodes are placed in the bilateral temporal regions, and the measuring electrodes are placed in the bilateral frontal regions.

[0022] The excitation signal parameters include excitation frequency, excitation current amplitude, sampling parameters, and DFT points. The preferred frequencies for hemodialysis volume load monitoring are 30–60 kHz and currents are 50–200 μA; the preferred frequencies for pulmonary fluid monitoring are 10–20 kHz and currents are 50–200 μA; the preferred frequencies for abdominal fluid monitoring are 20–30 kHz and currents are 50–200 μA; and the preferred frequencies for intracranial fluid monitoring are 20–40 kHz and currents are 20–100 μA. In specific embodiments, hemodialysis volume load monitoring uses a frequency of 50 kHz and a current of 100 μA; pulmonary fluid monitoring uses a frequency of 16 kHz and a current of 100 μA; abdominal fluid monitoring uses a frequency of 25 kHz and a current of 100 μA; and intracranial fluid monitoring uses a frequency of 30 kHz and a current of 50 μA.

[0023] The early warning parameters include: individual baseline setting rules, primary fluid load early warning threshold, secondary fluid load early warning threshold, weight change percentage reference threshold, continuous abnormality judgment duration, and baseline update cycle.

[0024] All parameters for abdominal fluid monitoring are set according to the clinical standards for abdominal edema: the device collects the average stable abdominal impedance of the subject for 24 consecutive hours as the exclusive fluid baseline. A 7% deviation of abdominal impedance from the baseline triggers a mild Level 1 warning, and a 14% deviation triggers a severe Level 2 warning. A ≥2.3% weight gain in 48 hours is used as an auxiliary judgment standard. The warning is only activated after the indicator exceeds the limit for 12 minutes to avoid movement interference. At the same time, the abdominal fluid baseline is automatically updated by collecting data every 7 days.

[0025] In the hemodialysis scenario, in accordance with the standard operating procedure for blood purification (2021 version), the equipment is set to trigger a level 1 (mild) volume overload warning if the abdominal impedance deviates from the baseline by 7%, and trigger a level 2 (severe) warning if it deviates by 14%. At the same time, a weight gain of ≥2.3% in 48 hours is introduced as an auxiliary criterion, and an alarm is only triggered after 12 minutes of continuous over-limit to avoid interference from changes in body position. The abdominal impedance baseline is automatically updated every 7 days in the pre-dialysis resting state.

[0026] Lung fluid monitoring is based on the emergency expert consensus on the application of non-invasive hemodynamic monitoring using bioelectrical impedance analysis in acute and critical care. A 7% deviation in device settings indicates subclinical pulmonary congestion (Level 1 warning), and a 14% deviation indicates the risk of overt pulmonary edema (Level 2 warning). A 48-hour weight gain of ≥2.3% is used as evidence of volume overload. Respiratory motion artifacts are filtered out by continuous overload for 12 minutes, and baseline data is resampled at the same time point every week.

[0027] Intracranial fluid monitoring follows the Chinese expert consensus on monitoring and treatment of increased intracranial pressure. The device only uses abdominal impedance monitoring as an auxiliary reference for whole-body volume management. Setting an impedance deviation of 7% or 14% only indicates that whole-body volume fluctuations may indirectly affect intracranial volume. The 48-hour weight gain ≥2.3% and 12-minute delay alarm logic are intended to help identify the risk of systemic fluid overdose. The baseline is updated every 7 days.

[0028] S2, Signal Excitation, Acquisition and Impedance Calculation: Based on the main control and communication modules, AC excitation, ADC sampling and DFT demodulation are completed, and the impedance amplitude, real part and imaginary part of the impedance are calculated.

[0029] The main control and communication modules write the excitation frequency, excitation current amplitude, ADC sampling, and DFT parameters to the register via the Serial Peripheral Interface (SPI). An internal waveform generator generates a sinusoidal excitation signal, which is applied to the two excitation electrodes via a high-bandwidth loop and a constant current drive unit; the two measurement electrodes acquire the differential response voltage across the tissue.

[0030] The voltage and current response signals are input to a buffer, a programmable gain amplifier, and an anti-aliasing filter for signal processing to keep the signals within the allowable range of the ADC. The analog signals are then converted into discrete digital sequences by a 16-bit ADC, and the real and imaginary parts of the voltage and current channels are extracted separately by a DFT module. The complex impedance is calculated by the main control and communication modules based on the real and imaginary parts of the voltage and current outputs from the voltage and current channels, using the following formula: Z = V / I = (V Re +jV Im ) / (I Re +jI Im ) = R + jX, Among them, the real and imaginary parts of the voltage and current outputs of the voltage channel and current channel are V, respectively. Re V Im I Re and I Im R is the real part of the impedance related to tissue conductivity loss, X is the imaginary part of the impedance related to the energy storage characteristics corresponding to cell membrane capacitance, and j is the imaginary unit, j=√-1.

[0031] The formula for calculating impedance amplitude is: |Z| = √(R) 2 +X 2 The formula for calculating the phase is: θ = arctan(X,R).

[0032] Before measurement, a calibration resistor with a known resistance value, Rcal, can be connected to obtain the measured value Zcal of the calibration resistor. Preferably, the resistance value of the calibration resistor Rcal can be selected in the range of 100Ω to 1kΩ, and the complex calibration coefficient K=Rcal / Zcal is calculated. The human body impedance correction value is Zc=K·Zm, where Zm is the original measured impedance. After completing the impedance calculation, the main control and communication modules package R, X, |Z|, θ, timestamp, and scene identifier, and send them to the host computer system via serial-to-USB conversion.

[0033] S3. Data Preprocessing and Feature Construction: The host computer system applies a moving average of M (M=50) sampling points to the continuous impedance sequences of the real, imaginary, amplitude, and phase data channels calculated in step S2. This means taking the current point and the previous 49 points at each step, and then performing a low-pass filter to reduce random noise and short-term fluctuations. The cutoff frequency of the low-pass filter is configured based on the sampling frequency and the rate of change of the target body fluid, allowing slowly changing components of the body fluid to pass through while suppressing high-frequency interference; this is common knowledge and will not be elaborated further.

[0034] Within an analysis window containing N (N≤M) sampling points, the mean, variance, maximum, minimum, peak-to-peak value, and rate of change of the impedance's real and imaginary parts and amplitude are extracted. Simultaneously, the mean and rate of change of the phase are extracted. The impedance amplitude rate of change is expressed as: k Z =(|Z N |-|Z1|) / [(N-1)Δt], where Z N To analyze the impedance amplitude at the Nth sampling point within the analysis window, it is necessary to ensure that N ≤ M / 2 and that there are at least two sliding operations to achieve statistical significance. Z1 represents the impedance amplitude at the first sampling point within the analysis window, and Δt is the sampling time interval between two adjacent sampling points. A Fast Fourier Transform is performed on the mean-reduced impedance sequence to obtain the dominant frequency f. dom Spectral peak P max and the proportion of low-frequency energy E L =Σf∈F L |F(f)| 2 / Σf|F(f)| 2 E L This represents the ratio of low-frequency signal energy to the total energy of the entire spectrum, reflecting the proportion of slow physiological components; F L For the set of low-frequency frequencies, F L ={1kHz, 5kHz}. The impedance characteristics mentioned above are aligned with age, height, weight, BMI, electrode placement, excitation frequency, and excitation current according to sample number and timestamp, and then concatenated to construct a scene-based feature vector x.

[0035] S4. Contextualized Extremely Random Tree Prediction: Contextualized Extremely Random Tree Prediction: Integrates impedance features, weight, and age information, and calls the extreme random tree model for prediction.

[0036] Extremely randomized tree models were established for four detection scenarios: hemodialysis volume load monitoring, pulmonary fluid monitoring, and intracranial fluid monitoring. The extremely randomized tree model, as an optional implementation, was constructed as follows: Scenario-specific feature vectors x and clinical reference labels y were collected at the same time point; missing values ​​were removed or imputed, and the training and validation sets were divided according to a preset ratio, preferably 7:3; a subset of input features was randomly selected at each node of each tree, and candidate splitting thresholds were randomly generated; then, node splitting was completed based on classification purity or regression error; multiple independent decision trees were constructed, with majority voting used for classification tasks and the average output of each tree used for regression tasks. The input feature subset included three types of parameters: impedance time-domain features, impedance frequency-domain features, and individual physiological features. Specifically, these included the mean, variance, maximum, minimum, peak-to-peak value, and rate of change of the real and imaginary parts of impedance, amplitude, and phase; as well as the dominant frequency, peak frequency, low-frequency energy proportion, age, height, weight, and BMI. The criteria for determining the candidate splitting threshold are as follows: for classification tasks, classification purity (Gini coefficient or information gain) is used as the criterion; for regression tasks, regression error (mean squared error) is used as the criterion. The threshold that maximizes the performance gain after node splitting is selected to complete the node split. Multiple independent decision trees are constructed. For classification tasks, majority voting is used, and for regression tasks, the average output of each tree is used. Furthermore, the number of trees, maximum depth, and minimum number of leaf node samples are determined through cross-validation. A 5-fold cross-validation combined with grid search is used, with the optimal combination selected based on the validation set classification accuracy or regression mean squared error as the optimization objective. The typical value ranges for these hyperparameters are: 50–200 trees, 8–20 maximum tree depths, and 5–20 minimum number of leaf node samples. The trained model file is stored and run on the host computer system.

[0037] For each of the four detection scenarios, a corresponding scenario-based dataset is constructed. The scenario-based dataset includes a structured sample set containing original impedance parameters, time-domain and frequency-domain features, basic information of the test subject, and clinical reference labels. This dataset supports the training and validation of extreme random tree models for each scenario. Clinical reference labels include, for example, normal abdominal fluid status, mild abdominal fluid retention (Level 1 warning), and moderate to severe ascites accumulation (Level 2 warning), which can be specifically categorized according to the clinical standards of the corresponding detection scenario. Each row of the scenario-based dataset represents a test sample and includes at least the sample number, test subject number, detection time, detection scenario, excitation frequency, excitation current, real part R of impedance, imaginary part X, amplitude |Z|, phase θ, and the impedance mean, variance, peak-to-peak value, rate of change, dominant frequency, peak spectral value, and low-frequency energy proportion obtained through feature extraction. It also includes the test subject's age, height, weight, BMI, basic information, and corresponding clinical reference labels. During the model training phase, the extreme random tree model is constructed and validated based on the scenario-specific dataset. During the actual detection phase, the input feature vector is generated according to the field order and feature dimensions consistent with the dataset, and the trained model for the corresponding scenario is called to predict the body fluid state.

[0038] Reference labels for hemodialysis volume overload models can be obtained from changes in body weight before and after dialysis, actual ultrafiltration volume, or clinical volume overload grading; reference labels for pulmonary fluid models can be obtained from lung ultrasound B-line, chest X-ray, or CT grading; reference labels for abdominal fluid models can be obtained from abdominal ultrasound or CT results; and reference labels for intracranial fluid models can be obtained from head CT, MRI, or intracranial pressure-related examination results. Model inputs include complex impedance characteristics and anthropometric information such as age, height, weight, and BMI. Model outputs show the trend of fluid status changes at the corresponding site, the probability of fluid abnormalities, or the risk level. Remaining dialysis time, edema type, or specific complication risk not supported by training labels are not considered essential outputs.

[0039] S5. Results Output and Long-Term Risk Warning: Outputs body fluid status and grading prompts, and constructs individualized impedance thresholds. The host computer system displays R(t), X(t), |Z(t)|, θ(t), and relative impedance change curves, and displays the body fluid change trend and risk level output by the scenario-based extreme random tree model.

[0040] The classifications for each scenario are as follows: Hemodialysis volume overload monitoring scenario: Normal state corresponds to the patient's volume overload being within the target dry weight range; Level 1 warning corresponds to a slight increase in volume overload with a low risk of fluid retention; Level 2 warning corresponds to a significant increase in volume overload with a clear risk of fluid overload. Pulmonary fluid monitoring scenario: Normal state corresponds to lung tissue water content within the physiological range; Level 1 warning corresponds to a slight increase in pulmonary fluid, requiring continuous observation; Level 2 warning corresponds to a significant increase in pulmonary fluid content with a high risk of pulmonary edema. Abdominal fluid monitoring scenario: Normal state corresponds to abdominal fluid volume within the physiological range; Level 1 warning corresponds to a slight accumulation of abdominal fluid; Level 2 warning corresponds to a significant increase in ascites, meeting the criteria for clinical intervention. Intracranial fluid monitoring scenario: Normal state corresponds to intracranial fluid in physiological equilibrium; Level 1 warning corresponds to a slight increase in intracranial fluid; Level 2 warning corresponds to a significant increase in intracranial fluid, with a high risk of cerebral edema.

[0041] Long-term continuous testing refers to continuous testing for no less than 12 hours, preferably covering a dialysis interval of 24 to 72 hours.

[0042] For interdialysis monitoring, the average impedance amplitude Z0 within a preset stable period of 15–30 minutes after dialysis is used as the individual baseline, and the corresponding weight W0 at the end of dialysis is used as the weight baseline. The current relative impedance change is ΔZ = (Zt - Z0) / Z0 × 100%, and the current relative weight change is ΔW = (Wt - W0) / W0 × 100%. Using paired data from at least three consecutive monitoring days, ΔZ = aΔW + b is fitted to convert the primary and secondary weight thresholds TW1 and TW2 into impedance thresholds TZ1 = aTW1 + b and TZ2 = aTW2 + b. Preferably, the TW1 and TW2 weight thresholds are set according to the standard operating procedures for blood purification, and can be set to 3% and 5% respectively, and can be adjusted by medical staff according to the target weight and individual circumstances. Since the increase in body fluid during interdialysis usually corresponds to a decrease in impedance, when a<0, if ΔZ>TZ1, it is a normal state; if TZ2<ΔZ≤TZ1, a first-level warning is output; if ΔZ≤TZ2, a second-level warning is output.

[0043] like Figure 3 As shown, taking hemodialysis volume load monitoring as an example, bioelectrical impedance signals from the patient's upper arm are continuously collected after dialysis begins. The impedance amplitude at the start of dialysis is used as the reference value for this monitoring session, and the relative change in impedance at different time points is calculated. As the dialysis process continues, the relative change in impedance generally shows a gradual increasing trend. This is because ultrafiltration continuously removes water from the body during hemodialysis, causing a gradual decrease in the content of extracellular fluid in the upper arm tissue, and a corresponding reduction in the tissue's conductive pathways, which manifests as an increase in the measured impedance amplitude. Figure 3The continuous trend of change in the data indicates that the detection method described in this invention can reflect the dynamic changes in the state of tissue fluid during dialysis, and is not limited to a single measurement before and after dialysis.

[0044] Furthermore, the relative change in impedance and its rate of change over time can serve as parameters characterizing changes in volume overload. Specifically, the relative change in impedance reflects the cumulative change in the current tissue fluid state relative to the initial state during dialysis, while the slope of the curve reflects the rate of change in fluid state over a given time period. These parameters, along with individual information such as patient age, height, and weight, can form a contextualized feature vector, which is then input into an extreme random tree prediction model for assessing volume overload. Simultaneously, the continuous impedance change trend can provide a data foundation for subsequent individualized threshold establishment and abnormal change identification.

[0045] Based on the above method, this invention also discloses a multi-scenario body fluid bioelectrical impedance detection system. For example... Figure 2 The diagram shown is a system block diagram of the bioimpedance detection device of the present invention, which includes: a four-electrode array, an AC excitation and signal acquisition module, a power supply module, a main control and communication module, and a host computer system.

[0046] A four-electrode array is used to achieve high-precision human tissue impedance detection, comprising two current-excitation electrodes and two voltage-measuring electrodes. A microampere-level AC excitation current is applied to the tissue under test through independent excitation electrodes, and the response voltage across the tissue is acquired using high-impedance measuring electrodes. This effectively isolates the interference from electrode contact resistance and skin impedance, eliminates contact errors, and improves the accuracy and stability of bioimpedance detection, providing high-precision raw signals for subsequent complex impedance calculation and body fluid characteristic analysis.

[0047] The electrical properties of biological tissues can be described by an equivalent AC circuit comprising intracellular fluid resistance Ri, cell membrane resistance RM, cell membrane capacitance CM, and extracellular fluid resistance RE. The cell membrane resistance RM and cell membrane capacitance CM are connected in parallel to form the membrane impedance Z. M The expression is: Z M =RM||(1) / (jω*CM)=(RM) / (1+jω*RM*CM).

[0048] In the circuit, the two membrane units are connected in series through the intracellular fluid resistance Ri; therefore, the cell impedance Z seRies The calculation formula for Z is: seRies =Z M +Ri+Z M =2Z M +Ri.

[0049] The extracellular fluid resistance RE is connected in parallel across the entire cell impedance. Therefore, the total impedance is the parallel combination of the cell impedance and the extracellular fluid resistance, expressed as: Z total=RE||Z seRies =(RE*Z seRies ) / |RE+Z seRies |, Find a common denominator for the fractions, multiply both the numerator and denominator by the common denominator term (1+jω*RM*CM) to eliminate the inner fractions, and simplify to obtain the equivalent total impedance of biological tissue (Z). total The final expression for ) is: Z total =RE[2RM+Ri(1+jω*RM*CM)] / [(RE+Ri)(1+jω*RM CM)+2RM].

[0050] Finally, the impedance measurement result can be written as: Z total =RE(Z total )+jIm(Z total ), where R=RE(Z total X = Im(Z) represents the real part of the impedance, reflecting the resistive component of the tissue's conductive pathway; total The imaginary part of the impedance is denoted as |Z(t)|, reflecting energy storage components such as cell membrane capacitance. The formulas for calculating its amplitude |Z(t)| and phase θ are: |Z(t)| = sqrt(RE[Z(t)]^2 + Im[Z(t)]^2), θ = arctg(Im(Z(t)) / (RE[Z(t)]^2)). total ) / RE(Z total In this context, j is the imaginary unit, and ω is the angular frequency of the test. Therefore, changes in the water content of biological tissues will alter Ri, RM, CM, and RE, and consequently, Z. total By measuring the total impedance Z total The amplitude and phase of impedance values ​​are analyzed to assess changes in the water content of biological tissues. Taking hemodialysis volume load monitoring as an example, during dialysis, water is removed, and the extracellular fluid in the upper arm tissue decreases, resulting in a gradual increase in the impedance amplitude relative to pre-dialysis levels. During the interdialysis period, as water re-accumulates, the impedance amplitude gradually decreases relative to the post-dialysis baseline. This change can be used as a feature of body fluid changes to input into a scenario-based prediction model and for individualized threshold comparisons.

[0051] The above equivalent circuit is used to explain the correspondence between complex impedance and changes in body fluid; the actual detection parameters are obtained by the main controller reading the voltage and current DFT calculation results stored in the internal register of the analog front-end chip, without the need to directly measure the parameters of each equivalent element Ri, RM, CM and RE.

[0052] The AC excitation and signal acquisition module is used for sinusoidal excitation signal generation, tissue signal pickup, and digital demodulation. Centered on an analog front-end chip, this module configures the operating parameters of each functional unit through internal registers. Its internal functional units include: a digital waveform generator, a high-speed digital-to-analog converter (DAC), a constant current drive path, an input buffer, a programmable gain amplifier (PGA), an anti-aliasing filter, a 16-bit DAC, and a DFT module. Specifically, the AC excitation and signal acquisition module applies a sinusoidal excitation to the tissue under test, controls the analog differential voltage obtained from the measurement electrodes within the allowable range of the ADC, and converts it into a digital sequence. The DFT module performs complex demodulation on the voltage and current channels, outputting the real and imaginary parts. The single-ended analog input voltage range of the analog front-end is 0.2–2.1V, and the effective differential range of the ADC changes according to the PGA gain parameters configured in the registers.

[0053] During data calibration, the system switches to a calibration resistor with a known resistance value, obtains the measured value, and calculates the complex calibration coefficient. During human body measurements, the calibration coefficient is multiplied by the original complex impedance to compensate for gain, phase, and measurement path errors. The host computer system further employs a 50-point moving average and low-pass filtering to suppress random noise and high-frequency interference.

[0054] The power supply module is used to supply power to the system. It obtains 5V power from the computer's USB interface, and supplies power to the main control and communication modules and peripheral digital and analog circuits through a 3.3V voltage regulator and filter circuit; this embodiment does not provide ±5V power.

[0055] The main control and communication modules are used for system measurement control, data acquisition, and data interaction, with the STM32F411RETL selected as the core main control chip. Specifically, the main control configures the register (AD5940) and measurement sequence through the serial peripheral interface SPI to control measurement initiation, channel switching, data reading, abnormal state detection, and data packaging. The acquired complex impedance data is sent to the serial-to-USB module via serial port and then transmitted to the host computer system. Furthermore, the main control writes a complete set of measurement action instructions into the internal SRAM of the AD5940 register via SPI; the chip hardware sequencer then automatically executes the entire measurement process sequentially without the main control unit needing to intervene at each step, achieving high-precision timing control, reducing the main control burden, and reducing power consumption.

[0056] The host computer system receives timestamps, detection scenarios, impedance real part R, imaginary part X, amplitude |Z|, and phase θ from the main control and communication modules. It then performs data storage, 50-point moving average, low-pass filtering, feature extraction, feature vector construction, extreme random tree model invocation, threshold comparison, and result display. Real-time curves include at least the curves of R versus time, X versus time, impedance amplitude versus time, phase versus time, and relative impedance change.

[0057] In summary, this invention provides a real-time detection method for multiple body fluids in the human body based on bioelectrical impedance. It constructs a mathematical model describing the relationship between the electrical impedance of biological tissues and body fluids based on the principle of bioelectrical impedance, thereby assessing changes in body fluids through transdermal tissue electrical impedance detection. Furthermore, it utilizes a microcontroller and automatic control principles to design a lightweight electrical impedance detection system, achieving multifunctional detection of changes in human body fluids, demonstrating significant advancements.

[0058] The above embodiments are merely illustrative of the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. Any modifications made to the technical solutions based on the technical concept proposed in this invention shall fall within the scope of protection of this invention.

Claims

1. A method for real-time detection of multiple human body fluids based on bioelectrical impedance, characterized in that, Includes the following steps: S1. Test preparation and parameter configuration: Complete electrode placement and personalized system parameter settings; S2, Signal Excitation, Acquisition and Impedance Calculation: Based on the main control and communication modules, AC excitation, ADC sampling and DFT demodulation are completed, and the impedance amplitude, real part and imaginary part of impedance are calculated. S3. Data preprocessing and feature construction: The host computer system uses a moving average of M sampling points to calculate the continuous impedance sequence of each data channel of the impedance real part, imaginary part, amplitude and phase output in step S2, and performs low-pass filtering. S4. Contextualized Extremely Random Tree Prediction: Integrates impedance features, weight, and age information, and calls the extreme randomly tree model for prediction; S5. Results Output and Long-Term Risk Warning: Output body fluid status and graded prompts to construct individualized impedance thresholds.

2. The method for real-time detection of multiple human body fluids based on bioelectrical impedance as described in claim 1, characterized in that, In step S1, the parameter configuration includes: basic information of the test subject, excitation parameters, and warning parameters; the basic information of the test subject includes age, gender, height, weight, BMI, and clinical diagnostic information; the excitation signal parameters include: excitation frequency, excitation current amplitude, sampling parameters, and DFT points; the warning parameters include: individual baseline setting rules, primary fluid load warning threshold, secondary fluid load warning threshold, weight change percentage reference threshold, continuous abnormality judgment duration, and baseline update cycle.

3. The method for real-time detection of multiple human body fluids based on bioelectrical impedance as described in claim 1, characterized in that, In step S2, the main control and communication module writes the excitation frequency, excitation current amplitude, ADC sampling and DFT parameters to the register through the serial peripheral interface; the internal waveform generator generates a sinusoidal excitation signal and applies it to the two excitation electrodes. Two measuring electrodes are used to acquire the differential response voltage.

4. The method for real-time detection of multiple human body fluids based on bioelectrical impedance as described in claim 3, characterized in that, In step S2, the voltage and current response signals are respectively input to a buffer, a programmable gain amplifier, and an anti-aliasing filter for conditioning; the analog signals are then converted into discrete digital sequences by a 16-bit ADC, and the real and imaginary parts of the voltage and current outputs of the voltage and current channels are extracted by the DFT module.

5. The method for real-time detection of multiple human body fluids based on bioelectrical impedance as described in claim 4, characterized in that, The complex impedance is calculated by the main control and communication modules based on the real and imaginary parts of the voltage and current output from the voltage and current channels, using the following formula: Z=V / I=(V Re +jV Im ) / (I Re +jI Im )=R+jX, Among them, the real and imaginary parts of the voltage and current outputs of the voltage channel and current channel are V, respectively. Re V Im I Re and I Im R is the real part of the impedance related to tissue conductivity loss, X is the imaginary part of the impedance related to the energy storage characteristics corresponding to cell membrane capacitance, j is the imaginary unit, j=√-1, and the impedance amplitude is |Z|=√(R) 2 +X 2 The phase is θ = arctan(X,R).

6. The method for real-time detection of multiple human body fluids based on bioelectrical impedance as described in claim 1, characterized in that, In step S3, the cutoff frequency of the low-pass filter is configured according to the sampling frequency and the rate of change of the target body fluid; within the analysis window containing N sampling points, the mean, variance, maximum value, minimum value, peak-to-peak value and rate of change of the impedance real part, imaginary part and amplitude are extracted, and the mean and rate of change of the phase are extracted at the same time.

7. The method for real-time detection of multiple human body fluids based on bioelectrical impedance as described in claim 1, characterized in that, In step S4, extreme random tree models are established for four detection scenarios. The four detection scenarios include: hemodialysis volume load monitoring, pulmonary fluid monitoring, abdominal fluid monitoring, and intracranial fluid monitoring. The construction process of the extreme random tree model is as follows: collect the scenario-based feature vector x and clinical reference label y at the same time point; remove or impute missing values, and divide the training set and validation set according to a preset ratio; randomly select a subset of input features at each node of each tree and randomly generate candidate split thresholds, and then complete the node split based on classification purity or regression error; construct multiple independent decision trees, with majority voting for classification tasks and the average value of the outputs of each tree for regression tasks.

8. The method for real-time detection of multiple human body fluids based on bioelectrical impedance as described in claim 1, characterized in that, In step S5, the host computer system displays R(t), X(t), |Z(t)|, θ(t) and the relative change curve of impedance, and displays the fluid change trend and risk level output by the scenario-based extreme random tree model.

9. A real-time detection system for multiple human body fluids based on bioelectrical impedance, characterized in that, include: The system comprises a four-electrode array, an AC excitation and signal acquisition module, a power supply module, a main control and communication module, and a host computer system. A four-electrode array is used to achieve high-precision human tissue impedance detection, including two current-excited electrodes and two voltage-measuring electrodes; The AC excitation and signal acquisition module is used for sinusoidal excitation signal generation, signal pickup, and digital demodulation. The AC excitation and signal acquisition module uses registers to configure the operating parameters of each functional unit. Each functional unit includes: digital waveform generator, high-speed digital-to-analog converter, constant current drive path, input buffer, programmable gain amplifier, anti-aliasing filter, 16-bit analog-to-digital converter, and DFT module. The power supply module is used to supply power to the system; The main control and communication modules are used for system measurement and control, data acquisition, and data interaction; The host computer system is used to receive the timestamps, detection scenarios, real part R, imaginary part X, amplitude |Z|, and phase θ from the main control and communication modules, and to complete data storage, 50-point moving average, low-pass filtering, feature extraction, feature vector construction, extreme random tree model calling, threshold comparison, and result display.

10. A real-time detection system for multiple human body fluids based on bioelectrical impedance as described in claim 9, characterized in that, In the main control and communication modules, the main control uses the serial peripheral interface SPI to configure registers and measurement sequences, control measurement startup, channel switching, data reading, abnormal status detection and data packaging, and transmit the obtained complex impedance data to the host computer system.

Images