A user body fluid balance monitoring method, system, device and medium

Abstract

This application relates to a method, system, device, and medium for monitoring user body fluid balance, belonging to the field of biomedical engineering technology. The monitoring method includes: acquiring user body fluid state data collected by sensors in real time and performing timestamp synchronization and drift correction to obtain a calibrated multimodal data stream; constructing a three-dimensional body fluid state tensor from the calibrated multimodal data stream; constructing a body fluid dynamic balance model based on the three-dimensional body fluid state tensor, and performing dynamic balance analysis and anomaly detection to obtain anomaly feature vectors; determining the intervention level based on fuzzy decision rules; generating a corresponding candidate intervention instruction set based on the anomaly feature vectors and the intervention level; inputting the candidate intervention instruction set into the body fluid dynamic balance model to predict changes in user body fluid state, and selecting the optimal intervention instruction based on the prediction results; and driving the execution device corresponding to the optimal intervention instruction to adjust body fluid parameters. This application can achieve comprehensive management and precise regulation of user body fluid state.

Description

Technical Field

[0001] This application relates to the field of biomedical engineering technology, and in particular to a method, system, device and medium for monitoring user body fluid balance. Background Technology

[0002] With the continuous advancement of technology, the demand for fluid balance monitoring in the medical and health field is increasing. Fluid balance is one of the basic conditions for the human body to maintain normal physiological functions. Fluid imbalance not only affects the metabolism of water and electrolytes, but also directly affects the function of multiple organs such as the cardiovascular system and kidneys. For example, abnormal fluctuations in the concentration of electrolytes such as sodium and potassium in the body may lead to arrhythmia, kidney failure, and even endanger life. In postoperative care, intensive care, and chronic disease management, fluid imbalance is one of the most common complications, often requiring accurate real-time monitoring and timely intervention to prevent the condition from worsening.

[0003] Currently, most common methods for monitoring body fluids rely on laboratory testing and manual analysis, which often suffers from problems such as data collection lag and poor accuracy. For example, traditional methods mainly monitor electrolyte concentrations in the blood through venipuncture, but these methods cannot provide continuous, real-time monitoring data, nor can they comprehensively reflect the dynamic changes in body fluid distribution. For monitoring aspects such as fluid distribution and tissue fluid changes in patients, current technologies still rely on equipment such as ultrasound, which cannot achieve long-term continuous monitoring and accurate detection. Especially for patients with chronic diseases or those in the postoperative recovery period, relying solely on static body fluid parameter monitoring cannot reflect the real-time trends in electrolyte concentration changes and their impact on body fluid distribution, potentially leading to missed opportunities for optimal intervention. Summary of the Invention

[0004] In order to achieve comprehensive management and precise regulation of users' body fluid status, this application provides a method, system, device and medium for monitoring users' body fluid balance.

[0005] Firstly, this application provides a method for monitoring a user's body fluid balance, employing the following technical solution: A method for monitoring a user's fluid balance, the monitoring method comprising: Real-time acquisition of user body fluid status data collected by sensors; the body fluid status data includes blood parameters, tissue fluid data, and excretion data; The user's bodily fluid state data is time-stamped and drift-corrected to obtain a calibrated multimodal data stream; The calibrated multimodal data stream is constructed as a three-dimensional fluid state tensor; A dynamic equilibrium model of body fluids is constructed based on the three-dimensional body fluid state tensor, and dynamic equilibrium analysis and anomaly detection are performed to obtain an anomaly feature vector; the anomaly feature vector includes the level of body fluid imbalance and key abnormal parameters. Based on the aforementioned abnormal feature vector, the intervention level is determined using fuzzy decision rules. Based on a preset strategy library, a corresponding set of candidate intervention instructions is generated according to the abnormal feature vector and the intervention level; The candidate intervention instruction set is input into the body fluid dynamic balance model to predict the changes in the user's body fluid state after execution, and the optimal intervention instruction is selected based on the prediction results. The device corresponding to the optimal intervention command is driven to adjust the body fluid parameters.

[0006] By adopting the above technical solutions, comprehensive real-time monitoring and analysis of users' fluid balance is achieved. By establishing a dynamic balance model and using fuzzy decision rules to determine the intervention level, the system can automatically generate precise intervention instructions based on the abnormal characteristics of different fluid states and drive the execution device to perform personalized fluid regulation. The closed-loop feedback mechanism enables the system to dynamically optimize the intervention plan, continuously improve the accuracy of monitoring and treatment, and ensure the restoration of fluid balance.

[0007] Secondly, this application provides a user fluid balance monitoring system, which adopts the following technical solution: A user fluid balance monitoring system, the monitoring system comprising: The data acquisition module is used to acquire user body fluid status data collected by sensors in real time; the body fluid status data includes blood parameters, tissue fluid data and excretion data; The data calibration module is used to perform timestamp synchronization and drift correction on the user's body fluid state data to obtain a calibrated multimodal data stream; The state tensor construction module is used to construct the calibrated multimodal data stream into a three-dimensional fluid state tensor; An abnormal feature vector generation module is used to construct a dynamic equilibrium model of body fluids based on the three-dimensional body fluid state tensor, and to perform dynamic equilibrium analysis and anomaly detection to obtain an abnormal feature vector; the abnormal feature vector includes the level of body fluid imbalance and key abnormal parameters. An intervention level determination module is used to determine the intervention level based on the abnormal feature vector and fuzzy decision rules. The candidate intervention instruction generation module is used to generate a corresponding set of candidate intervention instructions based on a preset strategy library, according to the abnormal feature vector and the intervention level. The optimal intervention instruction determination module is used to input the candidate intervention instruction set into the body fluid dynamic balance model, predict the changes in the user's body fluid state after execution, and select the optimal intervention instruction based on the prediction results. An intervention command driving module is used to drive the execution device corresponding to the optimal intervention command to adjust body fluid parameters.

[0008] Thirdly, this application provides a computer device, which adopts the following technical solution: A computer device includes a memory, a processor, and a computer program stored in the memory, the processor executing the computer program to perform the steps of the method as described in the first aspect.

[0009] Fourthly, this application provides a computer-readable storage medium, which adopts the following technical solution: A computer-readable storage medium storing a computer program that can be loaded by a processor and executed as in any of the methods in the first aspect.

[0010] In summary, this application includes at least one of the following beneficial technical effects: it not only possesses the ability for real-time monitoring and precise intervention, but also provides personalized treatment plans through predictive and simulation analysis, significantly improving clinical treatment efficiency. For scenarios such as postoperative recovery and chronic disease monitoring, the system can respond promptly when a user experiences fluid imbalance, reducing the risk of serious complications and improving patient recovery outcomes. Furthermore, the intelligence and automation of this technical solution allow medical personnel to focus more on the management of high-risk patients, while enabling precise intervention for patients with general fluid imbalances, optimizing the allocation and use of medical resources. Attached Figure Description

[0011] Figure 1 This is a schematic diagram of the first process of a user fluid balance monitoring method according to one embodiment of this application.

[0012] Figure 2 This is a schematic diagram of the second process of a user fluid balance monitoring method according to one embodiment of this application.

[0013] Figure 3 This is a schematic diagram of the third process of a user fluid balance monitoring method according to one embodiment of this application.

[0014] Figure 4 This is a schematic diagram of the fourth process of a user fluid balance monitoring method according to one embodiment of this application.

[0015] Figure 5 This is a schematic diagram of the fifth process of a user fluid balance monitoring method according to one embodiment of this application.

[0016] Figure 6 This is a schematic diagram of the sixth process of a user fluid balance monitoring method according to one embodiment of this application. Detailed Implementation

[0017] To make the purpose, technical solution, and advantages of this application clearer, the following description is provided in conjunction with the appendix. Figures 1-6 The present application will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the application.

[0018] This application discloses a method for monitoring a user's body fluid balance.

[0019] Reference Figure 1 A method for monitoring a user's fluid balance, the monitoring method comprising: Step S101: Acquire user body fluid status data collected by sensors in real time; Among them, body fluid status data includes blood parameters, tissue fluid data, and excretion data; through implantable sensors, wearable devices, and excretion monitoring units, the user's blood parameters, tissue fluid data, and excretion data are collected in real time. The acquisition of these data can directly reflect the user's body fluid status and help assess whether it is in a balanced or unbalanced state.

[0020] Specifically, implanted sensors monitor physiological indicators in real time, such as sodium ion concentration, potassium ion concentration, and hemoglobin content in the blood. Blood electrolyte levels are directly related to body fluid balance and reflect the user's water and salt metabolism status. Skin impedance sensors and terahertz imaging technology monitor the distribution of tissue fluid in the user. This data can provide information about the distribution of body fluid and tissue swelling, such as pulmonary edema caused by heart failure or peripheral edema. The excretion monitoring unit collects data such as urine flow rate and urine electrolyte concentration. This data can help identify whether there is too much or too little body fluid and the excretion of electrolytes.

[0021] Step S102: Timestamp synchronization and drift correction are performed on the user's body fluid status data to obtain a calibrated multimodal data stream; Data collected by different sensors may have time differences or data errors due to sensor drift. By using timestamp synchronization and drift correction, the accuracy and consistency of the data can be ensured.

[0022] Specifically, there may be time differences between sensors. Algorithms (such as dynamic time warping, interpolation algorithms, etc.) are used to synchronize the time of different data sources, so that various types of data have a unified time axis. For example, interpolation methods can be used to synchronize the time of blood sodium concentration and urine electrolyte concentration data at different times.

[0023] Furthermore, sensor drift can cause the collected data to gradually deviate from the actual values ​​over time. Drift correction can be performed on sensors using calibration algorithms (such as weighted averaging or sensor calibration) to ensure data accuracy. For example, applying a correction weight of 0.6 to blood parameter sensors and a correction weight of 0.1 to excretion parameter sensors ensures more accurate multimodal fusion of data sources.

[0024] Step S103: Construct the calibrated multimodal data stream into a three-dimensional fluid state tensor; Specifically, different body fluid data (blood parameters, tissue fluid data, and excretion data) and their time series are transformed into a three-dimensional body fluid state tensor. The first dimension is the data type (e.g., blood, tissue fluid, excretion data); the second dimension is the time series, recording data at different times; and the third dimension is a terahertz imaging-based spatial distribution matrix of body fluids, where each body fluid data point corresponds to a spatial location, representing the distribution of body fluids in different parts of the body, such as the lower limbs and pleural cavity, providing spatial distribution information.

[0025] Among these technologies, body fluid spatial distribution imaging can display the spatial distribution of a user's body fluids at different points in time, such as lower limb edema and pericardial effusion. It can be understood that by providing a comprehensive and operable three-dimensional data structure, multi-dimensional information combining time, space, and body fluid type can be analyzed.

[0026] Step S104: Construct a dynamic equilibrium model of body fluids based on the three-dimensional body fluid state tensor, and perform dynamic equilibrium analysis and anomaly detection to obtain anomaly feature vectors; the anomaly feature vectors include the level of body fluid imbalance and key abnormal parameters. The three-dimensional body fluid state tensor includes physiological parameter dimensions, time series dimensions, and spatial distribution dimensions. The physiological parameter dimension includes blood, tissue fluid, and excretion data, while the spatial distribution dimension includes a body fluid spatial distribution matrix based on terahertz imaging. By constructing a body fluid dynamic equilibrium model, the data in the three-dimensional body fluid state tensor are input into the model to analyze the changing trend of body fluid equilibrium, perform anomaly detection, and generate feature vectors.

[0027] Specifically, a model is built to capture the dynamic relationships between blood, tissue fluid, and excretion data, calculating trends in electrolyte concentration and body fluid distribution. The model incorporates the user's historical metabolic data to predict trends in body fluid changes. Through real-time monitoring of body fluid status, abnormal patterns are identified based on the model. For example, if the rate of decrease in blood sodium concentration exceeds the normal threshold, the model will detect this change and mark it as an abnormal feature vector, providing a basis for intervention by issuing early warnings of potential health risks.

[0028] The abnormal feature vectors include imbalance levels (such as mild imbalance, severe imbalance, etc.) and key abnormal parameters (such as the rate of change in blood sodium concentration). These feature vectors can help the system quickly assess the current body fluid status and generate early warnings.

[0029] Step S105: Determine the intervention level based on the abnormal feature vector and fuzzy decision rules; Among them, based on the information extracted from the abnormal feature vector, the intervention level is determined by combining fuzzy decision rules, and the severity of fluid imbalance is judged by intelligent judgment, so as to achieve precise graded intervention.

[0030] For example, if the balance index deviation exceeds 10% and the serum sodium concentration continues to decrease for more than 1 hour, a Level III intervention is triggered, indicating the need for an emergency warning; if the urine flow rate is less than 30 mL / h and accompanied by an abnormally high skin resistance, a Level II intervention is triggered, indicating that there may be a more serious fluid imbalance; if the balance index deviation is less than 5% but lasts for more than 30 minutes, a Level I intervention is triggered, reminding medical personnel to pay attention.

[0031] Step S106: Based on the preset strategy library, generate a corresponding set of candidate intervention instructions according to the abnormal feature vector and the intervention level; The pre-defined strategy library contains treatment strategies for various diseases, such as hyponatremia and hyperkalemia, and provides specific treatment plans based on different intervention levels. For example, for hyponatremia, the library may include plans for sodium supplementation or the use of related medications.

[0032] Specifically, based on the user's fluid status (such as decreased blood sodium concentration, abnormal urine electrolyte concentration, etc.) and intervention level, candidate intervention instruction plans are generated, including drug type, dosage, and infusion rate. For example, if a Level II or Level III intervention is triggered, the system automatically recommends drug type, dosage, and infusion method, such as sodium supplementation or hydration adjustment; if a Level I intervention is triggered, nutritional regulation, such as sodium restriction diet, can be implemented.

[0033] Step S107: Input the candidate intervention instruction set into the fluid dynamic balance model, predict the changes in the user's fluid state after execution, and select the optimal intervention instruction based on the prediction results. The fluid dynamic balance model predicts changes in the user's fluid status based on the current fluid state and candidate intervention commands. For example, after inputting the command to replenish saline solution, the system predicts the trends in serum sodium concentration, urine electrolyte concentration, etc. Based on the prediction results, the system automatically selects the most effective intervention to ensure maximum benefit in restoring fluid balance.

[0034] For example, if the system predicts that supplementing with saline solution will restore the blood sodium concentration to the normal range, it will select that option to execute; if other options are less effective, they will be automatically excluded.

[0035] Step S108: Drive the execution device corresponding to the optimal intervention command to adjust the body fluid parameters.

[0036] In this process, based on optimal intervention instructions, fluid regulation is performed by equipment (such as drug infusion pumps and fluid management devices). The system precisely adjusts the drug infusion rate and fluid flow rate according to the system's instructions to ensure accurate execution of the intervention. For example, if saline solution is selected, the drug infusion pump will start working according to the specified dosage and flow rate to ensure precise fluid delivery.

[0037] It should be noted that the device also needs to monitor the body fluid status in real time during the treatment and make adjustments based on real-time data as needed to ensure the therapeutic effect. Automated execution and real-time feedback adjustments can improve the accuracy of the treatment and ensure the smooth restoration of fluid balance.

[0038] In the above implementation, comprehensive real-time monitoring and analysis of the user's fluid balance is achieved. By establishing a dynamic balance model and using fuzzy decision rules to determine the intervention level, the system can automatically generate precise intervention instructions based on the abnormal characteristics of different fluid states and drive the execution device to perform personalized fluid regulation. The closed-loop feedback mechanism enables the system to dynamically optimize the intervention plan, continuously improve the accuracy of monitoring and treatment, and ensure the restoration of fluid balance.

[0039] Reference Figure 2 As a further implementation of the user fluid balance monitoring method, after the step of driving the execution device corresponding to the optimal intervention command to adjust the fluid parameters, the method further includes: Step S201: Start a timer for a preset time interval and record the initial abnormal feature vector as a baseline reference data; The system simulates the homeostasis recovery cycle of the human body's internal environment by setting a fixed time window (e.g., 30 days). This period is not arbitrarily chosen but is based on physiological research on the extracellular fluid renewal cycle; it is generally believed that the human extracellular fluid completes a full metabolic renewal process approximately every 20 to 30 days. Therefore, using this timeframe as the basic unit for monitoring and adjustment helps capture subtle trends in the body fluid system under long-term stable conditions, thus avoiding the risk of misjudgment due to short-term fluctuations. Furthermore, the extraction of the "initial abnormal feature vector" is not limited to the measurement value of a single indicator but is a multi-dimensional comprehensive representation vector. It contains a set of information from multiple dimensions, such as the level of imbalance, key biochemical parameters (e.g., serum sodium concentration), and functional indicators (e.g., urine osmolality). This information collectively constitutes the "state anchor point" of the individual's current health status, serving as the basic reference frame for all subsequent comparisons and inferences by the system.

[0040] Step S202: When the timer reaches the preset time interval, the updated body fluid status data of the user collected by the sensor is reacquired; Unlike traditional static diagnostic methods, this timed resampling approach effectively tracks the evolution of a patient's condition over time. To ensure consistent and accurate data acquisition, the sensor network used adheres to internationally recognized standards (such as IEEE 11073), supporting the simultaneous acquisition and synchronous transmission of multiple types of physiological signals. Specifically, diverse and heterogeneous data, including but not limited to blood biochemical parameters (electrolyte levels), tissue fluid dynamics (interstitial pressure), and excretory function (urine volume and composition analysis), are all considered, forming a comprehensive data foundation covering the entire spectrum of the body's water and electrolyte metabolism.

[0041] Step S203: Timestamp synchronization and drift correction are performed on the updated body fluid state data to generate an updated multimodal data stream; The timestamp synchronization relies primarily on the Network Time Protocol (NTP) or higher-level precision time services to unify the clock references within each subsystem and eliminate phase shifts caused by communication delays. To address potential baseline drift issues within the sensors themselves, a Kalman filter algorithm is introduced for real-time compensation. This method models the probability distribution of observation noise and system disturbances, constructing a state-space model describing sensor behavior. It iteratively estimates the true state value with each new data arrival, effectively denoising and correcting the original readings. The final output is a high-quality, time-aligned, and error-controllable sequence of multimodal physiological parameters.

[0042] Step S204: Construct the updated multimodal data stream into an updated three-dimensional body fluid state tensor; The design of the three-dimensional body fluid state tensor allows each specific physiological indicator to be precisely located in its time-parameter-space coordinate system, facilitating subsequent more refined state evolution analysis and anomaly identification.

[0043] Step S205: Based on the updated three-dimensional body fluid state tensor, load the body fluid dynamic equilibrium model, perform dynamic equilibrium analysis and anomaly detection, and generate the updated anomaly feature vector. Specifically, the pre-established fluid dynamic balance model combines the user's historical metabolic data to predict the updated fluid changes. After determining that there are signs of potential pathological changes, an alarm mechanism is triggered and a new round of "abnormal feature vectors" containing the latest imbalance level classification and the location of related abnormal parameters are generated.

[0044] Step S206: Compare the updated abnormal feature vector with the baseline reference data, and determine the trend of symptom data change based on the change in fluid imbalance level or the numerical difference of key abnormal parameters. Specifically, two complementary methods can be used for quantitative comparison. One is to examine the overall severity trend from a macro perspective, that is, by calculating the difference ΔL=LU between the imbalance level fields in the feature vectors before and after the two comparisons. LB can be used to intuitively reflect whether the condition is improving or worsening; on the other hand, it delves into the microscopic details and uses statistical methods to normalize the changes in various specific parameters to obtain the so-called "parameter difference" index. This scheme is essentially a Mahalanobis distance transformation, which can accurately reveal the relative deviation between two sets of data while removing the influence of the units of each variable.

[0045] Step S207: Determine whether the trend of symptom data changes indicates an increase in symptoms; if not, proceed to step S208; if yes, proceed to step S209. Step S208: Use the optimal intervention instruction and restart the timer to enter the monitoring cycle for the next preset time interval; Specifically, if the above comparison confirms no significant worsening trend, it means that the existing treatment is still sufficiently effective, and no immediate adjustments are needed. Simply continue with the existing intervention pace and continue observation. This approach not only saves unnecessary resource consumption but also reduces the risk of side effects from frequent changes in the treatment plan. It is worth noting that although maintaining the status quo may seem conservative, it is supported by extensive clinical experience and scientific validation, fully demonstrating the high adaptability of the automated management system to complex chronic disease management models.

[0046] Step S209: Based on the updated abnormal feature vector, re-execute the fuzzy decision rule to determine the new intervention level; Among them, "fuzzy decision rules" play an important bridging role between objective data and subjective experience. By defining a set of linguistic IF-THEN statement rule bases (e.g., "IF serum potassium concentration IS high risk AND imbalance level IS rapid deterioration THEN intervention level = emergency"), it allows computer programs to mimic the thinking patterns of doctors to make reasoning judgments.

[0047] It should be noted that such rules often originate from authoritative guidelines or consensus opinions of senior experts, possessing strong practical value and potential for widespread application. By assigning different membership function weight coefficients to each input factor, quantitative measurement results can be flexibly converted into qualitative descriptive labels, which can then be used to match corresponding levels of response plans.

[0048] Step S210: Based on the preset strategy library, generate a new set of candidate intervention instructions according to the updated abnormal feature vector and the new intervention level; The pre-defined strategy library is essentially a collection of intervention combinations encoded and stored in the form of decision trees or state machines, covering a variety of possibilities such as drug dosage adjustments and commands to start and stop mechanical assistive devices. Furthermore, considering the potential for multiple factors to interact in reality, the strategy library should also possess a degree of flexible expansion capability, allowing for further refinement and expansion as sufficient experience is accumulated from new cases in the future.

[0049] Step S211: Input the new candidate intervention instruction set into the fluid dynamic balance model, predict the changes in the user's fluid state after execution, and select the new optimal intervention instruction based on the prediction results. Step S212: Drive the execution device corresponding to the new optimal intervention command to adjust the body fluid parameters and update the baseline reference data to the updated abnormal feature vector.

[0050] The above implementation method constructs a novel medical support platform that not only closely aligns with clinical practice needs but also allows for continuous autonomous optimization and improvement. The innovation of this solution lies in breaking away from the traditional limitations of relying solely on human experience for intervention. Instead, it establishes a new diagnostic and treatment paradigm centered on time, based on multi-level parameter changes, and driven by intelligent decision-making. This significantly improves the quality of life and convenience for patients with chronic kidney disease, congestive heart failure, and related conditions, demonstrating broad application prospects and social significance.

[0051] Reference Figure 3 As one implementation of step S104, the steps of constructing a dynamic equilibrium model of body fluids based on a three-dimensional body fluid state tensor, performing dynamic equilibrium analysis and anomaly detection, and obtaining anomaly feature vectors include: Step S301: Perform multi-scale feature decomposition based on the time series dimension of the three-dimensional body fluid state tensor to extract physiological rhythm features of different frequency bands. Multiscale feature decomposition (MFLOP) involves performing time-frequency analysis on time-series data to extract physiological rhythm features across different frequency bands. Fluid metabolism is influenced by various physiological rhythms (such as heart rate, respiratory rhythm, and blood flow), which typically fall within the low-frequency (0.1-2 Hz) and high-frequency (2-10 Hz) ranges. Techniques such as wavelet transform can decompose time-series data into different frequency bands, helping to capture periodic changes closely related to fluid metabolism and further reveal the body's physiological state.

[0052] Specifically, wavelet transform is a commonly used time-frequency domain analysis method that can effectively decompose the time-frequency features in complex signals. By decomposing time series data into different frequency bands using wavelet transform, important low-frequency and high-frequency signals in body fluid metabolism can be highlighted. The low-frequency component usually reflects slow physiological changes (such as changes in fluid balance), while the high-frequency component reflects more rapid physiological fluctuations (such as changes in blood circulation and respiration).

[0053] For example, during monitoring, it is assumed that changes in serum sodium concentration exhibit regular fluctuations in the 0.1-2 Hz frequency band, which may be related to chronic water-electrolyte imbalances. Conversely, the 2-10 Hz frequency band reflects faster fluid regulation processes, such as minor short-term changes in body fluids. Wavelet transform effectively extracts key frequency bands related to fluid metabolism, improving the ability to detect dynamic changes in body fluids. By preserving the characteristics of these two frequency bands, changes in body fluid states can be accurately reflected, providing more precise data for subsequent dynamic equilibrium analysis.

[0054] Step S302: Construct a dynamic balance model of body fluids based on physiological rhythm characteristics and calculate the real-time balance index; In one embodiment of this application, the low-frequency and high-frequency features obtained through wavelet transform are concatenated with the original data and then input into an improved recurrent neural network (RNN) for time-series modeling to construct a fluid dynamic balance model. In this way, the model can effectively learn the temporal patterns of fluid metabolism and identify the contribution of different frequency band features to fluid balance.

[0055] For example, by inputting the spliced ​​data (including low-frequency and high-frequency features of blood sodium concentration) into an improved RNN, the model can learn the time dependence of these data and predict the trend of blood sodium concentration changes in the future, thus improving the prediction accuracy and sensitivity.

[0056] Step S303: Perform electrolyte concentration change trend analysis on the time series dimension of the three-dimensional body fluid state tensor to obtain the electrolyte concentration change trend results; Among them, by analyzing the trend of electrolyte concentration changes in the time series data in the three-dimensional body fluid state tensor, the rate of change of serum sodium, serum potassium and urinary sodium concentrations is calculated. This step helps to understand the dynamic changes of body fluid electrolytes and provides basic data for the detection and early warning of abnormal body fluid balance.

[0057] Specifically, trend analysis includes calculating the rate of change in serum sodium, serum potassium, and urinary sodium concentrations. The hourly change in serum sodium concentration is calculated using a sliding time window, and any change exceeding 2 mmol / L is marked as abnormal fluctuation. Urine electrolyte concentration data are then used to verify the authenticity of abnormal potassium metabolism and rule out sensor false alarms. Ultimately, the electrolyte concentration trend results include the hourly change in serum sodium concentration, the hourly change in serum potassium concentration, and the urinary sodium excretion rate.

[0058] Step S304: Based on the deviation between the real-time balance index and the preset index threshold, and combined with the electrolyte concentration change trend results, generate an abnormal feature vector containing the fluid imbalance level and key abnormal parameters.

[0059] Specifically, by analyzing the deviation between the real-time balance index and a preset threshold, combined with the trend of electrolyte concentration changes, an abnormal feature vector is generated. This vector includes not only the level of fluid imbalance (e.g., mild, moderate, or severe imbalance) but also key abnormal parameters (such as the rate of change in serum sodium concentration and the rate of urinary sodium excretion), providing a clear basis for intervention decisions.

[0060] Specifically, if the real-time balance index deviates from the normal range and the rate of change in serum sodium concentration exceeds the normal fluctuation range (e.g., exceeding 2 mmol / L), the system will generate an abnormal feature vector containing information such as the imbalance level and abnormal parameters (e.g., the rate of change in serum sodium concentration), prompting medical personnel to take corresponding intervention measures to improve the accuracy of fluid balance management.

[0061] The above embodiments effectively achieve real-time monitoring and anomaly detection of the user's fluid balance. By generating abnormal feature vectors, the system can accurately determine the level and key parameters of fluid imbalance, thereby providing a precise basis for subsequent intervention decisions. This technical solution can significantly improve the efficiency and accuracy of fluid balance management, providing strong support for clinical treatment.

[0062] Reference Figure 4 As a further implementation of the user fluid balance monitoring method, after the step of generating an abnormal feature vector containing the fluid imbalance level and key abnormal parameters, the method further includes: Step S401: Analyze stress indicators related to fluid imbalance from abnormal feature vectors and generate neural activation trigger signals; Specifically, the fluid imbalance level (such as hypertonic dehydration level III) and key abnormal parameters (such as serum sodium concentration >150 mmol / L and abnormally low antidiuretic hormone level) are structured and analyzed. Quantitative indicators that can reflect the macro-stress intensity and micro-neuroregulation needs are extracted through a multi-dimensional feature separation algorithm.

[0063] The macroscopic stress intensity is primarily determined by the overall severity of fluid imbalance, used to determine whether the activation threshold of the hypothalamus-pituitary-adrenal (HPA) axis has been reached. For example, when the imbalance level exceeds a preset threshold (e.g., level II), the neuroendocrine response initiation signal of the HPA axis can be triggered. Microscopic regulation requirements are based on key abnormal parameters, which are converted into activation codes for specific brain region neuronal groups using a parameter-neural pathway mapping table. For instance, when serum sodium concentration increases, the firing pattern of neurons activating the thirst center in the supraoptic nucleus of the hypothalamus can be encoded in binary form (e.g., 1011), while abnormal antidiuretic hormone corresponds to the activation code of neurons in the paraventricular nucleus (e.g., 0110). The final output neural activation trigger signal is a structured data packet containing spatial coordinate information of the target brain region (e.g., three-dimensional position in the MNI coordinate system), the type of target neurotransmitter (e.g., glutamatergic or γ-aminobutyric acid), and a baseline stimulus intensity scalar value. This information together constitutes the basic instruction set for the neuromodulation system to execute interventions.

[0064] Step S402: Based on the neural activation trigger signal, query the neural response mapping library and output the autonomic neural regulation target and expected regulation intensity; The query process employs a hierarchical matching strategy: In the initial matching stage, baseline response functions of all neural nuclei within the corresponding region are retrieved based on the target brain region coordinates in the trigger signal; in the dynamic correction stage, a stimulus intensity scalar is input into the response function to calculate the theoretical neuronal firing frequency increment Δf; the synergistic constraint stage further introduces a neural pathway inhibition matrix to eliminate conflicting regulatory commands (such as the physiological contradiction caused by simultaneously activating the sympathetic and vagus nerves). The final output is a set of clearly defined autonomic neural regulatory target triplets (including brain region coordinates, neural nucleus names, and neurotransmitter types) and their expected regulatory intensity (e.g., stimulating paraventricular nucleus glutamatergic neurons, resulting in a target firing frequency increase of 15 Hz).

[0065] Step S403: Input the autonomic nervous system regulation target and expected regulation intensity into the body fluid dynamic balance model to simulate the changes in body fluid state after neural feedback and generate a neural-humoral coupling compensation vector. The model is extended to a neuro-humoral bidirectional coupling system, reflecting not only the basic dynamics of fluid metabolism but also the feedback effect of neural regulation on the endocrine system. During the simulation, the regulatory target and intensity are first converted into specific physiological parameter changes, such as calculating the expected increase in antidiuretic hormone (ADH) concentration using the neuronal-endocrine transfer function. This increase is then injected as a new input into the model's renal tubular reabsorption module, iteratively calculating a new fluid balance state in conjunction with factors such as glomerular filtration rate (GFR), and further updating changes in plasma osmolality. Finally, by comparing the predicted fluid state values ​​before and after intervention, the system generates a neuro-humoral coupling compensation vector V_comp = [Δfluid imbalance level, Δkey abnormal parameter], for example, reducing the originally predicted dehydration level from III to II, and decreasing serum sodium concentration from 152 mmol / L to 146 mmol / L.

[0066] Step S404: Fuse the neuro-humoral coupling compensation vector with the abnormal feature vector, and update the fluid imbalance level and key abnormal parameters in the abnormal feature vector.

[0067] One approach is to use a Bayesian fusion mechanism to integrate multi-source biological data, thereby improving the accuracy and reliability of anomaly feature vectors. The fusion process first assigns weights to the original anomaly feature vector and the neural compensation vector based on the reliability of the data source. The weights of the original vectors are... The compensation vector weights are determined based on the sensor's measurement accuracy (e.g., ±2% error of the blood sodium electrode). This depends on the confidence level of the model's prediction (e.g., when R² > 0.85). =0.7). In the parameter-level fusion stage, key abnormal parameters (such as serum sodium concentration) are updated using a weighted average method to comprehensively consider the contributions of measured values ​​and model predictions. Furthermore, the fluid imbalance level is reclassified using the fused parameters; for example, a decision tree classifier is used to determine the new imbalance level based on the updated serum sodium concentration and ADH level. The updated abnormal feature vector not only retains the original fluid disturbance information but also indicates the contribution of neural compensation mechanisms, providing a more comprehensive physiological state assessment for subsequent intervention decisions.

[0068] In the above embodiments, by constructing a neuro-humoral bidirectional closed-loop feedback system, the traditional unidirectional humoral monitoring mechanism is upgraded to an intelligent physiological regulation framework with active intervention capabilities. Starting from biochemical indicators of humoral imbalance, the system generates neural activation signals, matches and queries neural responses, and conducts simulation verification using a neuro-humoral coupling model. Finally, it updates abnormal feature vectors through a Bayesian fusion mechanism, forming a complete "perception-analysis-intervention-feedback" closed-loop regulatory process. This mechanism makes the intervention process more closely resemble the body's own compensatory mechanisms, significantly improving the timeliness and safety of treatment.

[0069] Reference Figure 5 As one implementation of step S304, the step of generating an abnormal feature vector containing the fluid imbalance level and key abnormal parameters based on the deviation between the real-time balance index and a preset index threshold, combined with the electrolyte concentration change trend results, includes: Step S501: Calculate the absolute deviation value based on the real-time balance index and the preset index threshold; The Real-Time Balance Index (LBI) is a core indicator for measuring a user's fluid balance, reflecting the difference between the actual state of bodily fluids and the ideal balance state. By comparing the LBI with a preset normal threshold, an absolute deviation value can be calculated, representing the degree of difference between the current fluid state and the normal state. The magnitude of the absolute deviation value directly reflects whether there is an imbalance in bodily fluids; the larger the deviation value, the greater the degree of imbalance.

[0070] For example, assuming the real-time fluid balance index (LBI) is 10 and the preset normal threshold is 5, the calculated absolute deviation value is 5, indicating a significant deviation in fluid balance. Quantifying the balance deviation provides a quantitative assessment standard for subsequent anomaly detection and intervention, helping to accurately determine the degree of deviation from fluid balance.

[0071] Step S502: Based on the spatial distribution dimension of the three-dimensional body fluid state tensor, perform edge detection based on the body fluid spatial distribution matrix and mark the edema area; It is important to note that during fluid monitoring, the spatial distribution of body fluids is crucial for assessing edema areas. Edge detection algorithms (such as the Sobel operator) can be applied to the spatial distribution matrix of body fluids to extract regions of significant variation in the spatial data, i.e., edema areas. Edema typically occurs when excessive fluid accumulates in tissues, and labeling these edema areas is essential for determining the type and severity of fluid imbalance.

[0072] For example, in the body fluid distribution matrix (128x128 pixels), some areas show significant differences from the distribution of fluid in the surrounding tissues. Through edge detection, the system can mark these areas as potential edema areas (such as fluid accumulation around the heart or in the lower limbs). For example, ascites areas are distributed in a high-gradient block shape in the matrix and are marked as red covered areas (edema areas) by the algorithm.

[0073] Step S503: Based on the area ratio of the edema region, spatial correction is performed on the absolute deviation value to generate a spatially corrected deviation value. The area proportion of edema is closely related to the spatial distribution of fluid balance. By calculating the proportion of edema, the absolute deviation value can be spatially corrected, thereby compensating for the impact of edema on the overall fluid balance. The spatially corrected deviation value comprehensively considers the distribution of fluid in different parts of the body, which helps to more accurately assess the actual state of fluid balance.

[0074] For example, if the edema area accounts for 15%, the absolute deviation value is adjusted based on this proportion, correcting the originally calculated deviation value of 14 to 14.42. This avoids global assessment errors caused by fluid accumulation in local areas. The specific formula is: Spatial correction deviation value = Absolute deviation value × (1 + 0.2 × edema area percentage); where, according to clinical trial data, when the edema area is greater than 10%, the body's compensatory capacity decreases by 20%.

[0075] Understandably, by correcting the spatial deviation value, the actual situation of fluid imbalance can be reflected more accurately, avoiding the assessment error based solely on global data and improving the accuracy of imbalance judgment.

[0076] Step S504: Generate the electrolyte comprehensive weighting coefficient based on the electrolyte concentration change trend results; Changes in electrolyte concentrations (such as serum sodium, serum potassium, and urinary sodium) directly reflect the state of body fluids, and trends in electrolyte changes can help determine the type and severity of fluid imbalance. Based on the trend of electrolyte concentration changes (such as hourly changes) and risk level (e.g., larger changes in serum sodium concentration indicate higher risk), a comprehensive weighting coefficient is generated. This coefficient is used in subsequent abnormality score calculations.

[0077] For example, a patient's electrolyte data showed significant abnormalities: serum sodium concentration decreased by 2.5 mmol / L per hour (exceeding the high-risk threshold of 2.0 mmol / L, trigger weight 3), serum potassium concentration increased by 0.6 mmol / L per hour (low risk, weight 1), and urinary sodium excretion was only 60% of the baseline value (below the 50%-80% medium-risk range, reflecting impaired renal compensatory function). According to the preset weighting formula, serum sodium risk dominated the calculation (weighting 83%), combined with correction terms for serum potassium and urinary sodium excretion, ultimately generating an electrolyte comprehensive weighting coefficient of 3.62. This value is significantly higher than the baseline threshold (generally ≥2.5 indicates a high-risk state), indicating that the patient needs priority intervention for sodium ion disorder, and urinary sodium excretion needs to be monitored to assess renal perfusion status.

[0078] Understandably, the weighting coefficients generated by the trend of electrolyte concentration changes can provide quantitative weighting support for the assessment of fluid imbalance, so that the assessment of fluid imbalance is not only based on the overall deviation value, but also takes into account the specific situation of electrolyte fluctuations.

[0079] Step S505: Calculate the corresponding abnormal score based on the area ratio of the edema region, combined with the spatial correction deviation value and the electrolyte comprehensive weighting coefficient. Among them, the area ratio of edema represents the spatial distribution of fluid imbalance, the corrected balance index deviation reflects the overall fluid balance status, and the electrolyte comprehensive weight coefficient adjusts the score according to the degree of change of electrolytes in the body fluid. The combination of the three provides a comprehensive abnormal score to measure the severity of fluid imbalance.

[0080] For example, the specific abnormality scoring formula is: Abnormality Score = Spatial Correction Bias Value × Electrolyte Comprehensive Weighting Coefficient + 20 × Percentage of Edema Area. The clinical significance of the coefficient 20 is as follows: each 1% of edema area contributes 20 points, equivalent to a risk equivalent to a 2 mmol / L decrease in serum sodium per hour.

[0081] Understandably, by combining a comprehensive score that includes edema area and electrolyte changes, a more accurate and comprehensive assessment of fluid imbalance can be achieved. This scoring system can be dynamically adjusted to adapt to different changes in fluid state.

[0082] Step S506: Determine key abnormal parameters and fluid imbalance levels based on the abnormality score to obtain the abnormal feature vector.

[0083] By calculating the abnormality score and combining it with specific thresholds, the system can determine the level of fluid imbalance (e.g., severe, moderate, mild) and key abnormal parameters (e.g., serum sodium concentration, urinary sodium excretion rate, etc.) and generate abnormal feature vectors. These feature vectors can be used for subsequent intervention decisions and treatment plan formulation.

[0084] For example, an abnormal score of 40 or higher is considered a severe imbalance, an abnormal score between 20 and 40 is considered a moderate imbalance, and an abnormal score of 20 or lower is considered a mild imbalance. During the scoring process, parameters that contribute significantly to the score are selected as key abnormal parameters. For instance, if the rate of change in serum sodium contributes 58% (the dominant factor), the edema area contributes 5.4%, and other parameters (such as urinary sodium excretion) account for 36.6%, this indicates that sodium ion imbalance should be corrected first, while monitoring the progression of local edema.

[0085] In the above implementation, by assessing the degree of imbalance in a user's body fluid status in real time and accurately, multi-dimensional data is integrated into a single score, avoiding subjective judgment bias and providing a scientific and quantitative basis for medical intervention. Through multi-dimensional information fusion and intelligent analysis, this solution significantly improves the accuracy and timeliness of monitoring body fluid imbalance, helps reduce medical misdiagnosis, avoids complications caused by body fluid imbalance, and supports the development of personalized treatment plans.

[0086] Reference Figure 6 As a further implementation of the monitoring method, after the step of driving the execution device corresponding to the optimal intervention command to adjust the body fluid parameters in step S108, the method further includes: Step S601: Collect actual body fluid data of the user after the intervention; After the intervention command is executed, the user's body fluid status should change. In order to verify the effectiveness of the intervention, it is necessary to collect actual body fluid data in real time after the intervention. These data can reflect the changes in the user's body fluid status after the intervention and serve as the basis for subsequent comparison and optimization.

[0087] For example, assuming that after a user receives fluid replacement intervention, their blood sodium concentration recovers and their urine flow may increase as fluid balance is restored, the system will collect this new fluid data for subsequent error comparison and model adjustment.

[0088] Step S602: Determine the model prediction results of the optimal intervention instruction based on the fluid dynamic balance model; Before intervention, the fluid dynamics model generates a predicted fluid change outcome (i.e., the model prediction) based on the user's current fluid status and the model's prediction mechanism. This prediction represents the trajectory of fluid changes under ideal conditions. For example, it is assumed that after fluid replacement, the model predicts that serum sodium concentration will return to normal within 2 hours, and urine flow should gradually increase.

[0089] Understandably, by using the model's prediction results, the system can establish an ideal baseline for changes in body fluids, providing a benchmark for subsequent error comparisons and model parameter adjustments. This also helps determine whether an intervention is effective.

[0090] Step S603: Compare the user's actual body fluid data with the model prediction results to obtain the error comparison results; By comparing the collected actual body fluid data with the model's prediction results, the system can calculate the error between the actual body fluid state and the predicted result. The error comparison results reflect the deviation between the actual body fluid state and the expected state after intervention, revealing the degree of deviation in the intervention effect.

[0091] For example, assuming the model predicts that serum sodium concentration will recover from 120 mmol / L to 135 mmol / L, while the actual monitored serum sodium concentration is 130 mmol / L, the error comparison result will show a deviation of 5 mmol / L in serum sodium concentration. This error comparison result can help assess the effectiveness of the intervention. If the error between the actual body fluid data and the predicted result is large, it may indicate that the intervention effect is not as expected, or that the intervention measures need further adjustment.

[0092] Step S604: Adjust the parameter weights of the body fluid dynamic balance model according to the error comparison results, and update the sensor sampling strategy.

[0093] Based on the error comparison results, the system can adjust the parameter weights of the body fluid dynamic balance model to better reflect actual body fluid changes. For example, if the error between the model's prediction and actual body fluid data is large, it may be necessary to adjust the weights of certain variables in the model to improve the model's prediction accuracy. Simultaneously, based on the error results, the sensor's sampling strategy can also be optimized, such as adjusting the sampling frequency or selecting a higher-precision sensor.

[0094] For example, assuming the error comparison results show that the recovery rate of serum sodium concentration is slow, the system may adjust the weights of parameters related to sodium ion metabolism in the fluid dynamic balance model to improve the model's sensitivity to changes in serum sodium concentration. Simultaneously, the sensor sampling frequency may be adjusted to increase the monitoring frequency of changes in serum sodium concentration; when the system is in steady state (balance index fluctuation <1% for 2 hours), the terahertz imaging module may be turned off to reduce power consumption.

[0095] Understandably, by adjusting model parameter weights and updating sampling strategies, the system can dynamically optimize the model based on actual conditions, ensuring higher accuracy and adaptability in subsequent interventions. This feedback mechanism enhances the system's adaptive capabilities, ensuring more precise and effective future interventions.

[0096] In the above implementation, a real-time feedback mechanism is introduced after intervention. This mechanism compares actual body fluid data with predicted results and adjusts the parameters of the body fluid dynamic balance model and the sensor sampling strategy based on the error. This technical solution enables system self-optimization, allowing for dynamic adjustment of intervention measures and model predictions according to actual effects, significantly improving the accuracy and effectiveness of body fluid monitoring and intervention. Through this feedback loop mechanism, the system can continuously adapt to changes in the user's body fluid state, providing personalized and precise intervention plans to ensure efficient restoration of body fluid balance.

[0097] As a further implementation of the monitoring method, after obtaining the error comparison result, the method further includes: when the error value of the error comparison result exceeds a preset error threshold, switching to the backup sensor group and recalibrating the user's body fluid status data.

[0098] Specifically, when the error comparison results show that the deviation between the actual body fluid data and the predicted results exceeds the preset error threshold, the system will trigger the switching of the backup sensor group and recalibrate the user's body fluid status data to ensure that the system continuously provides accurate data support and avoids incorrect body fluid intervention commands due to sensor failure or data deviation.

[0099] For example, suppose the system compares the initial blood sodium concentration data collected by the sensors with the prediction model results and finds that the error value is greater than the preset error threshold (e.g., the error is greater than 5 mmol / L). This indicates that there may be a sensor malfunction or measurement error, leading to inaccurate body fluid data. In this case, the system automatically switches to the backup sensor group, re-collects data, and calibrates the body fluid status.

[0100] The above embodiments avoid data deviations caused by sensor malfunctions or environmental interference, ensuring the reliability of the body fluid regulation system. Through this self-correcting mechanism, the system can maintain normal operation when encountering abnormal data and promptly return to an effective working state, avoiding misleading interventions and potential health risks.

[0101] This application also discloses a user fluid balance monitoring system.

[0102] A user fluid balance monitoring system, the monitoring system comprising: The data acquisition module is used to acquire user body fluid status data collected by sensors in real time; body fluid status data includes blood parameters, tissue fluid data and excretion data; The data calibration module is used to perform timestamp synchronization and drift correction on the user's body fluid status data to obtain a calibrated multimodal data stream; The state tensor construction module is used to construct a three-dimensional fluid state tensor from the calibrated multimodal data stream; The abnormal feature vector generation module is used to construct a dynamic equilibrium model of body fluids based on a three-dimensional body fluid state tensor, and to perform dynamic equilibrium analysis and anomaly detection to obtain abnormal feature vectors. The abnormal feature vectors include the level of body fluid imbalance and key abnormal parameters. The intervention level determination module is used to determine the intervention level based on the abnormal feature vector and fuzzy decision rules. The candidate intervention instruction generation module is used to generate a set of corresponding candidate intervention instructions based on a preset strategy library, according to the abnormal feature vector and the intervention level. The optimal intervention instruction determination module is used to input the candidate intervention instruction set into the fluid dynamic balance model, predict the changes in the user's fluid state after execution, and select the optimal intervention instruction based on the prediction results. The intervention command drive module is used to drive the execution device corresponding to the optimal intervention command to adjust the body fluid parameters.

[0103] The user fluid balance monitoring system of this application embodiment can implement any of the above monitoring methods, and the specific working process of each module in the monitoring system can refer to the corresponding process in the above method embodiments.

[0104] In the several embodiments provided in this application, it should be understood that the provided methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for example, the division of a certain module is merely a logical functional division, and in actual implementation there may be other division methods, such as multiple modules can be combined or integrated into another system, or some features can be ignored or not executed.

[0105] This application also discloses a computer device.

[0106] A computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement a user fluid balance monitoring method as described above.

[0107] This application also discloses a computer-readable storage medium.

[0108] A computer-readable storage medium storing a computer program that can be loaded by a processor and executed as described above in any of the user fluid balance monitoring methods.

[0109] The computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in connection with an instruction execution system, apparatus, or device; the program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0110] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0111] The above are all preferred embodiments of this application and are not intended to limit the scope of protection of this application. Any feature disclosed in this specification (including the abstract and drawings) may be replaced by other equivalent or similar features unless specifically stated otherwise. That is, unless specifically stated otherwise, each feature is only one example of a series of equivalent or similar features.

Claims

1. A method for monitoring a user's fluid balance, characterized in that, The monitoring method includes: Real-time acquisition of user body fluid status data collected by sensors; the body fluid status data includes blood parameters, tissue fluid data, and excretion data; The user's bodily fluid state data is time-stamped and drift-corrected to obtain a calibrated multimodal data stream; The calibrated multimodal data stream is constructed as a three-dimensional fluid state tensor; A dynamic equilibrium model of body fluids is constructed based on the three-dimensional body fluid state tensor, and dynamic equilibrium analysis and anomaly detection are performed to obtain an anomaly feature vector; the anomaly feature vector includes the level of body fluid imbalance and key abnormal parameters. Based on the aforementioned abnormal feature vector, the intervention level is determined using fuzzy decision rules. Based on a preset strategy library, a corresponding set of candidate intervention instructions is generated according to the abnormal feature vector and the intervention level; The candidate intervention instruction set is input into the body fluid dynamic balance model to predict the changes in the user's body fluid state after execution, and the optimal intervention instruction is selected based on the prediction results. The device corresponding to the optimal intervention command is driven to adjust the body fluid parameters.

2. The method for monitoring user fluid balance according to claim 1, characterized in that, Following the step of driving the execution device corresponding to the optimal intervention command to adjust the body fluid parameters, the following is also included: Start a timer with a preset time interval and record the initial abnormal feature vector as baseline reference data; When the timer reaches the preset time interval, the updated body fluid status data of the user collected by the sensor is reacquired; The updated body fluid state data is timestamped and drift corrected to generate an updated multimodal data stream; The updated multimodal data stream is used to construct an updated three-dimensional fluid state tensor; Based on the updated three-dimensional body fluid state tensor, the body fluid dynamic equilibrium model is loaded, dynamic equilibrium analysis and anomaly detection are performed, and an updated anomaly feature vector is generated. The updated abnormal feature vector is compared with the baseline reference data, and the trend of symptom data change is determined based on the change in the level of fluid imbalance or the numerical difference of key abnormal parameters. Determine whether the trend of the symptom data indicates an increase in symptoms. If not, continue with the optimal intervention instruction and restart the timer to enter the monitoring cycle for the next preset time interval. If so, the fuzzy decision rule is re-executed based on the updated abnormal feature vector to determine a new intervention level; Based on the preset strategy library, a new set of candidate intervention instructions is generated according to the updated abnormal feature vector and the new intervention level; The new set of candidate intervention instructions is input into the fluid dynamic balance model to predict changes in the user's fluid state after execution, and a new optimal intervention instruction is selected based on the prediction results. The device corresponding to the new optimal intervention instruction is driven to adjust the body fluid parameters and update the baseline reference data to the updated abnormal feature vector.

3. The method for monitoring user fluid balance according to claim 1, characterized in that: The three-dimensional body fluid state tensor includes a physiological parameter dimension, a time series dimension, and a spatial distribution dimension. The physiological parameter dimension includes blood, tissue fluid, and excretion data, and the spatial distribution dimension includes a body fluid spatial distribution matrix based on terahertz imaging.

4. The method for monitoring user fluid balance according to claim 3, characterized in that, The steps for constructing a dynamic equilibrium model of body fluids based on the three-dimensional body fluid state tensor, performing dynamic equilibrium analysis and anomaly detection, and obtaining anomaly feature vectors include: Multi-scale feature decomposition is performed based on the time series dimension of the three-dimensional body fluid state tensor to extract physiological rhythm features of different frequency bands; Based on the aforementioned physiological rhythm characteristics, the dynamic balance model of body fluids is constructed, and the real-time balance index is calculated. Electrolyte concentration variation trend analysis was performed on the time series dimension of the three-dimensional body fluid state tensor to obtain the electrolyte concentration variation trend results. Based on the deviation between the real-time balance index and the preset index threshold, and combined with the electrolyte concentration change trend results, an abnormal feature vector containing the fluid imbalance level and key abnormal parameters is generated.

5. A method for monitoring user fluid balance according to claim 4, characterized in that, The process of generating anomaly feature vectors that include fluid imbalance levels and key abnormal parameters also includes: The stress indicators related to fluid imbalance are analyzed from the abnormal feature vectors to generate neural activation trigger signals; Based on the neural activation trigger signal, the neural response mapping library is queried, and the autonomic neural regulation target and expected regulation intensity are output. The autonomic nervous system regulation target and expected regulation intensity are input into the body fluid dynamic balance model to simulate the changes in body fluid state after neural feedback and generate a neural-body fluid coupling compensation vector. The neuro-humoral coupling compensation vector is fused with the abnormal feature vector to update the fluid imbalance level and key abnormal parameters in the abnormal feature vector.

6. The method for monitoring user fluid balance according to claim 4, characterized in that, The steps for generating an abnormal feature vector containing the fluid imbalance level and key abnormal parameters, based on the deviation between the real-time balance index and the preset index threshold, and combined with the electrolyte concentration change trend results, include: The absolute deviation value is calculated based on the real-time balance index and the preset index threshold. Based on the spatial distribution dimension of the three-dimensional body fluid state tensor, edge detection is performed based on the body fluid spatial distribution matrix, and edema areas are marked. Based on the area ratio of the edema region, the absolute deviation value is spatially corrected to generate a spatially corrected deviation value; Based on the electrolyte concentration change trend, an electrolyte comprehensive weighting coefficient is generated; Based on the area ratio of the edema region, the corresponding abnormal score is calculated by combining the spatial correction deviation value and the electrolyte comprehensive weighting coefficient. Based on the abnormality score, key abnormal parameters and fluid imbalance levels are determined, and an abnormal feature vector is obtained.

7. The method for monitoring user fluid balance according to claim 1, characterized in that, After the step of driving the execution device corresponding to the optimal intervention command to adjust the body fluid parameters, the method further includes: Collect actual bodily fluid data of users after the intervention is performed; The model prediction results for determining the optimal intervention instruction are based on the fluid dynamic balance model. The actual bodily fluid data of the user is compared with the prediction results of the model to obtain the error comparison results; The parameter weights of the body fluid dynamic balance model are adjusted based on the error comparison results, and the sensor sampling strategy is updated.

8. A user fluid balance monitoring system, characterized in that, The monitoring system includes: The data acquisition module is used to acquire user body fluid status data collected by sensors in real time; the body fluid status data includes blood parameters, tissue fluid data and excretion data; The data calibration module is used to perform timestamp synchronization and drift correction on the user's body fluid state data to obtain a calibrated multimodal data stream; The state tensor construction module is used to construct the calibrated multimodal data stream into a three-dimensional fluid state tensor; An abnormal feature vector generation module is used to construct a dynamic equilibrium model of body fluids based on the three-dimensional body fluid state tensor, and to perform dynamic equilibrium analysis and anomaly detection to obtain an abnormal feature vector; the abnormal feature vector includes the level of body fluid imbalance and key abnormal parameters. An intervention level determination module is used to determine the intervention level based on the abnormal feature vector and fuzzy decision rules. The candidate intervention instruction generation module is used to generate a corresponding set of candidate intervention instructions based on a preset strategy library, according to the abnormal feature vector and the intervention level. The optimal intervention instruction determination module is used to input the candidate intervention instruction set into the body fluid dynamic balance model, predict the changes in the user's body fluid state after execution, and select the optimal intervention instruction based on the prediction results. An intervention command driving module is used to drive the execution device corresponding to the optimal intervention command to adjust body fluid parameters.

9. A computer device, characterized in that: The method includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer program is stored that can be loaded by a processor and executed as described in any one of claims 1 to 7.

Images