A blood pressure and blood lipid remote real-time monitoring system and device based on edge computing

Abstract

The application discloses a blood pressure and blood lipid remote real-time monitoring system and device based on edge computing and relates to the technical field of medical health monitoring. The system comprises an information acquisition and processing module: a time-domain resolution multispectral photon counter is used to acquire photon dynamic observation data of subcutaneous tissue of a user; and on the edge side, a parameterized general biological tissue optical model and a biological-physical inverse solver are used to generate a user personal health digital twin model through an inverse optimization process and to solve real-time physiological parameters of the user. The application uses a time-domain resolution multispectral photon counter to acquire multi-dimensional photon dynamic observation data, uses a biological-physical inverse solver to suppress noise through model comparison, effectively resists motion artifacts, and improves the accuracy of data; by executing a first inverse optimization process, a personal health digital twin model is generated for each user, the influence of individual differences such as skin color and weight on the monitoring accuracy is reduced, and the authenticity of the monitoring result is improved.

Description

Technical Field

[0001] This invention relates to the field of medical and health monitoring technology, specifically to a remote real-time monitoring system and device for blood pressure and blood lipids based on edge computing. Background Technology

[0002] With the accelerating aging of society, the increasing prevalence of chronic diseases, and the growing awareness of public health management, the demand for continuous, remote, and real-time monitoring of human physiological parameters is becoming increasingly urgent. However, existing remote monitoring technologies, especially in the field of consumer-grade wearable devices, still have significant shortcomings: On the one hand, consumer-grade sensors are affected by factors such as motion artifacts and individual differences, resulting in low signal-to-noise ratios in raw data. Direct cloud transmission not only increases network burden but also fails to meet the needs of emergency-level real-time response. On the other hand, traditional systems place complex calculations entirely in the cloud, leading to a disconnect between analysis results and the data collection scenario. This means that data reliability cannot be determined by combining motion status, nor can it be intelligently interpreted by incorporating the user's personalized physiological baseline. At the same time, user experience bottlenecks such as the comfort of monitoring devices further restrict users' long-term reliance on them.

[0003] For example, invention patent CN115153460A discloses a multimodal physiological indicator monitoring and analysis system, including a smart wearable device and a smart terminal. The smart wearable device is configured to measure the user's body temperature, blood oxygen saturation, heart rate, and skin conductance data, and transmit them to the smart terminal. The smart terminal is configured to store, process, and display the physiological indicator status to the user. Through this invention, users can collect their own physiological indicator data in a portable, convenient, and accurate manner using the smart wearable device, and the smart terminal reflects the real-time status of the body's physiological indicators, meeting the user's health monitoring needs.

[0004] However, the above and similar technical solutions still have the following shortcomings: the lack of effective suppression of motion interference and physiological differences leads to high noise and poor stability of the collected raw data, resulting in reduced accuracy of subsequent analysis; the judgment criteria for health status rely too much on general standards, resulting in poor reliability of monitoring results in reflecting individual health; and the lack of local high-performance computing capabilities leads to reliance on remote servers for complex monitoring and analysis, causing system response delays. Summary of the Invention

[0005] The purpose of this invention is to provide a remote real-time monitoring system and device for blood pressure and blood lipids based on edge computing, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a remote real-time monitoring system for blood pressure and blood lipids based on edge computing, comprising:

[0007] Information acquisition and processing module: used to acquire dynamic photon observation data of the user's subcutaneous tissue through a time-domain resolved multispectral photon meter; and on the edge side, based on a parameterized general biological tissue optical model and a bio-physical inverse solver, to generate a digital twin model of the user's personal health through an inverse optimization process, and solve for the user's real-time physiological parameters;

[0008] Data monitoring module: Used at the edge to analyze and identify anomalies in users' real-time physiological parameters based on a data monitoring intelligent model, and trigger early warning signals;

[0009] Wireless transceiver module: used to transmit early warning signals and related physiological parameters generated at the edge to the remote monitoring platform, and to receive instructions from the platform;

[0010] Microenvironment management module: used to sense the user's skin microenvironment in real time, dynamically determine breathability needs, and intelligently control the switching and coordination of active and passive heat dissipation mechanisms;

[0011] Human-computer interaction module: used to display the user's real-time physiological parameters, health status information and warning prompts, and to receive user input.

[0012] Furthermore, the method for constructing the user's personal health digital twin model includes:

[0013] In the user's first use and calibration mode, in the resting state, acquire their photon dynamic observation data;

[0014] The bio-physical inverse solver is invoked, and a parameterized general biological tissue optical model is used as the forward model. The model parameters include structural parameters and physiological parameters. The structural parameters include melanin content, subcutaneous fat layer thickness, blood vessel wall elasticity, and basic optical properties of blood. The physiological parameters include intravascular pressure, blood flow velocity, and blood lipid concentration.

[0015] Preset structural and physiological parameters, execute the first inverse optimization process to obtain the user's personal structural parameters;

[0016] It stores the user's personal structural parameters and combines the parameterized general biological tissue optical model with the user's personal structural parameters to generate a digital twin model of the user's personal health.

[0017] Furthermore, the preset values ​​of the physiological parameters are within a physiologically reasonable range of resting prior values; the steps of the first reverse optimization process include:

[0018] The preset structural and physiological parameters are input into the forward model to calculate the dynamic prediction data of photons;

[0019] The dynamic observation data of photons in the resting state is compared with the dynamic prediction data of photons, the difference between the two is calculated by the loss function, and the structural parameters are adjusted accordingly.

[0020] Determine whether the optimization process has reached the first stopping condition;

[0021] If the stopping condition is met, the optimization will terminate, and the user's personal structure parameters will be obtained.

[0022] If the stopping condition is not met, repeat the above steps.

[0023] Furthermore, the method for solving the user's real-time physiological parameters includes:

[0024] The bio-physical inverse solver is invoked to use the user's personal health digital twin model as the forward model;

[0025] Real-time acquisition of dynamic photon observation data from users;

[0026] Preset physiological parameters, execute the second reverse optimization process, dynamically adjust the physiological parameters, and obtain the user's real-time physiological parameters.

[0027] Furthermore, the steps of the second reverse optimization process include:

[0028] Preset physiological parameters are input into the forward model to calculate real-time photon dynamic prediction data;

[0029] The real-time acquired photon dynamic observation data is compared with the real-time photon dynamic prediction data. The difference between the two is calculated through a loss function, and physiological parameters are adjusted accordingly.

[0030] Determine whether the difference meets the second stopping condition;

[0031] If the stopping condition is met, the optimization will terminate and the user's real-time physiological parameters will be obtained.

[0032] If the stopping condition is not met, repeat the above steps.

[0033] Furthermore, the method by which the intelligent data monitoring model analyzes and identifies anomalies in users' real-time physiological parameters includes:

[0034] Receive real-time physiological parameters from users, determine the user's current activity status, and add a timestamp and activity status label to each set of data;

[0035] Call the user's personal health digital twin model to obtain the expected range of physiological parameters for the user in the current state and generate a dynamic baseline;

[0036] The user's real-time physiological parameters are compared with a dynamic baseline to calculate the degree of deviation, and an instantaneous abnormality score is generated based on the magnitude and direction of the deviation.

[0037] The trend health status of real-time physiological parameters is assessed at multiple time scales to obtain a trend health status score;

[0038] Examine whether there are any abnormalities in the correlation between physiological parameters and obtain the correlation analysis results;

[0039] By combining the instantaneous anomaly score, trend health score, and correlation analysis results, a final anomaly judgment is made on the real-time physiological parameters, and a graded response is triggered.

[0040] Furthermore, the microenvironment management module specifically includes:

[0041] The temperature and humidity data of the user's skin contact surface are collected in real time through a bionic microarray base.

[0042] Based on temperature and humidity data and the user's current condition, determine the risk level of sweat accumulation;

[0043] The strategy is selected based on the risk level of sweat buildup, and the strategies include passive ventilation strategy, mixed ventilation strategy and active ventilation strategy.

[0044] It monitors temperature and humidity changes in real time and makes adaptive adjustments.

[0045] Furthermore, the time-domain resolved multispectral photon meter includes a multi-frequency coherent photon emission source and a photon detector array; the photon dynamic observation data includes photon flight time distribution, spatial correlation of photon count rate, and microsecond-level fluctuations in the polarization state of the scattered field; the bio-physical inverse solver integrates Monte Carlo simulation and deep learning models.

[0046] A remote real-time monitoring device for blood pressure and blood lipids based on edge computing, characterized in that: it uses any one of the aforementioned remote real-time monitoring systems for blood pressure and blood lipids based on edge computing.

[0047] Compared with the prior art, the beneficial effects of the present invention are:

[0048] A remote real-time monitoring system and device for blood pressure and blood lipids based on edge computing dynamically manages the temperature and humidity of the skin contact surface through a biomimetic microarray base and intelligent ventilation strategy of the microenvironment management module, solving the problems of sweat accumulation and insufficient breathability, and providing a stable and reliable data acquisition guarantee for the information acquisition and processing module. Through the intelligent data monitoring model, multi-dimensional collaborative analysis and intelligent context judgment of real-time physiological parameters are performed to improve the accuracy, real-time performance and reliability of physiological monitoring, while reducing the false alarm rate.

[0049] Meanwhile, by employing a time-domain resolved multispectral photon meter to collect multidimensional photon dynamic observation data, and utilizing a bio-physical inverse solver to suppress noise through model comparison, the accuracy of the data is effectively resisted and motion artifacts are improved. By executing the first inverse optimization process, a personal health digital twin model is generated for each user, reducing the impact of individual differences such as skin color and weight on monitoring accuracy and improving the authenticity of monitoring results. Through an edge computing architecture, the construction of the user's personal digital twin model, the inverse solution of real-time physiological parameters, and the analysis and early warning decision-making of the data monitoring intelligent model are all executed on local devices, ensuring low latency and high real-time performance. Attached Figure Description

[0050] Figure 1 This is a schematic diagram of the remote real-time blood pressure and blood lipid monitoring system and device of the present invention;

[0051] Figure 2 This is a schematic diagram of the method for constructing a user's personal health digital twin model according to the present invention;

[0052] Figure 3 This is a schematic diagram of the first reverse optimization process of the present invention;

[0053] Figure 4 This is a schematic diagram of the method for solving real-time physiological parameters of users according to the present invention;

[0054] Figure 5 This is a schematic diagram of the second reverse optimization process of the present invention;

[0055] Figure 6 This is a schematic diagram of the intelligent data monitoring model of the present invention;

[0056] Figure 7 This is a schematic diagram of the microenvironment management method of the present invention. Detailed Implementation

[0057] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0058] It should be noted that the core of this invention adopts an "edge computing" architecture. To address the latency issues faced by traditional "end-to-cloud" architectures in data processing, this invention relates to core computing tasks such as constructing and calling a user's personal health digital twin model and solving and analyzing real-time physiological parameters, all of which are deployed and executed on the user's local device (i.e., the edge side).

[0059] like Figure 1As shown, the present invention provides a technical solution: a remote real-time monitoring system and device for blood pressure and blood lipids based on edge computing, comprising:

[0060] Information acquisition and processing module: It is used to acquire dynamic photon observation data of the user's subcutaneous tissue through a time-domain resolved multispectral photon meter; and on the edge side, based on a parameterized general biological tissue optical model and a bio-physical inverse solver, it generates a digital twin model of the user's personal health through an inverse optimization process and solves the user's real-time physiological parameters.

[0061] like Figure 2 As shown, the present invention provides a method for constructing a user's personal health digital twin model;

[0062] Specifically:

[0063] In the user's first use and calibration mode, in the resting state, acquire their photon dynamic observation data;

[0064] The bio-physical inverse solver is invoked, and the parameterized general biological tissue optical model is used as the forward model. The model parameters include structural parameters and physiological parameters.

[0065] Preset structural and physiological parameters, execute the first inverse optimization process to obtain the user's personal structural parameters;

[0066] It stores the user's personal structural parameters and combines the parameterized general biological tissue optical model with the user's personal structural parameters to generate a digital twin model of the user's personal health.

[0067] It is important to note that the first step is to acquire dynamic photon observation data. First, the device's human-computer interface prompts the user to maintain a seated or lying position for at least 5 minutes to keep their heart rate, blood pressure, and other physiological parameters at stable resting levels. Then, the multi-frequency coherent photon emission source (such as a vertical-cavity surface-emitting laser array) in the time-domain resolved multispectral photon meter is activated to emit a precisely controllable near-infrared micro-photon stream into the subcutaneous tissue of the user's wrist or other areas. Next, a high-temporal-resolution photon detector array (such as a single-photon avalanche diode detector array) is used to capture the photons scattered by the subcutaneous tissue, and the following dynamic photon observation data are recorded: photon time-of-flight distribution, measured based on time-correlated single-photon counting technology, determining the time interval from photon emission to detection; spatial correlation of the photon count rate, calculated by analyzing the photon count rate received by different units in the detector array; and microsecond-level fluctuations in the polarization state of the scattered field, recorded by placing a polarization analysis element in front of the detector and using a high-temporal-resolution polarization state detection module.

[0068] The second step involves invoking the bio-physical inverse solver. First, a pre-built parameterized general biological tissue optical model, constructed using Monte Carlo simulation, is invoked. This model simulates the propagation of photons in multi-layered biological tissues (including epidermis, dermis, subcutaneous fat, blood vessels, and blood). The model parameters are then configured, including: structural parameters S (melanin content S1, subcutaneous fat layer thickness S2, blood vessel wall elasticity S3, and basic blood optical properties S4, such as the absorption coefficient bases of oxyhemoglobin and deoxyhemoglobin); and physiological parameters D (intravascular pressure D1, blood flow velocity D2, and blood lipid concentration D3). Next, the bio-physical inverse solver is activated. This solver is a hybrid computational architecture that integrates Monte Carlo simulation and deep learning models. Its training data comes from a large amount of data simulating the correspondence between tissue optical properties and photon dynamics using the Monte Carlo method.

[0069] The third step is to execute. Figure 3 The first reverse optimization process is illustrated below. First, based on the user's basic information and prior experience obtained from the human-computer interaction module, a set of values ​​is initialized for the structural parameter S. Simultaneously, the physiological parameter D is strictly constrained within its physiologically reasonable resting prior value range (e.g., blood pressure is preset to 80-120 mmHg). The preset initial parameters (S, D) are input into a general biological tissue optical model, and a set of predicted photon dynamic data (i.e., photon dynamic prediction data) is calculated and output through the embedded fast Monte Carlo simulator. Next, the photon dynamic observation data obtained in the first step is compared with the photon dynamic prediction data, and the difference between the two is quantitatively calculated using a loss function (e.g., mean squared error function). Then, an optimization algorithm (e.g., covariance matrix adaptive evolution strategy) is used. Based on the output value of the loss function, and under the condition that the physiological parameter D is constrained within the resting prior range, the structural parameter S is adjusted to reduce the loss value. The above steps are repeated until the optimization process reaches the first stopping condition, at which point the optimization terminates, and the current optimal user-specific structural parameter S* is output. The first stopping condition includes one or more of the following: the loss function converges, or the output value of the loss function is less than a preset threshold (e.g., 1×10⁻⁶). -6 (e.g., the change in loss value after multiple iterations is less than a threshold); parameter changes converge, and the change between the parameter vector generated in the current iteration and the parameter vector of the previous iteration (e.g., Euclidean norm) is less than a threshold; iteration reaches the upper limit, and the number of optimization iterations reaches the preset maximum number (e.g., 1000 times).

[0070] The fourth step is to generate a digital twin model of the user's personal health. The optimized user-specific structural parameters S* are encrypted and stored in a secure storage area on the local device. The user-specific structural parameters S* are then fixedly combined with a parameterized general biological tissue optical model; specifically, in the general model, the optimized structural parameters are fixed as constants, thereby generating a user-specific digital twin model of their personal health that can be used to solve physiological parameters in real time.

[0071] like Figure 4 As shown, the present invention provides a method for solving real-time physiological parameters of users;

[0072] Specifically:

[0073] The bio-physical inverse solver is invoked to use the user's personal health digital twin model as the forward model;

[0074] Real-time acquisition of dynamic photon observation data from users;

[0075] Preset physiological parameters, execute the second reverse optimization process, dynamically adjust the physiological parameters, and obtain the user's real-time physiological parameters.

[0076] It is important to note that by controlling the continuous operation of the time-resolved multispectral photon meter, its multi-frequency coherent photon emission source emits a micro-photon stream into the subcutaneous tissue, and the photon detector array synchronously captures the emitted photons; within a very short time window (e.g., 50ms), a set of dynamic photon observation data is collected and generated in real time.

[0077] The second reverse optimization process is as follows: Figure 5Specifically, as shown, the process is as follows: First, a set of initial values ​​is set for the physiological parameter D. These initial values ​​can be the user's most recent historical values ​​or the resting average of a group. The currently preset physiological parameter D is input into the loaded user's personal health digital model, and a set of real-time photon dynamic prediction data is calculated and output through the embedded fast Monte Carlo simulator. Next, the real-time photon dynamic observation data acquired in the second step is compared with the real-time photon dynamic prediction data, and the difference between the two is quantitatively calculated using a loss function (such as the mean squared error function). Then, an efficient optimization algorithm (such as stochastic gradient descent) is used to adjust the physiological parameter based on the loss value. The above steps are repeated until the optimization process reaches the second stopping condition, at which point the optimization is immediately terminated, and the current optimal real-time physiological parameter D* for the user is output. The second stopping condition is configured to meet real-time requirements, including one or more of the following: availability convergence, the loss function value drops to an acceptable threshold preset for real-time monitoring (this threshold is usually higher than the standard of the first reverse optimization process); real-time forced stopping, the number of optimization iterations reaches a maximum iteration limit set to ensure real-time performance (e.g., 20 times); results stabilize, the change in physiological parameter D is less than a clinically negligible threshold in several consecutive iterations.

[0078] Furthermore, the stable operation of this module also relies on the good skin contact interface maintained by the microenvironment management module. When the skin contact interface becomes damp and uncomfortable due to sweat accumulation or poor optical coupling, the signal-to-noise ratio of the acquired photon dynamic data will decrease significantly, directly affecting the accuracy of subsequent model construction and parameter solving. Therefore, the microenvironment management module provides an important guarantee for data acquisition for this module.

[0079] Data monitoring module: Used at the edge to analyze and identify anomalies in users' real-time physiological parameters based on a data monitoring intelligent model, and trigger early warning signals.

[0080] like Figure 6 As shown, the present invention provides an intelligent model for data monitoring;

[0081] Specifically:

[0082] Receive real-time physiological parameters from users, determine the user's current activity status, and add a timestamp and activity status label to each set of data;

[0083] Call the user's personal health digital twin model to obtain the expected range of physiological parameters for the user in the current state and generate a dynamic baseline;

[0084] The user's real-time physiological parameters are compared with a dynamic baseline to calculate the degree of deviation, and an instantaneous abnormality score is generated based on the magnitude and direction of the deviation.

[0085] The trend health status of real-time physiological parameters is assessed at multiple time scales to obtain a trend health status score;

[0086] Examine whether there are any abnormalities in the correlation between physiological parameters and obtain the correlation analysis results;

[0087] By combining the instantaneous anomaly score, trend health score, and correlation analysis results, a final anomaly judgment is made on the real-time physiological parameters, and a graded response is triggered.

[0088] It is important to note that the first step is data reception and status labeling. The system continuously receives real-time physiological parameters from the user, including blood pressure, blood flow velocity, and blood lipid concentration, from the biophysical solver. Simultaneously, it receives raw activity status data from the device's built-in motion sensors (such as accelerometers and gyroscopes). A lightweight activity status classifier (such as a decision tree-based classification model) is used to analyze the activity status data and determine in real-time whether the user is currently in an activity state such as "resting," "walking," "running," or "sleeping." A timestamp is added to each set of received real-time physiological parameters, and an activity status label is associated with it, forming a complete data record with contextual information.

[0089] The second step involves generating a personalized, dynamic baseline related to the current state based on the user's personal historical data. This involves accessing the user's personal health digital twin model stored locally, which contains the user's personal historical physiological data; retrieving physiological parameter records from history under similar states and time periods based on the currently identified activity state and time information; and calculating the user's personalized expected range in the current state using a statistical range estimation algorithm based on these historical records. As a preferred implementation, a statistical method based on historical data quantiles can be used to determine the range, such as an interval defined by the lower and upper quantiles of historical data: [Q] 25 -k×IQR,Q 75 +k×IQR], where Q 25 Q 75 The values ​​are 25th and 75th percentiles, IQR is the interquartile range, and k is the sensitivity coefficient (default 1.5).

[0090] The third step is instantaneous analysis, which quantifies the instantaneous deviation between real-time physiological parameters and the dynamic baseline. The real-time physiological parameter values ​​are compared with the obtained dynamic baseline (e.g., the median), and the standard deviation is calculated: Standard deviation = |Real-time physiological parameter - Median of dynamic baseline| / Standard deviation of dynamic baseline. Subsequently, this deviation is converted into an instantaneous abnormality score between 0 and 1 using a nonlinear mapping function. Preferably, the nonlinear mapping function can be configured to be less sensitive to small deviations and more sensitive to large deviations, such as an sigmoid function, whose output value gradually increases from 0 to 1 as the input deviation increases, effectively distinguishing between normal fluctuations and significant abnormalities in physiological parameters.

[0091] The fourth step is trend analysis, which assesses the health risks reflected by the long-term trends of physiological parameters across multiple time scales. Short-term trend analysis (e.g., analyzing changes over the past twenty minutes) and long-term trend analysis (e.g., analyzing baseline drift over the past three weeks) are performed on the physiological parameter sequences. Then, the results of the short-term and long-term trend analyses are weighted and fused to synthesize a comprehensive trend health score. In this weighted fusion, short-term trends can be assigned a higher weight (e.g., 0.7) to more sensitively reflect recent changes. Therefore, the trend health score = short-term trend score × 0.7 + long-term trend score × 0.3.

[0092] The fifth step is correlation analysis. This step verifies the consistency of anomalies through parameter correlation analysis to reduce false alarms. Based on historical user data, strong correlation rules between different physiological parameters are discovered using association rule mining techniques (such as the Apriori algorithm). During real-time monitoring, it checks whether the current combination of multiple physiological parameters violates established correlation rules. Based on rule violations, a correlation anomaly index is calculated: Correlation Anomaly Index = Number of Violated Rules / Total Number of Applicable Rules. This index provides an additional confidence dimension for anomaly detection. Furthermore, it can be configured to identify pending abnormal parameter combination patterns that correspond to known physiological or pathological states, including significantly elevated heart rate with normal blood oxygenation (physiological stress), elevated heart rate with decreased blood oxygenation (pathological state), and persistently elevated blood pressure with dyslipidemia (cardiovascular risk).

[0093] The sixth step involves synthesizing the results from all analytical dimensions to make a final decision and trigger corresponding tiered response actions. The instantaneous anomaly score, trend health score, and correlation analysis results (such as the correlation anomaly index) are used as input features. A classification decision model (such as a random forest classification model) is used for the final anomaly judgment, and a tiered response mechanism is triggered based on the risk score. The risk score is calculated as follows: Risk Score = 0.5 × Instantaneous Anomaly Score + 0.3 × Trend Health Score + 0.2 × Correlation Anomaly Index, corresponding to four risk levels: L1 (<0.3), L2 (0.3-0.6), L3 (0.6-0.8), and L4 (>0.8). The graded response mechanism specifically includes: L1 (prompt level) and L2 (advice level), which execute local prompts at the edge, such as displaying prompt information or providing health advice on the device interface. These levels of warning signals are marked as locally recorded only; L3 (warning level), which executes a local alarm (such as strong vibration) at the edge and generates a warning signal that requires remote notification. This signal includes the level of this warning, key real-time physiological parameters, and their contextual information (such as timestamps and activity status); L4 (emergency level), which executes an emergency alarm at the edge and simultaneously generates an emergency warning signal that requires remote assistance. This signal includes an emergency event identifier, a detailed summary of physiological parameters, and possible cause analysis.

[0094] Wireless transceiver module: used to transmit early warning signals and related physiological parameters generated at the edge to the remote monitoring platform, and to receive instructions from the platform.

[0095] It should be noted that this module executes the corresponding remote communication strategy according to the warning signal level generated by the data monitoring module, so as to realize the data and command interaction between the edge side and the remote monitoring platform. Specifically, the module is configured to: (1) listen to the warning signals from the data monitoring module: when a warning signal marked as L3 / L4 is identified, it automatically starts and establishes a connection with the remote monitoring platform; transmits the warning signal to the remote monitoring platform; for warning signals marked as L1 / L2, it maintains a low-power standby state by default and does not initiate remote transmission. (2) continuously or intermittently listen to the downlink commands from the remote monitoring platform: receive commands issued by the platform, such as model update files, warning threshold adjustment parameters, and remote diagnostic commands; distribute the received commands to the corresponding processing modules in the system, such as the data monitoring module and the information collection and processing module. (3) manage the wireless communication unit of the device (such as cellular network), use low-power Bluetooth or LPWAN technology to maintain the connection during non-transmission periods to save energy consumption, and quickly switch to high-speed mode when transmission is required.

[0096] Microenvironment management module: used to sense the user's skin microenvironment in real time, dynamically determine breathability needs, and intelligently control the switching and coordination of active and passive heat dissipation mechanisms.

[0097] like Figure 7 As shown, the present invention provides a microenvironment management method;

[0098] Specifically:

[0099] The temperature and humidity data of the user's skin contact surface are collected in real time through a bionic microarray base.

[0100] Based on temperature and humidity data and the user's current condition, determine the risk level of sweat accumulation;

[0101] The strategy is selected based on the risk level of sweat buildup, and the strategies include passive ventilation strategy, mixed ventilation strategy and active ventilation strategy.

[0102] It monitors temperature and humidity changes in real time and makes adaptive adjustments.

[0103] It is important to note that the first step involves acquiring the microenvironmental state of the skin contact surface in real time through a specific sensor structure. Data acquisition is achieved through a biomimetic microarray base, which is formed by an array of several miniature flexible polymer pillars. This structure creates multiple air and sweat circulation channels at the microscopic level. A miniature temperature and humidity sensor array integrated on the base continuously measures the temperature and humidity values ​​of the skin contact surface at a fixed sampling frequency (e.g., 10Hz). The acquired temperature and humidity data undergoes digital filtering (e.g., using a moving average filtering method) to eliminate occasional noise interference and ensure data stability and reliability. Simultaneously, real-time activity data of the user is acquired from motion sensors.

[0104] The second step involves comprehensively assessing the current risk level of sweat accumulation based on temperature and humidity data and real-time activity status data. A risk assessment logic unit is used to determine the risk: temperature and humidity data are filtered to eliminate instantaneous fluctuations; combined with activity status data, a multi-parameter fusion decision is made to classify the risk of sweat accumulation. Specific rules for determining the risk level of sweat accumulation include: a low-risk level is defined as follows: when the humidity value is consistently below a first threshold (e.g., 65%RH) and the activity level is low-intensity (e.g., sleep / rest); a medium-risk level is defined as follows: when the humidity value exceeds a second threshold (75%RH) or the skin temperature is higher than the first threshold (e.g., 35℃) and the activity level is medium-intensity (e.g., walking); and a high-risk level is defined as follows: when the humidity value exceeds a third threshold (e.g., 85%RH) or the skin temperature is higher than the second threshold (e.g., 37℃) and the activity level is high-intensity (e.g., running). This risk assessment logic unit can be implemented using a rule-based state machine.

[0105] The third step involves triggering corresponding heat dissipation strategies based on the risk level of sweat accumulation. When the risk level is low, a passive ventilation strategy is triggered, relying solely on the microscopic channel structure of the biomimetic microarray base for natural heat dissipation and sweat evaporation; the active ventilation unit remains closed. When the risk level is medium, a hybrid ventilation strategy is triggered, operating the micro piezoelectric fan in low-power mode to provide auxiliary airflow. When the risk level is high, an active ventilation strategy is triggered, operating the micro piezoelectric fan in rated power mode to actively disperse moisture through the channels of the biomimetic microarray base. The micro piezoelectric fan is integrated on the side of the device panel and uses the piezoelectric ceramic driving principle to achieve brushless vibration and generate airflow. The low-power mode refers to the fan operating within 20% to 40% of its maximum rated power; the rated power mode refers to the fan operating within 80% to 100% of its maximum rated operating power. Preferably, the power mode control can be achieved through pulse width modulation technology, precisely controlling the average input power of the fan by adjusting the duty cycle, thereby regulating the airflow.

[0106] The fourth step is adaptive adjustment and optimization. During the execution of the active or hybrid ventilation strategy, the rate of humidity change is continuously monitored, and the fan power is dynamically adjusted through a closed-loop feedback control unit. Preferably, the dynamic adjustment includes: automatically reducing the fan power by one level when the humidity drops by more than 10%RH within 1 minute (too rapid improvement); increasing the fan power by one level when the humidity drops by less than 5%RH within 2 minutes (insufficient improvement); and sending a prompt to the user to clean the device or adjust its wearing condition through the human-machine interface if the humidity fails to drop below the safe level of 75%RH after operating at the highest power level for more than 10 minutes (continuously ineffective).

[0107] This module not only actively keeps the contact surface dry and comfortable, but also ensures that the information acquisition and processing module can continuously obtain high-quality data.

[0108] Human-computer interaction module: used to display the user's real-time physiological parameters, health status information and warning prompts, and to receive user input.

[0109] It is important to note that this module includes: multimodal information display management, which dynamically manages interface elements through a display content scheduler; real-time values ​​of core physiological parameters are displayed under normal conditions; when the data monitoring module triggers an L1 / L2 level warning, a warning icon is displayed in the status bar while keeping the main interface unchanged; when an L3 level or higher warning is triggered, a full-screen priority display mechanism is activated, using a high-contrast color scheme to display warning details and recommended measures.

[0110] User input acquisition and processing. The input signal fusion processor uniformly parses multi-channel commands, including touchscreen gestures, physical buttons, and voice input. Specifically, this includes: receiving basic information (such as age, gender, height, weight, and underlying medical conditions) input by the user through a guided interface during initial device use; receiving real-time control commands from the user regarding monitoring functions, including instant measurement commands and function on / off switching; and receiving confirmation and feedback from the user regarding warning signals.

[0111] Interaction scenarios are adaptive. When the micro-environment management module detects high skin surface humidity, it automatically increases the priority of physical buttons and voice input; when the data monitoring module determines that the user is in a running or other exercise state, it activates the anti-mistouch algorithm and expands the touch target area; during nighttime hours, it automatically switches to a simplified interaction mode to reduce interaction complexity.

[0112] Intelligent feedback and guidance, through a multimodal feedback generator coordinating vibration, audio, and visual cues: routine operation feedback uses short vibrations combined with subtle interface animations; important warnings use specific vibration patterns (such as three short vibrations followed by one long one) combined with voice announcements. A progressive guidance process is provided for new users, guiding them through the device wearing, personalized information entry, and initial calibration procedures.

[0113] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended embodiments and their equivalents.

Claims

1. A remote real-time monitoring system for blood pressure and blood lipids based on edge computing, characterized in that, include: Information acquisition and processing module: used to acquire dynamic photon observation data of the user's subcutaneous tissue through a time-domain resolved multispectral photon meter; On the edge side, based on a parameterized general biological tissue optical model and a bio-physical inverse solver, a digital twin model of the user's personal health is generated through an inverse optimization process, and the user's real-time physiological parameters are solved. Data monitoring module: Used at the edge to analyze and identify anomalies in users' real-time physiological parameters based on a data monitoring intelligent model, and trigger early warning signals; Wireless transceiver module: used to transmit early warning signals and related physiological parameters generated at the edge to the remote monitoring platform, and to receive instructions from the platform; Microenvironment management module: used to sense the user's skin microenvironment in real time, dynamically determine breathability needs, and intelligently control the switching and coordination of active and passive heat dissipation mechanisms; Human-computer interaction module: used to display the user's real-time physiological parameters, health status information and warning prompts, and to receive user input operations; The method for constructing the user's personal health digital twin model includes: In M1, during initial user use and calibration mode, photon dynamic observation data are acquired in a resting state. M2. Call the bio-physical inverse solver and use the parameterized general biological tissue optical model as the forward model. The model parameters include structural parameters and physiological parameters. The structural parameters include melanin content, subcutaneous fat layer thickness, blood vessel wall elasticity, and basic optical properties of blood. The physiological parameters include intravascular pressure, blood flow velocity, and blood lipid concentration. M3, preset structural parameters and physiological parameters, execute the first reverse optimization process to obtain the user's personal structural parameters; M4 stores the user's personal structural parameters and combines the parameterized general biological tissue optical model with the user's personal structural parameters to generate a digital twin model of the user's personal health.

2. The remote real-time monitoring system for blood pressure and blood lipids based on edge computing according to claim 1, characterized in that: The preset values ​​of the physiological parameters are within the range of physiologically reasonable resting prior values; The steps of the first reverse optimization process include: N1. Input the preset structural and physiological parameters into the forward model to calculate the dynamic prediction data of photons; N2. Compare the photon dynamic observation data in the resting state with the photon dynamic prediction data, calculate the difference between the two through the loss function, and adjust the structural parameters. N3. Determine whether the optimization process has reached the first stopping condition; N31. If the stopping condition is met, the optimization will terminate and the user's personal structure parameters will be obtained. N32. If the stopping condition is not met, repeat the above steps.

3. The remote real-time monitoring system for blood pressure and blood lipids based on edge computing according to claim 1, characterized in that: The methods for solving the user's real-time physiological parameters include: P1. Call the bio-physical inverse solver and use the user's personal health digital twin model as the forward model; P2. Real-time acquisition of user photon dynamic observation data; P3. Preset physiological parameters, execute the second reverse optimization process, dynamically adjust the physiological parameters, and obtain the user's real-time physiological parameters.

4. The remote real-time monitoring system for blood pressure and blood lipids based on edge computing according to claim 3, characterized in that: The steps of the second reverse optimization process include: Q1. Input the preset physiological parameters into the forward model to calculate the real-time photon dynamic prediction data; Q2. Compare the real-time acquired photon dynamic observation data with the real-time photon dynamic prediction data, calculate the difference between the two through the loss function, and adjust the physiological parameters accordingly. Q3. Determine whether the difference meets the second stopping condition; Q31. If the stopping condition is met, the optimization will terminate and the user's real-time physiological parameters will be obtained. Q32. If the stopping condition is not met, repeat the above steps.

5. The remote real-time monitoring system for blood pressure and blood lipids based on edge computing according to claim 1, characterized in that: The methods used by the intelligent data monitoring model to analyze and detect anomalies in users' real-time physiological parameters include: R1 receives real-time physiological parameters from the user, determines the user's current activity status, and adds a timestamp and activity status label to each set of data. R2. Call the user's personal health digital twin model to obtain the expected range of physiological parameters for the user in the current state and generate a dynamic baseline; R3. Compare the user's real-time physiological parameters with the dynamic baseline, calculate the degree of deviation, and generate an instantaneous abnormality score based on the magnitude and direction of the deviation. R4. Perform trend health assessment on real-time physiological parameters at multiple time scales to obtain trend health scores; R5. Check whether there are any abnormalities in the correlation between physiological parameters and obtain the correlation analysis results; R6. Based on the combined instantaneous anomaly score, trend health score, and correlation analysis results, a final anomaly judgment is made on the real-time physiological parameters, and a graded response is triggered.

6. The remote real-time monitoring system for blood pressure and blood lipids based on edge computing according to claim 1, characterized in that: The microenvironment management module specifically includes: T1. Real-time temperature and humidity data of the user's skin contact surface are collected through a biomimetic microarray base; T2. Determine the risk level of sweat accumulation based on temperature and humidity data and the user's current condition; T3. Select strategies based on the risk level of sweat buildup, including passive ventilation, mixed ventilation, and active ventilation strategies. T4. Monitor temperature and humidity changes in real time and make adaptive adjustments.

7. The remote real-time monitoring system for blood pressure and blood lipids based on edge computing according to claim 1, characterized in that: The time-resolved multispectral photon meter includes a multi-frequency coherent photon emission source and a photon detector array; the photon dynamic observation data includes photon flight time distribution, spatial correlation of photon count rate, and microsecond-level fluctuations in the polarization state of the scattered field; the bio-physical inverse solver integrates Monte Carlo simulation and deep learning models.

8. A remote real-time monitoring device for blood pressure and blood lipids based on edge computing, characterized in that: The system described in any one of claims 1-7 is a remote real-time monitoring system for blood pressure and blood lipids based on edge computing.

Images