Thrombus prevention intelligent equipment linkage monitoring quality control method and system based on Internet of Things

Abstract

The invention provides a thrombus prevention intelligent equipment linkage monitoring quality control method and system based on the Internet of Things, and relates to the technical field of the Internet of Things, and the method comprises the steps: obtaining the physiological state data of a target monitoring object, carrying out the analysis processing of the data based on a preset thrombus risk assessment model, and determining the thrombus risk level. The model is established according to an association relationship between limb activity state data and blood circulation state data; matching a corresponding equipment linkage control strategy from a preset equipment control strategy library according to the thrombus risk level; according to a cooperative control relationship defined in the strategy, generating a linkage control instruction sequence for the plurality of Internet of Things devices; and finally, distributing the instruction sequence to a plurality of Internet of Things devices, and driving the devices to execute thrombus prevention intervention operation according to the time sequence dependency relationship and the execution condition constraint relationship. According to the intelligent thrombus prevention and monitoring system, intelligent thrombus prevention and monitoring of multi-device cooperative linkage are realized, and the accuracy and timeliness of prevention and intervention are improved.

Description

Technical Field

[0001] This invention relates to the field of Internet of Things (IoT) technology, and in particular to a method and system for quality control and linkage monitoring of smart devices for thrombosis prevention based on IoT. Background Technology

[0002] Thrombotic diseases are a common and serious health threat in clinical practice, especially for special populations such as those who are bedridden for extended periods, recovering from surgery, or suffering from chronic diseases in the elderly, where the incidence of complications such as deep vein thrombosis and pulmonary embolism is significantly increased. Traditional thrombosis prevention measures mainly rely on regular rounds by healthcare professionals and manual operation of preventative equipment, but this approach has significant limitations in terms of monitoring continuity, timely intervention, and resource utilization efficiency. With the rapid development of IoT, smart sensing, and big data analytics technologies, it has become possible to network and integrate various intelligent monitoring and intervention devices to build an intelligent thrombosis prevention and monitoring system. By collecting patients' physiological data in real time, combining it with intelligent analysis algorithms to assess thrombosis risk, and automatically controlling various preventative devices to perform intervention operations, the initiative and accuracy of thrombosis prevention can be effectively improved, reducing the workload of healthcare professionals and lowering the incidence of thrombotic diseases.

[0003] However, existing thrombosis prevention technologies still have many shortcomings. First, most existing monitoring systems operate independently with a single device, lacking effective data exchange and collaborative mechanisms between monitoring and intervention devices. This prevents dynamic adjustments to the collaborative working strategies of multiple devices based on the patient's real-time status, resulting in insufficient targeting and adaptability of preventive interventions and hindering precise, personalized prevention. Second, current technologies primarily rely on single-dimensional physiological indicators or simple threshold judgment methods for thrombosis risk assessment, failing to fully explore the deep correlation between limb activity and blood circulation status. This limits the accuracy and predictability of risk assessment, making it prone to missed or false alarms and affecting the timeliness of preventive interventions. Summary of the Invention

[0004] This invention provides a method and system for monitoring and controlling the linkage of intelligent devices for thrombosis prevention based on the Internet of Things, which can solve the problems in the prior art.

[0005] A first aspect of this invention provides a method for quality control and monitoring of thrombosis prevention smart devices based on the Internet of Things, comprising: Physiological status data associated with the target monitored object is acquired, and the physiological status data is analyzed and processed based on a preset thrombosis risk assessment model to determine the thrombosis risk level of the target monitored object. The thrombosis risk assessment model is established based on the correlation between the limb activity status data and the blood circulation status data. Based on the thrombosis risk level, a corresponding device linkage control strategy is matched from a preset device control strategy library. The device control strategy library includes multiple sets of device linkage control strategies for different thrombosis risk levels, and each set of device linkage control strategies defines the collaborative control relationship between multiple IoT devices. Based on the collaborative control relationship defined in the device linkage control strategy, a linkage control instruction sequence for multiple IoT devices is generated; the linkage control instruction sequence is distributed to the multiple IoT devices, driving the multiple IoT devices to perform thrombosis prevention intervention operations according to the time sequence dependency relationship and the execution condition constraint relationship.

[0006] The physiological state data are analyzed and processed based on a pre-defined thrombosis risk assessment model to determine the thrombosis risk level of the target monitored individual, including: Feature extraction is performed on the limb activity state data to obtain a limb activity feature vector, which represents the limb movement pattern of the target monitored object within a preset time window; Feature extraction is performed on the blood circulation status data to obtain a blood circulation feature vector, which characterizes the blood flow status of the target monitored object within the preset time window; The limb activity feature vector and the blood circulation feature vector are fused to obtain a fused feature vector. The fusion process is based on the physiological correlation between the limb activity state data and the blood circulation state data. The fused feature vector is input into the thrombosis risk assessment model, and the thrombosis risk assessment model outputs a thrombosis risk score based on the matching relationship between the fused feature vector and the preset thrombosis risk determination rules. The thrombosis risk level is determined based on the thrombosis risk score and a preset risk level classification standard, wherein the risk level classification standard maps the thrombosis risk score to the corresponding thrombosis risk level.

[0007] The thrombosis risk assessment model outputs a thrombosis risk score based on the matching relationship between the fused feature vector and the preset thrombosis risk determination rules, including: The fusion feature vector is subjected to dimensionality analysis to extract multiple feature components associated with thrombosis risk. Each feature component corresponds to a different physiological state dimension. The weighted values ​​of each feature component are obtained by weighting the multiple feature components with the preset feature weight matrix in the thrombosis risk determination rule. The weighted values ​​of each feature component are summed to obtain the initial risk score; Historical risk event sequences and historical physiological state sequences are extracted from the historical thrombosis risk data of the target monitored object. The frequency and time interval of risk events in the historical risk event sequences are statistically analyzed, and the occurrence frequency of risk events is calculated. After normalizing the occurrence frequency of risk events, it is used as the risk trend intensity value. The risk trend intensity value is then compared with a preset correction coefficient benchmark value to obtain the risk correction coefficient. The initial risk score is multiplied by the risk correction coefficient to obtain the corrected risk score. The corrected risk score is then weighted and summed with a dynamic adjustment factor determined based on the current physiological state of the target monitored object to obtain the thrombosis risk score.

[0008] Based on the thrombosis risk level, a corresponding device linkage control strategy is matched from a preset device control strategy library. The device control strategy library includes multiple sets of device linkage control strategies for different thrombosis risk levels. Each set of device linkage control strategies defines the collaborative control relationship between multiple IoT devices, including: Using the thrombosis risk level as a search condition, a set of candidate device linkage control strategies corresponding to the thrombosis risk level is retrieved from the device control strategy library. The set of candidate device linkage control strategies includes at least one set of device linkage control strategies. Obtain the current device configuration information of the target monitored object, calculate the matching degree between the device functional requirements of each group of device linkage control strategies and the device configuration information, and obtain the matching degree score of each group of device linkage control strategies; Based on the matching score, the device linkage control strategy with the highest matching score is selected from the candidate device linkage control strategy set as the target device linkage control strategy; The device collaborative control relationship data structure is obtained by parsing the target device linkage control strategy. The device collaborative control relationship data structure defines the triggering relationship and execution dependency relationship between each IoT device participating in the linkage.

[0009] Based on the collaborative control relationships defined in the device linkage control strategy, a sequence of linkage control instructions for multiple IoT devices is generated, including: The device trigger sequence and device execution sequence are obtained from the device linkage control strategy. The device trigger sequence defines the triggering order and triggering conditions of each IoT device, and the device execution sequence defines the execution actions and execution parameters of each IoT device. A temporal dependency graph is constructed based on the device triggering sequence. The temporal dependency graph uses each IoT device as a node and the triggering order and triggering condition as directed edges to represent the temporal triggering dependency relationship between each IoT device. The temporal dependency graph is topologically sorted to obtain a device triggering time chain. The device triggering time chain determines the execution order of each IoT device according to the temporal triggering dependency. Based on the execution actions and execution parameters defined in the device execution sequence, corresponding control instructions are generated for each IoT device in the device triggering timing chain. The control instructions include device identifier, execution action instruction code, and execution parameter value.

[0010] The method further includes: Collect device operation data and intervention response data generated by the multiple IoT devices during the thrombosis prevention intervention operation; The thrombosis risk assessment model is dynamically calibrated based on the equipment operation data and the intervention response data, and the thrombosis risk level of the target monitored object is re-determined according to the calibrated thrombosis risk assessment model, thus forming a closed-loop quality control mechanism.

[0011] A second aspect of the present invention provides an Internet of Things-based intelligent device linkage monitoring and quality control system for thrombosis prevention, comprising: The first unit is used to acquire physiological state data associated with the target monitored object, analyze and process the physiological state data based on a preset thrombosis risk assessment model, and determine the thrombosis risk level of the target monitored object. The thrombosis risk assessment model is established based on the correlation between the limb activity state data and the blood circulation state data. The second unit is used to match the corresponding device linkage control strategy from a preset device control strategy library according to the thrombosis risk level. The device control strategy library includes multiple sets of device linkage control strategies for different thrombosis risk levels, and each set of device linkage control strategies defines the collaborative control relationship between multiple Internet of Things devices. The third unit is used to generate a sequence of linkage control instructions for multiple IoT devices based on the collaborative control relationship defined in the device linkage control strategy; distribute the sequence of linkage control instructions to the multiple IoT devices, and drive the multiple IoT devices to perform thrombosis prevention intervention operations according to the time sequence dependency relationship and the execution condition constraint relationship.

[0012] A third aspect of the present invention, An electronic device is provided, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0013] Fourth aspect of the embodiments of the present invention, A computer-readable storage medium is provided, having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0014] The beneficial effects of this application are as follows: This invention establishes a thrombosis risk assessment model based on the correlation between limb activity status data and blood circulation status data. This model can comprehensively analyze the physiological status data of the target monitored object, accurately determine the thrombosis risk level, realize intelligent and precise assessment of thrombosis risk, and improve the accuracy and timeliness of thrombosis risk warning.

[0015] This invention stores multiple sets of device linkage control strategies for different thrombosis risk levels through a preset device control strategy library. Each set of strategies defines the collaborative control relationship between multiple IoT devices. The corresponding device linkage control strategy is automatically matched according to the thrombosis risk level, realizing intelligent collaborative work of multiple devices, avoiding the problem of limited intervention effect of a single device, and improving the overall effect of thrombosis prevention and intervention.

[0016] This invention generates a sequence of linkage control instructions and distributes them to multiple Internet of Things (IoT) devices, driving the devices to perform thrombosis prevention intervention operations according to time-series dependencies and execution condition constraints. This automates and standardizes the prevention intervention process, reduces the workload of manual intervention and human error, ensures the timeliness and effectiveness of thrombosis prevention measures, and improves the overall level of monitoring and quality control. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the IoT-based intelligent device linkage monitoring and quality control method for thrombosis prevention, as described in an embodiment of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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.

[0019] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0020] Figure 1 This is a flowchart illustrating the IoT-based intelligent device linkage monitoring and quality control method for thrombosis prevention according to an embodiment of the present invention. Figure 1As shown, the method includes: Physiological status data associated with the target monitored object is acquired, and the physiological status data is analyzed and processed based on a preset thrombosis risk assessment model to determine the thrombosis risk level of the target monitored object. The thrombosis risk assessment model is established based on the correlation between the limb activity status data and the blood circulation status data. Based on the thrombosis risk level, a corresponding device linkage control strategy is matched from a preset device control strategy library. The device control strategy library includes multiple sets of device linkage control strategies for different thrombosis risk levels, and each set of device linkage control strategies defines the collaborative control relationship between multiple IoT devices. Based on the collaborative control relationship defined in the device linkage control strategy, a linkage control instruction sequence for multiple IoT devices is generated; the linkage control instruction sequence is distributed to the multiple IoT devices, driving the multiple IoT devices to perform thrombosis prevention intervention operations according to the time sequence dependency relationship and the execution condition constraint relationship.

[0021] In one optional implementation, the physiological state data is analyzed and processed based on a preset thrombosis risk assessment model to determine the thrombosis risk level of the target monitored object, including: Feature extraction is performed on the limb activity state data to obtain a limb activity feature vector, which represents the limb movement pattern of the target monitored object within a preset time window; Feature extraction is performed on the blood circulation status data to obtain a blood circulation feature vector, which characterizes the blood flow status of the target monitored object within the preset time window; The limb activity feature vector and the blood circulation feature vector are fused to obtain a fused feature vector. The fusion process is based on the physiological correlation between the limb activity state data and the blood circulation state data. The fused feature vector is input into the thrombosis risk assessment model, and the thrombosis risk assessment model outputs a thrombosis risk score based on the matching relationship between the fused feature vector and the preset thrombosis risk determination rules. The thrombosis risk level is determined based on the thrombosis risk score and a preset risk level classification standard, wherein the risk level classification standard maps the thrombosis risk score to the corresponding thrombosis risk level.

[0022] After activation, the thrombosis risk assessment system continuously collects physiological data from the monitored subject using its sensor units. Limb activity data is acquired through triaxial accelerometers installed on the subject's lower leg, thigh, and ankle, recording acceleration changes along each axis at a sampling frequency of 50 times per second. Within a preset time window of 2 hours, the system accumulates 360,000 raw data points. Blood circulation data is acquired using a photoplethysmography (PPG) sensor and a Doppler flow detector. The PPG sensor is attached to the dorsalis pedis artery, collecting pulse waveform data 100 times per second, while the Doppler flow detector is installed in the popliteal vein region of the lower leg, recording blood flow velocity values ​​20 times per second.

[0023] The extraction process of limb activity feature vectors is carried out based on the collected acceleration data. The system divides the 2-hour time window into 240 sub-windows, each 30 seconds long, and performs statistical analysis on the acceleration data within each sub-window. The average, maximum, minimum, standard deviation, and rate of change of acceleration values ​​are calculated for the X, Y, and Z axes. For a specific sub-window, the average X-axis acceleration is 0.15 m / s², the maximum is 1.23 m / s², the minimum is -0.08 m / s², the standard deviation is 0.34 m / s², and the rate of change, expressed as the sum of the differences between adjacent sampling points, is 18.7 m / s². The system further extracts the amplitude features of the composite acceleration, obtaining the composite acceleration by taking the square root of the sum of the squares of the three-axis accelerations. The average amplitude of the composite acceleration in this sub-window is 0.89 m / s². The system counted 127 instances within this sub-window where the acceleration amplitude exceeded the threshold of 0.5 m / s², representing 8.5% of the total sampling points in that window. The system also extracted the zero-crossing rate of the acceleration signal, i.e., the number of times the signal changes from positive to negative or vice versa; the zero-crossing rate for this sub-window was 3.2 times per second. The feature values ​​from all 240 sub-windows were arranged chronologically to form a limb activity feature vector containing 5760 elements, which comprehensively represents the limb movement patterns of the monitored subject over a two-hour period.

[0024] The extraction of blood circulation feature vectors is performed on pulse waveform data and blood flow velocity data. The pulse waveform data contains 3000 sampling points within each 30-second sub-window. The system identifies each complete pulse wave cycle and extracts the peak amplitude, trough amplitude, rise time, and fall time. Within one sub-window, 52 complete pulse wave cycles were detected. The average peak amplitude of these cycles was 2.8 volts, the average trough amplitude was 0.6 volts, the average rise time was 0.18 seconds, and the average fall time was 0.45 seconds. The system calculates the peak-to-peak value of the pulse wave, i.e., the difference between the peak and trough. The average peak-to-peak value for this sub-window was 2.2 volts. The system also extracts the slope features of the pulse waveform. The rising slope is calculated by dividing the peak-to-peak value by the rise time. The average rising slope for this sub-window was 12.2 volts per second. The falling slope is calculated by dividing the peak-to-peak value by the fall time. The average falling slope for this sub-window was 4.9 volts per second. The blood flow velocity data within this sub-window includes 600 sampling points. The system calculated the average blood flow velocity to be 8.3 cm / s, the maximum to be 14.7 cm / s, the minimum to be 3.1 cm / s, and the standard deviation to be 2.6 cm / s. The system extracts the fluctuation index of the blood flow velocity, which is the ratio of the standard deviation to the mean. The fluctuation index for this sub-window is 0.31. The system also calculates the percentage of time the blood flow velocity is below a preset threshold of 5 cm / s. In this sub-window, the duration of low-velocity blood flow is 8 seconds, accounting for 26.7%. The circulatory feature values ​​of all 240 sub-windows are arranged in chronological order to form a blood circulation feature vector containing 4320 elements.

[0025] The fusion processing module receives limb activity feature vectors and blood circulation feature vectors and performs correlation analysis. The physiological correlation law established by the system is based on the medical principle that limb activity directly affects blood circulation. For each sub-window, the system pairs the limb activity intensity index with the blood circulation activity index. Limb activity intensity is represented by the product of the average amplitude of synthetic acceleration and the proportion of activity; the activity intensity value of a certain sub-window is 0.89 multiplied by 0.085, which equals 0.076. Blood circulation activity is represented by the product of the average blood flow velocity and the slope of the pulse wave; the circulation activity of this sub-window is 8.3 multiplied by 12.2, which equals 101.3. The system calculates the ratio of activity intensity to circulation activity as a coordination index; the coordination index of this sub-window is 101.3 divided by 0.076, which equals 1333. The system also calculates the changes in activity intensity and cyclic activity between adjacent sub-windows. If the activity intensity of the previous sub-window is 0.092 and the current sub-window is 0.076, the change is -0.016. If the cyclic activity of the previous sub-window is 118.5 and the current is 101.3, the change is -17.2. The system extracts the time delay feature between the two changes. If the cyclic activity decreases after a delay of 2 sub-windows (1 minute), this delay value is recorded as 1 minute. The fusion processing also includes the identification of abnormal states. If the activity intensity of a sub-window is close to zero (less than 0.01) but the cyclic activity is significantly low (less than 50), the system marks the window as a high-risk state. The fused feature vector integrates the original features and associated features, containing a total of 12,000 elements.

[0026] Long-term features were extracted, and the total number of sub-windows with activity intensity below 0.02 within the entire 2-hour window was 156, accounting for 65%, indicating that the monitored subject was in a prolonged state of inactivity. The model statistically identified 143 sub-windows with cyclic activity consistently below 60, accounting for 59.6%. The model calculated 128 sub-windows where both activity intensity and cyclic activity were low, accounting for 53.3%. Trend features were extracted, dividing the 2-hour period into four 30-minute segments. The average activity intensity for each segment was calculated to be 0.082, 0.071, 0.063, and 0.058, showing a continuous downward trend with a decrease of 0.008 per segment. The average cyclic activity for each segment was 106.3, 95.7, 88.4, and 81.2, also showing a downward trend with a decrease of approximately 8.5 per segment. The model scores each item according to preset judgment rules: a long-term quiescent state accounting for more than 60% of the total score is 30 points; a persistently low circulatory activity accounting for more than 50% of the total score is 28 points; the simultaneous occurrence of two low values ​​accounting for more than 50% of the total score is 25 points; a persistently declining trend in activity intensity accounts for 12 points; a persistently declining trend in circulatory activity accounts for 15 points; a delayed response time of more than 1 minute accounts for 8 points; and 17 high-risk sub-windows are detected, accounting for 17 points. The model sums up the scores of each item to obtain a thrombosis risk score of 135 points.

[0027] The risk grading standard maps the thrombosis risk score to five levels: 0-40 for low risk, 41-80 for lower risk, 81-120 for moderate risk, 121-160 for higher risk, and 161-200 for high risk. The monitored subject's score of 135 falls within the higher risk range. The system outputs this level as the final thrombosis risk assessment result and triggers the corresponding early warning mechanism and intervention recommendation generation process.

[0028] In one optional implementation, the thrombosis risk assessment model outputs a thrombosis risk score based on the matching relationship between the fused feature vector and a preset thrombosis risk determination rule, including: The fusion feature vector is subjected to dimensionality analysis to extract multiple feature components associated with thrombosis risk. Each feature component corresponds to a different physiological state dimension. The weighted values ​​of each feature component are obtained by weighting the multiple feature components with the preset feature weight matrix in the thrombosis risk determination rule. The weighted values ​​of each feature component are summed to obtain the initial risk score; Historical risk event sequences and historical physiological state sequences are extracted from the historical thrombosis risk data of the target monitored object. The frequency and time interval of risk events in the historical risk event sequences are statistically analyzed, and the occurrence frequency of risk events is calculated. After normalizing the occurrence frequency of risk events, it is used as the risk trend intensity value. The risk trend intensity value is then compared with a preset correction coefficient benchmark value to obtain the risk correction coefficient. The initial risk score is multiplied by the risk correction coefficient to obtain the corrected risk score. The corrected risk score is then weighted and summed with a dynamic adjustment factor determined based on the current physiological state of the target monitored object to obtain the thrombosis risk score.

[0029] After receiving the fused feature vector, the thrombosis risk assessment model initiates a dimension parsing procedure. This procedure decomposes the fused feature vector according to a predefined dimension mapping table. The fused feature vector contains 512 numerical elements, each ranging from 0 to 1. The dimension mapping table divides these 512 dimensions into 18 physiological state dimension groups, each containing several continuous or discrete dimension indices. The blood flow velocity dimension group corresponds to dimensions 1 to 28, the blood viscosity dimension group to dimensions 29 to 56, the coagulation factor concentration dimension group to dimensions 57 to 92, the platelet aggregation rate dimension group to dimensions 93 to 120, and the vessel wall integrity dimension group to dimensions 121 to 156. The parsing procedure reads the values ​​of the fused feature vector within each dimension group and calculates the mean of all dimension values ​​within each group as the feature component of that physiological state dimension. The fusion feature vector of a certain monitored object has values ​​of 0.62, 0.58, 0.65, 0.61, etc. in the 28 dimensions of the blood flow velocity dimension group, and the feature component of this dimension group is calculated to be 0.614.

[0030] The extracted feature components are stored in a feature component vector, which contains 18 elements, corresponding to 18 physiological state dimensions. The thrombosis risk assessment rule incorporates a feature weight matrix, a symmetric 18x18 matrix. Diagonal elements represent the base weight values ​​for each dimension, while off-diagonal elements represent the interaction weight values ​​between different dimensions. The base weight value for blood flow velocity is set to 0.15, for blood viscosity to 0.18, for coagulation factor concentration to 0.12, and for platelet aggregation rate to 0.14. The interaction weight value between blood flow velocity and blood viscosity is set to 0.08, representing the synergistic effect on thrombosis risk when both are abnormal. The weighted calculation process involves multiplying each element in the feature component vector by all elements in the corresponding row of the weight matrix and then summing the results. Multiplying the blood flow velocity feature component 0.614 by its basic weight 0.15 yields 0.0921. Adding the sum of the products of the interaction weights of this feature component with other dimensions, we get the weighted value of the blood flow velocity dimension, 0.137.

[0031] After weighted calculation of all 18 feature components, 18 weighted values ​​are obtained. A summation operation is then performed to accumulate these 18 weighted values. The weighted values ​​for a monitored subject are, in order: 0.137, 0.142, 0.089, 0.096, 0.104, 0.078, 0.065, 0.091, 0.083, 0.072, 0.058, 0.063, 0.071, 0.049, 0.054, 0.047, 0.041, and 0.038. The summation yields an initial risk score of 1.378. This initial risk score reflects the monitored subject's current static thrombosis risk level.

[0032] The historical thrombosis risk data retrieval module extracts the complete monitoring records of the target subject for the past 90 days from the database. The historical risk event sequence records all time points marked as high-risk. One monitored subject experienced 12 high-risk events within 90 days, with timestamps of days 3, 8, 15, 19, 27, 34, 41, 48, 56, 63, 72, and 84. The statistical program calculates the time interval between two adjacent risk events, obtaining 11 interval values: 5 days, 7 days, 4 days, 8 days, 7 days, 7 days, 7 days, 8 days, 7 days, 9 days, and 12 days. The frequency of risk events is 12, and the total monitoring duration is 90 days; dividing these two values ​​gives a risk event frequency of 0.133 times per day.

[0033] Normalization maps the frequency of risk events to a standard range of 0 to 1. A reference upper limit for the frequency of risk events is set at 0.2 times per day. The actual frequency of 0.133 is divided by the reference upper limit of 0.2, yielding a normalized value of 0.665, which is the risk trend strength value. The preset correction coefficient baseline value is determined based on large-scale clinical data statistics and set at 0.5. The risk trend strength value of 0.665 is compared with the correction coefficient baseline value of 0.5 to obtain a risk correction coefficient of 1.33. A coefficient greater than 1 indicates that the monitored subject has a significant risk recurrence trend, requiring an upward adjustment of the initial risk score.

[0034] The corrected risk score was obtained by multiplying the initial risk score of 1.378 by a risk correction coefficient of 1.33, resulting in a score of 1.833. The dynamic adjustment factor was determined based on the real-time changes in the monitored subject's current physiological state. The real-time acquisition module obtained physiological parameters such as heart rate variability, blood pressure fluctuation amplitude, and body temperature change rate over the past 5 minutes. The heart rate variability was 68 milliseconds, below the normal range of 80 to 120 milliseconds, indicating weakened autonomic nervous system regulation. The blood pressure fluctuation amplitude was 18 mmHg, exceeding the normal fluctuation range by 10 mmHg. The body temperature change rate was 0.3 degrees Celsius per hour, exceeding the normal rate of 0.1 degrees Celsius per hour. According to the preset dynamic adjustment rule table, a decrease in heart rate variability corresponds to an adjustment weight of 0.12, an increase in blood pressure fluctuation corresponds to an adjustment weight of 0.15, and a faster body temperature change corresponds to an adjustment weight of 0.08. The weighted sum of these three factors yielded a dynamic adjustment factor of 0.35.

[0035] The weighted summation calculation combines the corrected risk score of 1.833 with the dynamic adjustment factor of 0.35 according to a preset weight ratio. The weight coefficient for the corrected risk score is set to 0.85, and the weight coefficient for the dynamic adjustment factor is set to 0.15, ensuring that their sum is 1. Multiplying the corrected risk score of 1.833 by the weight coefficient of 0.85 yields 1.558, and multiplying the dynamic adjustment factor of 0.35 by the weight coefficient of 0.15 yields 0.053. The two products are added together to obtain the final thrombosis risk score of 1.611. This score falls within the medium-to-high risk range on a scale of 0 to 3, triggering the system to send an early warning notification to medical staff, recommending in-depth thrombosis screening for the monitored individual and adjustment of the anticoagulation therapy regimen.

[0036] In one optional implementation, a corresponding device linkage control strategy is matched from a preset device control strategy library according to the thrombosis risk level. The device control strategy library includes multiple sets of device linkage control strategies for different thrombosis risk levels. Each set of device linkage control strategies defines the collaborative control relationship between multiple IoT devices, including: Using the thrombosis risk level as a search condition, a set of candidate device linkage control strategies corresponding to the thrombosis risk level is retrieved from the device control strategy library. The set of candidate device linkage control strategies includes at least one set of device linkage control strategies. Obtain the current device configuration information of the target monitored object, calculate the matching degree between the device functional requirements of each group of device linkage control strategies and the device configuration information, and obtain the matching degree score of each group of device linkage control strategies; Based on the matching score, the device linkage control strategy with the highest matching score is selected from the candidate device linkage control strategy set as the target device linkage control strategy; The device collaborative control relationship data structure is obtained by parsing the target device linkage control strategy. The device collaborative control relationship data structure defines the triggering relationship and execution dependency relationship between each IoT device participating in the linkage.

[0037] In a thrombosis risk monitoring system, after determining the thrombosis risk level of the target monitored object, the system needs to retrieve the appropriate device linkage control strategy from a pre-established device control strategy library. The device control strategy library is established during the system initialization phase. This library is stored hierarchically according to thrombosis risk levels, including four main categories: low risk, medium risk, high risk, and extremely high risk. Each risk level stores multiple sets of device linkage control strategies. Each strategy is defined in a structured data format, containing core data fields such as strategy identifier, applicable risk level range, required device type list, device functional requirements description, device trigger relationship definition, and execution dependency configuration.

[0038] The system uses the currently assessed thrombosis risk level as the primary search criterion for strategy matching. Assuming the target monitored individual is assessed as high-risk, the system locates the high-risk level category in the device control strategy library and extracts all strategy records under that category. In practice, the high-risk level may include five different sets of device linkage control strategies: Strategy A, Strategy B, Strategy C, Strategy D, and Strategy E. These five sets of strategies constitute a candidate set of device linkage control strategies, each designed differently for different device configuration scenarios. Strategy A may require the configuration of four types of devices: a smart mattress pressure sensor, a lower limb intermittent compression device, a smart infusion monitor, and a nursing station alarm terminal. Strategy B may require the configuration of four types of devices: a smart mattress pressure sensor, a wearable activity monitor, a smart lighting system, and a nursing station alarm terminal. Other strategies also define their respective combinations of device functional requirements.

[0039] After obtaining the candidate strategy set, the system needs to read the current actual device configuration information of the target monitored object. This device configuration information is stored in the monitored object's device binding record table, recording all IoT devices currently associated with the object and their functional attributes. For example, suppose the device configuration information of a monitored object shows that it is currently configured with one smart mattress pressure sensor, one lower limb intermittent compression device, one smart infusion monitor, and one access port for a nursing station alarm terminal. The system converts this device configuration information into a device functional feature vector for subsequent matching degree calculation.

[0040] For each set of candidate strategies, the system performs a matching degree calculation. The matching degree calculation process includes two levels: device type matching verification and device functional completeness verification. In the device type matching verification, the system compares the device types required by the strategy with the actual device types configured. Taking strategy A as an example, this strategy requires four types of devices: a smart mattress pressure sensor, a lower limb intermittent pressure device, a smart infusion monitor, and a nursing station alarm terminal. The actual configuration perfectly matches all four types of devices, thus the device type matching item receives full marks. In the device functional completeness verification, the system further checks whether the specific functional attributes of each type of device meet the requirements of strategy execution. Strategy A requires the smart mattress pressure sensor to have pressure distribution monitoring, body position change recognition, and a prolonged resting state alarm function. The system checks the functional attribute table of the actually configured smart mattress pressure sensor and confirms that the device supports the above three functions, thus confirming the device's functional completeness.

[0041] By performing type matching and functional completeness checks on each of the four types of devices required by Strategy A, the system statistically shows that the device type matching rate and device functional completeness rate for this strategy are both 100%. The system calculates the overall matching score for this strategy according to preset scoring rules, with device type matching accounting for 60% of the weight and device functional completeness accounting for 40%. After weighted calculation, the overall matching score for Strategy A is 98 points. The same method is used to calculate the matching score for Strategy B. Because the wearable activity monitor required by Strategy B is missing in the actual device configuration, the device type matching rate is lower, and the final overall matching score for Strategy B is only 65 points. The matching scores for Strategy C, Strategy D, and Strategy E are calculated to be 72, 58, and 61 points, respectively.

[0042] The system compares and ranks the matching scores of all strategies in the candidate strategy set, identifying the strategy with the highest matching score as the target device linkage control strategy. In this case, strategy A has the highest matching score of 98, therefore the system selects strategy A as the final target device linkage control strategy. After selecting the target strategy, the system extracts the device collaborative control relationship data structure from the strategy's data records. This data structure is organized in the form of a directed graph, where nodes represent the various IoT devices participating in the linkage, and directed edges represent the triggering relationships and execution dependencies between devices.

[0043] For the device collaborative control relationship data structure of Strategy A, the smart mattress pressure sensor is defined as the main triggering source device. When the smart mattress pressure sensor detects that the monitored subject has maintained the same body position for more than two hours, the sensor triggers the first-level linkage event. The execution object of the first-level linkage event is the lower limb intermittent pressure device. The triggering relationship is defined as the smart mattress pressure sensor sending a start command to the lower limb intermittent pressure device. After receiving the start command, the lower limb intermittent pressure device immediately executes the pressure cycle program, providing intermittent air pressure massage to the monitored subject's lower limbs to promote blood circulation. At the same time, the smart mattress pressure sensor also triggers the second-level linkage event, sending a body position warning message to the nursing station alarm terminal to notify nursing staff to pay attention to the body position status of the monitored subject.

[0044] During the lower limb intermittent compression procedure, the intelligent infusion monitor continuously monitors the patient's infusion status. If the monitor detects an abnormal infusion rate or a risk of blockage in the infusion tubing, a third-level linkage event is triggered. This event defines the intelligent infusion monitor sending a pause command to the lower limb intermittent compression device, requiring it to suspend its current compression operation to prevent interference with the infusion process. Simultaneously, the monitor sends an infusion anomaly alarm message to the nursing station alarm terminal, requesting intervention from nursing staff. This execution dependency ensures the safety and effectiveness of multi-device collaborative operation, avoiding medical risks caused by operational conflicts between devices.

[0045] The parsed device collaborative control relationship data structure is loaded into the system's device linkage control engine. The control engine monitors the status changes of each device in real time according to the triggering relationship and execution dependency relationship defined in the data structure. When the triggering condition is met, the device linkage operation is executed sequentially according to the predefined relationship chain, realizing intelligent collaborative work between multiple devices and providing comprehensive proactive prevention and nursing support for high-risk thrombosis patients.

[0046] In one optional implementation, generating a sequence of linkage control instructions for multiple IoT devices based on the collaborative control relationships defined in the device linkage control strategy includes: The device trigger sequence and device execution sequence are obtained from the device linkage control strategy. The device trigger sequence defines the triggering order and triggering conditions of each IoT device, and the device execution sequence defines the execution actions and execution parameters of each IoT device. A temporal dependency graph is constructed based on the device triggering sequence. The temporal dependency graph uses each IoT device as a node and the triggering order and triggering condition as directed edges to represent the temporal triggering dependency relationship between each IoT device. The temporal dependency graph is topologically sorted to obtain a device triggering time chain. The device triggering time chain determines the execution order of each IoT device according to the temporal triggering dependency. Based on the execution actions and execution parameters defined in the device execution sequence, corresponding control instructions are generated for each IoT device in the device triggering timing chain. The control instructions include device identifier, execution action instruction code, and execution parameter value.

[0047] The process of parsing the device trigger sequence and device execution sequence from the device linkage control strategy is implemented using XML structured parsing technology. The device linkage control strategy is stored in a structured data format. The device trigger sequence part includes a trigger device identifier field, a trigger condition type field, a trigger threshold field, and a trigger priority field. The device execution sequence part includes an execution device identifier field, an execution action type field, an execution parameter set field, and an execution delay time field. The parsing engine reads the trigger sequence node of the strategy file and extracts the trigger condition for the temperature sensor device with the identifier SENSOR_001, which is a temperature value greater than 28 degrees Celsius, and a trigger priority of 1. At the same time, it extracts the execution action for the air conditioning device with the identifier AC_002, which is to start the cooling mode, and the execution parameters include a target temperature of 26 degrees Celsius, a fan speed level of 3, and an execution delay time of 5 seconds. The parsing process stores the trigger sequence data in a trigger data structure array and the execution sequence data in an execution data structure array. Each structure contains complete device identifier, condition parameters, and action parameter information.

[0048] The temporal dependency graph based on device trigger sequences is constructed using an adjacency list data structure. An array of node objects is created, with each node corresponding to an IoT device. Each node object contains device identifier, device type, and trigger status attributes. Node_A is created for the temperature sensor SENSOR_001, Node_B for the air conditioner AC_002, and Node_C for the curtain controller CURTAIN_003. Directed edge relationships are established according to the trigger order defined in the trigger sequence. The relationship where the temperature sensor triggers the air conditioner to start is represented by a directed edge Edge_AB from Node_A to Node_B. The edge object contains a trigger condition attribute storing the conditional expression for a temperature greater than 28 degrees Celsius. The relationship where the air conditioner starts triggers the curtain to close is represented by a directed edge Edge_BC from Node_B to Node_C. The edge object contains a trigger condition attribute storing the conditional expression for the air conditioner's operating status as "on". Each node maintains an outgoing edge list recording all directed edges originating from that node. Node_A's outgoing edge list contains Edge_AB, and Node_B's outgoing edge list contains Edge_BC. Each node maintains an in-degree counter, with Node_A having an in-degree of 0, Node_B having an in-degree of 1, and Node_C having an in-degree of 1. After the graph is constructed, a temporal dependency network is formed, consisting of 3 nodes and 2 directed edges.

[0049] Topological sorting of the temporal dependency graph is performed using a queue-based implementation of the Kahn algorithm. A queue (Queue) and a result list (List) are created. All nodes are traversed, and nodes with an in-degree of 0 are added to the queue. Node_A, with an in-degree of 0, is added to the queue. Node_A is removed from the queue and added to the end of the result list. The outgoing edges of Node_A are traversed to find Edge_AB, and the target node Node_B is obtained. The in-degree count of Node_B is decremented by 1, making its in-degree 0, and it is added to the queue. Node_B is then removed from the queue and added to the end of the list. The outgoing edges of Node_B are traversed to find Edge_BC, and the target node Node_C is obtained. The in-degree of Node_C is decremented by 1, making its in-degree 0, and it is added to the queue. Node_C is then removed from the queue and added to the end of the list. At this point, the Queue is empty and all nodes have been processed. The nodes stored in the List are in the order of Node_A, Node_B, and Node_C, which constitutes the device triggering sequence chain. This sequence chain clearly defines the strict timing relationship: the temperature sensor triggers first, the air conditioning device executes second, and the curtain controller acts last.

[0050] Generating corresponding control commands for each IoT device in the device triggering timing chain requires combining the execution actions and parameters in the device execution sequence. For the temperature sensor SENSOR_001, the first device in the timing chain, its execution action is defined as data acquisition. The generated control command includes the device identifier field filled with SENSOR_001, the execution action command code field filled with 0x01 indicating start acquisition, and the execution parameter value field filled with an acquisition interval of 2 seconds and an acquisition accuracy of 0.1 degrees Celsius. For the air conditioner AC_002, the second device in the timing chain, its execution action is defined as cooling start. The generated control command includes the device identifier field filled with AC_002, the execution action command code field filled with 0x11 indicating cooling mode, and the execution parameter value field filled with a target temperature value of 26, a fan speed value of 3, and a swing mode value of 1 indicating horizontal swing. For the curtain controller CURTAIN_003, the third position in the timing chain, the execution sequence defines its action as closing the curtains. The generated control command includes a device identifier field filled with CURTAIN_003, an execution action command code field filled with 0x21 indicating a closing action, and an execution parameter value field filled with a closing speed value of 50 for medium speed and a closing position value of 100 for fully closed. All control commands are assembled into a linkage control command sequence according to the timing chain. The first command in the sequence is a temperature sensor acquisition command, the second command is an air conditioning cooling command with a 5-second delay parameter, and the third command is a curtain closing command with a 10-second delay parameter. The command sequence is encapsulated in byte stream format. Each command includes a start identifier byte 0xAA, a 4-byte device identifier field, a 2-byte command code field, a 2-byte parameter length field, a variable-length parameter value field, and a 2-byte checksum field. The complete linkage control command sequence is distributed to each target device for execution via the IoT communication protocol.

[0051] In one optional implementation, the method further includes: Collect device operation data and intervention response data generated by the multiple IoT devices during the thrombosis prevention intervention operation; The thrombosis risk assessment model is dynamically calibrated based on the equipment operation data and the intervention response data, and the thrombosis risk level of the target monitored object is re-determined according to the calibrated thrombosis risk assessment model, thus forming a closed-loop quality control mechanism.

[0052] During the thrombosis prevention intervention, the system continuously acquires real-time data from multiple IoT devices through a data acquisition module. The data acquisition module establishes data communication connections with various IoT devices, using a message queue telemetry transmission protocol with a data transmission frequency of once every 5 seconds. For pneumatic therapy devices, the acquired equipment operation data includes the airbag inflation pressure, inflation cycle duration, deflation cycle duration, and continuous working time. Specific examples include an inflation pressure of 120 mmHg, an inflation cycle of 12 seconds, a deflation cycle of 48 seconds, and a continuous working time of 45 minutes. For smart mattresses, the system collects tilt angle adjustment values, adjustment frequency, and patient position change response time. Example values ​​are: tilt angle adjusted from 0 degrees to 30 degrees, adjusted every 2 hours, and a position change response time of 8 seconds.

[0053] The collection of intervention response data involves real-time changes in the patient's physiological parameters. Wearable sensors are used to acquire data on the rate of change of lower limb blood flow velocity, the change in lower limb circumference, and the amplitude of skin temperature fluctuations. In one monitored subject, after 30 minutes of pneumatic compression therapy, the blood flow velocity in the left lower limb increased from a baseline of 18 cm / s to 26 cm / s, the calf circumference decreased from 38.2 cm to 37.8 cm, and the skin temperature increased from 32.1 degrees Celsius to 33.4 degrees Celsius. This data is uploaded in real-time to a data processing server via a wireless transmission module, with the data packet size controlled within 512 bytes to ensure a transmission latency of no more than 200 milliseconds.

[0054] The data preprocessing module cleans and standardizes the collected raw data. For equipment operation data, it identifies and filters outlier values, classifying data points as exceeding the normal operating range of the equipment. For example, records of pressure values ​​exceeding 150 mmHg or falling below 80 mmHg on a pneumatic therapy device are marked as outliers. For time periods with missing data, a linear interpolation method using values ​​from preceding and following time points is used to complete the data. If a signal interruption within a certain time period results in the loss of data from three consecutive sampling points, the system extracts the preceding valid value (24.5) and the following valid value (26.3) from the missing points, calculating the interpolation sequence as 24.9, 25.3, and 25.7.

[0055] The dynamic calibration process is based on the correlation analysis between equipment operation data and intervention response data. The system establishes a feature mapping relationship, linking the operating parameters of the pneumatic therapy device with changes in blood flow velocity in the patient's lower limbs. Analysis revealed that in one patient, when the pressure was set to 110 mmHg, the increase in blood flow velocity was 6.2 cm / s; when the pressure was adjusted to 130 mmHg, the increase in blood flow velocity increased to 9.8 cm / s. By collecting 100 similar intervention records, the system identified a non-linear increasing relationship between pressure parameters and blood flow improvement; this relationship curve tends to flatten after the pressure value reaches 135 mmHg.

[0056] The calibration parameters of the thrombosis risk assessment model were adjusted based on feedback from actual intervention effects. In the original model, the weighting coefficient for sedentary behavior was set to 0.35. Analysis of intervention response data from 200 patients revealed that patients who sat for more than 4 hours experienced a 42% reduction in thrombosis risk after receiving intermittent pneumatic compression therapy, while patients who sat for 2 to 3 hours only experienced an 18% reduction. Based on this difference, the system increased the weighting coefficient for sedentary behavior to 0.42 and introduced subdivided intervals for sedentary duration, setting 2 to 3 hours, 3 to 4 hours, and more than 4 hours as different assessment levels, with corresponding risk contribution values ​​of 0.25, 0.38, and 0.56, respectively.

[0057] Regarding the assessment factor of postural change frequency, the original model used a single threshold method, classifying high risk as failure to turn over within 2 hours. Analysis of the smart mattress's adjustment records and patients' thrombosis patterns revealed that the angle of postural change also affected the preventative effect. One patient adjusted their position by only 15 degrees every 2 hours and developed pre-thrombotic warning signs after 3 days, while another patient adjusted their position by 35 degrees every 2 hours and maintained normal indicators for 7 days. The calibrated model combines postural change frequency and angle amplitude, setting a comprehensive scoring mechanism: a frequency of once every 2 hours with an angle greater than 30 degrees scores 10 points, while a frequency of once every 3 hours or an angle less than 20 degrees scores 4 points.

[0058] The continuous input of real-time data drives iterative updates to the model parameters. The system is configured to re-evaluate parameters every 500 accumulated valid intervention records. Within a certain calibration period, the system found that patients in a specific age group showed higher sensitivity to drug prophylaxis than other groups; patients aged 65 to 75 showed a 32% faster improvement in coagulation parameters after using anticoagulants compared to patients aged 55 to 64. Based on this, the model adjusted its age-based segmented assessment strategy, designating 65 to 75 years as an independent assessment interval, with a predictive value of 1.32 for the effect of drug intervention within this interval.

[0059] The calibrated thrombosis risk assessment model re-evaluated the risk level of the target monitored individuals. One patient was initially assessed as medium risk with a risk score of 62. After 7 days of continuous intervention and data collection, the system recorded a 28% increase in lower limb blood flow velocity, a 15% decrease in coagulation factor levels, and a 95% compliance rate with postural changes. Inputting these actual response data into the calibrated model, the risk score was recalculated to 48, and the risk level was adjusted to low risk. This adjustment triggered a corresponding change in the intervention strategy: the frequency of pneumatic compression therapy was reduced from three times daily to twice daily, while the medication dosage remained unchanged but the monitoring frequency was adjusted from twice daily to once daily.

[0060] The integrity of the closed-loop quality control mechanism is achieved through cyclical verification of data flow. A reassessed risk level generates new intervention instructions. Multiple IoT devices receive these instructions and execute the adjusted operational plan. During execution, a new round of device operation data and intervention response data is collected again, forming a continuously iterative data feedback loop. The system sets quality control monitoring indicators. When the model's prediction accuracy falls below 85% for three consecutive calibration cycles, a deep calibration process is automatically triggered, expanding the training dataset to 1.5 times its original size and recalculating the weight distribution of all evaluation factors. In one monitoring cycle, if the model's risk prediction bias for postoperative patients reaches 18%, the system immediately initiates deep calibration, retraining the model using historical data from 1500 postoperative patients. After calibration, the prediction bias is reduced to 7%.

[0061] This invention relates to an IoT-based intelligent device-linked monitoring and quality control system for thrombosis prevention, comprising: The first unit is used to acquire physiological state data associated with the target monitored object, analyze and process the physiological state data based on a preset thrombosis risk assessment model, and determine the thrombosis risk level of the target monitored object. The thrombosis risk assessment model is established based on the correlation between the limb activity state data and the blood circulation state data. The second unit is used to match the corresponding device linkage control strategy from a preset device control strategy library according to the thrombosis risk level. The device control strategy library includes multiple sets of device linkage control strategies for different thrombosis risk levels, and each set of device linkage control strategies defines the collaborative control relationship between multiple Internet of Things devices. The third unit is used to generate a sequence of linkage control instructions for multiple IoT devices based on the collaborative control relationship defined in the device linkage control strategy; distribute the sequence of linkage control instructions to the multiple IoT devices, and drive the multiple IoT devices to perform thrombosis prevention intervention operations according to the time sequence dependency relationship and the execution condition constraint relationship.

[0062] A third aspect of the present invention provides an electronic device, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0063] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0064] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0065] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for quality control and monitoring of intelligent devices for thrombosis prevention based on the Internet of Things, characterized in that, include: Physiological status data associated with the target monitored individual is acquired, and the physiological status data is analyzed and processed based on a preset thrombosis risk assessment model to determine the thrombosis risk level of the target monitored individual. The thrombosis risk assessment model is established based on the correlation between the limb activity status data and the blood circulation status data. Based on the thrombosis risk level, a corresponding device linkage control strategy is matched from a preset device control strategy library. The device control strategy library includes multiple sets of device linkage control strategies for different thrombosis risk levels, and each set of device linkage control strategies defines the collaborative control relationship between multiple IoT devices. Based on the collaborative control relationship defined in the device linkage control strategy, a linkage control instruction sequence for multiple IoT devices is generated; the linkage control instruction sequence is distributed to the multiple IoT devices, driving the multiple IoT devices to perform thrombosis prevention intervention operations according to the time sequence dependency relationship and the execution condition constraint relationship.

2. The method according to claim 1, characterized in that, The physiological state data are analyzed and processed based on a pre-defined thrombosis risk assessment model to determine the thrombosis risk level of the target monitored individual, including: Feature extraction is performed on the limb activity state data to obtain a limb activity feature vector, which represents the limb movement pattern of the target monitored object within a preset time window; Feature extraction is performed on the blood circulation status data to obtain a blood circulation feature vector, which characterizes the blood flow status of the target monitored object within the preset time window; The limb activity feature vector and the blood circulation feature vector are fused to obtain a fused feature vector. The fusion process is based on the physiological correlation between the limb activity state data and the blood circulation state data. The fused feature vector is input into the thrombosis risk assessment model, and the thrombosis risk assessment model outputs a thrombosis risk score based on the matching relationship between the fused feature vector and the preset thrombosis risk determination rules. The thrombosis risk level is determined based on the thrombosis risk score and a preset risk level classification standard, wherein the risk level classification standard maps the thrombosis risk score to the corresponding thrombosis risk level.

3. The method according to claim 2, characterized in that, The thrombosis risk assessment model outputs a thrombosis risk score based on the matching relationship between the fused feature vector and the preset thrombosis risk determination rules, including: The fusion feature vector is subjected to dimensionality analysis to extract multiple feature components associated with thrombosis risk. Each feature component corresponds to a different physiological state dimension. The weighted values ​​of each feature component are obtained by weighting the multiple feature components with the preset feature weight matrix in the thrombosis risk determination rule. The weighted values ​​of each feature component are summed to obtain the initial risk score; Historical risk event sequences and historical physiological state sequences are extracted from the historical thrombosis risk data of the target monitored object. The frequency and time interval of risk events in the historical risk event sequences are statistically analyzed, and the frequency of risk events is calculated. After normalizing the frequency of risk events, it is used as the risk trend intensity value. The risk trend intensity value is then compared with a preset correction coefficient benchmark value to obtain the risk correction coefficient. The initial risk score is multiplied by the risk correction coefficient to obtain the corrected risk score. The corrected risk score is then weighted and summed with a dynamic adjustment factor determined based on the current physiological state of the target monitored object to obtain the thrombosis risk score.

4. The method according to claim 1, characterized in that, Based on the thrombosis risk level, a corresponding device linkage control strategy is matched from a preset device control strategy library. The device control strategy library includes multiple sets of device linkage control strategies for different thrombosis risk levels. Each set of device linkage control strategies defines the collaborative control relationship between multiple IoT devices, including: Using the thrombosis risk level as a search condition, a set of candidate device linkage control strategies corresponding to the thrombosis risk level is retrieved from the device control strategy library. The set of candidate device linkage control strategies includes at least one set of device linkage control strategies. Obtain the current device configuration information of the target monitored object, calculate the matching degree between the device function requirements of each group of device linkage control strategies and the device configuration information, and obtain the matching degree score of each group of device linkage control strategies; Based on the matching score, the device linkage control strategy with the highest matching score is selected from the candidate device linkage control strategy set as the target device linkage control strategy; The device collaborative control relationship data structure is obtained by parsing the target device linkage control strategy. The device collaborative control relationship data structure defines the triggering relationship and execution dependency relationship between each IoT device participating in the linkage.

5. The method according to claim 1, characterized in that, Based on the collaborative control relationships defined in the device linkage control strategy, a sequence of linkage control instructions for multiple IoT devices is generated, including: The device trigger sequence and device execution sequence are obtained from the device linkage control strategy. The device trigger sequence defines the triggering order and triggering conditions of each IoT device, and the device execution sequence defines the execution actions and execution parameters of each IoT device. A temporal dependency graph is constructed based on the device triggering sequence. The temporal dependency graph uses each IoT device as a node and the triggering order and triggering condition as directed edges to represent the temporal triggering dependency relationship between each IoT device. The temporal dependency graph is topologically sorted to obtain a device triggering time chain. The device triggering time chain determines the execution order of each IoT device according to the temporal triggering dependency. Based on the execution actions and execution parameters defined in the device execution sequence, corresponding control instructions are generated for each IoT device in the device triggering timing chain. The control instructions include device identifier, execution action instruction code, and execution parameter value.

6. The method according to claim 1, characterized in that, The method further includes: Collect device operation data and intervention response data generated by the multiple IoT devices during the thrombosis prevention intervention operation; The thrombosis risk assessment model is dynamically calibrated based on the equipment operation data and the intervention response data, and the thrombosis risk level of the target monitored object is re-determined according to the calibrated thrombosis risk assessment model, thus forming a closed-loop quality control mechanism.

7. A smart device-linked monitoring and quality control system for thrombosis prevention based on the Internet of Things, used to implement the method as described in any one of claims 1-6, characterized in that, include: The first unit is used to acquire physiological state data associated with the target monitored object, analyze and process the physiological state data based on a preset thrombosis risk assessment model, and determine the thrombosis risk level of the target monitored object. The thrombosis risk assessment model is established based on the correlation between the limb activity state data and the blood circulation state data. The second unit is used to match the corresponding device linkage control strategy from a preset device control strategy library according to the thrombosis risk level. The device control strategy library includes multiple sets of device linkage control strategies for different thrombosis risk levels, and each set of device linkage control strategies defines the collaborative control relationship between multiple Internet of Things devices. The third unit is used to generate a sequence of linkage control instructions for multiple IoT devices based on the collaborative control relationship defined in the device linkage control strategy; distribute the sequence of linkage control instructions to the multiple IoT devices, and drive the multiple IoT devices to perform thrombosis prevention intervention operations according to the time sequence dependency relationship and the execution condition constraint relationship.

8. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 6.

Images