Intelligent personalized nursing path management system for patients in department of rheumatism and immunology

Abstract

The invention relates to the technical field of nursing in the rheumatism and immunology department, in particular to an intelligent patient personalized nursing path management system in the rheumatism and immunology department, which comprises a joint activity-inflammation data acquisition unit, a path generation and dose binding unit, a cross-modal verification feedback unit and a closed-loop priority regulation and control unit, multi-modal data such as joint angles and inflammatory factors are collected through detection and fusion of a nine-axis inertial sensor and a biomarker, the morning stiffness period is recognized through a time sequence segmentation model, the dosage of prednisone is dynamically adjusted in combination with a hormone dosage response rule base, and the effectiveness of rehabilitation actions is verified based on a dynamic time warping algorithm. When the medicine taking time deviation exceeds one hour or the action execution is abnormal, a sensor is triggered to enhance the acquisition and dose increment mechanism, and meanwhile, the priority of a nursing instruction is optimized through a conflict arbitration mechanism, so that the full-process intelligence of data acquisition, path generation, verification feedback and priority regulation is realized, and an accurate nursing scheme is provided for patients in the rheumatism immunology department.

Description

Technical Field

[0001] The present invention relates to the technical field of rheumatology and immunology nursing, and specifically, to an intelligent personalized nursing path management system for rheumatology and immunology patients. Background Art

[0002] Rheumatology and immunology nursing is an important technology. In the clinical nursing of rheumatology and immunology diseases, accurately monitoring joint movement function, dynamically adjusting drug dosage, and personalized rehabilitation training path are crucial for controlling the progression of inflammation and improving the quality of life of patients.

[0003] With the development of wearable sensors and biometric detection technologies, the ability to collect nursing data in real time has been significantly improved. However, the traditional nursing system has a core problem of insufficient dynamic correlation analysis ability in practical applications. Existing solutions often rely on single-dimensional data, ignoring the complex coupling relationship between joint range of motion, inflammatory factor levels, and pharmacokinetics. When the morning stiffness time of a patient fluctuates or the inflammatory factors suddenly increase, the traditional system cannot quickly identify abnormalities based on multi-source data. The root cause lies in the lack of the ability to model the "inflammatory load - joint function - drug response" dynamic chain. The adjustment of drug dosage is mostly based on fixed thresholds, without fully considering the impact of physiological characteristics such as the patient's individual body weight and liver function on drug metabolism. On the other hand, the formulation of the rehabilitation training path does not accurately match the patient's morning stiffness period and the peak plasma drug concentration, resulting in a significant reduction in the training effect due to inappropriate timing. This lag in data correlation and model construction reduces the treatment synergy, and the nursing plan responds slowly to changes in the individual physiological state, making it difficult to adapt to the characteristics of rheumatology and immunology diseases with "large disease course fluctuations and significant individual differences". As a result, the traditional nursing mode lacks dynamics and individuation, causing waste of nursing resources. To solve this technical problem, we thus provide an intelligent personalized nursing path management system for rheumatology and immunology patients. Summary of the Invention

[0004] The purpose of the present invention is to provide an intelligent personalized nursing path management system for rheumatology and immunology patients to solve the problems raised in the above background art.

[0005] 1. Since the traditional system relies on single-dimensional data and lacks the ability of dynamic correlation analysis of multi-source data and cannot quickly identify the abnormal state of patients, in this case, a multi-modal data is collected by the joint movement - inflammation data acquisition unit, and the dynamic activity score and inflammatory load index are calculated, which can comprehensively and accurately evaluate the patient's state based on multi-dimensional data and improve the ability to identify abnormalities.

[0006] 2. Since the traditional system adjusts drug doses based on fixed thresholds without considering individual physiological characteristics and formulates the rehabilitation training path at an inappropriate time, in this case, the path generation and dose binding unit combines a time series segmentation model and a hormone dose response rule base to generate personalized rehabilitation training intervals and dynamically adjust drug doses, which can accurately match individual needs and improve treatment synergy and effectiveness.

[0007] To achieve the above objectives, a personalized nursing path management system for intelligent rheumatology and immunology patients is provided, including the following units: The joint activity-inflammation data acquisition unit collects the three-dimensional joint angle change sequence through a nine-axis inertial sensor, outputs the dynamic activity score after Kalman filtering, and detects interleukin-6 and tumor necrosis factor- through a chip , and calculates the inflammation load index in combination with the current daily dose of prednisone; The path generation and dose binding unit inputs the dynamic activity score into the time series segmentation model, identifies the starting time window of morning stiffness, and generates the recommended interval for the next day's rehabilitation training. According to the 24-hour fluctuation curve of the inflammation load index, it matches the preset hormone dose response rule base. When the inflammation load index exceeds the threshold for 6 consecutive hours, it triggers the instruction for gradient adjustment of the daily dose of prednisone; The cross-modal verification feedback unit obtains the actual medication time through an intelligent medicine box. When the deviation from the instruction for gradient adjustment of the daily dose of prednisone exceeds hours, it activates the enhanced acquisition mode of the joint sensor. At the same time, during the rehabilitation training period, it compares the dynamic time warping distance between the dynamic activity score and the standard action template in real time. If the distance exceeds the threshold and the inflammation load index rises synchronously , it sends a hormone dose increment signal to the path generation and dose binding unit.

[0008] As a further improvement of this technical solution, the specific implementation method of the time series segmentation model in the path generation and dose binding unit includes: Construct a morning stiffness feature extractor based on a temporal convolutional network, and set its dilation coefficient to , is the network layer number, , and the convolution kernel width is 3; Perform phase segmentation on the dynamic activity score sequence. When the standard deviation of the joint angle in the adjacent 30-minute window drops by ≥40% and the duration is >2 hours, it is determined as the starting window of morning stiffness; The formula for generating the recommended interval for the next day's rehabilitation training is: ; where is the predicted value of the starting time of morning stiffness, , is the empirical buffer duration, is the recommended interval for the next day's rehabilitation training.

[0009] As a further improvement of this technical solution, the construction method of the hormone dose response rule library includes: Establish a calculation formula for the dynamic threshold: ; where is the dynamic threshold at time , is the patient's baseline inflammation load index, is the inflammation load index, represents any moment within 24 hours, and the inflammation load index ; where is the current daily dose of prednisone, is the dose influence factor, is the prednisone half-life decay coefficient, is the time after the last dose, is interleukin-6, is tumor necrosis factor- , which is used to reflect the immediate inflammation state and the drug intervention effect. The patient's baseline inflammation load index is obtained according to the individual's historical benchmark value of the patient and is used to reflect the basic inflammation level of the individual's physiological characteristics; The generation rule of the gradient adjustment instruction is: When is greater than for 6 consecutive hours, increase the prednisone dose by 5 mg every 24 hours until the cumulative increment reaches 15 mg; When is less than 0.8 for 12 consecutive hours, reduce the prednisone dose by 2.5 mg every 24 hours, and increase the joint sensor sampling rate to 15 Hz after the dose reduction.

[0010] As a further improvement of this technical solution, the path generation and dose binding unit further includes a multi-constraint dose optimization mechanism: a. Before the gradient adjustment instruction takes effect, verify the following constraint conditions: b. The cumulative daily dose of prednisone ≤ 1 mg / kg, calculated according to the patient's weight; c. The fluctuation range of alanine aminotransferase, a liver function index, within the past 7 days is less than 30%; If any one of a, b, and c is not satisfied, start the alternative plan decision-making process: Replace the 5 mg prednisone increment with an equivalent intravenous injection plan of interleukin-6 inhibitor; and adjust the rehabilitation training period to 2 hours after the peak blood drug concentration.

[0011] As a further improvement of this technical solution, the dynamic time warping distance calculation method in the cross-modal verification and feedback unit includes: Construct a joint angle change template library for standard rehabilitation movements. Each template contains a sequence of three-dimensional Euler angles for 15 key frames, which are a set of angle values used to describe the attitude of a rigid body in three-dimensional space; The formula for calculating the dynamic time warping distance between the patient's movement and the template in real time is: ; where is the path cheap penalty coefficient, is the final dynamic time warping distance, is the optimal alignment path, is the alignment point pair, is the patient's joint angle vector, is the template joint angle vector, is the Euclidean distance; When is greater than the dynamic time warping distance threshold and the rising rate of the inflammation load index is greater than or equal to 2 pg / ml / h, it is determined that the movement execution is abnormal. The dynamic time warping distance threshold is a critical value used to determine the degree of difference between the patient's rehabilitation movement and the standard template.

[0012] As a further improvement of this technical solution, the activation logic of the joint sensor enhanced acquisition mode includes: When the deviation of the medicine-taking time is greater than 1 hour, increase the sampling rate of the nine-axis inertial sensor from 10 Hz to 20 Hz, activate the myoelectric signal acquisition module around the joint, synchronously analyze the muscle activation delay time. In the enhanced mode, generate a joint stability index every 5 minutes. When the joint stability index is less than 0.7, send a training period compression instruction to the path generation and dose binding unit, and reduce the single duration by 25%; Among them, the method for generating the joint stability index is as follows: Fuse the nine-axis sensor data and myoelectric signals to construct the joint stability index. The nine-axis sensor data includes the angular velocity variance and the acceleration fluctuation coefficient , and the myoelectric signals include the muscle activation synchrony . Then the calculation formula of the joint stability index is: ; where = actual co-contraction index / ideal co-contraction index, which is used to measure the coordination of the activation of the muscle groups around the joint. The actual co-contraction index is obtained by calculating the activation time difference and intensity ratio of the antagonist muscles through surface myoelectric signals, and the ideal co-contraction index is obtained based on the benchmark value established from the data of healthy people.

[0013] As a further improvement of this technical solution, the triggering mechanism of the hormone dose increment signal includes the following steps: Establish the mapping relationship between the increment signal intensity and multi-modal data: When is in the interval and the synchronous increase rate of the inflammation burden index ≥ 15%, a low-intensity increment signal is triggered. The is the low-intensity increment threshold, which is used to trigger the inflammation burden range for low-dose increment. The low-intensity increment signal is +2.5 mg / 6 h; When and the synchronous increase rate of the inflammation burden index ≥ 25%, a high-intensity increment signal is triggered. The high-intensity increment signal is +5 mg / 6 h. The is the high-intensity increment threshold, which is used to trigger the inflammation burden critical value for high-dose increment; During the period when the increment signal takes effect, the real-time motion monitoring mode of the patient-side APP is forcibly enabled: If it is detected that the sudden drop in joint range of motion > 30%, immediately suspend dose adjustment and initiate an emergency video call link; If the joint range of motion returns to more than 90% of the baseline value, resume the increment process and shorten the monitoring interval to once every 15 minutes.

[0014] As a further improvement of this technical solution, it further includes a closed-loop priority regulation unit, specifically as follows: Establish a conflict arbitration mechanism for nursing instructions. When there is a time conflict between hormone dose adjustment and the rehabilitation training period, calculate the expected inflammation inhibition efficiency of hormone adjustment and the joint function improvement efficiency of rehabilitation training respectively; If the expected inflammation inhibition efficiency of hormone adjustment is greater than 1.2 times the joint function improvement efficiency of rehabilitation training, postpone the training and execute dose increment, and at the same time generate a compensatory training plan; If the joint function improvement efficiency of rehabilitation training is greater than or equal to the expected inflammation inhibition efficiency of hormone adjustment, divide the original 60-minute training into two 25-minute modules and one 10-minute module, and place them 0.5 hour before and 1.5 hours after dose adjustment respectively. The two 25-minute modules are the main training for 50 minutes, and the 10-minute module is the buffer time for 10 minutes, which is used for preparation and relaxation before and after training.

[0015] As a further improvement of this technical solution, the calculation formulas for the expected inflammation inhibition efficiency of hormone adjustment and the joint function improvement efficiency of rehabilitation training are as follows: The expected inflammation inhibition efficiency of hormone adjustment: ; where is the difference between the current inflammation burden index and the average value in the previous 6 hours, is the predicted value of prednisone blood drug concentration, which is calculated based on the pharmacokinetic model; The joint function improvement efficiency of rehabilitation training: ; wherein, is the change rate of the dynamic activity score before and after training, is the muscle activation coordination degree, which is calculated by the phase synchronization of the electromyogram signal.

[0016] As a further improvement of the technical solution, the closed-loop priority regulation unit updates the parameters of the time series segmentation model every 30 days, specifically as follows: Re-fit the inflation coefficient of the time series segmentation model based on historical data, and adjust the convolution kernel width using the Bayesian optimization algorithm; Modify the threshold of the hormone dose response rule base according to the patient's individual pharmacokinetic characteristics, and the modification formula is: ; wherein, is the new generation of inflammation burden threshold, is the current cycle threshold, is the day inflammation burden peak value, and are the weights assigned according to historical data.

[0017] Compared with the prior art, the beneficial effects of the present invention: In the intelligent personalized nursing path management system for rheumatology and immunology patients, the path generation and dose binding unit uses the time series segmentation model to accurately identify the morning stiffness period, and combines the hormone dose response rule base to realize the dynamic adjustment of prednisone dose, so as to improve the matching degree of the rehabilitation training timing, shorten the lag time of drug dose adjustment. The cross-modal verification feedback unit triggers the enhanced acquisition of sensors and dose increment signals through the medication time monitoring and dynamic time warping algorithm, shortens the warning response time for abnormal joint stability, improves the recognition accuracy of abnormal action execution, and enhances the dynamic adaptability and treatment effect of rheumatology and immunology disease nursing. Brief Description of the Drawings

[0018] Figure 1 is the working flow chart of the present invention; Figure 2 is the overall block diagram of the present invention.

[0019] The meanings of the various labels in the figure are as follows: 1. Joint activity - inflammation data acquisition unit; 2. Path generation and dose binding unit; 3. Cross-modal verification feedback unit; 4. Closed-loop priority regulation unit. Detailed Embodiments

[0020] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention.

[0021] Patients with rheumatic immune diseases often face the dual contradictions of 'inflammation control' and 'joint function protection'. There are three major pain points in the existing nursing system: static threshold rigidity, multi-modal data fragmentation, and lack of arbitration for treatment conflicts. The present invention realizes the morning stiffness time series prediction based on a fully convolutional neural network, the binding of hormone dose driven by the dynamic threshold of inflammation load, and the priority intelligent arbitration under cross-modal verification through a three-dimensional closed-loop of biomechanics-biochemical indicators-medication behavior, and provides a personalized nursing path management system for intelligent rheumatology patients. Please refer to Figure 1 - Figure 2 As shown, it includes the following units: The joint activity-inflammation data acquisition unit 1 collects the three-dimensional joint angle change sequence through a nine-axis inertial sensor, and outputs a dynamic activity score after Kalman filtering. The nine-axis sensor can comprehensively capture the three-dimensional movement of the joint, and Kalman filtering can effectively suppress noise and improve the accuracy of the activity score. And it detects interleukin-6 and tumor necrosis factor- , combines with the current daily dose of prednisone to calculate the inflammation load index. Interleukin-6 and tumor necrosis factor- are key inflammatory factors. Combining with the medication dose can quantify the inflammation load. Using electrochemiluminescence immunoassay, continuous collection of interstitial fluid is carried out through a subcutaneous microdialysis probe, and the detection period is 1 hour. The chip is built-in with a temperature compensation module to eliminate the influence of body temperature fluctuation on the detection result. The inflammation load index ; where is the current daily dose of prednisone, is the dose influence factor, is the prednisone half-life decay coefficient, is the time after the last medication, is interleukin-6, is tumor necrosis factor- , which is used to reflect the immediate inflammation state and the effect of drug intervention. The patient's baseline inflammation load index is obtained according to the historical reference value of the patient's individual, and is used to reflect the basic inflammation level of the individual's physiological characteristics.

[0022] The path generation and dose binding unit 2 inputs the dynamic activity score into the time series segmentation model to identify the starting time window of morning stiffness. The specific implementation method of the time series segmentation model in the path generation and dose binding unit 2 includes: Construct a morning stiffness feature extractor based on a temporal convolutional network, and its dilation coefficient is set to , is the network layer sequence number, , the convolution kernel width is 3, capturing the activity trends at different time scales, performing phase segmentation on the dynamic activity scoring sequence, with a window size of 30 minutes and a step size of 5 minutes, calculating the change rate of the standard deviation of joint angles in adjacent windows. When the standard deviation drops by ≥40% and the duration lasts > 2 hours, it is marked as the starting window of morning stiffness, and the recommended interval for the next day's rehabilitation training is generated. The formula for generating the recommended interval for the next day's rehabilitation training is: ; where is the predicted value of the starting time of morning stiffness, , is the empirical buffer duration, is the recommended interval for the next day's rehabilitation training. According to the 24-hour fluctuation curve of the inflammation load index, it matches the preset hormone dose response rule library. The hormone dose response rule library contains dose adjustment thresholds corresponding to different inflammation load levels, and the thresholds are personalized according to the patient's weight and disease severity. When the inflammation load index exceeds the threshold for 6 consecutive hours, calculate the average value of the inflammation load index, and compare it with the rule library threshold in real time to trigger the prednisone daily dose gradient adjustment instruction. Continuous inflammation load exceeding the standard requires timely intervention, and the gradient adjustment avoids side effects caused by sudden dose changes.

[0023] Among them, the construction method of the hormone dose response rule library includes: Establish the calculation formula for the dynamic threshold: ; where is the time 's dynamic threshold, is the patient's baseline inflammation load index, is the inflammation load index, represents any moment within 24 hours; The generation rule of the prednisone daily dose gradient adjustment instruction is: When is greater than for 6 consecutive hours, increase the prednisone dose by 5mg every 24 hours until the cumulative increment reaches 15mg; When is less than 0.8 for 12 consecutive hours, reduce the prednisone dose by 2.5mg every 24 hours, and after the dose reduction, trigger the joint sensor sampling rate to increase to 15Hz. The rationality of the dose adjustment is improved through multi-constraint verification, while ensuring the inflammation control effect.

[0024] Among them, the multi-constraint verification improvement also includes the multi-constraint dose optimization mechanism: a. Before the gradient adjustment instruction takes effect, verify the following constraint conditions: b. The cumulative daily dose of prednisone ≤ 1 mg / kg. Calculated based on the patient's weight, a prednisone dose exceeding 1 mg / kg will increase the risk of liver injury. The cumulative dosage needs to be dynamically restricted based on weight as follows: Obtain the patient's weight data in real time and calculate the maximum allowable dose. ; Record the sum of the historical daily prednisone doses. When the cumulative dose after this adjustment is greater than , trigger a constraint alarm to avoid hormone side effects caused by drug overdose.

[0025] c. The fluctuation range of alanine aminotransferase, a liver function index, within the past 7 days is less than 30%. When liver function is abnormal, the hormone metabolism ability decreases. It is necessary to monitor the fluctuation of transaminase to evaluate liver tolerance as follows: Collect the patient's blood samples twice a week, detect the value of alanine aminotransferase, and calculate the fluctuation range within the past 7 days, that is ; Among them, is the fluctuation range of alanine aminotransferase within the past 7 days, used to evaluate the stability of liver function, is the current (most recent test) alanine aminotransferase test value of the patient, reflecting the immediate liver function status, is the baseline value of alanine aminotransferase of the patient within the past 7 days, taking the average value or the test value during the stable period within this time period. When , it is determined that the liver function is unstable, and the increase in hormone dose is prohibited to identify the risk of liver injury in advance.

[0026] If any one of a, b, and c is not satisfied, start the alternative plan decision-making process. When hormone adjustment is restricted, switch to a biologic agent to maintain inflammation control, and at the same time optimize the training time. According to the drug equivalent conversion table (such as 1 mg prednisone ≈ 0.4 mg interleukin-6 inhibitor), calculate the alternative dose and send it to the doctor's order system. Predict the time to reach the peak blood drug concentration of the interleukin-6 inhibitor through a population pharmacokinetic model, replace the 5 mg prednisone increment with an equivalent intravenous injection plan of the interleukin-6 inhibitor, and adjust the rehabilitation training period to 2 hours after the peak blood drug concentration to avoid the physical fatigue at the initial stage of drug onset from affecting the training effect and reduce the training interruption caused by drug side effects.

[0027] The cross-modal verification feedback unit 3 obtains the actual medication time through the smart medicine box. When the deviation from the prednisone daily dose gradient adjustment instruction exceeds hours, activate the enhanced acquisition mode of the joint sensor, fuse the data of the accelerometer, gyroscope, and magnetometer, improve the attitude solution accuracy through the Kalman filter algorithm, extend the single data acquisition window from 10 minutes to 30 minutes to cover the critical time period of drug onset. At the same time, during the rehabilitation training period, the activation logic of the enhanced acquisition mode of the joint sensor includes: When the deviation of the drug-taking time is greater than 1 hour, the deviation of the drug-taking time may cause fluctuations in blood drug concentration. It is necessary to enhance monitoring to capture changes in joint function. Increase the sampling rate of the nine-axis inertial sensor from 10 Hz to 20 Hz to enhance the ability to capture subtle joint movements. Activate the electromyogram signal acquisition module around the joint, and use a band-pass filter (20 - 450 Hz) to extract the characteristics of electromyogram activity. Calculate the muscle activation delay time, which is the time difference from the start of the movement to when the electromyogram signal exceeds the threshold, and is used to identify early muscle function abnormalities. In the enhanced mode, generate a joint stability index every 5 minutes. It is necessary to quantitatively evaluate the control ability of the joint during movement to provide a basis for training adjustment. Integrate the nine-axis sensor data and electromyogram signals to construct the joint stability index. The nine-axis sensor data includes the variance of angular velocity and the coefficient of acceleration fluctuation , and the electromyogram signal includes muscle activation synchrony . Then the calculation formula of the joint stability index is: ; where = actual co-contraction index / ideal co-contraction index, which is used to measure the coordination of the activation of the muscle groups around the joint. The actual co-contraction index is obtained by calculating the activation time difference and intensity ratio of the antagonist muscles through surface electromyogram signals, and the ideal co-contraction index is obtained based on the benchmark value established from the data of healthy people. Integrate the current window data every 5 minutes and generate the stability index through edge computing. When the joint stability index is less than 0.7, trigger the training period compression logic, automatically calculate the new training duration, and send a training period compression instruction to the path generation and dose binding unit 2, reducing the single duration by 25%.

[0028] Compare the dynamic time warping distance between the dynamic activity score and the standard action template in real time. If the distance exceeds the threshold and the inflammation load index rises synchronously , then send a hormone dose increment signal to the path generation and dose binding unit 2. The calculation method of the dynamic time warping distance includes: Construct a joint angle change template library for standard rehabilitation actions. The rehabilitation physician wears a nine-axis inertial sensor to perform standard actions, and collect more than 100 repeated action data. Use the clustering algorithm to segment the continuous action sequence, extract the three-dimensional Euler angle sequence of 15 key frames, take the average value of the three-dimensional Euler angles of each key frame, and construct a standard template matrix containing 15 key frames × 3 angles, which is a set of angle values used to describe the rigid body posture in three-dimensional space. Among them, the three-dimensional Euler angle includes the pitch angle, yaw angle, and roll angle, providing a reliable benchmark for action evaluation; There may be a time scale difference between the patient's action and the standard template. It is necessary to achieve elastic alignment through the dynamic time warping algorithm. The formula for calculating the dynamic time warping distance between the patient's action and the template in real time is: ; where is the path cheap penalty coefficient, which is used to control the penalty intensity of time axis stretching. is the final dynamic time warping distance, which measures the difference between the patient's movement and the standard template. The larger the value, the more serious the deviation. is the optimal alignment path, which represents the optimal time correspondence between the patient's movement sequence and the template sequence. is the alignment point pair. For the th frame of the patient's movement sequence and the th frame of the template sequence, the matching point. is the patient's joint angle vector, that is, the three-dimensional Euler angle of the patient's joint at the th frame. is the template joint angle vector, that is, the preset Euler angle of the standard rehabilitation movement at the th frame. is the Euclidean distance, which is used to calculate the instantaneous difference between the patient and the template in the joint angle space. The threshold is determined through the test data of the healthy population. A sliding window is adopted, with a window size of 2 hours and a step size of 30 minutes, and the change rate of the inflammation load index is calculated. ; among them, represents the concentration value of interleukin-6 detected at the current moment , which represents the current inflammation level. is the concentration value of interleukin-6 at the current moment pushed forward 2 hours as the concentration value of interleukin-6 detected as the historical reference baseline. When is greater than the dynamic time warping distance threshold and the rising rate of the inflammation load index is greater than or equal to 2 pg / ml / h, it is determined that the action execution is abnormal. The dynamic time warping distance threshold is the critical value for determining the difference between the patient's rehabilitation action and the standard template, which improves the recognition accuracy of the dynamic time warping algorithm for rehabilitation actions.

[0029] The triggering mechanism of the hormone dose increment signal includes the following steps: Establish the mapping relationship between the increment signal intensity and the multi-modal data. The rising rate of the inflammation load and the joint function status need to be evaluated collaboratively to accurately match the dose adjustment intensity: Calculate the rising rate of the inflammation load. The exponential smoothing method is used to calculate the dynamic rising rate of the inflammation load index. ; among them, is the dynamic rising rate of the inflammation load index. is the inflammation load index at the moment. is the inflammation load index pushed forward 6 hours before the moment. Compare the current rate with the preset threshold in real time. When When the synchronous increase rate of the inflammation burden index ≥ 15%, a low-intensity incremental signal is triggered, and the is the low-intensity incremental threshold, which is used to trigger the inflammation burden range for low-dose increment. The low-intensity incremental signal is +2.5 mg / 6 h; When and the synchronous increase rate of the inflammation burden index ≥ 25%, a high-intensity incremental signal is triggered. The high-intensity incremental signal is +5 mg / 6 h, and the is the high-intensity incremental threshold, which is the critical value of the inflammation burden for triggering high-dose increment; During the period when the incremental signal takes effect, the real-time motion monitoring mode of the patient-side APP is forcibly turned on. During the dose adjustment period, the changes in joint function need to be closely monitored to prevent acute injuries caused by drug side effects or aggravated inflammation: Temporarily increase the sampling rate of the joint sensor to 50 Hz, collect the joint range of motion every 2 minutes, calculate the difference from the baseline value. If it is detected that the sudden drop in the joint range of motion > 30%, immediately suspend the dose adjustment and start the emergency video call link. If the joint range of motion returns to more than 90% of the baseline value, resume the incremental process and shorten the monitoring interval to once every 15 minutes, shortening the average time for adjustment and recovery, and improving the treatment safety and clinical efficacy.

[0030] The closed-loop priority control unit 4 is as follows: There will be a conflict between hormone adjustment and rehabilitation training in terms of time. It is necessary to quantitatively evaluate the expected benefits of the two interventions to determine the priority. Compare the hormone adjustment instruction and the rehabilitation training plan on the time axis. When the following conditions are met, a conflict is determined. Among them, the hormone adjustment instruction includes the execution time and the duration, and the rehabilitation training plan includes the start time and the end time: ; where is the execution time, is the start time, is the duration, is the end time. Establish a conflict arbitration mechanism for nursing instructions. When there is a time conflict between the hormone dose adjustment and the rehabilitation training period, calculate the expected inflammation inhibition efficiency of hormone adjustment and the joint function improvement efficiency of rehabilitation training respectively. The calculation formulas for the expected inflammation inhibition efficiency of hormone adjustment and the joint function improvement efficiency of rehabilitation training are as follows: The expected inflammation inhibition efficiency of hormone adjustment: ; where, is the difference between the current inflammation burden index and the average value in the previous 6 hours, is the predicted value of prednisone blood drug concentration, calculated based on the pharmacokinetic model; The joint function improvement efficiency of rehabilitation training: ; where, is the change rate of the dynamic activity score before and after training, is the muscle activation coordination degree, which is calculated through the phase synchronization of the electromyogram signal. Specifically, the acquisition of the muscle activation coordination degree is as follows: First, use the electromyogram signal acquisition module around the joint to collect multi-channel electromyogram signals, preprocess them through band-pass filtering and Kalman filtering, perform Hilbert transform on the preprocessed signals to obtain the instantaneous phase, calculate the phase difference between adjacent channels, quantify the phase synchronization of a single pair of muscles through the phase locking value, and then set weights based on the proportion of muscle physiological functions, and perform weighted averaging on multiple pairs of phase locking values to obtain the muscle activation coordination degree.

[0031] If the expected inflammation inhibition efficiency of hormone adjustment is greater than 1.2 times the joint function improvement efficiency of rehabilitation training, postpone the training and execute a dose increase. Postpone the rehabilitation training until 2 hours after the end of hormone adjustment. At the same time, generate a compensatory training plan and increase 15 minutes of joint flexibility training; If the joint function improvement efficiency of rehabilitation training is greater than or equal to the expected inflammation inhibition efficiency of hormone adjustment, divide the original 60-minute training into two 25-minute modules and one 10-minute module, and place them 0.5 hour before and 1.5 hours after the dose adjustment respectively. The two 25-minute modules are the main training for 50 minutes, and the 10-minute module is the buffer time for 10 minutes, which is used for preparation and relaxation before and after training. Dynamically adjust the time window for efficiency calculation according to the patient's current inflammation state to avoid loss of treatment effect due to improper sequence. Postponing training may affect the rehabilitation progress, and an equivalent compensation plan needs to be designed to maintain treatment continuity. Select targeted training actions from the preset template library based on the patient's current joint function state, and calculate the compensation coefficient according to the postponed duration ; is the postponed duration, which is used to adjust the training intensity. Disperse the compensatory training within 24 hours, 10 - 15 minutes each time, to avoid fatigue caused by concentrated training.

[0032] The closed-loop priority control unit 4 updates the parameters of the time series segmentation model every 30 days, specifically as follows: Re-fit the inflation coefficient of the time series segmentation model based on historical data, and use the Bayesian optimization algorithm to adjust the width of the convolution kernel; Modify the threshold of the hormone dose response rule base according to the patient's individual pharmacokinetic characteristics. The modification formula is: ; where, is the new generation inflammation load threshold, is the current cycle threshold, is the day inflammation load peak value, and is the weight assigned according to historical data. Among them, the new-generation inflammation load threshold refers to the inflammation load threshold used for the next cycle after being corrected by historical data. It is an updated value that fuses the peak inflammation load data in the past 30 days based on the current cycle threshold. The current cycle threshold refers to the inflammation load critical value used for hormone dose adjustment decisions within the current 30-day cycle, and is used to determine whether the inflammation load index triggers dose adjustment.

[0033] The present invention collects multi-modal data such as joint angles and inflammatory factors through the fusion of a nine-axis inertial sensor and biomarker detection, identifies the morning stiffness period through a time series segmentation model, dynamically adjusts the prednisone dose in combination with a hormone dose response rule base, and verifies the effectiveness of rehabilitation actions based on the dynamic time warping algorithm. When the deviation of the medication-taking time exceeds 1 hour or the action execution is abnormal, a sensor enhanced acquisition and dose increment mechanism is triggered. At the same time, the priority of nursing instructions is optimized through a conflict arbitration mechanism, realizing the full-process intelligence of "data acquisition - path generation - verification feedback - priority regulation", and providing a precise nursing plan for patients in the rheumatology and immunology department.

[0034] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and descriptions in the specification are only preferred examples of the present invention and are not used to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of the present invention claimed is defined by the appended claims and their equivalents.

Claims

1. The intelligent personalized nursing path management system for rheumatology and immunology patients is characterized in that, It includes the following units: The joint movement - inflammation data acquisition unit (1) collects the three - dimensional joint angle change sequence through a nine - axis inertial sensor, outputs the dynamic activity score after Kalman filtering, and detects interleukin - 6 and tumor necrosis factor - , and calculates the inflammation load index in combination with the current daily dose of prednisone; The path generation and dose binding unit (2) inputs the dynamic activity score time series into the segmentation model, identifies the starting time window of morning stiffness, and generates the recommended interval for the next day's rehabilitation training. It matches the preset hormone dose response rule base according to the 24-hour fluctuation curve of the inflammation load index. When the inflammation load index exceeds the threshold for 6 consecutive hours, it triggers the prednisone daily dose gradient adjustment instruction; The cross-modal verification feedback unit (3) obtains the actual medication-taking time through the intelligent medicine box. When the deviation from the prednisone daily dose gradient adjustment instruction exceeds hours, it activates the enhanced acquisition mode of the joint sensor. At the same time, during the rehabilitation training period, it compares the dynamic time warping distance between the dynamic activity score and the standard action template in real time. If the distance exceeds the threshold and the inflammation load index rises synchronously , it sends a hormone dose increment signal to the path generation and dose binding unit (2).

2. The intelligent personalized nursing path management system for rheumatology and immunology patients according to claim 1, wherein The specific implementation method of the time series segmentation model in the path generation and dose binding unit (2) includes: Construct a morning stiffness feature extractor based on a temporal convolutional network, with its dilation coefficient set to , being the network layer number, , and the convolutional kernel width is 3; Performing phase segmentation on the dynamic activity score sequence. When the standard deviation of joint angles in adjacent 30-minute windows decreases by ≥ 40% and the duration is > 2 hours, it is determined as the starting window of morning stiffness; The formula for generating the recommended interval for the next day's rehabilitation training is: ; where is the predicted value of the starting time of morning stiffness, , is the empirical buffer duration, is the recommended interval for the next day's rehabilitation training.

3. The personalized nursing path management system for intelligent rheumatology and immunology department patients according to claim 1, wherein, The construction method of the hormone dose response rule base includes: Establish the calculation formula for the dynamic threshold: ; where is the dynamic threshold at time , is the patient's baseline inflammation burden index, is the inflammation burden index, represents any moment within 24 hours, and the inflammation burden index ; where is the current daily prednisone dose, is the dose impact factor, is the prednisone half-life decay coefficient, is the time since the last dose, is interleukin-6, is tumor necrosis factor- , which is used to reflect the immediate inflammation state and the drug intervention effect. The patient's baseline inflammation burden index is obtained based on the patient's individual historical benchmark value and is used to reflect the basic inflammation level of the individual's physiological characteristics; The generation rule of the gradient adjustment instruction is: When it is greater than for 6 consecutive hours, increase the prednisone dose by 5 mg every 24 hours until the cumulative increase reaches 15 mg; When less than 0.8 for 12 consecutive hours reduce the prednisone dose by 2.5 mg every 24 hours, and increase the sampling rate of the joint sensor to 15 Hz after the dose reduction.

4. The personalized nursing path management system for intelligent rheumatology and immunology patients according to claim 1, wherein The path generation and dose binding unit (2) also includes a multi-constraint dose optimization mechanism: a. Before the gradient adjustment instruction takes effect, verify the following constraint conditions: b. The cumulative daily dose of prednisone ≤ 1 mg / kg, calculated according to the patient's weight; c. The fluctuation range of alanine aminotransferase, a liver function index, within the past 7 days is less than 30%; If any one of a, b, and c is not satisfied, start the alternative plan decision-making process: Replace the 5 mg prednisone increment with an equivalent intravenous injection plan of interleukin-6 inhibitor; and adjust the rehabilitation training period to 2 hours after the peak blood drug concentration.

5. The personalized nursing path management system for intelligent rheumatology and immunology patients according to claim 1, characterized in that, The dynamic time warping distance calculation method in the cross-modal verification and feedback unit (3) includes: Construct a joint angle change template library for standard rehabilitation movements. Each template contains a three-dimensional Euler angle sequence of 15 key frames, which are a set of angle values used to describe the rigid body posture in three-dimensional space; The formula for calculating the dynamic time warping distance between the patient's movement and the template in real time is: ; among them, is the path cheap penalty coefficient, is the final dynamic time warping distance, is the optimal alignment path, is the alignment point pair, is the patient joint angle vector, is the template joint angle vector, is the Euclidean distance; When is greater than the dynamic time warping distance threshold and the rising rate of the inflammation load index is greater than or equal to 2 pg / ml / h, it is determined that the action execution is abnormal. The dynamic time warping distance threshold is a critical value for determining the difference between the patient's rehabilitation action and the standard template.

6. The personalized nursing path management system for intelligent rheumatology and immunology department patients according to claim 1, characterized in that, The activation logic of the joint sensor enhanced acquisition mode includes: When the medicine-taking time deviation is greater than 1 hour, increase the sampling rate of the nine-axis inertial sensor from 10 Hz to 20 Hz, activate the myoelectric signal acquisition module around the joint, synchronously analyze the muscle activation delay time. In the enhanced mode, generate a joint stability index every 5 minutes. When the joint stability index is less than 0.7, send a training period compression instruction to the path generation and dose binding unit (2), and reduce the single duration by 25%; Among them, the method for generating the joint stability index is as follows: Fusing nine-axis sensor data and electromyogram signals to construct a joint stability index, where the nine-axis sensor data includes angular velocity variance and acceleration fluctuation coefficient , and the electromyogram signals include muscle activation synchrony . Then the calculation formula for the joint stability index is: ; where = actual co-contraction index / ideal co-contraction index, which is used to measure the coordination of muscle activation around the joint. The actual co-contraction index is obtained by calculating the activation time difference and intensity ratio of antagonist muscles through surface electromyogram signals, and the ideal co-contraction index is obtained based on the benchmark value established from the data of healthy people.

7. The intelligent personalized nursing path management system for rheumatology and immunology patients according to claim 5, characterized in that, The triggering mechanism of the hormone dose increment signal includes the following steps: Establish the mapping relationship between the increment signal intensity and multi-modal data: When is in the interval and the synchronous increase rate of the inflammation load index ≥ 15%, a low-intensity incremental signal is triggered. The is the low-intensity incremental threshold, which is used to trigger the inflammation load range of low-dose increment. The low-intensity incremental signal is +2.5 mg / 6 h; When and the synchronous increase rate of the inflammation burden index ≥ 25%, a high-intensity incremental signal is triggered, and the high-intensity incremental signal is +5 mg / 6 h. The is the high-intensity incremental threshold, which is the inflammation burden critical value for triggering a high-dose increment; During the effective period of the increment signal, force the real-time motion monitoring mode of the patient-side APP to be turned on: If a sudden drop in joint range of motion > 30% is detected, immediately suspend the dose adjustment and start the emergency video call link; If the joint range of motion returns to more than 90% of the baseline value, resume the increment process and shorten the monitoring interval to once every 15 minutes.

8. The intelligent personalized nursing path management system for rheumatology and immunology patients according to claim 1, wherein It also includes a closed-loop priority regulation unit (4), specifically as follows: Establish a nursing instruction conflict arbitration mechanism. When there is a time conflict between hormone dose adjustment and rehabilitation training period, calculate the expected inflammation inhibition efficiency of hormone adjustment and the joint function improvement efficiency of rehabilitation training respectively; If the expected inflammation inhibition efficiency of hormone adjustment is more than 1.2 times the joint function improvement efficiency of rehabilitation training, postpone the training and execute a dose increment, while generating a compensatory training plan; If the joint function improvement efficiency of rehabilitation training is greater than or equal to the expected inflammation inhibition efficiency of hormone adjustment, divide the original 60-minute training into two 25-minute modules and one 10-minute module, which are respectively placed 0.5 hour before and 1.5 hours after the dose adjustment. The two 25-minute modules are the main training for 50 minutes, and the 10-minute module is the buffer time for 10 minutes, which is used for preparation and relaxation before and after training.

9. The intelligent personalized nursing path management system for rheumatology and immunology patients according to claim 8, wherein, The calculation formulas for the expected inflammation inhibition efficiency of hormone adjustment and the joint function improvement efficiency of rehabilitation training are as follows: The expected inflammation inhibition efficiency of hormone adjustment: ; wherein, is the difference between the current inflammation burden index and the mean value in the previous 6 hours, is the predicted value of prednisone blood drug concentration, calculated based on the pharmacokinetic model; The joint function improvement efficiency of rehabilitation training: ; wherein, is the change rate of the dynamic activity score before and after training, is the muscle activation coordination degree, which is calculated by the phase synchrony of the electromyogram signal.

10. The intelligent personalized nursing path management system for rheumatology and immunology patients according to claim 8, characterized in that, The closed-loop priority control unit (4) updates the parameters of the time series segmentation model every 30 days, specifically as follows: Re-fit the inflation coefficient of the time series segmentation model based on historical data, and adjust the convolution kernel width using the Bayesian optimization algorithm; Modify the threshold of the hormone dose response rule base according to the patient's individual pharmacokinetic characteristics, and the modification formula is: ; wherein, is the new generation of inflammation load threshold, is the current cycle threshold, is the peak value of inflammation load on the th day, and are the weights assigned according to historical data.

Images