Posture and wear monitoring system for motion signals of artificial joints of four limbs

Abstract

The invention relates to the technical field of medical instrument data monitoring, and provides a limb artificial joint motion signal posture and wear monitoring system which comprises a motion posture recognition module, a motion signal processing module and a motion signal processing module. On the basis of the collected motion signal data, the motion posture is judged and analyzed according to a motion posture judgment model, and the type of the motion posture is recognized; wherein the motion posture types comprise a high-load motion posture, a low-load motion posture and a static posture; and the monitoring frequency control module is used for extracting motion characteristics. The monitoring frequency control module dynamically adjusts the wear monitoring frequency according to the motion posture and the motion intensity, introduces a joint fatigue factor to comprehensively measure the motion intensity, improves the monitoring frequency during high-load motion, timely captures rapid wear changes, reduces or stops monitoring during low-load motion or static, reduces resource waste, and improves the working efficiency. Accurate and efficient monitoring is realized, and the monitoring accuracy and timeliness are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of medical device data monitoring, and specifically to a posture and wear monitoring system for the movement signals of artificial joints of the four limbs. Background Art

[0002] With the rapid development of medical technology, total joint arthroplasty of the four limbs has become an effective means for treating severe joint diseases, such as osteoarthritis, rheumatoid arthritis, traumatic arthritis, etc., greatly improving the joint function and quality of life of patients. Inevitably, wear problems will occur during the long-term use of artificial joints, which will not only affect the service life of the joints, but may also lead to complications such as joint loosening, pain, and infection. In severe cases, even a second operation may be required; However, most of the existing monitoring technologies rely on two methods: regular imaging examinations or fixed monitoring frequencies. Although the method of regular imaging examinations can reflect the wear condition of the joints to a certain extent, imaging examinations are static detection means, which can only provide the morphological information of the joints at specific time points and cannot reflect the real-time wear process of the joints during daily activities. The method of using a fixed monitoring frequency fails to fully consider the differences in the joint wear rates of patients in different movement postures. When the human body is performing high-load exercises, the artificial joints bear greater pressure and friction, and the wear rate will increase significantly. While in low-load exercises or a static state, the joint wear is relatively slow. Using a unified monitoring frequency cannot timely capture the rapid wear changes of the joints during high-load exercises, easily leading to missed diagnoses and delayed treatments of wear conditions, resulting in the accuracy and effectiveness of the monitoring results being affected; Therefore, the present invention provides a posture and wear monitoring system for the movement signals of artificial joints of the four limbs. Summary of the Invention

[0003] In order to make up for the deficiencies of the prior art and solve at least one of the technical problems proposed in the background art, the monitoring frequency can be adjusted according to the actual movement postures of patients, early signs of joint wear can be detected in a timely manner, the change trend of the wear degree and the remaining service life can be predicted, providing accurate and reliable basis for doctors to formulate personalized treatment plans and rehabilitation programs, thereby improving the use effect of artificial joints and the quality of life of patients; The technical solution adopted by the present invention to solve its technical problems is: a posture and wear monitoring system for the movement signals of artificial joints of the four limbs, including: A movement posture recognition module: collecting the movement signal data of the artificial joints, and based on the collected movement signal data, judging and analyzing the movement posture according to the movement posture judgment model to identify the movement posture type; Among them, the movement posture types include: high-load movement postures, low-load movement postures, and static postures; Motion monitoring frequency control module: Extract motion features, introduce joint fatigue factors for data fusion processing according to the motion features to obtain a motion intensity coefficient, combine the motion intensity coefficient with the real-time wear monitoring frequency, update the subsequent wear monitoring frequency for different motion posture types, and process the monitoring of artificial joints based on the new wear monitoring frequency to output the wear degree; Wear prediction module: By continuously judging whether the wear degree exceeds the wear degree warning value, if it exceeds the wear degree warning value, obtain the corresponding warning time point, and predict the remaining service life of the artificial joint in the future time period.

[0004] The beneficial effects of the present invention are as follows: The motion posture recognition module of the present invention collects multi-source data through a multi-modal sensor array, fuses information such as acceleration, angular velocity, myoelectric intensity, and pressure, reduces the one-sidedness of a single sensor, and uses an adaptive feature weighting algorithm to adapt to the non-linear and non-stationary characteristics of signals, improving the accuracy of posture recognition; The monitoring frequency control module of the present invention dynamically adjusts the wear monitoring frequency according to the motion posture and motion intensity, introduces joint fatigue factors to comprehensively measure the motion intensity, increases the monitoring frequency during high-load motion to timely capture rapid wear changes, reduces or stops monitoring during low-load motion or at rest, reduces resource waste, realizes precise and efficient monitoring, and improves the accuracy and timeliness of monitoring; The wear prediction module of the present invention compares the real-time wear degree with the preset value, obtains the warning and alarm time points, analyzes the change trend of the basic wear data, predicts the remaining service life of the artificial joint, and based on this, doctors can formulate personalized treatment plans, such as deciding whether to perform surgery in advance, adjusting the rehabilitation plan, and can also remind patients to pay attention to the motion intensity and have regular check-ups. It can not only improve the use effect of artificial joints and improve the quality of life of patients, but also help medical institutions reasonably allocate medical resources and make preparations for surgery in advance, playing an important role in medical decision-making and patient management. Description of the Drawings

[0005] The present invention will be further described below with reference to the drawings.

[0006] Figure 1 is a framework diagram of a posture and wear monitoring system for a four-limbed artificial joint motion signal of the present invention; Figure 2 is a step flow chart of a posture and wear monitoring method for a four-limbed artificial joint motion signal of the present invention. Detailed Embodiments

[0007] In order to make the technical means, creative features, achieved purposes, and functions of the present invention easy to understand, the present invention will be further described below in conjunction with specific embodiments.

[0008] Example 1 Please refer to Figure 1 As shown, a posture and wear monitoring system for limb artificial joint motion signals according to an embodiment of the present invention specifically includes: Motion posture recognition module 10: Collect motion signal data of the artificial joint, and based on the collected motion signal data, perform judgment and analysis on the motion posture according to the motion posture judgment model to identify the motion posture type; Among them, the motion posture types include: high-load motion posture, low-load motion posture, and static posture; For the motion posture recognition module, the execution process is as follows: It should be noted in advance that when implanting the joint, the monitoring computer screen can display the position and angle of the joint prosthesis, which can provide immediate joint posture information and support for the surgeon; Multimodal sensors (such as accelerometers, gyroscopes, etc.) on the artificial joint need to be deployed on the main motion axes of the joint. The accurate installation of the prosthesis is the premise for the sensor to correctly collect kinematic and mechanical signals. If the position or angle of the prosthesis deviates, it may cause sensor data distortion, thereby affecting the accuracy of the subsequent motion posture judgment model; Install a multimodal sensor array on the limb artificial joint to collect motion signal data of the artificial joint. The sensors include: accelerometers, gyroscopes, electromyography sensors, and pressure sensors; the artificial joint motion signals include: mechanical signals, kinematic signals, and electromyography signals; the signal data includes but is not limited to: acceleration data, angular velocity data, electromyography intensity data, and pressure data; Optionally, kinematic signal acquisition: Use high-precision triaxial accelerometers and gyroscopes, which are respectively deployed on the main motion axes of the artificial joint to monitor the linear acceleration and angular velocity changes of the joint in three-dimensional space in real time; Electromyography signal acquisition: At the muscle bellies of the main muscle groups that drive the artificial joint movement, arrange electromyography sensors in the form of a differential electrode array, and use bioelectric amplification and filtering technologies to capture the weak electrical signals generated during muscle contraction. After feature extraction, the muscle activation degree (obtaining electromyography intensity), contraction timing, and cooperative force generation mode can be quantified; Mechanical signal acquisition: implant a micro pressure sensor at the interface between the artificial joint and the bone, the contact part of the joint cavity, or integrate it into the key stress areas of the joint prosthesis, and based on the piezoresistive or piezoelectric effect, perceive the pressure distribution and stress changes of the joint during loading, friction, and impact in real time; By arranging the sensors on the limb around the artificial joint, the signals related to the movement of the artificial joint can be collected comprehensively and accurately; It should be noted that for the data collected by each sensor, a synchronous acquisition circuit and software algorithm need to be combined to synchronize the data of each sensor in time; Among them, the synchronous acquisition circuit includes but is not limited to: Synchronous control circuit: The synchronous control logic is implemented using a Field Programmable Gate Array (FPGA) or a Microcontroller Unit (MCU). Through programming configuration, the FPGA or MCU can generate a synchronous trigger signal according to the clock signal and send it to each sensor simultaneously to ensure that they start sampling at the same time; Software algorithms include but are not limited to: Timestamp marking algorithm: At the software level, an accurate timestamp is added to the data collected by each sensor and stored and transmitted together with the data; Preprocess the data of the artificial joint motion signals collected by different sensors. Among them, the preprocessing process includes: Denoising processing, normalization processing; Specifically, the denoising processing includes: Using a complementary filtering algorithm to fuse and denoise the data of the accelerometer and gyroscope. The accelerometer data is more accurate in the low-frequency band but is easily affected by vibration and shock. The gyroscope data is stable in the high-frequency band but has an integration drift problem. The complementary filtering algorithm combines the advantages of both. By fusing the data of the accelerometer and gyroscope, it relies on the accelerometer data at low frequencies and the gyroscope data at high frequencies, thereby effectively removing noise and reducing drift; For the electromyogram signal, the method of wavelet packet decomposition and reconstruction is used for denoising. Wavelet packet decomposition can decompose the signal into different frequency sub-bands. According to the frequency characteristics of the electromyogram signal, the appropriate sub-band is selected for reconstruction to remove noise and interference signals while retaining the characteristic information of the electromyogram signal; The pressure sensor data is interfered by the external environment. The median filtering algorithm can be selected to denoise it. Median filtering is a non-linear filtering method that sorts the data within the window and takes the median value as the output, which can effectively remove impulse noise and salt-and-pepper noise while retaining the edge information of the signal; Normalize all the data after denoising processing. The method of normalization processing can choose the minimum-maximum normalization method; Extract and fuse the features of the artificial joint motion signal data after preprocessing; For the feature extraction stage, the features extracted from the processed data are divided into time-domain features, frequency-domain features, time-frequency features, and combined features; Time-domain features include statistical features such as mean, variance, peak value, etc., which are used to describe the basic statistical characteristics and fluctuations of the signal; Frequency-domain features are obtained through Fourier transform and can reflect the distribution of data in the frequency domain, which helps to distinguish different motion postures; Time-frequency features use wavelet transform to analyze the common characteristics of data in the time domain and frequency domain, providing a more detailed analysis of time variation and frequency characteristics, and having a good effect on processing non-stationary signals (such as electromyogram signals); For the feature fusion stage, a feature fusion method based on Principal Component Analysis (PCA) is adopted; Input the fused features into the motion posture judgment model to output the real-time motion posture type; Furthermore, the training process of the motion posture judgment model includes: Obtain the historical data set, including the signals collected by the accelerometer, gyroscope, electromyography sensor, and pressure sensor at different time points, as well as the corresponding actual motion posture types; Divide the historical data set into a training set, a validation set, and a test set, with a division ratio of 70% for the training set, 15% for the validation set, and 15% for the test set; Preprocess the sensor signal data in the training set, validation set, and test set, including denoising and normalization; extract and fuse features from the preprocessed data; Use the data in the training set to calculate the importance score of each feature as the initial feature weight vector; during the model training process, for each node sample set in the training set, calculate the discrimination of each feature in the sample set, combine the adaptive adjustment factor, calculate the adaptive weight according to the feature discrimination and the initial weight, and perform normalization processing; Optionally, obtain the sum of the Gini impurity reduction values of each feature in all decision trees in the random forest through the Gini impurity reduction method as the feature importance, and then normalize the feature importance score to obtain the initial feature weight; measure the discrimination by calculating the variance ratio of the feature in different category samples, introduce an adaptive adjustment factor (set by those skilled in the art based on previous research experience), multiply the adaptive factor, the initial weight, and the discrimination to obtain the adaptive weight, and perform normalization processing on the adaptive weight; Thus, it is possible to perform adaptive feature weighting in the random forest algorithm, improve the model's discrimination ability for different features, and thus improve the accuracy of limb artificial joint motion posture recognition; Select the random forest model, preferentially select the features with higher weights for node splitting according to the adaptive weight of each feature, and train the model with the training set data; input the validation set data into the training model, and evaluate the model performance according to the difference between the model output result and the actual label (the evaluation metrics can include at least one of accuracy, precision, and recall). If the performance does not meet the expectation, adjust the model parameters (such as the number of decision trees and the maximum depth of the random forest), and repeat the training process until the model reaches the best performance on the validation set to obtain the motion posture judgment model; In summary, the motion posture recognition module dynamically recognizes the current motion posture type by deploying a multi-modal sensor array around the limb artificial joint, collecting the mechanical signals, kinematic signals, and electromyography signals of the artificial joint in real time, extracting time-domain, frequency-domain, and time-frequency domain features through synchronous acquisition and data preprocessing, and finally inputting them into the random forest model based on adaptive feature weighting; It has the following effects: integrating multi-source data of acceleration, angular velocity, myoelectric intensity, and pressure, covering the mechanics, kinematics, and muscle activation states of joint movement, reducing the one-sidedness of a single sensor, and improving the accuracy of posture recognition; Through an adaptive feature weighting algorithm, that is, combining Gini impurity and discrimination to dynamically adjust feature weights, adapting to the non-linear and non-stationary characteristics of signals in different motion scenarios, and ensuring robustness under complex motion postures; Identifying the types of motion postures can provide a core basis for monitoring strategies. When in a static posture, it reduces the acquisition of invalid data. If it is identified as a high-load or low-load motion posture, it triggers the motion intensity evaluation process, dynamically adjusts the wear monitoring frequency (increasing the frequency for high load and decreasing the frequency for low load), and conducts on-demand monitoring to ensure high-frequency real-time tracking in high-risk scenarios and energy-saving efficiency in low-risk scenarios; Monitoring frequency control module 20: Extracts motion features, based on the motion features, introduces a joint fatigue factor for data fusion processing to obtain a motion intensity coefficient, combines the motion intensity coefficient with the real-time wear monitoring frequency, updates the subsequent wear monitoring frequency for different types of motion postures, and processes the monitoring of artificial joints based on the new wear monitoring frequency to output the wear degree; For the monitoring frequency control module, the execution process is as follows: Specifically, the process of updating the wear monitoring frequency is as follows: If the motion posture type is a static posture, stop the wear monitoring; If the motion posture type is a high-load motion posture or a low-load motion posture, update and adjust the wear monitoring frequency based on the motion intensity; Specifically, set a sliding time window, and obtain pressure data, linear acceleration data, and myoelectric intensity data within each time window; Take the mean value of the pressure data within the time window as the joint bearing strength value, and take the peak value of the pressure data within the time window as the joint instantaneous load value; Synthesize the linear acceleration data in three-dimensional space to obtain the synthesized linear acceleration, take the mean value of the synthesized linear acceleration within the time window as the motion amplitude value, and take the peak value of the synthesized linear acceleration within the time window as the vibration amplitude value; Take the root mean square of the myoelectric intensity data within the time window as the muscle activation intensity value; Normalize the joint bearing strength value, joint instantaneous load value, motion amplitude value, vibration amplitude value, and muscle activation intensity value to obtain a motion intensity feature vector; The normalization process can select min-max normalization; Train a differentiated dynamic weight matrix based on the historical dataset, so as to reflect the importance of different data types under the corresponding motion posture types. In this embodiment, they are mainly high-load motion postures and low-load motion postures; The training process of the weight matrix is as follows: Based on the multi-modal sensor data related to high-load motion postures and low-load motion postures in the historical dataset, including pressure data, acceleration data, angular velocity data, myoelectric intensity data, etc., and the actual values or labels of the corresponding motion intensity; Determine the features participating in the weight calculation, the joint bearing strength value (pressure mean), the joint instantaneous load value (pressure peak), the motion amplitude value (synthetic linear acceleration mean), the vibration amplitude value (synthetic linear acceleration peak), the muscle activation intensity value (root mean square of myoelectric intensity); Adopt appropriate machine learning methods (such as linear regression, support vector machine, neural network) or statistical methods to construct a model with the evaluation of motion intensity as the goal, with the above features as the input and the motion intensity-related indicators as the output; During the model training process, adjust the weights of each feature through an optimization algorithm (such as gradient descent) so that the model output can accurately reflect the actual motion intensity under different motion postures; for high-load motion postures and low-load motion postures, train the corresponding weight matrices respectively to reflect the importance differences of different data types under different postures; Exemplarily, under high-load motion postures, the pressure data and myoelectric intensity data may have higher weights; while under low-load motion postures, the weight of the linear acceleration data may be relatively lower; Verification and adjustment: Use the validation set to evaluate the performance of the trained weight matrix, and adjust the model parameters or weight matrix according to the evaluation results (such as mean square error, correlation coefficient, etc.) until the weight matrix can accurately reflect the importance of different data types under the corresponding motion posture types and meet the motion intensity evaluation requirements; Calculate the joint fatigue factor through the spectral characteristics of the myoelectric signal. The calculation process of the joint fatigue factor is as follows: Extract the proportion of the middle-frequency band energy of the myoelectric signal, extract the minimum middle-frequency band energy and the maximum middle-frequency band energy in the historical data, and perform a difference calculation to obtain the historical middle-frequency band energy extreme difference. Calculate the difference between the proportion of the middle-frequency band energy and the small middle-frequency band energy, and then calculate the ratio with the historical middle-frequency band energy extreme difference to obtain the joint fatigue factor; Among them, the specific frequency range of the middle frequency band can be 50-150 Hz; The function of introducing the joint fatigue factor is as follows: The exercise intensity not only depends on physical data such as external pressure and acceleration, but also the muscle fatigue state is a key factor affecting joint wear. When the muscle is fatigued, its protective and supporting effects on the joint will weaken, resulting in an increase in the additional stress borne by the joint, thereby accelerating joint wear. By introducing the joint fatigue factor, this internal factor of muscle fatigue can be incorporated into the exercise intensity assessment system, making the measurement of exercise intensity more comprehensive and accurate, and thus more precisely reflecting the actual impact of exercise on the joint; When updating the wear monitoring frequency based on the exercise intensity, the joint fatigue factor is used as an adjustment parameter. When the joint fatigue factor is relatively high, it indicates that the joint is in a relatively vulnerable state and the wear risk increases. It is necessary to increase the monitoring frequency to more timely detect changes in joint wear; Data fusion is performed by combining the exercise intensity feature vector, the dynamic weight matrix, and the joint fatigue factor to calculate the exercise intensity coefficient. The calculation process of the exercise intensity coefficient is as follows: Perform weighted summation on the exercise intensity feature vector and the weight vector to obtain the basic exercise intensity. Based on the basic exercise intensity and the joint fatigue factor, calculate the weight to obtain the exercise intensity coefficient k; the weight coefficients of the basic exercise intensity and the joint fatigue factor are calculated by the entropy weight method, and the entropy weight method is common knowledge that can be implemented by those skilled in the art; Obtain the real-time wear monitoring frequency and calculate the new wear monitoring frequency after update; For high-load exercise postures, the calculation formula is: ; where is the new wear monitoring frequency, and k is the exercise intensity coefficient; it is necessary to increase the monitoring frequency for high-load exercise postures; For low-load exercise postures, the calculation formula is: ; where is the new wear monitoring frequency, and k is the exercise intensity coefficient; it is necessary to decrease the monitoring frequency for low-load exercise postures; It should be noted that there is a maximum and minimum range for setting the monitoring frequency, and the updated monitoring frequency needs to be within this range. If the calculated monitoring frequency is not within this range, then select the extreme value of the corresponding maximum or minimum range; Second specifically, the process of outputting the wear degree is as follows: Calculate the wear amount using a classical wear model; optionally, the classical wear model is the Archard model; Among them, the wear amount The calculation formula is: ; where M is the material wear coefficient, which is a preset parameter set according to the characteristics of artificial joint materials, represents the average pressure in the i-th monitoring period, Denote the sliding distance in the $i$-th monitoring period, and $T$ is the monitoring period; The sliding distance is the vector sum of the rotational displacement and the total translational displacement of the artificial joint; Integrate the angular velocity over time in the monitoring period to obtain the cumulative rotation angle of the joint. The product of the cumulative rotation angle and the radius of the joint contact surface (determined by the design parameters of the artificial joint) is the rotational displacement. The accelerometer collects three-dimensional linear acceleration signals, calculates the three-dimensional translational displacements through two integrations respectively, and synthesizes the three-dimensional translational positions to obtain the total translational displacement; For the wear amount calculated for each monitoring period, when comparing adjacent periods, it is used as the wear increment of this period relative to the previous period, that is, the wear increment in the $i$-th monitoring period is the wear amount in the $i$-th monitoring period subtracted by the wear amount in the $(i - 1)$-th monitoring period ; Calculate the wear increment for each monitoring period , and accumulate the historical wear increments starting from the initial monitoring period to obtain the real-time wear degree. The calculation formula is: ; where, is the real-time wear degree, is the basic wear degree at the initial installation, $n$ is the monitoring period, and $i$ represents the $i$-th monitoring period; In summary, the monitoring frequency control module: First, set the sliding time window according to the type of motion posture, obtain motion characteristics such as the joint bearing strength value, joint instantaneous load value, motion amplitude value, vibration amplitude value, and muscle activation strength value, and obtain the motion intensity feature vector through normalization processing. Then, train a differentiated dynamic weight matrix based on the historical data set, calculate the joint fatigue factor through the spectral characteristics of the electromyogram signal, fuse the motion intensity feature vector, weight matrix, and joint fatigue factor to calculate the motion intensity coefficient, and update the wear monitoring frequency accordingly. The monitoring frequency is increased for high-load motion postures, decreased for low-load motion postures, and stopped for static postures. At the same time, the Archard model is used to calculate the wear amount, and the wear increments of real-time motion are accumulated to obtain the real-time wear degree; Thus, the wear monitoring frequency can be dynamically adjusted according to different motion postures and motion intensities to achieve on-demand monitoring. The monitoring frequency is increased during high-load motion, which can capture rapid wear changes in a timely manner, reduce missed diagnoses and delayed treatments. The monitoring frequency is decreased or stopped during low-load motion or static states, which can reduce unnecessary data acquisition and achieve energy-saving and efficient monitoring. Introducing the joint fatigue factor makes the evaluation of motion intensity more comprehensive and accurate, and thus can more accurately reflect the actual impact of motion on joint wear, improving the accuracy and effectiveness of monitoring; Wear prediction module 30: By judging in real time whether the wear degree exceeds the wear degree warning value, if it exceeds the wear degree warning value, obtain the corresponding warning time point, and predict the remaining service life of the artificial joint in the future period; For the wear prediction module, the execution process is as follows: During the wear monitoring of the artificial joint, obtain the real-time wear degree, compare it with the wear degree warning value. If the real-time wear degree is less than the wear warning value, continue the wear monitoring; If the real-time wear degree is greater than or equal to the wear degree warning value, obtain the corresponding warning time point; and obtain the time point corresponding to the wear degree early warning value as the early warning time point; Among them, it should be noted that the wear degree warning value is the mean value of the wear degree early warning value and the wear degree threshold value; the wear degree early warning value is a preset wear degree reference value, which is used to indicate that the wear state of the artificial joint begins to enter the range that needs attention. For example, when the wear degree is close to the early warning value, it means that there may be an abnormal change trend in the joint wear. Although it has not reached the dangerous level, it is necessary to increase the monitoring frequency and closely observe the development of the wear situation in order to detect potential problems in time. The wear degree threshold value is the limit value reached by the wear of the artificial joint. Once the wear degree reaches or exceeds this threshold value, the performance of the artificial joint will be seriously affected, and problems such as joint loosening, increased pain, and functional limitation may occur, and even a second operation may be required to replace the joint; Optionally, the monitoring frequency can be increased after the warning time point, so as to closely monitor the development of the wear situation, which helps to detect potential problems early and prevent the occurrence of serious joint complications; Obtain the wear degree corresponding to the early warning time point to the warning time point as the basic wear degree data; Conduct a trend analysis on the basic wear degree data, including judging whether the basic wear degree data is linearly changing or non-linearly changing; Optionally, use the methods of statistical fitting and residual analysis to judge, and perform linear fitting and non-linear fitting on the basic wear degree data respectively; Conduct a univariate linear regression on the basic wear data, fit a univariate linear regression equation, and calculate the sum of squared linear residuals; At the same time, conduct a quadratic polynomial fitting on the basic wear data and calculate the sum of squared non-linear residuals; In addition to quadratic polynomial fitting, exponential fitting can also be used; Compare the linear fitting and non - linear fitting through the AIC information criterion. Here, the AIC information criterion, that is, the Akaike information criterion, is a standard for measuring the goodness of fit of a statistical model. Since it was founded and developed by the Japanese statistician Hirotugu Akaike, it is also called the Akaike information criterion. It is based on the concept of entropy and can balance the complexity of the estimated model and the goodness of fit of this model to the data; If the sum of squared non - linear residuals is less than the sum of squared linear residuals, and the AIC value corresponding to the non - linearity is smaller, then it is determined to be a non - linear change. If the sum of squared residuals shows a random distribution and no obvious curvature, then it is determined to be a linear change; It should be explained that the sum of squared residuals is an index to measure the degree of difference between the predicted value of the model and the actual value. When fitting the basic wear degree data, whether it is linear fitting or non - linear fitting, predicted values will be generated. The difference between the predicted value and the actual wear degree data is the residual. Summing the squares of all residuals gives the sum of squared residuals. When the sum of squared non - linear residuals is less than the sum of squared linear residuals, it means that the predicted values of the non - linear model are generally closer to the actual wear data than those of the linear model, that is, the non - linear model has a better fitting effect on the data. If the sum of squared residuals shows a random distribution and no obvious curvature, it indicates that the linear model can reasonably describe the variation law of the data. In this case, the trend of data change can be represented by a simple linear relationship, so it is determined to be a linear change; For the basic wear degree showing a linear change, it means that within the time period from the warning time point to the alarm time point, the wear rate of the artificial joint is relatively stable, and the future wear situation can be predicted and the remaining life can be calculated according to the corresponding stable wear rate; Obtain the linear prediction equation through the least - squares method fitting. The slope of the linear prediction equation is the wear rate. Calculate the difference between the wear degree threshold and the wear degree value corresponding to the warning time point, and then calculate the ratio of the difference to the wear rate to obtain the remaining service life; For the basic wear degree showing a non - linear change, when the basic wear degree shows a non - linear change, the wear rate is not constant and may change with time; Obtain the non - linear prediction equation through quadratic polynomial fitting or exponential fitting. According to the actual basic wear degree data, by comparing the sum of squared residuals and the AIC values of different non - linear models, select the model with the best fitting effect. For example, select the model with a smaller sum of squared residuals and a lower AIC value; Based on the non - linear prediction equation, taking the wear degree threshold as known, calculate the corresponding predicted time value, and then calculate the difference between the predicted time value and the alarm time value to obtain the remaining service life; Based on the remaining service life, doctors can formulate personalized treatment plans according to the remaining life, such as whether early surgery is needed, adjust the rehabilitation plan, and can remind patients to pay attention to the exercise intensity, reduce high-load activities, conduct regular reviews, predict the replacement time, and reasonably allocate resources; In summary, the wear prediction module first compares the real-time wear degree with the preset wear warning value. If it exceeds the warning value, it obtains the corresponding warning time point and warning time point, and extracts the basic wear data between the two time points; then, through statistical fitting, residual analysis, and AIC information criterion, it judges whether the wear trend is linear or non-linear. If it is a linear change, it fits the wear rate by the least square method and calculates the remaining life. If it is a non-linear change, it selects the non-linear model with the best fitting effect to predict the time to reach the wear threshold, and finally outputs the remaining service life; Based on the dual determination of the wear warning value and the warning value, the abnormal wear stage is identified in advance. Then, according to the change trend (linear or non-linear) of the basic wear degree data, the corresponding fitting method is used to predict the remaining service life of the artificial joint, enabling doctors and patients to know the joint status in advance and providing a key basis for subsequent medical decisions.

[0009] Embodiment 2 The present application also provides a method for monitoring the posture and wear of the motion signal of a limb artificial joint. This method is applicable to the system for monitoring the posture and wear of the motion signal of this kind of limb artificial joint. Please refer to Figure 2 As shown, it is a schematic flowchart of a method for monitoring the posture and wear of the motion signal of a limb artificial joint provided by the present application, including the following steps: Step S100: Collect the motion signal data of the artificial joint, and based on the collected motion signal data, perform judgment and analysis on the motion posture according to the motion posture judgment model to identify the motion posture type; Step S200: High-load motion postures, low-load motion postures, and static postures; Monitoring frequency control module: Extract motion features, according to the motion features, introduce a joint fatigue factor for data fusion processing to obtain a motion intensity coefficient, combine the motion intensity coefficient with the real-time wear monitoring frequency, and for different motion posture types, update the subsequent wear monitoring frequency, and process the monitoring of the artificial joint based on the new wear monitoring frequency, and output the wear degree; Step S300: Judge in real time whether the wear degree exceeds the wear degree warning value. If it exceeds the wear degree warning value, obtain the corresponding warning time point and predict the remaining service life of the artificial joint in the future period.

[0010] The basic principles, main features and advantages of the present invention have been shown and described above. Those skilled in the art should understand that the present invention is not limited by the above embodiments. What is described in the above embodiments and the specification only illustrates the principles of 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. A posture and wear monitoring system for limb artificial joint movement signals, characterized in that: Including: A motion posture recognition module: collecting motion signal data of an artificial joint, and based on the collected motion signal data, judging and analyzing the motion posture according to a motion posture judgment model to identify the type of motion posture; Among them, the types of motion postures include: high-load motion postures, low-load motion postures, and static postures; A monitoring frequency control module: extracting motion features, according to the motion features, introducing a joint fatigue factor for data fusion processing to obtain a motion intensity coefficient, combining the motion intensity coefficient with the real-time wear monitoring frequency, for different types of motion postures, updating the subsequent wear monitoring frequency, and processing the monitoring of the artificial joint based on the new wear monitoring frequency to output the wear degree; A wear prediction module: by real-time judging whether the wear degree exceeds the wear degree warning value, if it exceeds the wear degree warning value, obtaining the corresponding warning time point and predicting the remaining service life of the artificial joint in the future period.

2. The posture and wear monitoring system for limb artificial joint movement signals according to claim 1, characterized in that: The process of identifying the type of motion posture is: Collecting motion signal data of the artificial joint and performing data preprocessing; extracting and fusing features from the motion signal data of the artificial joint after preprocessing, inputting the fused features into the motion posture judgment model, and outputting the real-time motion posture type; Among them, when training the motion posture judgment model using a historical data set, the random forest algorithm is used, and the training features are calculated in combination with the adaptive feature weighting algorithm.

3. The posture and wear monitoring system for limb artificial joint movement signals according to claim 1, characterized in that: The process of obtaining motion features is: Motion features include: joint bearing strength value, joint instantaneous load value, motion amplitude value, vibration amplitude value, and muscle activation strength value; Setting a sliding time window, and obtaining pressure data, linear acceleration data, and electromyogram intensity data within each time window; Taking the mean value of the pressure data within the time window as the joint bearing strength value, and taking the peak value of the pressure data within the time window as the joint instantaneous load value; Synthesizing the linear acceleration data in three-dimensional space to obtain a synthesized linear acceleration, taking the mean value of the synthesized linear acceleration within the time window as the motion amplitude value, and taking the peak value of the synthesized linear acceleration within the time window as the vibration amplitude value; Taking the root mean square of the electromyogram intensity data within the time window as the muscle activation strength value.

4. The posture and wear monitoring system for limb artificial joint movement signals according to claim 1, characterized in that: The process of obtaining the motion intensity coefficient is: Normalizing the motion features to obtain a motion intensity feature vector; obtaining a historical data set and training a differentiated weight matrix based on the historical data set; Performing weighted fusion by combining the motion intensity feature vector and the weight matrix to obtain a basic motion intensity; calculating the weight of the basic motion intensity and the joint fatigue factor to obtain the motion intensity coefficient.

5. The posture and wear monitoring system for limb artificial joint movement signals according to claim 1, characterized in that: The process of obtaining the joint fatigue factor: Extracting the proportion of the middle frequency band energy corresponding to the electromyogram signal in the motion signal data of the artificial joint, extracting the minimum middle frequency band energy and the maximum middle frequency band energy in the historical data, and performing a difference calculation to obtain the historical middle frequency band energy extreme difference, performing a difference calculation between the middle frequency band energy proportion and the small middle frequency band energy, and then performing a ratio calculation with the historical middle frequency band energy extreme difference to obtain the joint fatigue factor.

6. The posture and wear monitoring system for limb artificial joint movement signals according to claim 1, characterized in that: The process of updating the wear monitoring frequency is: If the motion posture type is a static posture, stop wear monitoring; If the motion posture type is a high-load motion posture or a low-load motion posture, obtain the real-time wear monitoring frequency , calculate the new wear monitoring frequency after update; For high-load exercise postures, the calculation formula is: ; where is the new wear monitoring frequency, and k is the exercise intensity coefficient; For a low-load exercise posture, the calculation formula is as follows: ; where is the new wear monitoring frequency, and k is the exercise intensity coefficient.

7. The posture and wear monitoring system for limb artificial joint movement signals according to claim 1, characterized in that: The process of outputting the wear degree is: Collect the motion signal data of the artificial joint according to the new wear monitoring frequency, and calculate the wear amount using the Archard model in the classical wear model based on the collected motion signal data of the artificial joint; calculate the wear increment of the real-time motion, accumulate the historical wear increments, and output the wear degree.

8. The posture and wear monitoring system for limb artificial joint movement signals according to claim 1, characterized in that: Predict the remaining service life of the artificial joint in the future period. The process is as follows: Obtain the time point corresponding to the wear degree warning value as the warning time point, and obtain the wear degree corresponding to the warning time point to the warning time point as the basic wear degree data; Conduct a trend analysis of the basic wear degree data, including judging whether the basic wear degree data is linearly changing or non-linearly changing; For the basic wear degree showing a linear change, fit a linear prediction equation. The slope of the linear prediction equation is the wear rate. Calculate the difference between the wear degree threshold and the wear degree value corresponding to the warning time point, and calculate the ratio of the difference to the wear rate to obtain the remaining service life; For the basic wear degree showing a non-linear change, fit a non-linear prediction equation; Based on the non-linear prediction equation, take the wear degree threshold as known, calculate the corresponding predicted time value, and calculate the difference between the predicted time value and the warning time value to obtain the remaining service life.

9. The attitude and wear monitoring system for the movement signals of artificial joints of four limbs according to claim 8, characterized in that: Conduct a trend analysis. The process is as follows: Use the methods of statistical fitting and residual analysis to judge, and conduct linear fitting and non-linear fitting on the basic wear degree data respectively; Conduct a univariate linear regression on the basic wear data, fit a univariate linear regression equation, and calculate the sum of squared linear residuals; at the same time, conduct a quadratic polynomial fitting on the basic wear data and calculate the sum of squared non-linear residuals; Compare the linear fitting and non-linear fitting through the AIC information criterion; if the sum of squared non-linear residuals is less than the sum of squared linear residuals and the AIC value corresponding to the non-linearity is smaller, it is determined to be a non-linear change. If the residuals are randomly distributed and there is no obvious curvature, it is determined to be a linear change.

Images