A urinary surgery postoperative rehabilitation effect monitoring system and method

Abstract

The application discloses a kind of urological postoperative rehabilitation effect monitoring system, method, specifically relates to the field of health informatics, to solve the problem that existing rehabilitation monitoring technology cannot identify action mode progressive deterioration trend;It is through the rehabilitation training movement data sequence collected by the patient intelligent terminal device, constructs personal action state space and extracts principal component subspace, calculates new data reconstruction error and analyzes its bimodal or multimodal characteristics of statistical distribution to determine the progressive drift of action mode, maps the peak value of distribution as the dominant action mode and determines the development stage by analyzing the time sequence competition relationship of dominant right, finally assesses risk level based on development stage and generates early warning information, realize the dynamic monitoring and early warning of rehabilitation effect.

Description

Technical Field

[0001] This invention relates to the field of health informatics technology, and in particular to a system and method for monitoring the postoperative rehabilitation effect in urology. Background Technology

[0002] In urological clinical practice, the management of postoperative rehabilitation is crucial, as its effectiveness directly impacts patients' functional recovery level and quality of life. Currently, postoperative rehabilitation monitoring can be conducted based on telemedicine and mobile health technologies. This typically relies on patients recording their rehabilitation training progress using smart terminal devices, and may utilize built-in sensors to collect some movement data. Existing solutions focus on static assessments of the compliance of individual rehabilitation training movements or trend tracking of patient-reported symptom indicators, aiming to improve patient compliance and provide clinicians with a remote observation window.

[0003] However, existing technologies have significant drawbacks: they cannot effectively identify and warn of the progressive deterioration trend of movement patterns formed by patients during long-term rehabilitation training. This hidden pattern drift phenomenon manifests as a systematic deviation in the overall pattern of training movements, even if they do not significantly exceed the standard in a single assessment. This leads to a continuous decline in training effectiveness and even negative effects. Because the existing monitoring system lacks the ability to analyze the evolutionary trajectory of implicit patterns in temporal behavioral data, clinical rehabilitation guidance faces the problem of insufficient foresight and difficulty in effectively intervening before the functional recovery bottleneck occurs. Summary of the Invention

[0004] This invention addresses the technical problems existing in the prior art by providing a system and method for monitoring the postoperative rehabilitation effect in urology.

[0005] The technical solution of the present invention to solve the above-mentioned technical problems is as follows: A method for monitoring the postoperative rehabilitation effect in urology includes: S1. Obtain rehabilitation training exercise data sequences at multiple time points collected by the patient through a smart terminal device; S2. Construct a personal movement state space based on rehabilitation training exercise data sequences, and extract principal component subspaces representing the patient's personalized movement patterns through principal component analysis; S3. Calculate the reconstruction error of the newly acquired rehabilitation training exercise data in the principal component subspace; S4. Analyze whether the statistical distribution of the reconstruction error shows bimodal or multimodal characteristics. If it shows bimodal or multimodal characteristics, it is determined that there is a progressive drift of the action mode. S5. When there is progressive drift in action patterns, different peaks in the statistical distribution are mapped to different dominant action patterns, and the development stage of progressive drift in action patterns is determined by analyzing the temporal competition relationship of dominance among dominant action patterns. S6. Based on the developmental stages of progressive drift in movement patterns, assess the risk level of rehabilitation effectiveness and generate corresponding early warning information.

[0006] Furthermore, acquire rehabilitation training exercise data sequences from multiple time points collected by the patient through a smart terminal device, including: Collect rehabilitation training exercise data sequences using accelerometers and gyroscopes in smartphones or wearable devices; The rehabilitation training exercise data sequence includes acceleration and angular velocity data collected at multiple consecutive time points; Both acceleration and angular velocity data are contained in data components along the three orthogonal axes in a three-dimensional spatial coordinate system.

[0007] Furthermore, a personal movement state space is constructed based on rehabilitation training exercise data sequences, and principal component analysis is used to extract principal component subsets representing the patient's personalized movement patterns, including: The acceleration and angular velocity data at each time point in the rehabilitation training exercise data sequence are combined to form the motion state vector in the individual motion state space; The state space matrix of the individual action state space is constructed by combining the action state vectors at multiple time points. Principal component analysis is performed on the state space matrix, and the top principal components whose cumulative contribution rate exceeds the preset contribution rate threshold are selected to form the principal component subspace.

[0008] Furthermore, the reconstruction error of the newly acquired rehabilitation training exercise data in the principal component subset space is calculated, including: The newly collected rehabilitation training exercise data is used to construct a new motion state vector; Calculate the projection vector of the new action state vector onto the principal component subspace; The reconstruction error is obtained by calculating the Euclidean distance between the new action state vector and the projection vector.

[0009] Furthermore, the statistical distribution of the reconstruction error is analyzed to determine whether it exhibits bimodal or multimodal characteristics. If it does, a progressive drift in the action pattern is identified, including: The reconstruction errors calculated over a continuous time period are collected to form a reconstruction error sequence; The probability density distribution curve of the reconstruction error is obtained by kernel density estimation of the reconstruction error sequence; The number of peaks in the probability density distribution curve is detected, and when two or more significant peaks are detected, it is determined that the curve exhibits bimodal or multimodal characteristics. Analyze the trend of the distance between adjacent peaks. When the distance between peaks shows a continuous increasing trend, it confirms the existence of progressive drift in the action pattern.

[0010] Furthermore, the probability density distribution curve of the reconstruction error is obtained by kernel density estimation of the reconstruction error sequence, which includes: using a Gaussian kernel function to perform kernel density estimation on the reconstruction error sequence, and generating a smooth probability density distribution curve by calculating the bandwidth parameter based on the standard deviation of the reconstruction error sequence.

[0011] Furthermore, when progressive drift of action patterns exists, different peaks in the statistical distribution are mapped to different dominant action patterns, and the development stages of progressive drift of action patterns are determined by analyzing the temporal competition for dominance among dominant action patterns, including: Based on the significant peaks in the probability density distribution curve, each significant peak is mapped to an independent dominant action pattern; Calculate the peak area corresponding to each dominant action pattern as the dominance strength of the corresponding dominant action pattern within the current time window; The dominance strength of each dominant action pattern is repeatedly calculated within multiple consecutive time windows to form a dominance strength sequence for each dominant action pattern. Analyze the temporal competition relationship of dominance among the dominant action patterns of different dominant action patterns, including identifying the contrast between the growth trend and the decline trend in the dominance intensity sequence; The development stage of the progressive drift of the action pattern is determined by comparing the growth trend and the decline trend.

[0012] Furthermore, the analysis of the temporal competition relationship of dominance among the dominant action patterns of different dominant action patterns includes: applying moving average smoothing to the dominance strength sequence of each dominant action pattern, using linear regression analysis to extract the growth trend and decay trend respectively, and quantifying the relative change trend by calculating the difference in the slope of the trend line.

[0013] Furthermore, based on the developmental stages of progressive movement pattern drift, the risk level of rehabilitation effectiveness is assessed and corresponding early warning information is generated, including: The developmental stages of progressive drift in movement patterns are matched with a preset risk level mapping table to determine the corresponding risk level of rehabilitation effect. Select the appropriate early warning level based on the risk level of rehabilitation outcomes; Based on the warning level, generate warning information containing specific rehabilitation guidance and suggestions; The generated warning information is displayed through the user interface of smart terminal devices; The preset risk level mapping table includes: defining the risk level mapping relationship between the nascent stage of progressive movement pattern drift and the medium-risk stage, and the stable stage and the high-risk stage, forming a risk level mapping table for assessing the risk level of rehabilitation effect.

[0014] On the other hand, the present invention provides a postoperative rehabilitation effect monitoring system for urological surgery, comprising: The sequence acquisition module is used to acquire rehabilitation training exercise data sequences at multiple time points collected by the patient through a smart terminal device; The spatial extraction module is used to construct a personal movement state space based on rehabilitation training exercise data sequences, and to extract principal component subspaces that represent the patient's personalized movement patterns through principal component analysis. The error calculation module is used to calculate the reconstruction error of newly acquired rehabilitation training exercise data in the principal component subspace. The drift determination module is used to analyze whether the statistical distribution of the reconstruction error exhibits bimodal or multimodal characteristics. When bimodal or multimodal characteristics are present, it is determined that there is a progressive drift in the action mode. The stage determination module is used to map different peaks in the statistical distribution to different dominant action patterns when there is progressive drift in action patterns, and to determine the development stage of progressive drift in action patterns by analyzing the temporal competition relationship of dominance among dominant action patterns. The early warning assessment module is used to assess the risk level of rehabilitation effectiveness and generate corresponding early warning information based on the developmental stage of progressive drift in movement patterns.

[0015] The beneficial effects of this invention are: 1. By constructing a personal movement state space and a principal component subspace, the system can extract a feature subspace representing personalized movement patterns from long-term rehabilitation training data of patients, thereby achieving a precise characterization of movement pattern evolution. By calculating the reconstruction error of newly acquired data in the principal component subspace and analyzing its statistical distribution characteristics, the system can keenly capture the progressive drift phenomenon in movement patterns, overcoming the limitations of traditional single-action assessment. This allows potential deterioration trends to be identified before the training effect undergoes substantial decline. By mapping the multi-peak features in the statistical distribution to different dominant movement patterns and analyzing the temporal competition relationship of their dominance, the system can dynamically track each development stage of movement pattern evolution, thus providing continuous and fine-grained assessment basis for the dynamic monitoring of rehabilitation effects.

[0016] 2. Risk level assessment and early warning information generation based on the developmental stages of progressive movement pattern drift enable the rehabilitation monitoring process to have a proactive intervention capability. By establishing a mapping relationship between developmental stages and risk levels, the system can automatically generate graded early warning information and targeted rehabilitation guidance suggestions according to the severity of pattern evolution, effectively improving the personalization and accuracy of rehabilitation management. Through in-depth analysis of temporal behavioral data, long-term dynamic monitoring of rehabilitation training quality is achieved. This avoids the over-reliance on the compliance of single movements in traditional methods and can prevent the occurrence of functional recovery bottlenecks by identifying pattern drift early, thereby significantly improving the initiative and effectiveness of postoperative rehabilitation management. Attached Figure Description

[0017] Figure 1 This is a flowchart of a method for monitoring the postoperative rehabilitation effect in urology according to the present invention; Figure 2 This is a schematic diagram of the structure of a postoperative rehabilitation effect monitoring system for urology according to the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Example 1: Figure 1 This invention provides a method for monitoring the postoperative rehabilitation effect in urology, comprising: S1. Obtain rehabilitation training exercise data sequences at multiple time points collected by the patient through a smart terminal device; S2. Construct a personal movement state space based on rehabilitation training exercise data sequences, and extract principal component subspaces representing the patient's personalized movement patterns through principal component analysis; S3. Calculate the reconstruction error of the newly acquired rehabilitation training exercise data in the principal component subspace; S4. Analyze whether the statistical distribution of the reconstruction error shows bimodal or multimodal characteristics. If it shows bimodal or multimodal characteristics, it is determined that there is a progressive drift of the action mode. S5. When there is progressive drift in action patterns, different peaks in the statistical distribution are mapped to different dominant action patterns, and the development stage of progressive drift in action patterns is determined by analyzing the temporal competition relationship of dominance among dominant action patterns. S6. Based on the developmental stages of progressive drift in movement patterns, assess the risk level of rehabilitation effectiveness and generate corresponding early warning information.

[0020] S1. Obtain rehabilitation training exercise data sequences from multiple time points collected by the patient through a smart terminal device. Specifically, this is implemented as follows: In acquiring rehabilitation training exercise data sequences from multiple time points collected by patients via smart terminal devices, data is first collected using the accelerometer and gyroscope sensors built into smartphones or wearable devices. The accelerometer measures the linear acceleration of the device in three-dimensional space, while the gyroscope measures the rotational angular velocity. In practice, the sampling frequency needs to be pre-set. This setting is based on a comprehensive consideration of the typical frequency range of human rehabilitation training movements and the optimal performance of the smart terminal device's sensors. For example, the frequency of human rehabilitation training movements is typically between 1 Hz and 5 Hz, while the available sampling frequency range of smart terminal device sensors is generally between 10 Hz and 100 Hz. Therefore, by balancing the characteristic frequency of the movement with device performance, the sampling frequency is set to 50 Hz. The process of determining this value includes analyzing the periodic characteristics of common rehabilitation training movements and testing the data quality of the device sensors at different sampling frequencies. Finally, a sampling frequency value that can adequately capture movement details without exceeding the device's processing capabilities is selected. During the data collection process, the smart terminal device is fixed to a specific part of the patient's body where rehabilitation training movements are performed. The choice of device fixing position is determined based on the type of rehabilitation training. For example, when monitoring pelvic floor muscle rehabilitation training, the device can be fixed to the patient's waist, and when monitoring upper limb rehabilitation training, the device can be fixed to the patient's wrist. This ensures that the sensor coordinate system and the human motion coordinate system maintain a relatively consistent positional relationship. Fixing methods include using elastic straps or special fixing sleeves to prevent the device from sliding or rotating during movement.

[0021] The rehabilitation training exercise data sequence includes acceleration and angular velocity data collected at multiple consecutive time points. The data acquisition time corresponding to each time point is uniformly marked by the system clock of the smart terminal device, forming a data point sequence arranged strictly in chronological order. The timestamp accuracy reaches the millisecond level to ensure the accuracy of the time-series data. In the continuous time series, the time interval between adjacent time points is determined by a preset sampling frequency. For example, when the sampling frequency is 50 Hz, the time interval between adjacent time points is 0.02 seconds. The stability of the time interval is ensured by the real-time scheduling function of the device's operating system, avoiding fluctuations in the acquisition interval due to system load. Both acceleration and angular velocity data are contained as data components along three orthogonal axes in a three-dimensional spatial coordinate system. The three-dimensional spatial coordinate system is defined using the right-hand coordinate system rule, where the X-axis is parallel to the long side of the smart terminal device screen, the Y-axis is parallel to the short side of the screen, and the Z-axis is perpendicular to the screen plane. The orientation of the coordinate system is calibrated at the start of acquisition using the device's built-in orientation sensor to ensure the consistency of the data component orientations. Each data component is accompanied by a clear positive or negative direction identifier during acquisition. The direction identifier is set based on the coordinate system definition. For example, when the device accelerates along the positive X-axis, the recorded acceleration component is a positive value, and when it accelerates in the opposite direction, the recorded acceleration component is a negative value. The determination of positive and negative directions is achieved through the output polarity of the sensor hardware.

[0022] When collecting rehabilitation training exercise data sequences through smart terminal devices, it is necessary to ensure the continuity and integrity of sensor data acquisition. During data acquisition, the operating system of the smart terminal device reads the raw measurement values ​​from the accelerometer and gyroscope sensors in real time and converts the measurement values ​​into physical quantities in standard International System of Units (SI) through the sensor data interface. The unit for acceleration data is meters per second squared, and the unit for angular velocity data is radians per second. The unit conversion is based on the calibration parameters provided by the sensor manufacturer. For example, if the raw output value of the accelerometer is a digital quantity, it is converted into a physical quantity value by multiplying it by a sensitivity coefficient. The sensitivity coefficient is determined through the device's factory calibration process. The converted data is stored in the device's temporary buffer according to a preset data structure, forming a data sequence arranged in chronological order. The storage format of each data point includes a timestamp field and a corresponding sensor measurement value field. The timestamp field records the precise moment of data acquisition, and the sensor measurement value field contains acceleration components and angular velocity components in three axes. The data structure design ensures that each field has a fixed byte length to facilitate subsequent parsing and processing.

[0023] During data acquisition, necessary preprocessing of sensor data is required to ensure data quality. Preprocessing includes outlier detection and data smoothing. Outlier detection is achieved by setting reasonable ranges for physical quantities, based on the physiological limits of human rehabilitation training movements. For example, the acceleration value of human rehabilitation training movements typically does not exceed 10 m / s², and the angular velocity value typically does not exceed 20 radians per second. When outliers exceeding these ranges are detected, the system automatically marks them and replaces them using linear interpolation of adjacent data points. The threshold for outlier detection is determined by analyzing historical rehabilitation training data. Data smoothing employs a moving average filtering method. The size of the filtering window is determined based on a combination of the sampling frequency and the frequency of motion characteristics. For example, when the sampling frequency is 50 Hz, the filtering window can be set to 5 data points, corresponding to a 0.1-second time window. The window size is chosen based on the frequency components of the motion signal, ensuring that high-frequency noise is filtered out while retaining motion characteristics. The filtering calculation is achieved by taking the arithmetic mean of the data points within the window. The preprocessed rehabilitation training motion data sequence can then be used for subsequent processing steps.

[0024] The storage and management of rehabilitation training exercise data sequences employs a circular buffer mechanism. The buffer size is calculated based on the monitoring duration and sampling frequency. For example, when continuous monitoring of a 30-second training process is required at a sampling frequency of 50 Hz, the buffer needs to hold data from 1500 consecutive time points. The buffer size is set based on a balance between memory resources and data retention requirements, calculated using a preset maximum monitoring duration parameter. When the buffer is full, the oldest data is overwritten by the newly acquired data, ensuring that the complete exercise data sequence within the latest time period is always retained.

[0025] S2. Construct a personal movement state space based on rehabilitation training exercise data sequences, and extract principal component subsets representing the patient's personalized movement patterns through principal component analysis. The specific implementation is as follows: In the step of constructing a personal motion state space based on rehabilitation training exercise data sequences, the acceleration and angular velocity data at each time point in the rehabilitation training exercise data sequence are first combined to form a motion state vector in the personal motion state space. The construction of the motion state vector involves extracting the data components of the acceleration data and angular velocity data collected at each time point in the rehabilitation training exercise data sequence along the three orthogonal axes in the three-dimensional coordinate system, and then connecting these data components in a preset order to form a multidimensional vector. For example, the motion state vector at each time point contains six data components, three of which correspond to the acceleration data components along the X-axis, Y-axis, and Z-axis, and the other three correspond to the angular velocity data components along the X-axis, Y-axis, and Z-axis. The value of each data component in the vector is directly derived from the sensor data collected at the corresponding time point in the rehabilitation training exercise data sequence. The order of the data components remains fixed during the construction process; for example, they are always combined in the order of acceleration X-axis component, acceleration Y-axis component, acceleration Z-axis component, angular velocity X-axis component, angular velocity Y-axis component, and angular velocity Z-axis component. The number of dimensions of the motion state vector is determined by the data type and the number of spatial directions of the sensors used. For example, when using an accelerometer and a gyroscope sensor, and each sensor outputs three spatial direction data, the motion state vector has six dimensions. After the vector is constructed, each motion state vector represents the patient's complete motion state at a single time point. The physical unit of each data component in the vector is consistent with the original sensor data: the unit of acceleration data components is meters per second squared, and the unit of angular velocity data components is radians per second.

[0026] Next, the motion state vectors from multiple time points are combined to construct the state space matrix of the individual motion state space. The individual motion state space is a multi-dimensional vector space defined by the state space matrix. Motion state vectors are the vector elements in the individual motion state space, and the state space matrix is ​​its matrix representation. The dimension of the individual motion state space is determined by the dimension of the motion state vectors. The state space matrix is ​​constructed by arranging the motion state vectors from the continuous time series in chronological order as rows of the matrix. For example, each motion state vector is a row, and each data component in the vector is a column element of that row. The number of rows in the matrix corresponds to the number of time points used, and the number of columns corresponds to the dimension of the motion state vectors. The selection of time points is based on the completeness and continuity of the rehabilitation training exercise data sequence. For example, all continuous time points within a complete rehabilitation training cycle are selected to ensure that the matrix comprehensively reflects the patient's movement pattern characteristics. Strict consistency in the chronological order must be maintained during the construction of the state space matrix; that is, the order of the rows in the matrix must completely correspond to the chronological order of the data collection. Once the matrix is ​​constructed, its dimensions are determined by the number of time points and the vector dimensions. For example, when using 100 time points and a six-dimensional action-state vector, the state space matrix has a dimension of 100 rows and 6 columns. As the input data for subsequent principal component analysis, the state space matrix needs to ensure that there are no missing or outlier values. Therefore, data quality checks were performed in the previous step of constructing the action-state vector, for example, by using range checks to exclude data points that exceed a reasonable physical range.

[0027] When performing principal component analysis on a state-space matrix, the matrix data must first be standardized to eliminate dimensional differences between dimensions. Standardization is achieved by calculating the arithmetic mean and standard deviation of each data dimension in the state-space matrix. For example, for each column of data in the state-space matrix, the arithmetic mean and standard deviation of all elements in that column are first calculated. Then, each element in that column is subtracted from its arithmetic mean and divided by its standard deviation. The standardized state-space matrix has an arithmetic mean of zero and a standard deviation of one for each data dimension. The arithmetic mean and standard deviation are calculated based on all element values ​​of that data dimension in the matrix. For example, for a state-space matrix with 100 rows and 6 columns, the arithmetic mean of each column is the sum of the 100 element values ​​divided by 100, and the standard deviation of each column is the square root of the sum of the squares of the differences between the element values ​​and the arithmetic mean, divided by 99. Standardization ensures that the values ​​of different data dimensions are of the same order of magnitude, preventing certain dimensions from dominating the principal component analysis results due to their larger numerical ranges.

[0028] Then, the covariance matrix of the standardized state space matrix is ​​calculated. The covariance matrix is ​​calculated by multiplying the standardized state space matrix by its transpose and then dividing by the number of rows minus one. For example, for a standardized 100x6 state space matrix, first calculate its transpose to obtain a 6x100 matrix, then multiply the original matrix by the transpose to obtain a 6x6 matrix, and finally divide by 99 to obtain the covariance matrix. The dimension of the covariance matrix is ​​the same as the dimension of the action-state vector. For example, when the action-state vector is six-dimensional, the covariance matrix is ​​a 6x6 square matrix. Each element in the covariance matrix represents the covariance value between two data dimensions, and the diagonal elements represent the variance values ​​of each data dimension.

[0029] Next, eigenvalue decomposition is performed on the covariance matrix. Eigenvalue decomposition is achieved by solving for the eigenvalues ​​and eigenvectors of the covariance matrix. Eigenvalues ​​represent the magnitude of the variance contribution of each principal component, and eigenvectors represent the direction of each principal component. The specific calculation process of eigenvalue decomposition includes constructing the characteristic polynomial of the covariance matrix, solving for the roots of the characteristic polynomial to obtain the eigenvalues, and then solving for the corresponding eigenvector for each eigenvalue. The eigenvalues ​​are arranged in descending order, and the corresponding eigenvectors are the directions of the principal components. The contribution rate of each principal component is calculated by dividing its corresponding eigenvalue by the sum of all eigenvalues; the sum of the contribution rates of all principal components is 100%. The cumulative contribution rate is obtained by summing the contribution rates of each principal component in descending order of eigenvalues. For example, the cumulative contribution rate of the first principal component equals its contribution rate, the cumulative contribution rate of the second principal component equals the contribution rate of the first principal component plus the contribution rate of the second principal component, and so on.

[0030] A preset contribution rate threshold is used as the selection criterion when selecting principal components. The preset contribution rate threshold is set based on a comprehensive consideration of historical rehabilitation training data analysis and clinical needs. For example, by analyzing the contribution rate distribution patterns of principal components in a large amount of patient rehabilitation training data, it was found that when the cumulative contribution rate reaches 85%, the main features of the original movement pattern can be effectively preserved. Therefore, the preset contribution rate threshold is set to 85%. The specific value of the preset contribution rate threshold can be adjusted according to different rehabilitation stages or different patient groups. The adjustment is based on factors including the type of rehabilitation training, patient age, and rehabilitation progress. For example, a higher threshold, such as 90%, may be set for patients in the early stages of rehabilitation to retain more movement details, while a lower threshold, such as 80%, may be set for patients in the stable phase to simplify the model structure. The principal component selection process starts with the principal component with the largest contribution rate and selects sequentially until the cumulative contribution rate of the selected principal components first exceeds the preset contribution rate threshold. These selected principal components constitute the principal component subspace. Each principal component in the principal component subspace corresponds to a feature vector, and the space spanned by these feature vectors is the principal component subspace representing the patient's personalized movement pattern. The dimension of the principal component subspace is equal to the number of principal components selected. For example, when three principal components are selected, the principal component subspace is a three-dimensional space. Once the principal component subspace is constructed, it can be used in subsequent steps to calculate the reconstruction error.

[0031] S3. Calculate the reconstruction error of the newly acquired rehabilitation training exercise data in the principal component subspace, specifically as follows: In the step of calculating the reconstruction error of newly acquired rehabilitation training exercise data in the principal component subspace, the newly acquired rehabilitation training exercise data is first constructed into a new motion state vector. This newly acquired rehabilitation training exercise data originates from a sequence of rehabilitation training exercise data collected via a smart terminal device at a later time point than the time point used to construct the personal motion state space in step S1. The newly acquired time point is later than the time point of the data used to construct the personal motion state space. The construction process of the new motion state vector is based on the latest rehabilitation training exercise data collected by the smart terminal device, which includes acceleration data output by the accelerometer sensor and angular velocity data output by the gyroscope sensor at the current time point. During construction, the components of the acceleration data in the X-axis, Y-axis, and Z-axis directions of the newly acquired rehabilitation training exercise data, as well as the components of the angular velocity data in the X-axis, Y-axis, and Z-axis directions of the three-dimensional coordinate system, are extracted and combined into a multi-dimensional vector in the exact same order as during the construction of the personal motion state space. For example, the new motion state vector contains six data components, arranged in the following order: acceleration X-axis component, acceleration Y-axis component, acceleration Z-axis component, angular velocity X-axis component, angular velocity Y-axis component, and angular velocity Z-axis component. The value of each data component comes directly from the raw data collected by the sensor. The physical units of the data components remain the same: the unit for acceleration data components is meters per second squared, and the unit for angular velocity data components is radians per second. The dimension of the new motion state vector is consistent with the dimension of the motion state vector in the personal motion state space. For example, when the personal motion state space uses a six-dimensional motion state vector, the new motion state vector also adopts a six-dimensional structure. After the new motion state vector is constructed, it is necessary to verify whether the values ​​of each data component in the vector are within a reasonable physical range. For example, the absolute value of the acceleration data component usually does not exceed 10 meters per second squared, and the absolute value of the angular velocity data component usually does not exceed 20 radians per second. If any data components are found to exceed these ranges, the data needs to be re-collected.

[0032] Next, the projection vector of the new action state vector onto the principal component subspace is calculated. The calculation of the projection vector requires the use of eigenvectors from the principal component subspace, which are derived from the eigenvector set obtained through principal component analysis of the individual action state space. The calculation process begins by standardizing the new action state vector. Standardization uses the arithmetic mean and standard deviation of the data in each column of the state space matrix during the construction of the individual action state space. For example, for each data component of the new action state vector, the arithmetic mean of the corresponding dimension is subtracted, and then divided by the standard deviation of the corresponding dimension. The values ​​of the arithmetic mean and standard deviation are derived from the statistics of the state space matrix during the construction of the individual action state space; these statistics were calculated and saved during the construction of the principal component subspace. The mean of each data component of the standardized new action state vector is zero, and the standard deviation is one. Then, the projection components of the standardized new action state vector onto each principal component direction in the principal component subspace are calculated. The calculation of the projection components is achieved by performing a dot product operation between the standardized new action state vector and the corresponding eigenvector of each principal component. The specific process of the dot product operation is to multiply each data component of the standardized new action state vector by the corresponding data component of the eigenvector, and then sum all the product results. For example, when the principal component subspace contains three principal components, three projection components need to be calculated separately, with each projection component corresponding to a principal component direction. The set of all projection components constitutes a projection vector, and the dimension of the projection vector is the same as the number of principal components in the principal component subspace. For example, when the principal component subspace selects three principal components, the projection vector is a three-dimensional vector. The projection vector represents the coordinate values ​​of the new action state vector in the principal component subspace coordinate system, reflecting the representation of the new data in the personalized action pattern space.

[0033] The reconstruction error is then obtained by calculating the Euclidean distance between the new action state vector and the projected vector. Calculating the Euclidean distance requires reconstructing the projected vector back to the original data space. This reconstruction is achieved by linearly combining the projected vector with the eigenvectors of the principal component subspace. The linear combination involves multiplying each component of the projected vector by the corresponding eigenvector of the principal component, and then summing all the products to obtain the reconstructed action state vector. The reconstructed action state vector has the same dimension as the original new action state vector; for example, if the new action state vector is six-dimensional, the reconstructed action state vector will also be six-dimensional. The reconstructed action state vector needs to be de-standardized to restore the original dimensions. De-standardization uses the arithmetic mean and standard deviation of each data dimension of the state space matrix during the construction of the individual action state space. For example, for each data component of the reconstructed action state vector, it is multiplied by the standard deviation of the corresponding data dimension and then added to the arithmetic mean of the corresponding data dimension. The de-standardized reconstructed action state vector and the new action state vector reside in the same physical dimension space. The Euclidean distance is calculated by summing the squares of the differences between each corresponding data component of the new motion state vector and the reconstructed motion state vector, and then taking the square root of the sum. The difference is calculated for each data component; for example, the first data component of the new motion state vector is subtracted from the first data component of the reconstructed motion state vector to obtain the first difference, the second data component is subtracted from the second data component of the reconstructed motion state vector to obtain the second difference, and so on until all data components are calculated. The sum of squares is obtained by adding the squares of each difference, and the square root is taken using a mathematical square root function to obtain the final Euclidean distance value. The unit of the Euclidean distance is consistent with the unit of the original data; when the data components are in meters per second squared, the unit of the Euclidean distance is also meters per second squared. The Euclidean distance, as a reconstruction error, reflects the degree of difference between the newly acquired rehabilitation training exercise data and the patient's personalized movement pattern; a larger reconstruction error value indicates a greater deviation between the new data and the personalized movement pattern. After the reconstruction error is calculated, this value will be used in subsequent steps to analyze the statistical distribution characteristics of the reconstruction error.

[0034] S4. Analyze whether the statistical distribution of the reconstruction error exhibits bimodal or multimodal characteristics. If it does, it indicates a gradual shift in the action mode. The specific implementation is as follows: In analyzing whether the statistical distribution of reconstruction errors exhibits bimodal or multimodal characteristics, the reconstruction errors calculated over a continuous time period are first collected to form a reconstruction error sequence. The collection of the reconstruction error sequence is based on multiple reconstruction error values ​​obtained by calculating the reconstruction errors of newly acquired rehabilitation training exercise data in the principal component subset space within a continuous time period. These reconstruction error values ​​are arranged in chronological order to form the sequence. The length of the continuous time period is determined according to the rehabilitation training cycle and monitoring requirements; for example, the length of the continuous time period can be set to 30 minutes. This time length is based on the duration of a typical rehabilitation training session, ensuring that the sequence contains a sufficient number of reconstruction error values ​​for statistical analysis. Each reconstruction error value in the reconstruction error sequence corresponds to a specific time point, and the interval between time points is determined by the data acquisition frequency. For example, when the data acquisition frequency is 50 Hz, the time interval between adjacent reconstruction error values ​​is 0.02 seconds. The reconstruction error sequence is stored using an array data structure. The elements in the array store the reconstruction error values ​​in chronological order, and the length of the array is equal to the number of reconstruction error values ​​within the continuous time period. After the reconstruction error sequence is collected, it is necessary to check the integrity and continuity of the sequence. For example, it is necessary to verify whether there are missing values ​​or outliers in the sequence. Missing values ​​are handled by linear interpolation, and outliers are handled by threshold filtering based on standard deviation. For example, when the reconstruction error value exceeds three times the standard deviation of the sequence mean, it is considered an outlier and replaced.

[0035] Kernel density estimation of the reconstruction error sequence yields the probability density distribution curve of the reconstruction error. Kernel density estimation uses a Gaussian kernel function, which takes the form of a bell-shaped curve, the width of which is controlled by a bandwidth parameter. The bandwidth parameter is calculated based on the standard deviation of the reconstruction error sequence; for example, the bandwidth parameter equals the standard deviation of the reconstruction error sequence multiplied by an adjustment factor. The value of the adjustment factor is determined according to the sequence length and data distribution characteristics; for example, when the sequence contains 1000 reconstruction error values, the adjustment factor can be set to 1.06. The specific calculation process of kernel density estimation involves selecting multiple equally spaced points within the reconstruction error value range. For each point, its density estimate is calculated by substituting each reconstruction error value into the Gaussian kernel function, summing the results, and dividing by the product of the sequence length and the bandwidth parameter. For example, selecting 100 equally spaced points within the reconstruction error value range, the density estimate for each point equals the sum of the Gaussian kernel function values ​​of all reconstruction error values ​​divided by 1000 multiplied by the bandwidth parameter. The probability density distribution curve is generated by connecting the density estimates of these points to form a smooth curve, with the horizontal axis representing the reconstruction error value and the vertical axis representing the probability density value. The smoothness of the probability density distribution curve is determined by the bandwidth parameter. The larger the bandwidth parameter, the smoother the curve; the smaller the bandwidth parameter, the more detailed the curve. The specific value of the bandwidth parameter is optimized and selected through cross-validation methods, such as determining the optimal value of the bandwidth parameter by minimizing the mean square integral error.

[0036] Peak detection involves detecting the number of peaks in the probability density distribution curve. When two or more significant peaks are detected, the curve is considered to exhibit bimodal or multimodal characteristics. Peak detection is achieved by identifying local maxima on the probability density distribution curve, defined as points where the density value is greater than that of their neighbors. The determination of significant peaks is based on a peak height threshold, which is set based on the overall height and distribution characteristics of the probability density distribution curve. For example, the peak height threshold can be set to 10% of the maximum density value of the curve. This percentage is determined by analyzing the comparison between peaks and background noise in historical reconstruction error data. The specific process of peak detection includes scanning all points on the probability density distribution curve, identifying all local maxima, and then filtering out points whose peak height is lower than the peak height threshold. The remaining points are the significant peaks. The number of peaks equals the number of significant peaks. For example, detecting two significant peaks indicates a bimodal characteristic, while detecting three or more significant peaks indicates a multimodal characteristic. Peak detection also needs to consider the minimum distance constraint between peaks, for example, setting the minimum distance between peaks to 5% of the reconstruction error range to avoid detecting false peaks that are too close together.

[0037] Analyzing the trend of the distance between adjacent peaks confirms the existence of asymptotic drift in action patterns when the distance between peaks shows a continuous increasing trend. The distance between peaks is calculated based on the coordinate difference of adjacent significant peaks on the horizontal axis of the probability density distribution curve. For example, the distance between the first and second peaks is equal to the horizontal coordinate of the second peak minus the horizontal coordinate of the first peak. The trend of the distance between peaks is analyzed by calculating the distance between peaks within multiple consecutive time windows and observing their changing patterns. For example, the reconstruction error sequence is divided into multiple overlapping time windows, each 5 minutes long with a 1-minute overlap. Kernel density estimation and peak detection are performed within each time window, and the distance between peaks is recorded. The determination of a continuously increasing trend is achieved by calculating the slope of the distance between peaks sequence. The slope calculation uses a linear regression method. The input to the linear regression is the time window number and the corresponding distance between peaks, and the output is the slope of the regression line. When the slope is positive and statistically significant, it is considered a continuously increasing trend. Statistical significance is determined through hypothesis testing, such as using a t-test to calculate the p-value of the slope. When the p-value is less than 0.05, the trend is considered statistically significant. Analysis of the trend in distance changes between peaks also needs to consider the fluctuation range of the distance value. For example, a distance change threshold can be set at 10% of the initial distance value; a trend is only confirmed when the distance increases beyond this threshold.

[0038] S5. When progressive drift of action patterns exists, different peaks in the statistical distribution are mapped to different dominant action patterns, and the development stage of progressive drift of action patterns is determined by analyzing the temporal competition relationship of dominance among dominant action patterns. The specific implementation is as follows: In the step of mapping different peaks in the statistical distribution to different dominant motion patterns when there is progressive drift in motion patterns, the first step is to map each significant peak in the probability density distribution curve to an independent dominant motion pattern. Significant peaks are identified from local maxima on the probability density distribution curve determined by peak detection methods; the density values ​​of these local maxima exceed a preset peak height threshold. Each significant peak corresponds to a local high-density region on the probability density distribution curve, representing the range of concentrated reconstruction error values. The mapping process is achieved by assigning a unique identifier to each significant peak; for example, the first detected significant peak is mapped to the first dominant motion pattern, and the second detected significant peak is mapped to the second dominant motion pattern. The definition of a dominant motion pattern is based on the peak's position on the probability density distribution curve; for example, the range of reconstruction error values ​​corresponding to the peak reflects a specific motion characteristic pattern. During the mapping process, it is necessary to ensure that each significant peak maps to only one dominant motion pattern to avoid duplicate mapping. The number of dominant motion patterns is equal to the number of significant peaks in the probability density distribution curve; for example, when two significant peaks are detected, two independent dominant motion patterns are defined. The attributes of the dominant action pattern include peak position, peak height, and peak width, which are extracted from the probability density distribution curve and stored as pattern features.

[0039] The peak area corresponding to each dominant motion pattern is calculated as the dominance strength of that dominant motion pattern within the current time window. The peak area is calculated based on the integral value of the probability density distribution curve over the region surrounding the peak. The integration interval is determined by finding the valley points on both sides of the peak; these valley points are local minimum points on the probability density distribution curve. For example, for each significant peak, the probability density distribution curve is numerically integrated from the left valley point to the right valley point. The integration method uses the trapezoidal rule, dividing the curve into multiple small trapezoidal regions and summing their areas. The numerical value of the peak area represents the probability weight of the dominant motion pattern appearing within the current time window; a larger area value indicates a higher degree of dominance of the pattern in the motion data. The dimension of dominance strength is probability units, with a value ranging from 0 to 1. The sum of the dominance strengths of all dominant motion patterns is 1. The calculation of dominance strength requires consideration of the normalization of the probability density distribution curve to ensure that the total integrated area equals 1. The accuracy of the peak area calculation is determined by the integration step size; for example, the integration interval can be divided into 100 equal division points for approximate calculation.

[0040] The dominance strength of each dominant movement pattern is repeatedly calculated within multiple consecutive time windows, forming a dominance strength sequence for each dominant movement pattern. The division of consecutive time windows is based on a fixed time length and overlap ratio, for example, each time window is 5 minutes long, with adjacent time windows overlapping by 1 minute. The time window length is set based on the typical cycle of rehabilitation training movements and data analysis needs, for example, by analyzing the average duration of movement pattern changes in historical rehabilitation data to determine the window length. For each time window, the probability density distribution curve is regenerated and significant peaks are detected, then the dominance strength of the dominant movement pattern for each mapping is calculated. The dominance strength sequence is constructed by arranging the dominance strength values ​​of each dominant movement pattern within consecutive time windows in chronological order. Each data point in the sequence contains a timestamp and the corresponding dominance strength value; the timestamp corresponds to the start time of the time window. The length of the dominance strength sequence is equal to the number of time windows; for example, when the monitoring period is 30 minutes and the time window length is 5 minutes, the sequence contains 6 data points. The dominance strength sequence is stored using an array structure, with each dominant movement pattern corresponding to an independent array. The consistency of pattern mapping between time windows needs to be addressed during sequence construction. For example, peak position matching can be used to ensure that the same dominant action pattern has the same identifier in different time windows.

[0041] This analysis examines the temporal competition for dominance among different dominant action patterns, including identifying the contrast between increasing and decreasing trends in the dominance intensity sequence. The temporal competition for dominance among dominant action patterns is revealed by analyzing the relative trends in the dominance intensity sequences of different dominant action patterns. Dominant action patterns exhibiting an increasing trend in their dominance intensity sequences are defined as emerging dominant action patterns, while those exhibiting a decreasing trend are defined as existing dominant action patterns. Emerging dominant action patterns represent newly forming action patterns, while existing dominant action patterns represent previously formed action patterns. The temporal competition for dominance refers to the waxing and waning relationship of the dominance intensity of different dominant action patterns over time. The analysis of relative trends begins by applying a moving average smoothing process to the dominance intensity sequence of each dominant action pattern. The window size for the moving average is determined based on the sequence sampling frequency and period of change; for example, when the time window interval is 5 minutes, the moving average window can be set to 3 data points. The specific value of the moving average window size is determined by analyzing the autocorrelation characteristics of the sequence, for example, selecting the smallest window size that preserves the main trend after smoothing the sequence. Moving averages are calculated by taking the arithmetic mean of consecutive data points in the sequence. For example, for a given point in the sequence, the average of its value and the two preceding points is calculated as the smoothed value. The smoothed sequence is used for trend analysis to reduce noise interference. Then, linear regression analysis is used to extract the growth and decline trends of the dominance strength sequence for each dominant action pattern. The input to linear regression is the time window number and the corresponding dominance strength value, and the output is the slope and intercept of the regression line. Linear regression uses the least squares method to fit the line, which is calculated by minimizing the squared difference between the actual and predicted values. A growth trend corresponds to a positive slope value, and a decline trend corresponds to a negative slope value. Trend comparison quantifies the relative change trend by calculating the difference in the slopes of the trend lines for different dominant action patterns. For example, the relative change is obtained by subtracting the slope of the second dominant action pattern from the slope of the first dominant action pattern. The significance of the relative change trend is evaluated through statistical tests, such as calculating the confidence interval of the slope. A trend is considered significant when the confidence interval does not contain zero.

[0042] The development stage of progressive drift in movement patterns is determined by comparing the growth and decline trends. The stage is determined based on the relative changing trends of the dominance intensity sequences of different dominant movement patterns. For example, when an emerging dominant movement pattern shows an increasing trend while the existing dominant movement pattern shows a declining trend, it is considered a development stage. The specific stages of development include the nascent stage, the development stage, and the stable stage. The nascent stage is characterized by the emergence of a new dominant movement pattern but with low dominance intensity. The development stage is characterized by the continuous increase in the dominance intensity of the new dominant movement pattern while the dominance intensity of the existing pattern decreases. The stable stage is characterized by the dominance intensity of each dominant movement pattern tending to stabilize. The determination rule is implemented by setting a threshold for the difference in trend slopes. For example, when the difference between the slope of the growth trend and the slope of the decline trend exceeds 0.05, it is considered a development stage. The threshold is established based on historical data analysis, such as determining the threshold range by statistically analyzing the correspondence between trend changes and clinical stages in multiple rehabilitation cases. The output of the development stage is a classification label, used in subsequent risk assessment steps.

[0043] The identification of emerging dominant action patterns and existing dominant action patterns is based on the trend analysis results of the dominance strength sequence. When the linear regression slope of a certain dominant action pattern is positive, it is identified as an emerging dominant action pattern, and when the linear regression slope of a certain dominant action pattern is negative, it is identified as an existing dominant action pattern.

[0044] The criterion for determining that the dominance intensity tends to be stable is that the absolute value of the linear regression slope of the dominance intensity sequence of all dominant movement patterns is lower than the preset stability determination slope threshold. The preset stability determination slope threshold is set based on the statistical results of the slope distribution when the movement pattern enters the stable period in historical rehabilitation data. For example, the preset stability determination slope threshold can be set to 0.01. When the absolute value of the linear regression slope of each dominant movement pattern is less than 0.01, the dominance intensity is determined to be stable, that is, the intensity of the dominant movement pattern no longer changes significantly by increasing or decreasing.

[0045] S6. Based on the developmental stages of progressive drift in movement patterns, assess the risk level of rehabilitation effectiveness and generate corresponding early warning information. The specific implementation is as follows: In the step of assessing the risk level of rehabilitation effect and generating corresponding early warning information based on the developmental stages of progressive movement pattern drift, the developmental stages of progressive movement pattern drift are first matched with a preset risk level mapping table to determine the corresponding rehabilitation effect risk level. The preset risk level mapping table is constructed based on the experience of clinical rehabilitation medicine experts and historical rehabilitation data analysis. Mapping rules are established by collecting and analyzing a large amount of data on the correspondence between the developmental stages of progressive movement pattern drift and clinical rehabilitation effects in post-urological surgery patients. The preset risk level mapping table is stored using a key-value pair data structure, where the key is the developmental stage of progressive movement pattern drift, and the value is the corresponding rehabilitation effect risk level. The developmental stages include the nascent stage, the developmental stage, and the stable stage; the rehabilitation effect risk levels include low risk, medium risk, and high risk. The matching process is achieved by querying the preset risk level mapping table. The input is the developmental stage of progressive movement pattern drift, and the output is the corresponding rehabilitation effect risk level. For example, when the input developmental stage is the nascent stage, querying the mapping table yields a low-risk rehabilitation effect risk level. The maintenance of the preset risk level mapping table includes regularly updating the mapping relationships. Updates are based on newly accumulated clinical rehabilitation data analysis and expert assessments. For example, historical rehabilitation data is re-analyzed and the mapping relationships are adjusted every six months to ensure assessment accuracy. Boundary cases need to be handled during the matching process. For instance, when the input development stage is not in the preset risk level mapping table, the default mapping rule is used to map the unknown development stage to a medium-risk level.

[0046] The appropriate warning level is selected based on the risk level of rehabilitation outcomes. The warning level is defined based on the severity of the risk, for example, low risk corresponds to Level 1, medium risk to Level 2, and high risk to Level 3. The selection of the warning level is achieved through conditional logic, where the input is the risk level of rehabilitation outcomes, and the output is the corresponding warning level. For example, Level 1 is selected when the risk level is low, Level 2 when it is medium, and Level 3 when it is high. The specific meanings of each warning level include: Level 1 indicates the need to maintain the current rehabilitation training plan; Level 2 indicates the need to adjust the intensity of rehabilitation training; and Level 3 indicates the need for immediate medical intervention. The setting of warning levels is based on clinical rehabilitation guidelines and risk management principles, such as determining the warning level classification by analyzing the safety and effectiveness of rehabilitation training under different risk levels. The selection process for warning levels needs to consider the specific scenario of rehabilitation training; for example, a more conservative warning level selection strategy may be adopted for patients in the early stages of rehabilitation.

[0047] Warning messages containing specific rehabilitation guidance suggestions are generated based on warning levels. The generation of these suggestions is based on the warning level and rehabilitation training guidelines, which are derived from clinical rehabilitation medicine guidelines and expert consensus. The warning message structure includes three parts: a warning level identifier, a risk description, and rehabilitation guidance suggestions. The warning level identifier uses a combination of color coding and text description; for example, a level 1 warning uses a green identifier and low-risk text, a level 2 warning uses a yellow identifier and medium-risk text, and a level 3 warning uses a red identifier and high-risk text. The risk description briefly explains the current progressive drift in movement patterns and its potential impact; for example, for the developmental stage, it describes a detected change in movement patterns requiring attention to training effectiveness. The rehabilitation guidance suggestions provide specific training adjustment guidance; for example, for a level 1 warning, it suggests maintaining the current training pace; for a level 2 warning, it suggests appropriately reducing training intensity and increasing rest intervals; and for a level 3 warning, it suggests immediately stopping training and contacting a rehabilitation physician. The warning message is generated using a template-filling method. The template is predefined according to the warning level, and the suggestions are adjusted based on the specific patient's condition during the filling process. The specific content of the rehabilitation guidance suggestions is determined based on clinical rehabilitation practice guidelines, such as referring to the standard operating procedures for postoperative rehabilitation training in urology.

[0048] The generated warning information is displayed through the user interface of the smart terminal device. The user interface design follows human-computer interaction principles to ensure that the information is presented clearly and understandably. Warning information is displayed in a combination of formats, including pop-up windows, status bar notifications, and sound alerts. For example, a level 1 warning only displays a green icon in the status bar, a level 2 warning displays a yellow pop-up window accompanied by a short sound alert, and a level 3 warning displays a full-screen red window accompanied by a continuous alarm sound. The displayed content includes all components of the warning information: the warning level indicator, risk description, and rehabilitation guidance suggestions. The user interface also provides interactive functions, such as a confirmation button to close the warning window and a details button to view more detailed explanations. The update frequency of the warning information is synchronized with the rehabilitation assessment cycle, for example, updating the displayed content every 5 minutes to ensure real-time information. The user interface must be responsive during the display process to avoid display delays due to resource consumption. The user interface design also needs to consider the needs of different user groups, such as providing a font magnification function for elderly patients.

[0049] The pre-defined risk level mapping table defines the risk levels for progressive movement pattern drift, corresponding to low risk in the nascent stage, medium risk in the development stage, and high risk in the stable stage. The nascent stage is defined as the emergence of a new dominant movement pattern with low dominance; the development stage is defined as the increasing dominance of the new dominant movement pattern while the dominance of existing patterns decreases; and the stable stage is defined as the dominance of each dominant movement pattern stabilizing. Risk levels are defined based on clinical rehabilitation risk assessment standards: low risk indicates normal rehabilitation without intervention; medium risk indicates the need for monitoring and adjustment of the training plan; and high risk indicates a risk of deterioration in rehabilitation outcomes requiring immediate attention. The pre-defined risk level mapping table is validated through backtesting of historical rehabilitation data, such as using patient rehabilitation data from the past year to test the accuracy and reliability of the mapping table, ensuring consistency between risk assessment results and clinical reality. The pre-defined risk level mapping table is applied to all urological postoperative rehabilitation patients using this monitoring system, ensuring consistency in assessment standards. The adjustment mechanism for the pre-defined risk level mapping table includes periodic revisions of the mapping relationship based on emerging clinical evidence, such as quarterly reviews of the mapping table's applicability by a rehabilitation expert team.

[0050] Example 2: Figure 2 A schematic diagram of a postoperative rehabilitation effect monitoring system for urology is provided. The system includes: The sequence acquisition module is used to acquire rehabilitation training exercise data sequences at multiple time points collected by the patient through a smart terminal device; The spatial extraction module is used to construct a personal movement state space based on rehabilitation training exercise data sequences, and to extract principal component subspaces that represent the patient's personalized movement patterns through principal component analysis. The error calculation module is used to calculate the reconstruction error of newly acquired rehabilitation training exercise data in the principal component subspace. The drift determination module is used to analyze whether the statistical distribution of the reconstruction error exhibits bimodal or multimodal characteristics. When bimodal or multimodal characteristics are present, it is determined that there is a progressive drift in the action mode. The stage determination module is used to map different peaks in the statistical distribution to different dominant action patterns when there is progressive drift in action patterns, and to determine the development stage of progressive drift in action patterns by analyzing the temporal competition relationship of dominance among dominant action patterns. The early warning assessment module is used to assess the risk level of rehabilitation effectiveness and generate corresponding early warning information based on the developmental stage of progressive drift in movement patterns.

[0051] All calculations involved in the embodiments are dimensionless numerical calculations, and the preset parameters and thresholds in the calculations are set by those skilled in the art according to the actual situation.

[0052] It should be noted that this invention can be deployed on the device itself to realize embedded applications, or it can run on a PC or other terminal with a user interface, thereby meeting various hardware environments and usage requirements.

[0053] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions according to the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. Computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wireless or wired transmission; wired transmission methods include optical fiber, twisted pair, coaxial cable, etc.; wireless transmission includes infrared, microwave, etc. Computer-readable storage media can be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more sets of available media. Available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media. Semiconductor media can be solid-state drives.

[0054] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0055] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0056] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0057] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0058] If a function is implemented as a software module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0059] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0060] In conclusion, the above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for monitoring the postoperative rehabilitation effect in urology, characterized in that, include: S1. Obtain rehabilitation training exercise data sequences at multiple time points collected by the patient through a smart terminal device; S2. Construct a personal movement state space based on rehabilitation training exercise data sequences, and extract principal component subspaces representing the patient's personalized movement patterns through principal component analysis; S3. Calculate the reconstruction error of the newly acquired rehabilitation training exercise data in the principal component subspace; S4. Analyze whether the statistical distribution of the reconstruction error shows bimodal or multimodal characteristics. If it shows bimodal or multimodal characteristics, it is determined that there is a progressive drift of the action mode. S5. When there is progressive drift in action patterns, different peaks in the statistical distribution are mapped to different dominant action patterns, and the development stage of progressive drift in action patterns is determined by analyzing the temporal competition relationship of dominance among dominant action patterns. S6. Based on the developmental stages of progressive drift in movement patterns, assess the risk level of rehabilitation effectiveness and generate corresponding early warning information.

2. The method for monitoring the postoperative rehabilitation effect in urology according to claim 1, characterized in that, Acquire rehabilitation training exercise data sequences from multiple time points collected by the patient through a smart terminal device, including: Collect rehabilitation training exercise data sequences using accelerometers and gyroscopes in smartphones or wearable devices; The rehabilitation training exercise data sequence includes acceleration and angular velocity data collected at multiple consecutive time points; Both acceleration and angular velocity data are contained in data components along the three orthogonal axes in a three-dimensional spatial coordinate system.

3. The method for monitoring the postoperative rehabilitation effect in urology according to claim 1, characterized in that, A personal movement state space is constructed based on rehabilitation training exercise data sequences, and principal component analysis is used to extract principal component subset spaces representing the patient's personalized movement patterns, including: The acceleration and angular velocity data at each time point in the rehabilitation training exercise data sequence are combined to form the motion state vector in the individual motion state space; The state space matrix of the individual action state space is constructed by combining the action state vectors at multiple time points. Principal component analysis is performed on the state space matrix, and the top principal components whose cumulative contribution rate exceeds the preset contribution rate threshold are selected to form the principal component subspace.

4. The method for monitoring the postoperative rehabilitation effect in urology according to claim 1, characterized in that, Calculate the reconstruction error of the newly acquired rehabilitation training exercise data in the principal component subset space, including: The newly collected rehabilitation training exercise data is used to construct a new motion state vector; Calculate the projection vector of the new action state vector onto the principal component subspace; The reconstruction error is obtained by calculating the Euclidean distance between the new action state vector and the projection vector.

5. The method for monitoring the postoperative rehabilitation effect in urology according to claim 1, characterized in that, Analyze whether the statistical distribution of the reconstruction error exhibits bimodal or multimodal characteristics. If it does, it indicates a progressive drift in the action pattern, including: The reconstruction errors calculated over a continuous time period are collected to form a reconstruction error sequence; The probability density distribution curve of the reconstruction error is obtained by kernel density estimation of the reconstruction error sequence; The number of peaks in the probability density distribution curve is detected, and when two or more significant peaks are detected, it is determined that the curve exhibits bimodal or multimodal characteristics. Analyze the trend of the distance between adjacent peaks. When the distance between peaks shows a continuous increasing trend, it confirms the existence of progressive drift in the action pattern.

6. The method for monitoring the postoperative rehabilitation effect in urology according to claim 5, characterized in that, The process of estimating the kernel density of the reconstruction error sequence to obtain the probability density distribution curve of the reconstruction error includes: using a Gaussian kernel function to estimate the kernel density of the reconstruction error sequence, and generating a smooth probability density distribution curve by calculating the bandwidth parameter based on the standard deviation of the reconstruction error sequence.

7. The method for monitoring the postoperative rehabilitation effect in urology according to claim 1, characterized in that, When progressive drift in action patterns exists, different peaks in the statistical distribution are mapped to different dominant action patterns. The development stages of progressive drift are determined by analyzing the temporal competition for dominance among dominant action patterns, including: Based on the significant peaks in the probability density distribution curve, each significant peak is mapped to an independent dominant action pattern; Calculate the peak area corresponding to each dominant action pattern as the dominance strength of the corresponding dominant action pattern within the current time window; The dominance strength of each dominant action pattern is repeatedly calculated within multiple consecutive time windows to form a dominance strength sequence for each dominant action pattern. Analyze the temporal competition relationship of dominance among the dominant action patterns of different dominant action patterns, including identifying the contrast between the growth trend and the decline trend in the dominance intensity sequence; The development stage of the progressive drift of the action pattern is determined by comparing the growth trend and the decline trend.

8. The method for monitoring the postoperative rehabilitation effect in urology according to claim 7, characterized in that, The analysis of the temporal competition relationship of dominance among the dominant action patterns includes: applying moving average smoothing to the dominance strength sequence of each dominant action pattern, using linear regression analysis to extract the growth trend and decay trend respectively, and quantifying the relative change trend by calculating the difference in the slope of the trend line.

9. The method for monitoring the postoperative rehabilitation effect in urology according to claim 1, characterized in that, Based on the developmental stages of progressive movement pattern drift, the risk level of rehabilitation outcomes is assessed and corresponding early warning information is generated, including: The developmental stages of progressive drift in movement patterns are matched with a preset risk level mapping table to determine the corresponding risk level of rehabilitation effect. Select the appropriate early warning level based on the risk level of rehabilitation outcomes; Based on the warning level, generate warning information containing specific rehabilitation guidance and suggestions; The generated warning information is displayed through the user interface of smart terminal devices; The preset risk level mapping table includes: defining the risk level mapping relationship between the nascent stage of progressive movement pattern drift and the medium-risk stage, and the stable stage and the high-risk stage, forming a risk level mapping table for assessing the risk level of rehabilitation effect.

10. A urological postoperative rehabilitation effect monitoring system, used to implement the urological postoperative rehabilitation effect monitoring method according to any one of claims 1-9, characterized in that, include: The sequence acquisition module is used to acquire rehabilitation training exercise data sequences at multiple time points collected by the patient through a smart terminal device; The spatial extraction module is used to construct a personal movement state space based on rehabilitation training exercise data sequences, and to extract principal component subspaces that represent the patient's personalized movement patterns through principal component analysis. The error calculation module is used to calculate the reconstruction error of newly acquired rehabilitation training exercise data in the principal component subspace. The drift determination module is used to analyze whether the statistical distribution of the reconstruction error exhibits bimodal or multimodal characteristics. When bimodal or multimodal characteristics are present, it is determined that there is a progressive drift in the action mode. The stage determination module is used to map different peaks in the statistical distribution to different dominant action patterns when there is progressive drift in action patterns, and to determine the development stage of progressive drift in action patterns by analyzing the temporal competition relationship of dominance among dominant action patterns. The early warning assessment module is used to assess the risk level of rehabilitation effectiveness and generate corresponding early warning information based on the developmental stage of progressive drift in movement patterns.

Images