Big data based sports injury risk prediction system

Abstract

The application discloses a sports injury risk prediction system based on big data, comprising a standardized data acquisition module, an individual dynamic baseline evaluation module, a sports injury risk prediction model construction module, a chronic cumulative injury risk prediction module, an acute injury risk prediction module and a sports injury risk comprehensive management module. The application relates to the technical field of data processing, in particular to a sports injury risk prediction system based on big data. The application innovatively proposes an individual dynamic baseline evaluation mechanism, realizes individualization and dynamic evaluation of sports injury risk, proposes a double-type risk prediction mechanism combining chronic cumulative injury risk prediction and acute injury risk prediction, improves the accuracy of sports injury risk prediction, and proposes a basic risk prediction architecture model based on a static injury mechanism hypergraph, a dynamic risk semantic hypergraph, shared private feature gate fusion and bidirectional time sequence gate feature extraction, thereby enhancing the accuracy of injury risk prediction results.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to a sports injury risk prediction system based on big data. Background Technology

[0002] The sports injury risk prediction system is an intelligent sports health prevention and control system for medical sports rehabilitation centers. Relying on big data analysis and artificial intelligence technology, it can uniformly collect and standardize the processing of multi-source related data from users, complete the extraction and multi-dimensional fusion analysis of sports injury-related characteristics, achieve accurate classification and early warning of sports injury risks, effectively reduce the probability of injury among athletes, and help athletes, coaches, and rehabilitation therapists avoid injury risks in advance, so as to achieve scientific exercise and safe training.

[0003] However, traditional sports injury risk prediction methods typically use fixed risk thresholds for judgment, which are difficult to adapt to the technical problem of individual differences among different athletes; traditional sports injury risk prediction systems usually use a single prediction model to output a comprehensive risk value, which is difficult to distinguish between chronic cumulative injury risk and acute sudden injury risk, resulting in technical problems of inaccurate prediction of long-term training load risk and instantaneous movement abnormality risk; existing models applicable to sports injury risk prediction are unable to fully express the complex correlation between features and have insufficient ability to identify long-term cumulative changes and short-term sudden abnormalities, resulting in technical problems of inaccurate injury risk prediction. Summary of the Invention

[0004] To address the above issues and overcome the shortcomings of existing technologies, this invention provides a sports injury risk prediction system based on big data. Addressing the problem that traditional sports injury risk prediction methods typically use fixed risk thresholds, making it difficult to adapt to individual differences among athletes, this invention innovatively proposes an individual dynamic baseline assessment mechanism. This mechanism acquires the chronic cumulative baseline state and acute sudden injury baseline state of the target athlete, and incorporates the baseline assessment results into the subsequent risk prediction process. This allows risk assessment to move beyond a single fixed threshold, combining the target athlete's historical training adaptation status, motor ability status, and current movement characteristics for individualized analysis. This improves the individual adaptability and accuracy of sports injury risk prediction, enhances the system's ability to identify risks in athletes with previous injuries, athletes at different training levels, and athletes in different movement states, reduces the probability of false alarms and false negatives, and achieves individualized, dynamic, and precise assessment of sports injury risk. Furthermore, addressing the problem that traditional sports injury risk prediction systems typically use a single prediction model to output a comprehensive risk value, making it difficult to distinguish between chronic cumulative injury risk and acute sudden injury risk, leading to inaccurate predictions of long-term training load risk and instantaneous movement abnormality risk, this invention innovatively… This invention proposes a dual-type risk prediction mechanism that combines chronic cumulative injury risk prediction and acute sudden injury risk prediction. Based on a unified basic risk prediction architecture model, two risk prediction models are trained separately: one for chronic cumulative injury and the other for acute sudden injury. This improves the accuracy of the system's risk prediction for both long-term cumulative injury and short-term sudden injury, enabling comprehensive intelligent management of sports injury risks through early prediction of chronic risks, timely warning of acute risks, and integrated management of sports injury risks. Addressing the technical problem that existing sports injury risk prediction models struggle to fully express the complex relationships between features and lack the ability to identify long-term cumulative changes and short-term sudden anomalies, leading to inaccurate injury risk predictions, this invention innovatively proposes a basic risk prediction architecture model based on a static injury mechanism hypergraph, a dynamic risk semantic hypergraph, shared private feature gating fusion, and bidirectional temporal gating feature extraction. Stable high-order correlation features and dynamic risk correlation features are extracted through hypergraph convolution, and chronic cumulative trends and acute sudden changes are comprehensively identified through gating fusion and bidirectional temporal modeling. This improves the model's ability to express complex sports injury risk factors and its temporal recognition capabilities, enhancing the accuracy, stability, and robustness of injury risk prediction results, and achieving accurate prediction of sports injury risks.

[0005] The technical solution adopted by the present invention is as follows: The sports injury risk prediction system based on big data provided by the present invention includes a standardized data acquisition module, an individual dynamic baseline assessment module, a sports injury risk prediction model construction module, a chronic cumulative injury risk prediction module, an acute sudden injury risk prediction module, and a comprehensive sports injury risk management module.

[0006] The standardized data acquisition module obtains standardized data for predicting sports injury risks through data collection and data preprocessing.

[0007] The individual dynamic baseline assessment module is used to obtain the chronic cumulative baseline status and acute sudden baseline status of the target athlete, respectively. Specifically, it constructs a chronic cumulative injury risk baseline assessment model and an acute sudden injury risk baseline assessment model based on a long short-term memory neural network, trains the model using standardized dynamic baseline assessment data of the reference athlete, and inputs the standardized dynamic baseline assessment data of the target athlete into the trained baseline assessment model to obtain the target chronic cumulative baseline assessment result and the target acute sudden baseline assessment result.

[0008] The sports injury risk prediction model construction module is used to build a basic risk prediction architecture model that can be applied to both chronic cumulative injury risk prediction and acute sudden injury risk prediction. Specifically, it is based on the acquisition of sports injury risk temporal features, static injury mechanism hypergraph, dynamic risk semantic hypergraph, hypergraph convolutional feature extraction, shared private feature gating fusion, bidirectional temporal gating feature extraction and injury risk prediction output to build a basic sports injury risk prediction model.

[0009] The chronic cumulative injury risk prediction module is used to predict the risk of chronic cumulative injury that will occur under the long-term training load of the target athlete. Specifically, it obtains the target chronic cumulative injury risk result based on the sports injury risk prediction model and the target chronic cumulative baseline assessment results.

[0010] The acute injury risk prediction module is used to predict the risk of acute injury to the target athlete during exercise; specifically, it obtains the target acute injury risk result based on the sports injury risk prediction model and the target acute injury baseline assessment results.

[0011] The sports injury risk comprehensive management module specifically determines the comprehensive assessment result of sports injury risk based on the target chronic cumulative injury risk result and the target acute sudden injury risk result, thereby realizing intelligent comprehensive management of sports injury risk.

[0012] Furthermore, the standardized data acquisition module specifically obtains raw data for sports injury risk prediction through data acquisition operations, and performs data preprocessing on the raw data for sports injury risk prediction to obtain standardized data for sports injury risk prediction.

[0013] The standardized data for predicting sports injury risk includes standardized data for reference athlete dynamic baseline assessment, standardized data for target athlete dynamic baseline assessment, standardized data for reference athlete injury risk prediction, and standardized data for target athlete injury risk prediction. The data preprocessing specifically involves performing data cleaning, normalization, and feature selection on the raw data to obtain the standardized data for predicting sports injury risk.

[0014] Furthermore, the individual dynamic baseline assessment module specifically includes the following steps:

[0015] The baseline assessment of chronic cumulative injury risk is specifically constructed based on a long short-term memory neural network to build a baseline assessment model for chronic cumulative injury risk. The baseline assessment data of chronic cumulative injury risk in the standardized dynamic baseline assessment data of the reference athlete is used as the training data for the model. The assessment model is iteratively trained for chronicity. Finally, the baseline assessment data of chronic cumulative injury risk in the standardized dynamic baseline assessment data of the target athlete is input into the trained baseline assessment model for chronic cumulative injury risk to obtain the target chronic cumulative baseline assessment result.

[0016] The baseline assessment of acute sudden injury risk is specifically carried out by constructing an acute sudden injury risk baseline assessment model based on a long short-term memory neural network. The acute sudden injury risk baseline assessment data in the standardized dynamic baseline assessment data of the reference athlete is used as the model training data for acute iterative training. Finally, the acute sudden injury risk baseline assessment data in the standardized dynamic baseline assessment data of the target athlete is input into the trained acute sudden injury risk baseline assessment model to obtain the target acute sudden baseline assessment result.

[0017] Furthermore, the sports injury risk prediction model construction module specifically includes the following steps:

[0018] The acquisition of temporal features of sports injury risk involves performing feature mapping on the dynamic temporal features in the model input data through convolutional mapping operations, and then concatenating the features of each embedding result to obtain the temporal features of sports injury risk.

[0019] To construct a static damage mechanism hypergraph, body parts are used as damage risk nodes. Damage mechanism feature vectors are constructed based on the basic body state and individual dynamic baseline state corresponding to the nodes. The K-Means clustering algorithm is used to divide nodes with similar damage mechanisms into the same hyperedge to obtain the static damage mechanism hypergraph.

[0020] The construction of dynamic risk semantic hypergraphs involves calculating the semantic similarity between different injury risk feature channels based on the temporal features of sports injury risk, and using the K-nearest neighbor algorithm to select the k nodes most similar to the current node to form dynamic hyperedges, thus obtaining a set of dynamic risk semantic hypergraphs.

[0021] Hypergraph convolution feature extraction specifically involves performing hypergraph convolution based on the static hypergraph association matrix to obtain static private risk features, and performing hypergraph convolution based on the dynamic risk semantic hypergraph to obtain dynamic shared risk features.

[0022] The shared private feature gating fusion is specifically based on dynamic shared risk features and static private risk features. The gating fusion coefficient is calculated through a Sigmoid gating unit. The dynamic shared risk features and static private risk features are then adaptively weighted and fused according to the gating fusion coefficient to obtain the fused risk features.

[0023] Bidirectional temporal gating feature extraction specifically involves inputting the fused risk features into a bidirectional temporal convolutional network to extract bidirectional temporal convolutional features, and then inputting the bidirectional temporal convolutional features into a bidirectional gated recurrent unit to obtain bidirectional gated temporal features.

[0024] The damage risk prediction output is specifically achieved by concatenating bidirectional gated temporal features with individual dynamic baseline state features, then inputting the concatenation into a multilayer perceptron, and finally outputting the damage risk prediction through the Softmax function.

[0025] Furthermore, the chronic cumulative injury risk prediction module specifically involves using the chronic cumulative injury risk prediction data from the reference athlete injury risk prediction data and the corresponding chronic cumulative baseline assessment results as training data to train the sports injury risk prediction model, thereby obtaining the chronic cumulative injury risk prediction model. Finally, the chronic cumulative injury risk prediction data from the target athlete injury risk prediction data and the corresponding target chronic cumulative baseline assessment results are input into the model to obtain the target chronic cumulative injury risk result.

[0026] Furthermore, the acute sudden injury risk prediction module specifically involves using the acute sudden injury risk prediction data from the reference athlete injury risk prediction data and the corresponding acute sudden baseline assessment results as training data to train the sports injury risk prediction model, thereby obtaining the acute sudden injury risk prediction model. Finally, the acute sudden injury risk prediction data from the target athlete injury risk prediction data and the corresponding target acute sudden baseline assessment results are input into the model to obtain the target acute sudden injury risk result.

[0027] Furthermore, the sports injury risk comprehensive management module specifically determines the comprehensive assessment result of sports injury risk based on the target chronic cumulative injury risk result and the target acute sudden injury risk result, and generates the corresponding sports injury risk management result, thereby realizing intelligent comprehensive management of sports injury risk.

[0028] The beneficial effects achieved by the present invention using the above solution are as follows:

[0029] (1) In view of the technical problem that traditional sports injury risk prediction methods usually use fixed risk thresholds for judgment, it is difficult to adapt to the individual differences of different athletes in terms of physical condition, training level, past injury status and recovery ability. This can easily lead to high-level athletes being misjudged as high-risk, ordinary athletes or athletes with previous injuries being missed, and the same risk index being insufficiently adaptable to different athletes. This invention innovatively proposes an individual dynamic baseline assessment mechanism, which obtains the chronic cumulative baseline status and acute sudden baseline status of the target athlete, and introduces the baseline assessment results into the subsequent risk prediction process. This makes risk judgment no longer solely dependent on a uniform fixed threshold, but rather combines the target athlete's own historical training adaptation status, motor ability status and current sports characteristics for individualized analysis. This can improve the individual adaptability and accuracy of sports injury risk prediction, enhance the system's risk identification ability for athletes with previous injuries, athletes with different training levels and athletes with different sports states, reduce the probability of false alarms and false alarms, and realize individualized, dynamic and precise assessment of sports injury risk.

[0030] (2) In view of the technical problem that traditional sports injury risk prediction systems usually use a single prediction model to output a comprehensive risk value, which makes it difficult to distinguish between chronic cumulative injury risk and acute sudden injury risk, resulting in inaccurate prediction of long-term training load risk and instantaneous movement abnormality risk, it is easy to cause the failure to identify chronic fatigue cumulative risk in advance and the failure to provide timely warning of acute movement abnormality risk. This invention innovatively proposes a dual-type risk prediction mechanism that combines chronic cumulative injury risk prediction and acute sudden injury risk prediction. Based on a unified basic risk prediction architecture model, a chronic cumulative injury risk prediction model and an acute sudden injury risk prediction model are trained respectively, and the target chronic cumulative injury risk result and the target acute sudden injury risk result are output respectively. This can improve the accuracy of the system in predicting the risk of long-term cumulative injury and short-term sudden injury, and realize the comprehensive intelligent management of sports injury risk, including early prediction of chronic risk, timely warning of acute risk and comprehensive intelligent management of sports injury risk.

[0031] (3) To address the technical problem that existing sports injury risk prediction models are unable to fully express the complex relationships between features and lack the ability to identify long-term cumulative changes and short-term sudden anomalies, resulting in inaccurate injury risk prediction, this invention innovatively proposes a basic risk prediction architecture model based on a static injury mechanism hypergraph, a dynamic risk semantic hypergraph, shared private feature gating fusion, and bidirectional temporal gating feature extraction. By extracting stable high-order correlation features and dynamic risk correlation features through hypergraph convolution, and by comprehensively identifying chronic cumulative trends and acute sudden changes through gating fusion and bidirectional temporal modeling, the model's ability to express complex sports injury risk factors and temporal identification ability is improved, enhancing the accuracy, stability, and robustness of injury risk prediction results, and realizing accurate prediction of sports injury risk. Attached Figure Description

[0032] Figure 1 A schematic diagram of the modules of the sports injury risk prediction system based on big data provided by the present invention;

[0033] Figure 2 A flowchart illustrating the module for building a sports injury risk prediction model;

[0034] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. Detailed Implementation

[0035] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0036] In the description of this invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the system or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0037] Example 1, see Figure 1 The sports injury risk prediction system based on big data provided by this invention includes a standardized data acquisition module, an individual dynamic baseline assessment module, a sports injury risk prediction model construction module, a chronic cumulative injury risk prediction module, an acute sudden injury risk prediction module, and a comprehensive sports injury risk management module.

[0038] The standardized data acquisition module obtains standardized data for predicting sports injury risks through data collection and preprocessing, and sends the data to the individual dynamic baseline assessment module, the chronic cumulative injury risk prediction module, and the acute sudden injury risk prediction module.

[0039] The individual dynamic baseline assessment module is used to acquire the chronic cumulative baseline status and acute sudden injury baseline status of the target athlete, respectively, and receive data sent by the standardized data acquisition module. Specifically, it constructs a chronic cumulative injury risk baseline assessment model and an acute sudden injury risk baseline assessment model based on a long short-term memory neural network, trains the model using standardized dynamic baseline assessment data of the reference athlete, and inputs the standardized dynamic baseline assessment data of the target athlete into the trained baseline assessment model to obtain the target chronic cumulative baseline assessment result and the target acute sudden injury baseline assessment result, and sends the data to the chronic cumulative injury risk prediction module and the acute sudden injury risk prediction module.

[0040] The sports injury risk prediction model construction module is used to build a basic risk prediction architecture model that can be applied to both chronic cumulative injury risk prediction and acute sudden injury risk prediction. Specifically, it is based on the acquisition of sports injury risk temporal features, static injury mechanism hypergraph, dynamic risk semantic hypergraph, hypergraph convolution feature extraction, shared private feature gating fusion, bidirectional temporal gating feature extraction, and injury risk prediction output to build a basic sports injury risk prediction model, and sends the data to the chronic cumulative injury prediction module and the acute sudden injury prediction module.

[0041] The chronic cumulative injury risk prediction module is used to predict the risk of chronic cumulative injury that will occur under the long-term training load of the target athlete. It receives data sent by the sports injury risk prediction model construction module and the individual dynamic baseline assessment module. Specifically, it obtains the target chronic cumulative injury risk result based on the sports injury risk prediction model and the target chronic cumulative baseline assessment result, and sends the data to the sports injury risk comprehensive management module.

[0042] The acute sudden injury risk prediction module is used to predict the risk of acute injury to the target athlete during exercise, and to receive data sent by the sports injury risk prediction model construction module and the individual dynamic baseline assessment module; specifically, based on the sports injury risk prediction model and the target acute sudden baseline assessment results, the target acute sudden injury risk result is obtained, and the data is sent to the sports injury risk comprehensive management module.

[0043] The sports injury risk comprehensive management module receives data from the chronic cumulative injury risk prediction module and the acute sudden injury risk prediction module. Specifically, it determines the comprehensive assessment result of sports injury risk based on the target chronic cumulative injury risk result and the target acute sudden injury risk result, thereby realizing intelligent comprehensive management of sports injury risk.

[0044] By performing the above operations, this invention addresses the technical problem that traditional sports injury risk prediction systems typically use a single prediction model to output a comprehensive risk value, making it difficult to distinguish between chronic cumulative injury risk and acute sudden injury risk. This leads to inaccurate predictions of long-term training load risk and instantaneous movement abnormality risk, and easily results in the inability to identify chronic fatigue cumulative risk in advance and the inability to provide timely warnings of acute movement abnormality risk. This invention innovatively proposes a dual-type risk prediction mechanism that combines chronic cumulative injury risk prediction and acute sudden injury risk prediction. Based on a unified basic risk prediction architecture model, chronic cumulative injury risk prediction models and acute sudden injury risk prediction models are trained separately, and target chronic cumulative injury risk results and target acute sudden injury risk results are output respectively. This can improve the accuracy of the system in predicting the risk of long-term cumulative injury and short-term sudden injury, and realize the comprehensive intelligent management of sports injury risk, including early prediction of chronic risk, timely warning of acute risk, and comprehensive management of sports injury risk.

[0045] Example 2, see Figure 1 This embodiment is based on the above embodiment. The standardized data acquisition module specifically performs data acquisition operations through wearable sensors, inertial measurement units, smart insoles, electromyography sensors, sports bracelets, mobile terminal applications and camera equipment to obtain raw data for sports injury risk prediction, and performs data preprocessing on the raw data for sports injury risk prediction to obtain standardized data for sports injury risk prediction.

[0046] The raw data for predicting sports injury risk includes reference athlete dynamic baseline assessment data, target athlete dynamic baseline assessment data, reference athlete injury risk prediction data, and target athlete injury risk prediction data.

[0047] The standardized data for predicting sports injury risk includes standardized data for reference athlete dynamic baseline assessment, standardized data for target athlete dynamic baseline assessment, and standardized data for reference athlete injury risk prediction.

[0048] Both the reference athlete dynamic baseline assessment data and the target athlete dynamic baseline assessment data include baseline assessment data for chronic cumulative injury risk and baseline assessment data for acute sudden injury risk;

[0049] Both the reference athlete injury risk prediction data and the target athlete injury risk prediction data include chronic cumulative injury risk prediction data and acute sudden injury risk prediction data; the reference athlete dynamic baseline assessment data also includes reference chronic cumulative baseline assessment results and reference acute sudden injury baseline assessment results; the chronic cumulative baseline assessment results include no-injury adaptation state, low fatigue adaptation state, normal load adaptation state, and chronic discomfort sensitivity state; the acute sudden injury baseline assessment results include basic motor ability state, moderate motor ability state, high motor ability state, and extreme motor ability state; the reference athlete injury risk prediction data also includes reference chronic cumulative injury risk results and reference acute sudden injury risk results; the chronic cumulative injury risk results and acute sudden injury risk results... Injury risk results are categorized into no risk, low risk, medium risk, and high risk. The baseline assessment data for chronic cumulative injury risk includes basic physical condition data, long-term training load data, fatigue recovery data, discomfort feedback data, and previous injury rehabilitation data. The basic physical condition data includes age, gender, height, weight, BMI, body fat percentage, muscle mass, baseline heart rate, maximum heart rate, resting heart rate, and cardiorespiratory endurance level. The long-term training load data includes training frequency, training duration, training distance, training intensity, training volume, training density, weekly training load, monthly training load, training load growth rate, and acute / chronic training load ratio. The fatigue recovery data includes heart rate recovery rate, sleep duration, sleep quality, recovery time, number of consecutive training days, and rest interval.

[0050] The discomfort feedback data includes pain duration, pain location, pain trend, and discomfort feedback.

[0051] The previous injury rehabilitation data includes the site of the previous injury, the type of the previous injury, the time of the injury, the severity of the injury, the rehabilitation period, the number of recurrences, the current recovery status, and the load-bearing capacity of the site of the previous injury.

[0052] The baseline assessment data for acute sudden injury risk includes basic physical condition data, range of motion data, plantar pressure and impact range data, and range of motion stability data.

[0053] The range of motion posture data includes the range of knee flexion and extension angles, the range of ankle dorsiflexion angles, the range of trunk forward tilt angles, the range of trunk left and right lateral tilt angles, the range of pelvic tilt angles, the range of take-off posture angles, the range of landing posture angles, the range of posture changes during sudden stops, and the range of posture changes during changes of direction.

[0054] The plantar pressure and impact range data include the total plantar pressure range, the medial plantar pressure range, the lateral plantar pressure range, the balanced pressure range of both feet landing, the pressure center trajectory range, the pressure center movement speed range, the landing impact peak range, the impact loading rate range, the emergency stop braking pressure range, and the change of direction support pressure range.

[0055] The motion stability range data includes gait stability range, cadence fluctuation range, stride fluctuation range, body center of gravity swing range, center of gravity movement speed range, center of gravity recovery time range after landing, posture swing amplitude range, and motion deviation amplitude range.

[0056] The chronic cumulative injury risk prediction data includes basic physical condition data, long-term training load data, fatigue recovery data, pain feedback data, and movement posture data.

[0057] The long-term training load data includes the training load growth trend, the ratio of acute to chronic training load, the training intensity change trend, and the training frequency change trend; the fatigue recovery data includes the number of consecutive training days, the number of times of insufficient rest, the fatigue score trend, the number of times of insufficient sleep, the heart rate recovery decline trend, the muscle fatigue accumulation trend, and the recovery insufficiency index; the pain feedback data includes the duration of pain, changes in the location of pain, the severity of pain after exercise, and the frequency of pain recurrence.

[0058] The acute injury risk prediction data includes basic body status data, movement posture data, movement cycle data, muscle activation data, and real-time environmental data. The movement posture data includes joint angles, joint angular velocities, trunk tilt angles, pelvic tilt angles, knee valgus angles, ankle rotation angles, hip joint range of motion, and upper limb swing angles. The movement cycle data includes gait cycle, stance phase time, swing phase time, airtime, ground contact time, movement duration, and movement phase markers. The muscle activation data includes electromyographic signals, muscle activation sequence, and muscle activation intensity. The real-time environmental data includes the hardness of the sports field, surface slippage, temperature, humidity, and slope.

[0059] The data preprocessing specifically involves performing data cleaning, normalization, and feature selection on the raw data to obtain standardized data for predicting sports injury risks.

[0060] The data cleaning process is used to improve the integrity, accuracy, and usability of the raw data for predicting sports injury risks; specifically, it involves removing outliers and deleting duplicate data from the raw data.

[0061] The outlier removal process specifically involves identifying data that exceeds the preset reasonable value range or significantly deviates from the statistical distribution characteristics based on the preset reasonable value range of each field, and then removing the identified outlier data.

[0062] The deduplication process involves identifying duplicate data based on the athlete's ID, the timestamp collected, and the activity segment identifier, and retaining one valid record among the duplicate data.

[0063] The normalization process is used to convert raw data of sports injury risk prediction from different sources, types, and scales into a unified data representation form, so as to improve the stability of subsequent individual dynamic baseline assessment and sports injury risk prediction model training. Specifically, firstly, label encoding technology is used to map the categorical fields in the raw data to unique integer values ​​according to their different categories, thereby realizing the quantitative conversion of categorical data. Then, the Z-Score standardization algorithm is used to normalize the continuous variables, so that each continuous variable is converted into a standardized variable with a mean of 0 and a standard deviation of 1, thereby achieving a uniform numerical scale between variables of different scales.

[0064] The feature selection is used to filter out features from standardized sports injury risk prediction data that are highly correlated with individual dynamic baseline assessment results, chronic cumulative injury risk prediction results, and acute sudden injury risk prediction results, thereby reducing redundant and noisy features and improving model training efficiency, prediction accuracy, and result interpretability. Specifically, the feature selection for chronic cumulative baseline assessment, acute sudden injury baseline assessment, chronic cumulative injury risk prediction, and acute sudden injury risk prediction is performed sequentially through correlation analysis.

[0065] The chronic cumulative baseline assessment feature selection is used to screen features that are relevant to the chronic cumulative baseline assessment results;

[0066] The acute outbreak baseline assessment feature selection is used to screen features that are related to the acute outbreak baseline assessment results.

[0067] The selection of features for predicting chronic cumulative injury risk is used to screen features that are related to the outcome of chronic cumulative injury risk.

[0068] The selection of features for predicting acute sudden injury risk is used to screen features that are related to the outcome of acute sudden injury risk.

[0069] Example 3, see Figure 1 This embodiment is based on the above embodiment. The individual dynamic baseline assessment module is used to acquire the chronic cumulative baseline status and acute sudden injury baseline status of the target athlete, respectively, to provide individualized reference for subsequent prediction of chronic cumulative injury risk and acute sudden injury risk; specifically, it includes the following steps:

[0070] Chronic cumulative injury risk baseline assessment is used to evaluate the individualized baseline status of chronic cumulative injury risk in athletes. Specifically, a chronic cumulative injury risk baseline assessment model is constructed based on a long short-term memory neural network. The chronic cumulative injury risk baseline assessment data in the standardized dynamic baseline assessment data of the reference athletes is used as the model training data for iterative training of the chronic assessment model. Finally, the chronic cumulative injury risk baseline assessment data in the standardized dynamic baseline assessment data of the target athletes is input into the trained chronic cumulative injury risk baseline assessment model to obtain the target chronic cumulative baseline assessment result.

[0071] The specific process of the chronic evaluation model iterative training involves using the cross-entropy loss function, taking the chronic cumulative baseline evaluation results as supervision labels, and iteratively updating the weight matrix and bias parameters of the evaluation model through backpropagation algorithm and gradient descent optimization method. The evaluation model parameters are continuously optimized through multiple rounds of iteration. When the preset maximum number of training times is reached or the loss function value converges to a set threshold, the iterative training stops.

[0072] The baseline assessment of acute sudden injury risk is used to evaluate the individualized baseline status of an athlete's acute sudden injury risk. Specifically, it involves constructing an acute sudden injury risk baseline assessment model based on a long short-term memory neural network, using the acute sudden injury risk baseline assessment data from the standardized dynamic baseline assessment data of a reference athlete as the model training data, and conducting acute iterative training of the model. Finally, the acute sudden injury risk baseline assessment data from the standardized dynamic baseline assessment data of the target athlete is input into the trained acute sudden injury risk baseline assessment model to obtain the target acute sudden injury baseline assessment result.

[0073] The acute assessment model iterative training process specifically employs a cross-entropy loss function, uses the acute outbreak baseline assessment results as supervision labels, and iteratively updates the weight matrix and bias parameters of the assessment model through backpropagation algorithm and gradient descent optimization method. The assessment model parameters are continuously optimized through multiple rounds of iteration. When the preset maximum number of training times is reached or the loss function value converges to a set threshold, the iterative training stops.

[0074] By performing the above operations, this invention addresses the technical problem that traditional sports injury risk prediction methods typically use fixed risk thresholds for judgment, making it difficult to adapt to individual differences among athletes in terms of physical condition, training level, past injury history, and recovery ability. This can easily lead to high-level athletes being misjudged as high-risk, ordinary athletes or athletes with previous injuries being missed, and the same risk indicator not being suitable for different athletes. This invention innovatively proposes an individual dynamic baseline assessment mechanism. This mechanism acquires the chronic cumulative baseline state and acute sudden-onset baseline state of the target athlete, and incorporates the baseline assessment results into the subsequent risk prediction process. This allows risk judgment to move beyond simply relying on a uniform fixed threshold, instead combining the target athlete's historical training adaptation status, motor ability status, and current sports characteristics for individualized analysis. This improves the individual adaptability and accuracy of sports injury risk prediction, enhances the system's ability to identify risks in athletes with previous injuries, athletes at different training levels, and athletes in different sports states, reduces the probability of false alarms and missed alarms, and achieves individualized, dynamic, and precise assessment of sports injury risk.

[0075] Example 4, see Figure 1 and Figure 2 This embodiment is based on the above embodiment. The sports injury risk prediction model construction module is used to construct a basic risk prediction architecture model that can realize the prediction of chronic cumulative injury risk and acute sudden injury risk; specifically, it includes the following steps:

[0076] The acquisition of temporal features of sports injury risk is used to transform data into a unified feature representation form to improve the consistency of subsequent hypergraph modeling and temporal feature extraction. Specifically, the dynamic temporal features in the model input data are mapped by convolutional mapping operations, and the embedding results are concatenated to obtain the temporal features of sports injury risk.

[0077] A static injury mechanism hypergraph is constructed to capture relatively stable high-order correlations in sports injury risk. Specifically, body parts are used as injury risk nodes, and injury mechanism feature vectors are constructed based on the basic body state and individual dynamic baseline state corresponding to the nodes. The K-Means clustering algorithm is used to group nodes with similar injury mechanisms to the same hyperedge, resulting in the static injury mechanism hypergraph. The formula used is as follows:

[0078] ;

[0079] ;

[0080] In the formula, This represents the damage mechanism feature vector of the i-th damage risk node. Indicates basic physical characteristics. This represents the dynamic baseline state characteristics of an individual. The correlation matrix represents the static damage mechanism hypergraph. Represents the clustering function. This represents a feature matrix composed of the feature vectors of the damage mechanism of the damage risk nodes;

[0081] A dynamic risk semantic hypergraph is constructed to capture high-order risk relationships that change over time during motion. Specifically, based on the temporal features of motion injury risk, the semantic similarity between different injury risk feature channels is calculated, and the K-nearest neighbor algorithm is used to select the k nodes most similar to the current node to form a dynamic hyperedge, resulting in a dynamic risk semantic hypergraph set. The formula used is as follows:

[0082] ;

[0083] In the formula, This represents the first dynamic risk semantic hypergraph. This represents the second dynamic risk semantic hypergraph. This represents the R-th dynamic risk semantic hypergraph. This represents the K-nearest neighbor algorithm function. Indicating the temporal characteristics of sports injury risk;

[0084] Hypergraph convolution feature extraction is used to extract high-order risk association features based on static damage mechanism hypergraphs and dynamic risk semantic hypergraphs, enabling the model to simultaneously learn stable damage mechanism relationships and dynamic risk semantic relationships. Specifically, hypergraph convolution is performed based on the static hypergraph association matrix to obtain static private risk features, and hypergraph convolution is performed based on the dynamic risk semantic hypergraph to obtain dynamic shared risk features. The formulas used are as follows:

[0085] ;

[0086] ;

[0087] In the formula, R represents the dynamic shared risk characteristics, and R represents the number of dynamic risk semantic hypergraphs. This represents the weight of the r-th dynamic risk semantic hypergraph. This represents the association matrix of the r-th dynamic risk semantic hypergraph. Represents the node degree matrix, Represents the hypermarginality matrix. Represents the hyperedge weight matrix. Represents the learnable parameter matrix, Represents a non-linear activation function. This indicates static private risk characteristics. This represents the static hyperedge weight matrix. Represents the learnable parameter matrix;

[0088] Shared private feature gating fusion is used to adaptively fuse dynamic shared risk features and static private risk features, avoiding the excessive dominance of strongly correlated features in risk prediction results. Specifically, based on dynamic shared risk features and static private risk features, a gating fusion coefficient is calculated using a Sigmoid gating unit. The dynamic shared risk features and static private risk features are then adaptively weighted and fused according to the gating fusion coefficient to obtain the fused risk features. The formula used is as follows:

[0089] ;

[0090] ;

[0091] In the formula, Indicates the gating fusion coefficient. and Indicates learnable weights, Indicates the bias parameter. This indicates element-wise multiplication. Indicates the characteristics of integration risk;

[0092] Bidirectional temporal gating feature extraction is used to extract long-term trend information and short-term dynamic change information from fused risk features, so as to enhance the model's ability to identify chronic cumulative trends and acute sudden anomalies. Specifically, the fused risk features are input into a bidirectional temporal convolutional network to extract bidirectional temporal convolutional features, and the bidirectional temporal convolutional features are input into a bidirectional gated recurrent unit to obtain bidirectional gated temporal features.

[0093] The damage risk prediction output is specifically achieved by concatenating bidirectional gated temporal features with individual dynamic baseline state features, then inputting the concatenation into a multilayer perceptron, and finally outputting the damage risk prediction through the Softmax function.

[0094] By performing the above operations, this invention addresses the technical problems of existing sports injury risk prediction models failing to fully express the complex relationships between features and lacking the ability to identify long-term cumulative changes and short-term sudden anomalies, leading to inaccurate injury risk prediction. It innovatively proposes a basic risk prediction architecture model based on a static injury mechanism hypergraph, a dynamic risk semantic hypergraph, shared private feature gating fusion, and bidirectional temporal gating feature extraction. This model extracts stable high-order correlation features and dynamic risk correlation features through hypergraph convolution, and comprehensively identifies chronic cumulative trends and acute sudden changes through gating fusion and bidirectional temporal modeling. This improves the model's ability to express complex sports injury risk factors and its temporal recognition capabilities, enhancing the accuracy, stability, and robustness of injury risk prediction results, and achieving accurate prediction of sports injury risks.

[0095] Example 5, see Figure 1This embodiment is based on the above embodiment. The chronic cumulative injury risk prediction module is used to assess the risk of chronic sports injury that may occur under long-term training load based on the chronic cumulative baseline assessment results of the athlete, so as to realize the early prediction of cumulative load-type injury. Specifically, the chronic cumulative injury risk prediction data and the corresponding chronic cumulative baseline assessment results in the injury risk prediction data of the reference athlete are combined to form training data, and the sports injury risk prediction model is trained to obtain the chronic cumulative injury risk prediction model. Finally, the chronic cumulative injury risk prediction data and the corresponding target chronic cumulative baseline assessment results in the injury risk prediction data of the target athlete are input into the model to obtain the target chronic cumulative injury risk result.

[0096] The training process of the chronic cumulative injury risk prediction model specifically involves using the cross-entropy loss function, taking the chronic cumulative injury risk result as a supervision label, and iteratively updating the weight matrix and bias parameters of the prediction model through backpropagation algorithm and gradient descent optimization method. The prediction model parameters are continuously optimized through multiple iterations. When the preset maximum number of training times is reached or the loss function value converges to a set threshold, the iterative training stops.

[0097] Example 6, see Figure 1 This embodiment is based on the above embodiment. The acute sudden injury risk prediction module is used to assess the risk of acute injury that an athlete may experience during exercise based on the athlete's acute sudden baseline assessment results, and to achieve an immediate warning of sudden sports injury risk. Specifically, the acute sudden injury risk prediction data and the corresponding acute sudden baseline assessment results in the reference athlete injury risk prediction data are used to form training data to train the sports injury risk prediction model, thereby obtaining the acute sudden injury risk prediction model. Finally, the acute sudden injury risk prediction data and the corresponding target acute sudden baseline assessment results in the target athlete injury risk prediction data are input into the model to obtain the target acute sudden injury risk result.

[0098] The training process of the acute sudden injury risk prediction model specifically involves using the cross-entropy loss function, taking the acute sudden injury risk results as supervision labels, and iteratively updating the weight matrix and bias parameters of the prediction model through backpropagation algorithm and gradient descent optimization method. The prediction model parameters are continuously optimized through multiple iterations. When the preset maximum number of training times is reached or the loss function value converges to a set threshold, the iterative training stops.

[0099] Example 7, see Figure 1This embodiment is based on the above embodiment. The sports injury risk comprehensive management module is used to integrate chronic and acute risk results to achieve sports injury risk management. Specifically, based on the target chronic cumulative injury risk result and the target acute sudden injury risk result, the comprehensive assessment result of sports injury risk is determined according to the risk level of the two injury risk results, and the corresponding sports injury risk management result is generated to achieve intelligent comprehensive management of sports injury risk.

[0100] The process for determining the comprehensive assessment result of sports injury risk specifically uses the higher risk level among the two injury risk results as the comprehensive assessment result.

[0101] Specifically, when the comprehensive assessment result of sports injury risk is no risk, a normal sports management result is generated, prompting the target athlete to maintain the current sports status.

[0102] When the comprehensive assessment result of the sports injury risk is low risk, a mild risk management result is generated, prompting the target athlete to pay attention to changes in exercise status and appropriately control exercise intensity.

[0103] When the comprehensive assessment result of the sports injury risk is medium risk, a medium risk management result is generated, prompting the target athlete to reduce exercise intensity, adjust training plan or reduce high-risk movements.

[0104] When the comprehensive assessment result of the sports injury risk is high, a high-risk management result is generated, prompting the target athlete to stop the current exercise and rest.

[0105] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0106] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

[0107] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. A big data based sports injury risk prediction system characterized in that: It includes a standardized data acquisition module, an individual dynamic baseline assessment module, a sports injury risk prediction model construction module, a chronic cumulative injury risk prediction module, an acute sudden injury risk prediction module, and a comprehensive sports injury risk management module; The standardized data acquisition module obtains standardized data for predicting sports injury risks through data collection and data preprocessing. The individual dynamic baseline assessment module constructs a baseline assessment model for chronic cumulative injury risk and an acute sudden injury risk based on a long short-term memory neural network. It trains the model using standardized dynamic baseline assessment data of a reference athlete and inputs the standardized dynamic baseline assessment data of the target athlete into the trained baseline assessment model to obtain the target chronic cumulative baseline assessment result and the target acute sudden baseline assessment result. The sports injury risk prediction model construction module constructs a sports injury risk prediction model based on sports injury risk temporal feature acquisition, static injury mechanism hypergraph, dynamic risk semantic hypergraph, hypergraph convolutional feature extraction, shared private feature gating fusion, bidirectional temporal gating feature extraction, and injury risk prediction output. The chronic cumulative injury risk prediction module is used to predict the risk of chronic cumulative injury that will occur under long-term training load for the target athlete. Based on the sports injury risk prediction model and the target chronic cumulative baseline assessment results, the target chronic cumulative injury risk result is obtained. The acute injury risk prediction module is used to predict the risk of acute injury that may occur to the target athlete during exercise. Based on the sports injury risk prediction model and the baseline assessment results of the target acute sudden injury, the target acute sudden injury risk results are obtained; The sports injury risk management module determines the comprehensive assessment result of sports injury risk based on the target chronic cumulative injury risk result and the target acute sudden injury risk result.

2. The big data based sports injury risk prediction system of claim 1, wherein: The individual dynamic baseline assessment module specifically includes the following steps: The baseline assessment of chronic cumulative injury risk is specifically constructed based on a long short-term memory neural network to build a baseline assessment model for chronic cumulative injury risk. The baseline assessment data of chronic cumulative injury risk in the standardized dynamic baseline assessment data of the reference athlete is used as the training data for the model. The assessment model is iteratively trained for chronicity. Finally, the baseline assessment data of chronic cumulative injury risk in the standardized dynamic baseline assessment data of the target athlete is input into the trained baseline assessment model for chronic cumulative injury risk to obtain the target chronic cumulative baseline assessment result. The baseline assessment of acute sudden injury risk is specifically carried out by constructing an acute sudden injury risk baseline assessment model based on a long short-term memory neural network. The acute sudden injury risk baseline assessment data in the standardized dynamic baseline assessment data of the reference athlete is used as the model training data for acute iterative training. Finally, the acute sudden injury risk baseline assessment data in the standardized dynamic baseline assessment data of the target athlete is input into the trained acute sudden injury risk baseline assessment model to obtain the target acute sudden baseline assessment result.

3. The big data based sports injury risk prediction system of claim 1, wherein: The sports injury risk prediction model construction module specifically includes the following steps: The acquisition of temporal features of sports injury risk involves performing feature mapping on the dynamic temporal features in the model input data through convolutional mapping operations, and then concatenating the features of each embedding result to obtain the temporal features of sports injury risk. To construct a static damage mechanism hypergraph, body parts are used as damage risk nodes. Damage mechanism feature vectors are constructed based on the basic body state and individual dynamic baseline state corresponding to the nodes. The K-Means clustering algorithm is used to divide nodes with similar damage mechanisms into the same hyperedge to obtain the static damage mechanism hypergraph. The construction of dynamic risk semantic hypergraphs involves calculating the semantic similarity between different injury risk feature channels based on the temporal features of sports injury risk, and using the K-nearest neighbor algorithm to select the k nodes most similar to the current node to form dynamic hyperedges, thus obtaining a set of dynamic risk semantic hypergraphs. Hypergraph convolution feature extraction specifically involves performing hypergraph convolution based on the static hypergraph association matrix to obtain static private risk features, and performing hypergraph convolution based on the dynamic risk semantic hypergraph to obtain dynamic shared risk features. The shared private feature gating fusion is specifically based on dynamic shared risk features and static private risk features. The gating fusion coefficient is calculated through a Sigmoid gating unit. The dynamic shared risk features and static private risk features are then adaptively weighted and fused according to the gating fusion coefficient to obtain the fused risk features. Bidirectional temporal gating feature extraction specifically involves inputting the fused risk features into a bidirectional temporal convolutional network to extract bidirectional temporal convolutional features, and then inputting the bidirectional temporal convolutional features into a bidirectional gated recurrent unit to obtain bidirectional gated temporal features. The damage risk prediction output is specifically achieved by concatenating bidirectional gated temporal features with individual dynamic baseline state features, then inputting the concatenation into a multilayer perceptron, and finally outputting the damage risk prediction through the Softmax function.

4. The big data based sports injury risk prediction system of claim 1, wherein: The chronic cumulative injury risk prediction module specifically involves using the chronic cumulative injury risk prediction data from the reference athlete injury risk prediction data and the corresponding chronic cumulative baseline assessment results as training data to train the sports injury risk prediction model, thereby obtaining the chronic cumulative injury risk prediction model. Finally, the chronic cumulative injury risk prediction data from the target athlete injury risk prediction data and the corresponding target chronic cumulative baseline assessment results are input into the model to obtain the target chronic cumulative injury risk result.

5. The big data based sports injury risk prediction system of claim 1, wherein: The acute sudden injury risk prediction module specifically involves using the acute sudden injury risk prediction data from the reference athlete injury risk prediction data and the corresponding acute sudden baseline assessment results as training data to train the sports injury risk prediction model, thereby obtaining the acute sudden injury risk prediction model. Finally, the acute sudden injury risk prediction data from the target athlete injury risk prediction data and the corresponding target acute sudden baseline assessment results are input into the model to obtain the target acute sudden injury risk result.

6. The big data based sports injury risk prediction system of claim 1, wherein: The sports injury risk comprehensive management module specifically determines the comprehensive assessment result of sports injury risk based on the target chronic cumulative injury risk result and the target acute sudden injury risk result, and generates the corresponding sports injury risk management result, thereby realizing intelligent comprehensive management of sports injury risk.

7. The big data based sports injury risk prediction system of claim 1, wherein: The standardized data acquisition module specifically obtains raw data for sports injury risk prediction through data acquisition operations, and performs data preprocessing on the raw data for sports injury risk prediction to obtain standardized data for sports injury risk prediction. The standardized data for predicting sports injury risk includes standardized data for reference athlete dynamic baseline assessment, standardized data for target athlete dynamic baseline assessment, standardized data for reference athlete injury risk prediction, and standardized data for target athlete injury risk prediction. The data preprocessing specifically involves performing data cleaning, normalization, and feature selection on the raw data to obtain the standardized data for predicting sports injury risk.

Images