Real-time, full episode information gathering method for orthopedic patients

Abstract

The application is a real-time whole-course information acquisition method for orthopedic patients, and belongs to the field of intelligent monitoring and risk management, and its core process comprises: personalized model construction: based on physiological baseline data of the patient and a predefined biomechanics causality graph, an initial state vector is generated; real-time state inference: a causality graph neural network is used to combine external observable data to infer the in-vivo biomechanics state which cannot be directly observed in real time; risk quantification and decision making: a risk influence index of the real-time state is calculated and compared with a preset threshold to generate an active information acquisition decision; closed-loop information acquisition: intervention tasks are pushed to the patient according to the decision, and newly acquired data is fed back to the network to continuously update the state. The application solves the hysteresis problem of traditional rehabilitation monitoring, and changes the previous passive and delayed management mode into active and real-time precise monitoring.

Description

Technical Field

[0001] This invention relates to the field of intelligent monitoring and risk management technology for orthopedic rehabilitation, specifically a method for real-time, full-course information collection for orthopedic patients. Background Technology

[0002] Information collection for orthopedic patients' rehabilitation process is a technology aimed at achieving precise and forward-looking management of patients; during the patient's rehabilitation process, the patient's biomechanical state is the core basis for assessing rehabilitation progress, predicting future trends, and developing personalized guidance plans;

[0003] However, in actual rehabilitation management, these crucial biomechanical states cannot be directly observed. Existing rehabilitation monitoring methods mainly rely on patients' subjective feedback and infrequent in-hospital examinations. This passive and delayed information acquisition model leads to significant delays in rehabilitation monitoring, makes it impossible to accurately predict patients' recovery trajectories, and hinders the provision of truly personalized rehabilitation guidance, thus affecting the accuracy and foresight of rehabilitation management.

[0004] The information disclosed in the background section above is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] The purpose of this invention is to provide a method for real-time, full-course information acquisition for orthopedic patients, in order to solve the problems mentioned in the background section; specifically, the technical solution of this invention includes the following steps:

[0006] S1. Personalized model construction steps: Obtain the patient's static physiological baseline data and predefined biomechanical causal graph; Based on the static physiological baseline data, calculate the initial weight of each state variable node in the biomechanical causal graph; Combine the initial weights with the predefined node basic state vector to generate a personalized initial state vector for each node.

[0007] S2. Real-time state inference steps: Collect externally observable data of the patient; use a causal graph neural network to infer the real-time biomechanical state of all unobserved nodes. The causal graph neural network uses a biomechanical causal graph as its topology and is initialized with a personalized initial state vector. It also uses externally observable data as input for evidence nodes.

[0008] S3. Risk Quantification and Decision-Making Steps: Based on real-time biomechanical status, calculate the risk influence index of each node; compare the risk influence index with the preset risk influence trigger threshold to generate proactive information collection decisions;

[0009] S4. Closed-loop information acquisition steps: In response to the proactive information acquisition decision, select and push intervention tasks to the patient; feed the newly acquired data through the execution of the intervention tasks back to the causal graph neural network to update the real-time biomechanical state.

[0010] Optionally, in the personalized model building step, the static physiological baseline data includes age data, gender data, weight data, and surgical type data.

[0011] Optionally, in the real-time state inference step, the training of the causal graph neural network is optimized using a composite loss function, which is composed of a weighted average of data fidelity loss and causal consistency loss.

[0012] Optionally, data fidelity loss is used to measure the difference between the model's predicted values ​​and the actual measured values ​​of evidence nodes; causal consistency loss is used to penalize influence connections that violate the pre-defined topological structure of the biomechanical causal graph.

[0013] Optionally, in the risk quantification and decision-making process, the step of calculating the risk impact index includes:

[0014] S31. Use real-time biomechanical state as initial conditions for forward reasoning to generate future rehabilitation trajectories;

[0015] S32. The Monte Carlo random deactivation technique is used to quantify the prediction uncertainty of each node;

[0016] S33. Calculate the sensitivity of long-term rehabilitation outcomes to node states by backpropagating the causal graph neural network.

[0017] S34. Multiply the prediction uncertainty by the sensitivity to obtain the risk influence index of the node.

[0018] Optionally, in the risk quantification and decision-making process, the generation of proactive information gathering decisions includes:

[0019] If the risk influence index of any node exceeds the risk influence trigger threshold, a decision to trigger information collection will be generated.

[0020] If the risk influence index of all nodes does not exceed the risk influence trigger threshold, a decision is generated to continuously execute the real-time state inference step.

[0021] Optionally, in the closed-loop information collection step, the selection of intervention tasks is based on the principle of maximizing risk reduction, and is chosen from a predefined intervention task library.

[0022] Optionally, the steps for selecting an intervention task include:

[0023] S41. Identify the set of nodes that can provide information for each optional intervention task;

[0024] S42. Summarize and calculate the risk influence index of all nodes in each node set to obtain the total risk influence;

[0025] S43. Select the intervention task with the greatest overall risk impact and push it as the optimal intervention task.

[0026] Compared with the prior art, the present invention has the following beneficial effects:

[0027] 1. It enables real-time and accurate inference of the patient's internal state, solving the problem of lag caused by traditional rehabilitation monitoring relying on subjective patient feedback and low-frequency in-hospital examinations; by constructing a biomechanical causal map and using a causal graph neural network to process external observable data, it can infer the internal biomechanical state that cannot be directly observed in real time, transforming the previous passive and delayed management mode into proactive and real-time accurate monitoring.

[0028] 2. An individualized rehabilitation mechanism model was constructed, which improved the personalization and accuracy of rehabilitation guidance. The method incorporates static physiological baseline data such as the patient's age, weight, and type of surgery in the initial stage, generating an initial state vector that conforms to the individual characteristics of each patient. This ensures that all subsequent inferences and predictions are based on a digital model starting point that is closer to the patient's real situation, providing a reliable foundation for achieving truly personalized rehabilitation guidance.

[0029] 3. A forward-looking risk quantification and decision-making mechanism was introduced, realizing the transformation from passive response to proactive risk management. This method can not only infer the current state, but also predict the future rehabilitation trajectory. By comprehensively assessing the uncertainty of the predicted nodes and their sensitivity to long-term rehabilitation outcomes, a risk influence index is calculated, thereby proactively identifying the key risk points that have the greatest impact on rehabilitation outcomes and realizing forward-looking risk warning.

[0030] 4. A closed-loop proactive information collection system has been established, which significantly improves the efficiency and targeting of information collection. When a key risk is identified, the system will select and push the task that can most effectively eliminate the current core uncertainty from the predefined intervention task library based on the principle of maximizing risk reduction. This goal-oriented closed-loop feedback mechanism ensures that every request for information from the patient is efficient and necessary, avoids blind interference, and realizes intelligent and precise management of the rehabilitation process. Attached Figure Description

[0031] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0032] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0033] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0034] Example 1:

[0035] Please see Figure 1 A method for real-time, full-course information collection for orthopedic patients includes the following steps:

[0036] S1. Personalized model construction steps: Obtain the patient's static physiological baseline data and predefined biomechanical causal graph; Based on the static physiological baseline data, calculate the initial weight of each state variable node in the biomechanical causal graph; Combine the initial weights with the predefined node basic state vector to generate a personalized initial state vector for each node.

[0037] S2. Real-time state inference steps: Collect externally observable data of the patient; use a causal graph neural network to infer the real-time biomechanical state of all unobserved nodes. The causal graph neural network uses a biomechanical causal graph as its topology and is initialized with a personalized initial state vector. It also uses externally observable data as input for evidence nodes.

[0038] S3. Risk Quantification and Decision-Making Steps: Based on real-time biomechanical status, calculate the risk influence index of each node; compare the risk influence index with the preset risk influence trigger threshold to generate proactive information collection decisions;

[0039] S4. Closed-loop information acquisition steps: In response to the proactive information acquisition decision, select and push intervention tasks to the patient; feed the newly acquired data through the execution of the intervention tasks back to the causal graph neural network to update the real-time biomechanical state.

[0040] This invention provides a method for real-time, full-course information collection for orthopedic patients. Its purpose is to address the problems of existing technologies, which, due to the inability to directly observe the patient's biomechanical state, lead to delayed rehabilitation monitoring and an inability to provide accurate predictions and personalized guidance. This embodiment constructs a personalized biomechanical model and combines it with a causal graph neural network for real-time inference and risk prediction. Ultimately, through closed-loop active information collection, it achieves precise and forward-looking management of the rehabilitation process.

[0041] The overall technical process of this embodiment, as a complete and self-consistent technical closed loop, includes the following steps:

[0042] The personalized model construction step S1 aims to create a digital rehabilitation mechanism model for each patient that conforms to their individual characteristics, serving as the foundational framework for all subsequent inferences and predictions. This step acquires the patient's static physiological baseline data and a predefined biomechanical causal graph. The biomechanical causal graph is a directed acyclic graph predefined by orthopedic medical experts as the basis of domain knowledge. Its function is to provide inviolable medical logical constraints for subsequent algorithms. Nodes represent key state variables in the rehabilitation process, and directed edges represent medical causal relationships between variables. Based on the static physiological baseline data, the initial weight of each state variable node in the biomechanical causal graph is calculated. In this embodiment, this calculation is implemented through a linear weighted model, designed to map discrete patient static information into initial state parameters that have a real impact on the model.

[0043] To integrate individual patient information into a general medical model, this embodiment uses the following formula to calculate the initial weights. :

[0044]

[0045] in, The initial weight of the i-th graph node is a dimensionless scalar, calculated by this formula.

[0046] The global baseline weight of a node is a dimensionless scalar that represents the average level of the general population and is preset in the system.

[0047] The total number of physiological baseline parameters, an integer, depends on the types of parameters collected;

[0048] The j-th physiological baseline parameter value after normalization is a dimensionless scalar obtained from the normalized raw data.

[0049] The weighting coefficient of the j-th physiological parameter is dimensionless; this weighting coefficient It was obtained in advance through offline multivariate regression analysis on a calibrated dataset covering numerous historical rehabilitation cases, reflecting the statistical strength of the influence of different physiological parameters on specific state variables;

[0050] This formula uses the patient's individual physiological parameters And the coefficient reflecting its importance Multiply and add to the baseline weights. Above, this ensures the calculated It encompasses both group commonalities and individual differences; by combining initial weights with predefined node base state vectors, a personalized initial state vector is generated for each node; node base state vector It is a learnable vector associated with the node type, which represents the general feature embedding of the node in general cases. Its dimension is determined by the network structure, and its initial value can be obtained through pre-training.

[0051] Personalized initial state vector The generation method is as follows:

[0052]

[0053] in, : The personalized initial state vector of the i-th node, which is a vector calculated by this formula, and serves as the initial input to the subsequent graph neural network;

[0054] The individualized scalar weights calculated in the preceding steps are dimensionless scalars, and their source is the same as above;

[0055] : The basic state vector associated with node i type, which is a vector and is predefined in the system or learned through training;

[0056] The physical significance of this operation lies in using individualized scalar weights for each patient. To scale or modulate a general fundamental feature vector This generates an initial state vector that incorporates both general domain knowledge and reflects individual differences; for example, the initial weights of cartilage stress-related nodes for a heavier patient. The initial state vector will be correspondingly higher, thus making its initial state vector... The larger norm allows the model to reflect the patient's individualized risk characteristics from the very beginning of the calculation;

[0057] The core of the real-time state inference step S2 is to use a causal graph neural network to infer the patient's complete, unobservable biomechanical state in real time based on sparse external data generated daily by the patient. This step collects the patient's externally observable data, such as joint flexion and extension angles collected in real time by wearable sensors or subjective pain scores recorded by a mobile app. The causal graph neural network is then used to infer the real-time biomechanical state of all unobserved nodes. In this embodiment, the causal graph neural network uses the biomechanical causal graph constructed in the previous step as its topology. This means that the network's connections are constrained by prior medical knowledge, avoiding inferences that violate medical logic from black-box models. Furthermore, the network is initialized using the personalized initial state vector generated in the previous step, ensuring that the starting point of the inference conforms to the patient's individual characteristics. Externally observable data is input into the network as evidence nodes. The graph neural network, through an information propagation mechanism, inversely infers the real-time values ​​of all unobserved nodes based on the values ​​of a few evidence nodes.

[0058] The risk quantification and decision-making step S3 aims to predict the patient's future recovery trend using a model and proactively identify key risk points affecting the prediction results. Based on the real-time biomechanical state obtained in the previous step, the risk influence index for each node is calculated. It is a custom composite indicator that quantifies the potential volatility of long-term rehabilitation outcomes caused by the uncertainty of a node's current state, thereby identifying the high-risk variables that require the most attention. It compares the risk impact index with a preset risk impact trigger threshold to generate proactive information collection decisions. (Risk impact trigger threshold) It is a value set by clinical experts based on an acceptable risk level. Its purpose is to determine whether the risk predicted by the current model exceeds the acceptable range, thereby deciding whether intervention is needed.

[0059] The closed-loop information collection step S4, based on the decision made in the previous stage, proactively requests key information from the patient that can minimize the current core risk, thus forming a complete closed-loop system from discovering the unknown to proactively clarifying it; in response to the proactive information collection decision, that is, when a certain node... Exceed When the system is activated, it selects to push intervention tasks to the patient; the newly collected data through the execution of the intervention tasks is fed back to the causal graph neural network to update the real-time biomechanical state; the newly collected data is used as new evidence nodes to input the model. Since this information is collected in a highly targeted manner to eliminate the highest risk, it can significantly reduce the uncertainty of relevant nodes, thereby effectively reducing the overall prediction risk and completing an intelligent closed loop.

[0060] Through the above-mentioned technical solutions, this invention constructs a complete system from personalized modeling, real-time state inference, prospective risk quantification to closed-loop proactive information collection. Compared with the passive and lagging mode of existing technologies that rely on subjective patient reports and low-frequency in-hospital examinations, this invention can infer the invisible biomechanical state of the patient in real time and accurately, predict the rehabilitation trajectory, and proactively identify and eliminate uncertainties in the prediction, thereby bringing significant technical effects. This greatly improves the accuracy and foresight of rehabilitation process management and can provide reliable data and decision support for achieving truly personalized rehabilitation guidance.

[0061] Example 2:

[0062] In the personalized model construction process, static physiological baseline data includes age data, gender data, weight data, and surgical type data.

[0063] This embodiment is a concretization of the personalized model construction steps in Embodiment 1. In this step, to ensure the accuracy and comprehensiveness of model personalization, static physiological baseline data includes age data, gender data, weight data, and surgical type data. These data are clinically recognized key factors affecting orthopedic rehabilitation outcomes. By using them as initial inputs to the model, the initial calibration accuracy of the personalized biomechanical causal map can be significantly improved. For example, age and weight directly affect tissue load-bearing capacity and recovery speed, while gender and surgical type are closely related to endocrine levels, tissue structure, and injury patterns. By explicitly specifying these specific physiological baseline data, the effectiveness and reliability of the personalized model calibration are ensured. Compared to using vague or incomplete data, this approach makes the initial weights... The calculations are more targeted, allowing the subsequent inferences and predictions of the entire model to start from a point that is closer to the patient's actual condition, thus improving the accuracy of the overall plan.

[0064] Example 3:

[0065] In the real-time state inference step, the training of the causal graph neural network is optimized using a composite loss function, which is composed of a weighted average of data fidelity loss and causal consistency loss.

[0066] Data fidelity loss is used to measure the difference between the model's predicted values ​​and the actual measured values ​​of evidence nodes; causal consistency loss is used to penalize influence connections that violate the pre-defined topological structure of the biomechanical causal graph.

[0067] This embodiment is a further optimization of the causal graph neural network training process in the real-time state inference step of Embodiment 1. To ensure that the inference results not only fit the observed data but also strictly adhere to the preset biomechanical causal laws, in this embodiment, the training of the causal graph neural network is optimized using a composite loss function, which is composed of a weighted average of data fidelity loss and causal consistency loss.

[0068] The purpose of this composite loss function is to simultaneously drive the network parameters to converge toward two objectives when minimizing the total loss using gradient descent: first, to fit the observed data; and second, to adhere to the causal structure. Its specific form is as follows:

[0069]

[0070] in, The total loss function is a dimensionless scalar and serves as the ultimate goal of network optimization.

[0071] Data fidelity loss is a dimensionless scalar quantity, calculated using the following formula;

[0072] The loss of causal consistency is a dimensionless scalar quantity, calculated by the following formula;

[0073] The causal constraint weight hyperparameter is dimensionless and is determined by performing a grid search on the validation set to achieve optimal generalization performance.

[0074] Data fidelity loss This is used to measure the difference between the model's predicted values ​​and the actual measured values ​​for evidence nodes; its form is derived from the mean squared error in statistics, and it is calculated as follows:

[0075]

[0076] in, The number of observable nodes in each training session, an integer, depends on the batch of input data;

[0077] The normalized true measurement value of the observable node is dimensionless and is obtained by collecting and processing data through sensors and other equipment.

[0078] The model's predictions for observable nodes are dimensionless and are calculated by forward propagation of the graph neural network.

[0079] Loss of causal consistency , is a custom loss function defined in this invention, used to penalize influential connections that violate the preset topological structure of the biomechanical causal graph; its purpose is to force the internal influential relationships learned by the network to be consistent with prior medical knowledge; the calculation method is as follows:

[0080]

[0081] in, The adjacency matrix of the predefined causal graph is a matrix that is predefined in the system. This indicates the existence of a causal edge from node j to i; otherwise, it is 0.

[0082] The weight matrix, dynamically generated by the attention mechanism during inference in a neural network, is a matrix calculated from the network's internal parameters. This represents the influence of node j on node i;

[0083] The Sigmoid function is a function that normalizes influence values ​​to a range of 0-1.

[0084] When there is no causal path from j to i in the graph At times, if the network mistakenly learns a strong influence... If it is larger, then The term is 1, resulting in This increases the penalty on network parameters, forcing them to correct the faulty connections in subsequent updates;

[0085] By introducing this composite loss function, this invention achieves a deep integration of data-driven and knowledge-driven approaches; compared with traditional neural network training methods that simply rely on data fitting, this scheme utilizes causal consistency loss. By incorporating prior knowledge of orthopedic medicine as a strong constraint into model training, the model effectively avoids generating inferences that violate biological norms. This not only significantly improves the model's inference accuracy and robustness under sparse observation data, but also enhances the model's interpretability and credibility in clinical applications.

[0086] Example 4:

[0087] In the risk quantification and decision-making process, the steps for calculating the risk impact index include:

[0088] S31. Use real-time biomechanical state as initial conditions for forward reasoning to generate future rehabilitation trajectories;

[0089] S32. The Monte Carlo random deactivation technique is used to quantify the prediction uncertainty of each node;

[0090] S33. Calculate the sensitivity of long-term rehabilitation outcomes to node states by backpropagating the causal graph neural network.

[0091] S34. Multiply the prediction uncertainty by the sensitivity to obtain the risk influence index of the node.

[0092] This embodiment is a detailed description of the process for calculating the risk impact index in the risk quantification and decision-making steps of Embodiment 1; this calculation process precisely combines the uncertainty of prediction with its potential impact on clinical outcomes, and includes the following steps:

[0093] Using real-time biomechanical states as initial conditions for forward reasoning to generate future rehabilitation trajectories; that is, using a trained graph neural network model to iterate over time to predict the evolution trend of all key state variables over a period of time.

[0094] Monte Carlo random deactivation is employed to quantify the prediction uncertainty of each node; Monte Carlo random deactivation is performed on the same input during the prediction phase. The forward propagation with random inactivation yields... Groups of different prediction results; quantify uncertainty by calculating the standard deviation of these results; node prediction uncertainty. The calculation formula is as follows:

[0095]

[0096] in, : No. The prediction uncertainty of each node, its dimensions and the predicted value Consistent, calculated using this formula;

[0097] The total number of random forward propagations, which is an integer, based on the experiment. The convergence condition is determined, typically between 30 and 50.

[0098] : No. During the second propagation, the node... The predicted value is a numerical value obtained from a single forward propagation of the model;

[0099] : For nodes of The average of the predictions is a numerical value. Second-rate We get the average.

[0100] By performing backpropagation on a causal graph neural network, the sensitivity of long-term rehabilitation outcomes to node states is calculated; long-term rehabilitation outcomes It is a predefined indicator by clinical experts that represents key long-term rehabilitation outcomes, such as expected maximum range of motion on day 90; sensitivity. Representative outcome indicators For nodes Current state value The degree of dependence is automatically calculated by performing standard gradient backpropagation on the trained network model;

[0101] Multiplying the forecast uncertainty by the sensitivity yields the node's risk impact index; the final risk impact index... The calculation logic is to multiply the uncertainty of a node by its sensitivity to long-term results, as shown in the following formula:

[0102]

[0103] in, :node The risk influence index, its dimensions and long-term results Consistent, calculated using this formula;

[0104] The uncertainty in node predictions calculated in the preceding steps has the same dimensions as... Consistent, sourced from the same source as above;

[0105] The sensitivity calculated in the preceding steps has the following dimensions: Source: same as above;

[0106] The logic behind this formula is that the risk of a node depends not only on its own uncertainty. The risk level is high, but it also depends on its high influence (sensitivity) on the final rehabilitation outcome; only when both are high does the node constitute a true core risk. Therefore, this embodiment defines in detail an innovative risk quantification method. Compared to traditional methods that only focus on predicted values ​​or simple uncertainties, this solution calculates a risk influence index. It can reveal the key risk points in the rehabilitation process more deeply; it not only tells us which state is uncertain, but more importantly, it tells us which uncertain state has the greatest impact on the final rehabilitation outcome, thus making subsequent proactive information collection more targeted and efficient, and realizing a qualitative change from passive response to proactive risk management.

[0107] Example 5:

[0108] In the risk quantification and decision-making process, the generation of proactive information gathering decisions includes:

[0109] If the risk influence index of any node exceeds the risk influence trigger threshold, a decision to trigger information collection will be generated.

[0110] If the risk influence index of all nodes does not exceed the risk influence trigger threshold, a decision is generated to continuously execute the real-time state inference step.

[0111] This embodiment is a concretization of the proactive information collection decision logic generated in the risk quantification and decision-making steps of Embodiment 1. This decision-making mechanism is simple and effective, ensuring that the system intervenes only when necessary. Specifically, the generation of proactive information collection decisions includes:

[0112] If the risk influence index of any node Exceeding the risk impact trigger threshold If this occurs, a decision is generated that triggers information collection; this indicates that at least one node in the system has predictive uncertainty and its potential negative impact on long-term outcomes that has reached a clinically unacceptable level, and immediate action must be taken to eliminate this risk.

[0113] If the risk influence index of all nodes None of them exceeded the risk impact trigger threshold. If the current model's prediction is within an acceptable risk range, the system will continue to perform silent monitoring in the background without disturbing the patient until a new risk emerges.

[0114] By establishing clear trigger thresholds and decision-making logic, this embodiment realizes an automated, non-invasive monitoring system. It avoids unnecessary and frequent information requests to patients, and only initiates intervention when truly critical risks are identified, thereby achieving a good balance between ensuring rehabilitation safety and improving user experience.

[0115] Example 6:

[0116] In the closed-loop information collection process, the selection of intervention tasks is based on the principle of maximizing risk reduction, and is chosen from a predefined intervention task library.

[0117] The steps for selecting an intervention task include:

[0118] S41. Identify the set of nodes that can provide information for each optional intervention task;

[0119] S42. Summarize and calculate the risk influence index of all nodes in each node set to obtain the total risk influence;

[0120] S43. Select the intervention task with the greatest overall risk impact and push it as the optimal intervention task.

[0121] This embodiment is a further refinement and optimization of how to select the optimal intervention task in the closed-loop information collection step of Embodiment 1. To ensure that the collected information can eliminate the currently identified core risks most efficiently, the selection of intervention tasks is based on the principle of maximizing risk reduction and is selected from a predefined intervention task library. The intervention task library is a pre-established database that stores a series of specific rehabilitation actions or questionnaires, and each task is pre-labeled by physical therapists and biomechanics experts, clarifying its strong correlation with one or more unobserved nodes in the causal graph.

[0122] The steps for selecting the optimal intervention task are based on the following decision model:

[0123] Identify the set of nodes that can provide information for each optional intervention task; that is, for each candidate task in the task library. Find the set of nodes that can directly or indirectly provide observation data for performing this task. This correspondence is pre-defined and stored in the system.

[0124] The risk influence index of all nodes within each node set is calculated to obtain the total risk influence; that is, for each candidate task... Calculate its corresponding node set The sum of the risk influence indices of all nodes within the system;

[0125] The intervention task with the greatest overall risk impact is selected as the optimal intervention task and pushed out; this selection process is expressed by the following formula:

[0126]

[0127] in, The selected optimal intervention task is identified by this formula.

[0128] : The set of all optional intervention tasks, which is a set derived from a predefined intervention task library;

[0129] The set of nodes that can provide information for task a is a pre-defined set.

[0130] The node risk influence index calculated in the previous steps is a numerical value, sourced from the same source as above;

[0131] The logic of this formula is to select the intervention task that can solve the node set with the largest sum of total risk impact in the current system in one go.

[0132] This embodiment ensures that every proactive information collection is highly efficient by maximizing risk reduction. It avoids blindly soliciting information from patients and instead precisely selects intervention tasks that can effectively address the root cause of the greatest uncertainty. This not only maximizes the efficiency of the closed-loop information collection process but also minimizes disruption to patients, thereby improving the intelligence and practicality of the entire system.

[0133] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for real-time, full-course information collection for orthopedic patients, characterized in that, Includes the following steps: S1. Personalized model construction steps: Obtain the patient's static physiological baseline data and predefined biomechanical causal graph; Based on the static physiological baseline data, calculate the initial weight of each state variable node in the biomechanical causal graph; By combining the initial weights with the predefined basic state vector of the nodes, a personalized initial state vector is generated for each node. S2. Real-time state inference steps: Collect externally observable data of the patient; use a causal graph neural network to infer the real-time biomechanical state of all unobserved nodes. The causal graph neural network uses a biomechanical causal graph as its topology and is initialized with a personalized initial state vector. It also uses externally observable data as input for evidence nodes. S3. Risk Quantification and Decision-Making Steps: Based on real-time biomechanical status, calculate the risk influence index of each node; compare the risk influence index with the preset risk influence trigger threshold to generate proactive information collection decisions; S4. Closed-loop information acquisition steps: In response to the proactive information acquisition decision, select and push intervention tasks to the patient; feed the newly acquired data through the execution of the intervention tasks back to the causal graph neural network to update the real-time biomechanical state; In the risk quantification and decision-making process, the steps for calculating the risk impact index include: S31. Use real-time biomechanical state as initial conditions for forward reasoning to generate future rehabilitation trajectories; S32. The Monte Carlo random deactivation technique is used to quantify the prediction uncertainty of each node; the formula for prediction uncertainty is: ； in, For the first The prediction uncertainty of each node This represents the total number of random forward propagations. For the first During the second propagation, the node... The predicted value, For nodes of The average of the predictions; S33. Calculate the sensitivity of long-term rehabilitation outcomes to node states by backpropagating the causal graph neural network. S34. Multiply the prediction uncertainty by the sensitivity to obtain the risk influence index of the node; the formula for the risk influence index is: ； in, For nodes The risk influence index, For the first The prediction uncertainty of each node The sensitivity of long-term rehabilitation outcomes to node status; The steps for selecting an intervention task include: S41. Identify the set of nodes that can provide information for each optional intervention task; S42. Summarize and calculate the risk influence index of all nodes in each node set to obtain the total risk influence; S43. Select the intervention task with the greatest overall risk impact and push it as the optimal intervention task. The formula for the optimal intervention task is: ； in, The selected optimal intervention task. For the set of all optional intervention tasks, The set of nodes that can provide information for performing task a. For nodes The risk impact index.

2. The method for real-time full-course information collection for orthopedic patients according to claim 1, characterized in that, In the personalized model construction process, static physiological baseline data includes age data, gender data, weight data, and surgical type data.

3. The method for real-time full-course information collection for orthopedic patients according to claim 1, characterized in that, In the real-time state inference step, the training of the causal graph neural network is optimized using a composite loss function, which is composed of a weighted average of data fidelity loss and causal consistency loss.

4. The method for real-time full-course information collection for orthopedic patients according to claim 3, characterized in that, Data fidelity loss is used to measure the difference between the model's predicted values ​​and the actual measured values ​​of evidence nodes; causal consistency loss is used to penalize influence connections that violate the pre-defined topological structure of the biomechanical causal graph.

5. The method for real-time full-course information collection for orthopedic patients according to claim 1, characterized in that, In the risk quantification and decision-making process, the generation of proactive information gathering decisions includes: If the risk influence index of any node exceeds the risk influence trigger threshold, a decision to trigger information collection will be generated. If the risk influence index of all nodes does not exceed the risk influence trigger threshold, a decision is generated to continuously execute the real-time state inference step.

6. The method for real-time full-course information collection for orthopedic patients according to claim 1, characterized in that, In the closed-loop information collection process, the selection of intervention tasks is based on the principle of maximizing risk reduction, and is chosen from a predefined intervention task library.

Images