Digital twinning driven perioperative period nursing scheme dynamic optimization method

Abstract

The invention relates to a digital twin-driven perioperative period nursing scheme dynamic optimization method, which comprises the following steps of: setting an intervention sensitive area associated with a current postoperative state of a patient, and delimiting and limiting a physical limiting condition of a nursing operation triggering threshold according to an operation type and an anatomical position of the patient; setting a time period from the completion of the operation to the recovery of the related physiological indexes to the steady state as an impact influence period, and prohibiting other nursing tasks which have physical effects on the same region or the physiological related region from entering an execution sequence; a damping feedback assembly is arranged to be embedded in a nursing feedback channel of the digital twin, and when the change rate is higher than a preset physiological tolerance threshold, a rhythm slowing mechanism is started, and the execution rhythm of a subsequent task is adjusted; a manual operation behavior monitoring node is arranged in a nursing terminal or a sickbed area; and based on the data collected by the monitoring nodes, constructing an execution behavior deviation judgment unit which is compared with a system recommendation path, and identifying a hidden correction behavior for a system strategy.

Description

Technical Field

[0001] This invention relates to a method for optimizing perioperative nursing care plans, specifically a digital twin-driven dynamic optimization method for perioperative nursing care plans. Background Technology

[0002] Currently, there are intelligent auxiliary systems for perioperative risk assessment and clinical decision-making, such as the one disclosed in CN111009322A. Although these systems have achieved intelligent integration in terms of structured data collection, risk analysis, intervention plans, and quality management throughout the entire preoperative-intraoperative-postoperative cycle, and have to some extent solved the problems of fragmentation and heterogeneity of traditional assessment tools, they still have several obvious technical shortcomings and practical application drawbacks in addressing the dynamic optimization of perioperative nursing pathways, especially in individualized nursing interventions driven by digital twins. First, these systems tend to focus on risk identification and static decision support, concentrating on preoperative data integration and postoperative follow-up quality control. However, they lack refined modeling of nursing behavior response mechanisms during the critical stages from intraoperative to postoperative recovery. Although the decision support models they construct have structured plan recommendation functions, they lack the ability to control the pace of nursing tasks, adjust the execution sequence, and dynamically analyze task dependencies. They cannot adjust, insert, or suppress tasks based on real-time fluctuations in the patient's postoperative condition. This design, lacking a closed-loop feedback mechanism between nursing care and physiological responses, is prone to problems such as excessively dense nursing schedules, overlapping tasks, or patients exceeding their workload thresholds during high-frequency clinical nursing procedures. This can lead to a mismatch between nursing care and patient tolerance, increasing the risk of postoperative complications. Furthermore, existing systems tend to focus on risk prediction rather than task execution management. Even when some nursing suggestions are pushed, they are mainly driven by static rule bases. Their task recommendations lack process awareness and cannot automatically buffer or break down the next task based on the physiological response generated by the previous task. This is precisely the dynamic scheduling capability that is most critical for postoperative nursing optimization.

[0003] Secondly, such systems typically lack a deep understanding mechanism for deviations in nurses' actual behavior. In real-world scenarios, nurses often fine-tune recommended pathways based on experience, such as performing certain palliative care earlier, postponing strenuous procedures, or skipping tasks that are temporarily unsuitable. However, existing systems cannot automatically identify these implicit corrective behaviors, nor can they distinguish whether their source is human arbitrariness or driven by the patient's state. They simply treat deviations as abnormal or erroneous pathways, thus failing to transform potential clinical experience into learning factors for pathway optimization. Essentially, such systems remain at the level of technical implementation of nursing recommendations, lacking a mechanism for semantic feedback of nursing behavior inputs and their use in model iteration, severely restricting the system's adaptability and evolutionary capabilities. Furthermore, perioperative management systems often lack modeling methods for the physical relationships and physiological impact cycles between tasks. For example, when turning over causes changes in local wound tension, the system cannot perceive its impact cycle, nor does it restrict the entry of nursing tasks that operate on that area again within the impact period into the execution sequence. This lack of management capability regarding the impact period easily leads to the accumulation of physical loads.

[0004] Furthermore, this technical solution lacks a mapping mechanism between behavioral path deviation and group consensus. In group nursing operations, multiple nurses often exhibit consistent deviation trends in the same task under specific pathological scenarios. This phenomenon has high practical optimization value and should be actively learned by the system and used for task priority adjustment. However, existing systems lack mechanisms for deviating behavior classification, reproduction, and multi-user trajectory comparison, and there is no dynamic weight decay function or strategy update model to support it. This results in a rigid system path that is difficult to integrate with the collective wisdom of real-world operations. At the same time, although this solution involves the connection between smart terminals and cloud platforms, its data feedback chain is still mainly based on structured forms and event records. It lacks real-time perception nodes at the behavioral level, such as monitoring methods for micro-behaviors like nursing action paths, execution interruptions, withdrawals, and sequence misalignments. This results in the system lacking the ability to align behavior between task execution and recommended strategies. When dealing with complex nursing operations, it cannot form precise interventions or real-time optimizations, and is prone to falling into the dilemma of recommendations being out of touch with reality. Summary of the Invention

[0005] The purpose of this invention is to provide a method for dynamic optimization of perioperative nursing plans driven by digital twins, thereby addressing some of the drawbacks and shortcomings pointed out in the background art.

[0006] The technical solution adopted by the present invention to solve the above-mentioned technical problems is as follows: a dynamic optimization method for perioperative nursing plans driven by digital twins, including: setting an intervention sensitive area associated with the patient's current postoperative status, wherein the intervention sensitive area is defined according to the patient's surgical type and anatomical location, and has physical limiting conditions that limit the trigger threshold of nursing operations, wherein the physical limiting conditions include changes in body position angle, pressure contact area and duration of action; For each nursing operation, a reaction cycle in the target area is predefined. The time period from the completion of the operation to the recovery of relevant physiological indicators to a steady state is set as the impact period. During the impact period, other nursing tasks that have a physical effect on the same area or its physiologically related areas are prohibited from entering the execution sequence. A damping feedback component is embedded in the nursing feedback channel of the digital twin to receive physiological parameter changes caused by the execution of nursing tasks. When the rate of change of the signal between consecutive nursing tasks exceeds a preset physiological tolerance threshold, the damping feedback component activates a rhythm slowing mechanism to adjust the execution rhythm of subsequent tasks. A set of manual operation behavior monitoring nodes are set up in the nursing terminal or bed area to record the operation path, action interruption, withdrawal behavior and duration generated by the nurses during the execution process. Based on the data collected by the monitoring nodes, an execution behavior deviation judgment unit is constructed to compare with the system recommended path to identify the nurses' implicit correction behavior to the system strategy.

[0007] Furthermore, the damping feedback component comprises: A physiological change monitoring interface is used to collect patients' physiological change data in real time after the execution of nursing tasks. The data includes heart rate, blood pressure, muscle tone and blood oxygen saturation. The intensity of change analysis unit is connected to the physiological change monitoring interface and is used to identify the intensity of physiological response after the nursing task based on the physiological change data, and output the response fluctuation signal. A nursing task buffer control unit, connected to the change intensity analysis unit, is used to dynamically adjust the execution rhythm of subsequent nursing tasks based on the response fluctuation signal, wherein: When the fluctuation signal is less than the first threshold, maintain the original task rhythm; When the fluctuation signal exceeds the second threshold, a task buffer delay instruction is inserted. When the fluctuation signals of two consecutive tasks both exceed the second threshold, the rhythm suppression state is activated. An interactive linkage interface, connected to the buffer control unit, is used to transmit the task adjustment information to the nursing operation interface in real time and provide interactive options for accepting or rejecting delayed execution. The twin status update interface, connected to the digital twin master system, is used to synchronize the rhythm adjustment results of nursing tasks and the results of manual selection to the twin system in order to update the patient status and acceptable care threshold.

[0008] Furthermore, the physiological change monitoring interface includes a nursing behavior recognition subunit, which is used to determine the start time and execution mode of the nursing behavior and synchronously collect physiological change data to align the nursing behavior with the physiological response; the change intensity analysis unit includes a critical point recognition logic structure, which is used to dynamically adjust the threshold recognition standard when a nonlinear change pattern is detected.

[0009] Furthermore, the nursing task buffer control unit has a task hierarchical reconstruction mechanism, which is used to break down the original high-load nursing task into multiple sub-tasks and execute them in stages under rhythm suppression. The breakdown scheme is derived from the individual patient tolerance data in the twin. The interactive linkage interface is connected to the nursing staff identification module and is used to dynamically adjust the execution strategy of the task postponement suggestion given by the system according to the executor's level and experience level.

[0010] Furthermore, the first threshold in the nursing task buffer control unit is that within a preset time period after the execution of the nursing task, the rate of change of the physiological parameters does not exceed 1.5 times the average rate of change of the parameters under the preoperative stable state, which is used to determine that the patient is in a physiologically stable state and allows the original nursing task rhythm to be maintained.

[0011] Furthermore, the second threshold is defined as the rate of change of the physiological parameter exceeding 2.5 times the average rate of change of the parameter under the preoperative stable state within a preset time period after the execution of the nursing task. This threshold is used to determine that the patient is in a state of high physiological response load, triggering a delayed task execution or rhythm suppression mode. The physiological parameters include heart rate, systolic blood pressure, diastolic blood pressure, blood oxygen saturation, or muscle tone, and the rate of change is the magnitude of the numerical change per unit time divided by the length of time.

[0012] Furthermore, the method for identifying implicit modification behaviors of nursing staff to system strategies includes: The nursing path suggestions generated by the digital twin system are recorded as recommended paths. The paths include the operation type of the nursing task, the recommended execution time, the sequence and frequency information. The action information of the nursing staff when actually performing the nursing task is collected by the acquisition device set on the nursing terminal, the hospital bed or the nursing equipment, including the actual operation time, operation type, execution sequence, and records of task skipping or repeated execution. The actual execution path is compared with the recommended path to identify whether there is a continuous deviation behavior, including task advancement, delay, frequency change, substitution behavior, or sequence misalignment. After identifying the above deviation behavior, it is determined whether it constitutes a stable operation mode deviation. If the preset trend conditions are met, it is identified as an implicit correction behavior of the nursing staff to the system suggestions.

[0013] Furthermore, the step of determining whether the implicit modification behavior constitutes the aforementioned behavior includes: The identified deviation behaviors are categorized into rhythm correction, alternative actions, skipped operations, and sequence rearrangement. Identification logic and processing rules are set for different types of deviations, enabling the system to optimize paths based on the classification results. The current nursing staff's performance is compared with the operation trajectories of multiple users in similar situations in the past. When multiple users have similar deviation patterns for the same recommended task, the task is marked as a weak recommendation and its priority is reduced. The system matches the current nursing staff's deviation behavior under specific pathological conditions with the historical behavioral trajectories of multiple nursing staff in similar clinical situations. When a consistent deviation trend is identified among most users for the same recommended task, the system marks the task as a weak recommendation and dynamically reduces its priority weight in the path generation model. To achieve this function, the following nursing behavior deviation weight decay function is defined: in: For a specific recommended nursing task at the current time The dynamic priority weighting factor has a range of values. The smaller the value, the higher the priority should be lowered by the system. This is the system adjustment coefficient, used to control the sensitivity of the impact of offset behavior on the path strategy; it is a positive real number, the larger the value, the faster the decay. The number of offset behavior types associated with the current recommended task (maximum of 4 types: rhythm, substitution, skip, rearrangement). For the first The importance weight set for the type of offset behavior in the system reflects the strength of its impact on the overall path quality; For the first Class offset behavior in time The intensity of the detected behavioral deviations, such as the degree of frequency deviation and the magnitude of operation advance, is represented after normalization. The offset behavior amplification index controls the nonlinearity of the behavior intensity with respect to the attenuation result, and is usually a real number greater than 1; The behavioral consensus factor represents the number of times a historical user has acted in the same context. Class offset behavior overlap, value range A larger value indicates a more prevalent offset behavior; By integrating the intensity, importance, and user consensus of different types of deviation behaviors, the path priority weight that the nursing task should be assigned at the current time point is calculated in real time. When a certain behavior deviation has high intensity, high consensus, and high type influence weight, the nursing task... It will drop rapidly; Conversely, if the offset is only occasional, of low intensity, or lacks user consensus, the task still maintains a high recommendation strength; this function can be embedded into the task priority queue in the path generation module to achieve dynamic convergence of care strategies and adaptive recommendation updates.

[0014] Furthermore, the identification of the implicit modification behavior includes: The system combines the patient's current physiological state data with the execution of deviation behaviors for cross-judgment. When nursing staff continuously avoid or adjust nursing tasks under a certain pathological state, the deviation behavior is attributed to patient state-driven corrective behavior and given priority for learning. After identifying three consecutive similar deviation behaviors, the system sends an active inquiry prompt to the nursing staff through the nursing terminal, requesting an explanation of the correction reason and allowing the nursing staff to choose from experience optimization, patient intolerance, or device limitation options. The feedback will be used to label the deviation behavior attributes.

[0015] The beneficial effects of this invention are as follows: By incorporating multi-source physiological state data of patients before, during, and after surgery into a twin model for real-time mapping, the rhythm, sequence, and method of nursing tasks can be dynamically adjusted according to the individual patient's tolerance, recovery pace, and sudden reactions. This avoids a one-size-fits-all fixed nursing procedure and significantly improves the accuracy and safety of nursing interventions. By embedding a damping feedback component, the system can automatically determine whether the patient is under excessive workload based on the intensity of physiological fluctuations between consecutive tasks, triggering strategies such as slowing down the pace, delaying tasks, or breaking down tasks in real time, effectively reducing the risk of secondary injury caused by excessive nursing interventions.

[0016] The system collects real-time data on nurses' actual operational behaviors and compares them with the system's recommended paths. It identifies implicit corrective behaviors such as pace adjustments, task skipping, and sequence rearrangements. Especially when operational trends deviate, the system proactively attributes the causes and learns from group experience, thereby improving the clinical relevance and practicality of the recommended paths. By precisely aligning nursing behavior nodes with the intensity of patients' physiological responses and combining this with a nonlinear change recognition algorithm, the system can not only assess task execution effectiveness but also adjust task pace, execution intensity, and even nursing strategies based on the patient's real-time status, truly achieving a closed-loop optimization mechanism centered on patient response. Through the innovative introduction of a deviation behavior weight decay function, the system can adjust the priority of each nursing task in the path generation model in real-time based on behavior intensity, the impact of deviation type, and the degree of group consensus, enabling the nursing system to have self-evolution and self-learning capabilities. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the digital twin-driven perioperative nursing optimization process of this invention.

[0018] Figure 2 This is a structural diagram showing the functional relationship of the damping feedback component of the present invention.

[0019] Figure 3 This is a diagram showing the relationship between the implicit modification and dynamic optimization functions of the nursing pathway in this invention.

[0020] Figure 4 This is a flowchart illustrating the dynamic control of postoperative care rhythm supported by digital twins in Embodiment 1 of the present invention.

[0021] Figure 5 This is a flowchart of the closed-loop process for digital twin nursing pathway recommendation and implicit correction in Embodiment 2 of the present invention. Detailed Implementation

[0022] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0023] Combined with appendix Figure 1 This invention discloses a digital twin-driven method for dynamic optimization of perioperative nursing care. It establishes intervention-sensitive zones associated with the patient's current postoperative state. These zones are defined based on the type of surgery performed and the corresponding anatomical location information during the postoperative recovery stage. The system dynamically retrieves preoperative imaging information, intraoperative operation records, and postoperative risk distribution maps based on the surgical procedure classification and postoperative stage to generate an individualized three-dimensional sensitive zone mapping. These sensitive zones are displayed in the digital twin model with labelable and highly responsive attributes, used to constrain the scope and intensity of specific nursing actions. Each intervention-sensitive zone is assigned at least one set of physical constraints to define whether the nursing action can be permitted or recommended by the system in the current state. These physical constraints include at least the change in body position angle, pressure contact area, and action duration, with the change in body position angle used to limit... The upper limit of the amplitude of somatic adjustment operations in nursing care is used to prevent risks such as traction, compression, or poor drainage of the suture area after surgery due to excessive angle. The pressure contact area is used to control the contact range of a single nursing operation on the patient's skin, wound, or tender area of ​​the body surface, so as to avoid inducing or aggravating postoperative tissue damage caused by pressure ulcers. The duration of the action is used to determine whether the duration of a specific nursing behavior (such as turning over, physical drainage, wiping, assisting sitting up, etc.) exceeds the tissue tolerance threshold within a unit of time. This parameter is dynamically updated in combination with real-time monitoring of patient muscle tone changes, spontaneous movement responses, and heart rate fluctuations, so that the system can output the tolerance judgment result of the current nursing operation based on twin simulation. Under the premise of meeting the physical constraints, the system allows subsequent nursing tasks to enter the candidate execution sequence, thereby forming a perioperative nursing optimization control mechanism based on dynamic perception and response constraints of interventional sensitive areas.

[0024] For each nursing procedure, a predefined reaction cycle in the patient's target area is defined. The reaction cycle refers to the entire time period from the initiation, development, recovery, and steady state of the local or systemic physiological response triggered by the nursing action. This reaction cycle is parameterized and modeled by the system by integrating historical case data, postoperative recovery models, and digital twin simulation results to generate a basic time interval. The target area can be the location of the postoperative wound, functionally sensitive areas (such as the thoracic and abdominal cavities, lower extremity venous access, and peripheral nerve distribution areas), or specific sites with increased reactivity after multiple interventions. After the nursing task is completed, the system monitors the changes in relevant physiological indicators (including but not limited to heart rate, blood pressure, respiratory rate, skin temperature, muscle tone, drainage volume, and blood oxygen saturation) in real time. When the monitored indicators gradually return to the preoperative steady-state level or enter the set physiological fluctuation safety threshold range, the reaction cycle of the current nursing procedure can be determined. The initial phase has been completed, providing a basis for the system's next path recommendation. The time period between the completion of the nursing operation and the recovery of physiological indicators to a steady state is defined as the impact period. During this impact period, the system activates the mutual exclusion scheduling mechanism for nursing tasks, removing all nursing tasks from the candidate task pool that will produce physical effects on the same target area or its physiologically related areas. Physical effects include direct contact, indirect traction, regional compression, liquid or gas intervention, and irritating operations, to avoid excessive local tissue load, superimposed stress responses, or chaotic monitoring indicators caused by continuous operations. At the same time, the definition of physiologically related areas is based on anatomical path mapping and the system's physiological response path table to achieve logical linkage judgment. This ensures that even if the physical contact sites of nursing tasks are different, if they have a synchronous impact on the same tissue system, they will be restricted by the system to enter the current execution sequence, thereby constructing a nursing operation rhythm control structure with the impact period as the core.

[0025] A damped feedback component is embedded in the nursing feedback channel of the digital twin. This component receives signals of physiological parameter changes generated by the patient after each nursing task is completed. These physiological parameters include, but are not limited to, heart rate, blood pressure, body temperature, blood oxygen saturation, muscle tone, respiratory rate, and subjective rating feedback. The damped feedback component has a continuous response recognition function, generating time-series data of parameter changes between multiple nursing tasks. When the system detects that the physiological response rate between consecutive tasks exceeds a preset physiological tolerance threshold, it determines that the patient's current physiological regulatory capacity for continuous nursing behavior is insufficient, indicating excessive tissue load, rhythm disorder, or potential risk signals. At this point, the damped feedback component activates a rhythm slowing mechanism, dynamically adjusting the execution rhythm of subsequent nursing tasks to buffer the time rhythm of the nursing tasks. Specific methods include delaying the task start time, introducing a rest period between tasks, or recommending a low-stimulation alternative task, ensuring that the patient's physiological response to the previous task recovers to an acceptable range before proceeding to the next intervention. Meanwhile, to identify the nursing staff's intention to adjust the system's recommended path at the execution level, a set of manual operation behavior monitoring nodes are deployed in the nursing terminal or bed area. These monitoring nodes include a manual recording button, an execution path tracker, an action trigger sensor, and a timestamp recording device, used to record key behavioral parameters of the nursing staff during task execution, including operation path selection, number of interrupted actions, task withdrawal records, and the duration of each operation. Based on this monitoring data, the system constructs an execution behavior deviation judgment unit. By comparing the actual operation path with the system's recommended path, it identifies whether there are any non-occasional, repetitive, or trending implicit correction behaviors. Implicit correction behaviors refer to behaviors in which the nursing staff adjusts, avoids, changes the order, or performs alternative executions of the system's recommended strategy without explicit feedback channels. If such deviation behaviors are determined to occur continuously within multiple cycles, the system will mark the nursing task as a potential optimizable path node and incorporate this deviation feature into subsequent path generation for iterative optimization.

[0026] Combined with appendix Figure 2The damping feedback component is embedded in the nursing feedback channel of the digital twin and consists of multiple functional units. Its structure includes: a physiological change monitoring interface for real-time collection of the patient's physiological change data after the nursing task is completed. This data includes, but is not limited to, heart rate, blood pressure, muscle tone, and blood oxygen saturation. The data collection can be performed using bedside monitoring devices, wearable sensing devices, or embedded nursing sensing nodes. The physiological change monitoring interface synchronously transmits the collected data to a connected change intensity analysis unit. This unit performs interval analysis and trend identification on the physiological data, calculating the intensity of the patient's physiological response after the nursing task is applied. The intensity can be expressed as the deviation amplitude, slope of change, or degree of cross-parameter coordinated fluctuation of the target indicator within a unit of time, and the result is output as a response fluctuation signal. The response fluctuation signal is transmitted to a connected nursing task buffer control unit. The buffer control unit sets dynamic adjustment logic for the task execution rhythm based on the response intensity value, wherein: when the fluctuation signal is lower than a first threshold set by the system, the response fluctuation signal is interrupted. The system maintains the current task rhythm as long as the patient's response to nursing tasks remains stable. When the fluctuation signal exceeds a second threshold, the system inserts a task buffer delay instruction, setting a minimum waiting window for the next nursing operation in the task sequence. If two consecutive tasks cause the fluctuation signal to exceed the second threshold, the patient is determined to be in a hypersensitive state, the system triggers a rhythm suppression state, suspends all recommendations for high-load nursing tasks, and activates a priority recommendation strategy for restorative nursing suggestions. The above rhythm adjustment information is transmitted in real time to the nursing operation interface via an interactive linkage interface connected to the buffer control unit. The interface provides clear interactive options for accepting or rejecting delayed execution, and nursing staff can make manual selections based on clinical judgment. The system synchronously records the selection results and operation behaviors. At the same time, the damping feedback component also includes a twin state update interface, which is used to synchronously upload the rhythm adjustment results of each nursing task and the results of the nursing staff's manual interaction selections to the digital twin main system to update the physiological response model, operation tolerance threshold, and rhythm sensitivity parameters of the corresponding virtual patient.

[0027] The physiological change monitoring interface in the damped feedback component further includes a nursing behavior recognition subunit. This subunit, during the actual execution of nursing tasks, determines the start time and execution method of the nursing behavior based on input signals from the operating terminal, action triggering mechanisms, or behavior path tracking results. This includes, but is not limited to, whether the operation is a planned task, a substitute behavior, or whether there are skipped steps or reverse order phenomena. It also records the actual position of the behavior in the nursing sequence using a timestamp, thus maintaining strict time synchronization with the patient's physiological change data collection. This ensures a one-to-one logical mapping relationship between nursing behavior triggering events and physiological response data, achieving precise alignment between nursing behavior and physiological response, and avoiding data collection errors. Delays, operational drift, or unexpected interventions can cause errors in the interpretation of physiological signals, thus improving the system's ability to determine causal relationships. Simultaneously, the change intensity analysis unit also includes a critical point identification logic structure to monitor and determine the presence of nonlinear change patterns in physiological change data. This logic structure can identify nonlinear change characteristics such as abrupt responses, hysteresis responses, biphasic fluctuations, plateau delays, and amplified perturbations. When such abnormal fluctuation patterns are identified, the system triggers a dynamic threshold adjustment mechanism, adjusting the originally set physiological parameter judgment threshold upwards, downwards, or recalculating the window interval based on the response curve trend, thereby avoiding false alarms, missed judgments, or erroneous delays during the nonlinear physiological response phase.

[0028] The nursing task buffer control unit possesses a task hierarchical reconstruction mechanism. When the system identifies a patient in a state of rhythm suppression, it structurally decomposes originally planned high-load, highly continuous, or physiologically impactful nursing tasks into multiple sub-tasks with lower operational intensity, shorter execution time, or less physiological burden. These sub-tasks are then re-embedded into the nursing task sequence in a progressive manner to reduce the overall intervention pressure on the patient during periods of high sensitivity. The decomposition scheme is derived from the patient's historical physiological tolerance data, postoperative response stability score, and nursing behavior response model within the digital twin. Before each task decomposition, the system calls upon the corresponding virtual patient model within the twin system to evaluate the simulated response of different sub-tasks, thereby determining whether the decomposed task combination is feasible and physiologically safe in the current state. This ensures that the hierarchically reconstructed task structure both meets nursing objectives and provides a buffer function for intervention rhythm. At the same time, the interactive linkage interface is connected to the nursing staff identification module. After the task postponement, dismantling, or buffering rhythm suggestions are generated, the system dynamically adjusts the suggested strategies based on dimensions such as the identity, professional title, clinical experience level, and execution authority level of the personnel currently performing the nursing task. When the system identifies the operation as being performed by a high-level or senior nursing staff, it allows for relaxation of the mandatory execution conditions of some delay instructions or provides more autonomous path adjustment options to ensure the flexibility of the human-machine collaborative decision-making mechanism. Without disrupting the system's rhythm control logic, the system respects the initiative of clinical experience in optimizing the nursing pathway. At the same time, the system incorporates the nursing staff's acceptance or modification of the task rhythm suggestions into the twin model feedback loop, which is further used to improve the behavior deviation identification mechanism and experience knowledge learning module. This enables multi-dimensional dynamic collaborative control of the perioperative nursing pathway among the twin system, physiological state, and manual execution.

[0029] In the nursing task buffer control unit, the first threshold is used to determine whether the patient is in an acceptable physiological stable state after the nursing task is performed, thereby deciding whether to adjust the pace of subsequent nursing tasks. Specifically, the first threshold is defined as follows: within a preset time period after the completion of the nursing task, the rate of change of one or more monitored physiological parameters does not exceed 1.5 times the average rate of change of the corresponding parameters in the patient's preoperative stable state. That is, a baseline steady-state model is formed by sampling physiological parameters in several preoperative resting states, and the average rate of change is extracted as a baseline indicator. Then, after the nursing task is performed, a response window is set to sample and monitor the rate of change of core parameters such as heart rate, blood pressure, muscle tone, and blood oxygen saturation in real time. If the obtained rate value is consistently maintained within 1.5 times the steady-state baseline value, the system determines that the patient is currently in a physiologically stable state, indicating that the patient's response to the current nursing task is within a controllable range and there is no significant risk of overload, stress activation, or regulatory imbalance. Therefore, the nursing task buffer control unit does not adjust the execution rhythm of subsequent tasks, allowing the nursing path to continue according to the originally set time rhythm, in order to ensure the implementation efficiency of the continuous nursing process. At the same time, this threshold strategy, through the combination of parametric modeling and dynamic window judgment, has good individual adaptability and cross-cycle scalability, and can dynamically adapt to and achieve safe rhythm control in various nursing scenarios and patient states.

[0030] The nursing task buffer control unit is equipped with a second threshold for rhythm regulation determination. This second threshold is used to identify whether the patient's short-term physiological response after the nursing task is in a high-load state. Specifically, it is defined as follows: within a preset time period after the completion of the nursing task, the monitored core physiological parameters are continuously sampled and their rates are calculated. If the obtained rate of change exceeds 2.5 times the average rate of change of the corresponding parameter in the patient's preoperative stable state, the system determines that the nursing operation has triggered a physiological shock or stress response beyond the normal tolerance range, triggering a task rhythm intervention mechanism, including generating a task delay execution instruction or activating a rhythm suppression mode. The physiological parameters include, but are not limited to, heart rate, systolic blood pressure, diastolic blood pressure, blood oxygen saturation, and muscle tone. The sampling method can be based on continuous monitoring. The system acquires data via backup or periodic interfaces. The rate of change is the magnitude of the current parameter's change in value per unit time divided by the time length, reflecting the immediate intensity of the parameter change. The system compares this immediate rate of change with the steady-state mean of the patient's multiple resting states before surgery to determine whether the relative deviation exceeds the aforementioned second threshold. Based on this, the system decides whether to postpone, buffer, or completely suspend the next nursing task to prevent irreversible physiological damage to the patient caused by the superposition of multiple tasks. At the same time, this threshold judgment mechanism, combined with the patient's individual physiological baseline data and real-time feedback model in the digital twin system, has dynamic adaptability and individual difference recognition capabilities, ensuring that the rhythm control is both responsive and clinically safe and redundant, thereby improving the stability and self-regulation of the perioperative nursing process.

[0031] Combined with appendix Figure 3This method identifies implicit modifications made by nursing staff to the system's recommended strategies. The specific steps are as follows: Individualized nursing pathway suggestions generated by the digital twin system are recorded as recommended paths. These recommended paths include nursing tasks generated for a specific patient's current postoperative state. Each task includes key scheduling parameters such as operation type, recommended execution time or time period, execution order, and repetition frequency, forming a standard nursing task sequence within the system. Subsequently, data acquisition devices deployed on nursing terminals, bedside areas, or intelligent nursing equipment are used to collect real-time data on the nursing staff's operational behavior during task execution. This operational data includes, but is not limited to, actual operation time, operation type, task order, task skipping, number of repetitions, operation duration, and whether there was any manual interruption. This information is packaged and mapped into the same data structure as the recommended path. The system then compares the actual execution path with the recommended path. The system compares each path item by item to identify persistent deviation behaviors. These deviation behaviors include differences in operation such as advancing or delaying task execution, changing execution frequency, replacing the original task with a non-recommended task, disrupting the recommended order, or skipping key steps. The judgment criteria are based on the temporal deviation between the behavior and the path, inconsistencies in execution strategies, and recurring behavioral correction characteristics. When the above deviation behaviors occur more than the minimum number of trend recognitions set by the system in a continuous nursing cycle, or meet the behavioral deviation amplitude threshold and time persistence standard, the system determines that the behavior constitutes a stable operational mode deviation and identifies it as an implicit correction behavior of the nursing staff to the system strategy. That is, the recommended path is not modified through explicit feedback or operation interface, but the tendency to adjust the path is expressed in the actual operation through behavior. The identification result will be pushed to the feedback module of the digital twin system for behavioral factor modeling and strategy adaptive correction in the subsequent path re-optimization process.

[0032] In determining whether a nurse's actions constitute implicit corrections to the system's recommended path, the system categorizes identified deviations into several types: rhythm corrections (adjusting the timing, intervals, or timing of nursing tasks); substitution actions (replacing system-suggested tasks with non-recommended tasks); skipped operations (directly skipping some system-defined steps); and sequence rearrangements (adjusting the execution order or structural arrangement of tasks within the path). Differentiated identification logic and processing rules are set for different categories of deviations, enabling the system to call appropriate path optimization strategy templates for response adjustments based on the behavior classification results. Subsequently, the system records the nurse's actual actions during the execution process. By comparing the behavioral trajectories of multiple nurses in similar clinical situations with historical data, including matching conditions such as patient type, surgical stage, physiological state, and operation frequency, the system identifies a convergent bias pattern among multiple users for the same recommended task. This is particularly evident when the same type of correction trend repeatedly appears under similar pathological conditions. The system then marks this nursing task as a weak recommendation and dynamically lowers its priority in the path generation model. Furthermore, a nursing behavior bias weight decay function is proposed to integrate the intensity, importance, and user consensus of different types of bias behaviors, thereby deriving the priority adjustment result of the recommended nursing task at the current time point. This function is defined as follows: in, For a specific recommended nursing task at the current time The dynamic priority weighting factor has a value range of [0,1]. The smaller the value, the more the task should be downgraded by the system. This is the system adjustment coefficient, used to control the sensitivity of the influence of offset behavior on the path strategy. It is a positive real number, and the larger the value, the higher the weight of the offset influence. The number of offset behavior types associated with this task, up to four categories: rhythm correction, alternative action, skip operation, and sequence rearrangement; For the first The importance weights of the offset behaviors set in the system are used to reflect the strength of their impact on path integrity and care safety; For the first Class offset behavior in time The offset intensity within the range represents the degree of deviation between the actual frequency, magnitude, or manner of behavior execution and the system's recommended value, and is then normalized. The offset behavior is amplified by an exponent, which controls the degree of nonlinear influence of high-intensity offsets on the final priority weights. It is usually set to a real number greater than 1. The behavioral consensus factor represents the historical consensus of multiple users in this nursing scenario regarding the first... The overlap of class-based offset behaviors ranges from [0,1]. A larger value indicates a greater degree of collective convergence. When a certain offset behavior has high offset strength, high importance weight, and significant user consensus, it will enhance the task's... If the value drops rapidly, the system identifies it as a path node with poor policy adaptability and prioritizes its adjustment. Conversely, if the offset behavior is only occasional, of low intensity, and lacks user consistency, then... Maintaining a high level and preserving its recommendation priority, the function can be embedded into the task priority queue of the path generation module as a basis for dynamic sorting weight, enabling the nursing path to have real-time adaptive recommendation update capability during execution, and realizing a closed-loop collaborative adjustment mechanism between nursing behavior, user practice and twin strategy.

[0033] The identification of implicit corrective behaviors by nursing staff is not only based on the comparison between the executed path and the system-recommended path, but also further combines the patient's current physiological state data with the characteristics of the deviation behavior for cross-judgment. Specifically, this includes synchronously collecting the patient's core physiological parameters (such as heart rate, blood pressure, respiratory rate, muscle tone, blood oxygen saturation, etc.) after the execution of the nursing task, and performing time correspondence analysis with the identified deviation behavior events. When the system continuously detects that nursing staff repeatedly perform deviation operations such as avoidance, substitution, delay, or sequence adjustment of similar nursing tasks under a certain pathological state, and the fluctuations of the patient's physiological state parameters at that time are similar, the system attributes the deviation behavior to patient state-driven corrective behavior. This is different from path adjustments caused solely by the operator's experience preferences. Driven behavior will be preferentially included in the learning sample library of the path generation model for iterative optimization of the system's path recommendation mechanism under similar physiological states. In addition, to enhance the system's understanding of the semantics of behavioral deviation, when the system continuously... When a stable deviation trend in the caregiver's performance on the same type of task is identified within three consecutive nursing cycles, the nursing terminal will proactively prompt the caregiver with a brief description of the deviation behavior and its system identification logic. The prompt will then ask the caregiver to select the reason behind the deviation behavior. The options include experience optimization (indicating that the operation is more reasonable based on clinical experience), patient intolerance (indicating that the patient cannot accept the original recommended task physiologically or psychologically), and device limitations (indicating that the task cannot be completed due to current conditions). The caregiver can complete the feedback operation by selecting via touch or confirming via voice. The system will bind this human feedback information with the behavior deviation event, label it as the attribute factor of the deviation behavior, and simultaneously transmit it to the path correction module of the digital twin main system. This is used for accurate attribution, behavior classification, and multi-source fusion modeling of subsequent recommendation strategies, thereby enhancing the system's ability to adapt to paths under the influence of multiple variables in complex nursing scenarios and realizing deep closed-loop learning and dynamic knowledge accumulation of the digital twin model in actual operation.

[0034] Example 1: Combined with appendix Figure 4In this embodiment, a male patient underwent laparoscopic cholecystectomy at 8:30 AM on the same day. Post-operatively, he was transferred to a smart nursing ward supported by a digital twin system. The system established an individual twin model of the patient and activated the perioperative nursing rhythm control mechanism. Three hours post-surgery, at 11:30 AM, the system scheduled the first bedside assisted turning task. Simultaneously with the task's commencement, the physiological change monitoring interface was activated, collecting real-time data on the patient's heart rate, systolic blood pressure, diastolic blood pressure, muscle tone, and blood oxygen saturation. Five minutes before turning, the physiological baseline was stable, with a heart rate of 82 beats / min, systolic blood pressure of 125 mmHg, and diastolic blood pressure of 78 mmHg. mHg, blood oxygen saturation 98%, muscle tone score 2 (based on the revised Ashworth scale); after the turning and care procedure began, within a 7-minute task window, the system recorded a rise in heart rate to 98 beats / min, a rise in systolic blood pressure to 138 mmHg, an increase in muscle tone to grade 3, and a drop in blood oxygen saturation to 95%. The change intensity analysis unit compared each indicator with the preoperative steady-state average rate of change, concluding that the comprehensive fluctuation signal value caused by this care task was 2.1 times the baseline rate, falling between the first threshold (1.5 times) and the second threshold (2.5 times). The buffer control unit determined that the original care rhythm could be maintained. The system entered observation mode; at 12:10, the system assigned the second nursing task—assisting the patient to sit up for respiratory function exercises. This task was challenging for the patient's abdominal pressure control, balance, and respiratory system load. After the nursing task began, monitoring data showed that the heart rate rapidly increased to 112 beats / min, systolic blood pressure rose to 151 mmHg, diastolic blood pressure was 91 mmHg, muscle tone reached grade 3 (mid-to-high), and blood oxygen dropped to 93%. The fluctuation signal value output by the intensity change analysis unit was 2.7 times, exceeding the second threshold. The buffer control unit immediately triggered the insertion task buffer delay instruction, postponing the next task involving abdominal load (such as...). (Washing the abdominal binder and changing the drainage bag), and pushing a pop-up window to the nursing terminal through the interactive interface, prompting the executor that the system suggests delaying the execution of the next task, that the patient is currently in a hyperresponsive period, and whether to agree to postpone for 10 minutes. The nursing staff clicks to accept, and the operation record is simultaneously uploaded to the twin status update interface; when the third nursing task—bedside sputum clearance and back percussion—was executed at 13:20, the patient had recovered to a relatively stable state. During the execution of the task, the heart rate once again spiked to 116 beats / min, the blood pressure rose to 157 / 93 mmHg, the muscle tone increased to grade 4, the blood oxygen dropped to 91%, and the fluctuation signal value rose to 2.When a patient's physiological workload exceeds the second threshold twice consecutively (eight times the normal range), the system determines that the patient has entered a state of high physiological load. It immediately activates a rhythm suppression mode, suspending all tasks requiring physical contact with the upper abdomen or back. The system then pushes low-load alternative tasks, such as environmental adjustments, static psychological reassurance, and music intervention, into the system's nursing pathway. Simultaneously, the continuous high-response record and the nursing staff's interactive selection results are synchronized to the digital twin main system. The twin updates the physiological tolerance adjustment status label for the patient 4–6 hours post-surgery, adjusting the nursing thresholds and rhythm recommendation standards for future pathway generation. This creates a closed-loop response mechanism from physiological monitoring, behavioral response, task rhythm regulation to model status updates, effectively reducing the risk of complications caused by continuous nursing care.

[0035] After recognizing that the patient's physiological parameter fluctuations exceeded the second threshold after two consecutive nursing tasks, the system determined that the patient had entered a high-load physiological state and officially activated the rhythm suppression mechanism. At this time, the nursing task buffer control unit activated the built-in task classification and reconstruction mechanism to evaluate the comprehensive nursing operation originally planned to be performed at the 5th hour postoperatively, which included bedside full-process positioning assistance, abdominal compression fixation bandage replacement, and cleaning of the drainage site. This task was classified as a high-load task in the system's labeling, as it involved multiple operation sites, had a long duration of action, and involved direct contact with the abdomen. The system called on the patient's preoperative tolerance assessment and actual monitoring data within 3 hours postoperatively from the digital twin model to form an individual tolerance curve, and based on this, the task was broken down into three sub-tasks: mild adjustment of the bed foot-head lateral position (expected load level L1), wiping around the drainage site and changing dressings (expected load level L). 2) And the fixation strap inspection and tightness adjustment (expected load level L1-L2), and time planning based on the individual physiological recovery window, setting the first sub-task to be executed 6 hours later, the second sub-task to be inserted at an opportune time with a buffer recovery period, and the third task to be retained or postponed by manual decision based on the physiological response after the first two executions. The task breakdown information is transmitted to the nursing terminal through the interactive linkage interface. The Class A nursing staff (experience level E4) currently responsible for the nursing task is identified through the identity recognition module. The system defaults to them having intermediate and advanced intervention permissions. Therefore, all three options in the pop-up window are open: accept the breakdown suggestion, merge execution, and request senior nursing staff to take over. The Class A nursing staff chooses to accept the breakdown and execute it in stages. The system records their behavior path and writes the feedback into the twin status update interface, forming the rhythm optimization node of the 6th hour after the patient's operation.

[0036] Meanwhile, the nursing behavior recognition subunit is further activated during the task execution process, performing real-time behavior confirmation at the start point of each subtask. The system confirms the behavior start point based on signals such as the nurse clicking the task start button, the activation of the hand trigger sensor, and changes in motion tracking image nodes, forming a one-to-one correspondence between operation and physiological response. The system collects real-time data on the patient's heart rate rising from 88 to 91 beats / min, systolic blood pressure rising from 124 to 129 mmHg, muscle tone maintained at grade 2, and blood oxygen saturation maintained at 97% within 5 minutes of the first subtask execution. The fluctuation rate is calculated as (91-88) / 5 = 0.6 beats / min, far below the 1.5 times critical value (0.9 beats / min) of the average rate under preoperative stable conditions, i.e., the first threshold. The system determines that the patient is in a physiologically stable state and continues to execute subsequent low-load tasks while maintaining the current rhythm. When executing the second subtask, the system monitors a short heart rate... Within a short period, the heart rate rose to 99 beats per minute, systolic blood pressure rose to 137 mmHg, muscle tone increased slightly, and blood oxygen dropped to 95%. The fluctuation rate was (99-88) / 5 = 2.2 beats per minute. The critical point recognition logic structure identified an asymmetrical increase in heart rate and blood pressure, exhibiting a slightly nonlinear jump curve. The system adjusted its judgment mechanism, raising the original second threshold from 2.5 times to 2.7 times to avoid misjudgment. This was determined to be an acceptable mild stress state, and the rhythm suppression mechanism was suspended, but the task decomposition path was retained. Finally, the nursing staff completed all sub-tasks 7 hours after surgery according to the nursing terminal prompts, and noted on the completion confirmation interface that the patient accepted the phased nursing care well and had no rejection or subjective discomfort feedback. This information was then synchronously written into the twin status record as experience feedback of the nursing path, and used to update the intervention beat tolerance parameters in the postoperative recovery model of this type of patient.

[0037] From a data perspective, this embodiment demonstrates the real-time linkage capability of multiple modules. For example, task decomposition forms differentiated paths based on specific physiological tolerance values, and the calculation of the rate of change is based on (Δphysiological value) / Δt. For example, the heart rate change rate is 2.2 beats / minute, and the blood pressure rise rate is (137-124) / 5=2.6mmHg / minute. Under the condition that neither exceeds the second threshold after dynamic adjustment (approximately 2.7 times the average rate), the system avoids the accidental initiation of rhythm suppression, thereby improving the sensitivity and safety redundancy of the system's rhythm regulation. At the same time, the nursing behavior recognition subunit establishes a precise mapping chain between nursing operations and physiological fluctuations, avoiding mis-associations and data drift, and ensuring the traceability of the judgment logic.

[0038] Example 2: Combined with appendix Figure 5Based on Example 1, at 08:00 on the first day after surgery, the system generated a recommended path based on the patient's surgical procedure, intraoperative blood loss, postoperative anesthesia recovery performance, and the standard nursing recovery template in the twin model. This path included four operations: 08:30 first assisted turning, 09:30 first disinfection of drainage tube outlet and dressing change, 10:30 encouraging sitting up for abdominal breathing exercises, and 11:00 guiding the patient to try drinking water. The system recorded this path as a recommended path and uploaded it to the ward nursing terminal. Each nursing task in the path included information such as operation type (e.g., assisted turning), recommended execution time, operation sequence, and suggested frequency (e.g., breathing training recommended 3 times a day), and embedded a task ID for subsequent comparison.

[0039] Nursing staff member B was on duty that day. Her behavioral data was collected by the operation recording subsystem connected to the nursing terminal and the motion tracking module on the bedside equipment. The system recorded that she actually completed the position assistance task ahead of schedule at 08:20 and skipped the dressing change operation at 09:30 because there was no obvious leakage from the drainage tube. She then performed the second task (drainage dressing change) at 10:00, while the breathing training task was postponed to 11:10 and the water drinking test was also postponed to 11:30. The system compared the recommended path with the actual execution path and identified multiple deviation behaviors, such as the task execution time being advanced (08:20 instead of 08:30), the task sequence being misaligned (drainage tube treatment before breathing training), the task being skipped and then made up (the task at 09:30 was skipped and then made up at 10:00), and the overall delay (the last two tasks were postponed). After detecting the continuous sequence deviation of the three tasks, the system further determined whether it constituted a stable operation mode deviation.

[0040] To determine whether the behavior constitutes a latent correction to stability, the system retrieves nursing pathway data from patients undergoing the same procedure within the past 7 days in the same ward and compares the operational behavior trajectories of 5 different nurses at the same 6-hour postoperative stage. It was found that 4 of the nurses moderately delayed the drainage tube treatment task, advanced the breathing training task, and frequently advanced the positional assistance task. Based on this, the system inferred that this type of deviation was not an isolated event but rather an experience-based correction logic formed by the operators in long-term practice. It was determined to meet the preset trend condition of convergent deviation among multiple users at similar task nodes, and thus identified the nurses' behavior as a latent correction to the system strategy. This judgment will be marked in the behavior labeling system of the nursing pathway strategy model and trigger the path adaptation module in the twin model to perform strategy correction for low-priority tasks. Furthermore, in the system statistics, the recommended time for the breathing training task was 10:30, but the average execution time was concentrated between 09:50 and 10:10, with a deviation frequency of 78%. Moreover, the deviation in execution order was a common behavior among most users. Therefore, the system adjusted the recommended execution time window for this task from the original 10:00–10:30 to 09:30–10:00. At the same time, the mandatory reminder mechanism for the drainage treatment task in the case of no postoperative exudation was turned off, so that the path recommendation is more in line with the actual execution logic of nursing staff.

[0041] Based on the data in this case, the system collected the task sequence performed by Nurse B as T1 - turning over (08:20), T2 - dressing change (10:00), T3 - breathing training (11:10), and T4 - water intake attempt (11:30). The time errors compared to the recommended path T1 (08:30) → T2 (09:30) → T3 (10:30) → T4 (11:00) were ±40 minutes, ±30 minutes, and ±40 minutes, respectively, with deviation rates reaching 33%~44%. The system was set to... The time tolerance threshold for stability deviation identification is ±15 minutes, and the task sequence misalignment threshold is Level 1 misalignment (e.g., the task sequence changes to T1→T3→T2→T4). Therefore, this combination meets the constituent elements of the deviation trend and is marked by the system as a stability implicit correction behavior event. This judgment is used to drive the twin model to directly advance the postural assistance time when generating this type of patient path next time, set the drainage-related tasks as conditionally triggered nursing tasks, and adjust the recommended rhythm parameters of the breathing training task.

[0042] The system further enters the stage of implicit correction behavior classification and priority decay calculation. In actual execution, nurse B delayed the drainage dressing change task from the recommended 09:30 to 10:00 and advanced the breathing training task to 09:50. The system initially identified that this behavior had deviation patterns such as rhythm correction (advance, delay), sequence rearrangement (change in operation order), and task skipping (dressing task was originally skipped and then made up). Subsequently, all nurses' deviation behaviors were classified into the following four categories: rhythm correction (e.g., advanced / delayed ≥15 minutes), substitute action (e.g., using other operations to replace the recommended task), skipped operation (task not executed and no system confirmation), and sequence rearrangement (execution order is different from the recommended logical chain). The system sets corresponding recognition logic for each type of deviation: for example, rhythm deviation is judged by a time error exceeding ±15 minutes, and sequence misalignment is judged by the difference of the longest common subsequence between the actual sequence and the recommended sequence >1.

[0043] The offset data was compared with the historical operation trajectory database within the digital twin system. The system retrieved the nursing trajectories of 162 patients who underwent the same procedure in the same hospital area and were in the 6-hour postoperative stage within the past 90 days. It was found that 128 of these patients (approximately 79%) exhibited the same task rhythm adjustment behavior in their nursing pathways. Among these, 71% had the rhythm advanced or delayed by ±20 minutes, 63% had the sequence misaligned by more than one level, and 58% had the drainage task skipped and then made up. Based on this highly consistent data, the system marked the drainage dressing change task on the morning of the first postoperative day as a weak recommendation and began the pathway priority decay calculation stage.

[0044] At this point, the nursing behavior offset weight decay function is invoked: Calculate using the following example parameters: Three types of deviation behaviors were detected: rhythm correction ( Skip operation () ), order rearrangement ( ); , , In the system evaluation of path integrity, rhythm correction has the greatest impact, followed by skipping, and rearrangement has a relatively minor impact. , , : These represent the offset intensity, which has been normalized (e.g., a 30-minute advance in tempo is an offset intensity of 0.75). , , The user consensus factor is derived from historical data, representing the degree of convergence among groups exhibiting divergent behaviors. The system adjustment coefficient controls the sensitivity of the offset response. A measured range of 2.0 to 6.0 is recommended, with 4.0 indicating medium sensitivity. The behavior amplification index enhances the nonlinear effect of high-intensity offsets in the overall priority calculation. A value of 1.5 to 2.5 is recommended, with 1.8 being a more robust value. Substitute into the calculation: Calculate sequentially: ; ; ; Multiply by the consensus factor: ; ; ; sum: ; Substitute again: Final result This indicates that the current recommendation priority weight of this task has dropped to about 67% of its original value. Although it has not been completely removed by the system, it has been significantly downgraded in the task scheduling priority ranking. In the subsequent path generation process, this task will be in a non-mandatory recommendation state, and will only be prompted when the nursing staff has spare time or the patient's condition permits. If the same offset trend is observed and more high consensus operations occur, this priority value will decrease further.

[0045] From the first to the second postoperative morning, the system recommended a combination of breathing training, abdominal binder examination, and water stimulation four times, with each session spaced six hours apart. In three of these sessions, nurse B delayed or disconnected the breathing training component, placing it at the end of the combined task or even skipping a task altogether. System monitoring data showed that all three instances occurred within the time window when the patient had mild abdominal distension, systolic blood pressure >140 mmHg, and muscle tone score >3. For example, the recommended task at 21:00 on the first postoperative day was to first perform breathing training (recommended time 21:00–21:15), followed by abdominal binder examination (21:15–21:30). The nursing staff actually postponed the breathing training to 21:45 and completed the remaining procedures between 21:00 and 21:15. The physiological data collected by the system showed that at 21:00, the patient's heart rate was 98 beats / min, systolic blood pressure was 146 mmHg, and muscle tone score was 3.5 (significant subjective resistance). By 21:45, the heart rate had decreased to 135 mmHg, heart rate to 91 beats / min, and muscle tone to 2.5, which was acceptable. No decrease in blood oxygen or significant discomfort occurred after the task was performed.

[0046] Subsequently, at 03:00 and 09:00 on the second postoperative day, similar pathways were again replaced by behavioral structures that performed other tasks earlier and delayed respiratory training. Each delay node corresponded to the patient's hypertension or elevated muscle tone. Based on the comparison between the collected physiological parameters and the execution pathway, the system determined that this deviation behavior had highly consistent pathological condition triggering characteristics and initiated a cross-judgment mechanism to bind and tag the deviation behavior with the corresponding physiological state. The system's triggering rule was set as follows: if the same deviation behavior occurs more than three times consecutively under the same pathological state (such as blood pressure > 140 mmHg or muscle tone score ≥ 3), it is initially classified as a state-driven correction behavior, and the system triggers a prompt pop-up window on the nursing terminal.

[0047] At this point, the system pops up a prompt on the nursing terminal: The system has detected that you have delayed the execution of the breathing training task three times in a row while the patient's blood pressure was high. Please select the reason to optimize the system's recommended logic, and provide three options: ① Experience optimization (based on your judgment that it is safer to stabilize blood pressure before training), ② Patient intolerance (such as the patient actively resisting cooperation), ③ Equipment limitations (such as the inability to activate the assistive cuff device immediately). Nursing staff B selected ② Patient intolerance and noted that the patient showed obvious resistance to breath-holding and actively requested a delay each time under high pressure. The system recorded this feedback and labeled it as the subjective attribute of the deviation behavior, and uploaded it to the behavior attribution module of the twin master system along with the three monitored behavior-state combinations.

[0048] The system thus forms three data loops: First, a behavior recognition logic chain, from task path deviation → behavior type classification → trend confirmation; second, a state correspondence mechanism, which dynamically captures physiological parameters and establishes a causal mapping with the deviation behavior; and third, nursing staff interaction feedback and label attribution, providing semantic input for model learning. Ultimately, this deviation behavior is identified by the system as a patient state-driven corrective behavior and is prioritized for inclusion in the strategy relearning of the digital twin model. In future path recommendations, if the system identifies patients with similar blood pressure / muscle tone states, it will postpone the recommended breathing training task or proactively provide an optional flexible execution window to avoid execution conflicts or patient discomfort caused by rigid sequencing, thereby improving the adaptability, safety, and naturalness of human-machine collaboration in the recommended path.

[0049] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A method for dynamic optimization of perioperative nursing plans driven by digital twins, characterized in that... include: An interventional sensitive area is set that is associated with the patient’s current postoperative status. The interventional sensitive area is defined according to the patient’s surgical type and anatomical location, and has physical limitations that limit the threshold that can be triggered by nursing operations. The physical limitations include changes in body position angle, pressure contact area and duration of action. For each nursing operation, a reaction cycle in the target area is predefined. The time period from the completion of the operation to the recovery of relevant physiological indicators to a steady state is set as the impact period. During the impact period, other nursing tasks that have a physical effect on the same area or its physiologically related areas are prohibited from entering the execution sequence. A damping feedback component is set up and embedded in the nursing feedback channel of the digital twin to receive physiological parameter change signals caused by the execution of nursing tasks. When the rate of change of the signal between consecutive nursing tasks exceeds a preset physiological tolerance threshold, the damping feedback component activates a rhythm slowing mechanism to adjust the execution rhythm of subsequent tasks. A set of manual operation behavior monitoring nodes are set up in the nursing terminal or bed area to record the operation path, action interruption, withdrawal behavior and duration generated by the nursing staff during the execution process. Based on the data collected by the monitoring nodes, an execution behavior deviation determination unit is constructed to compare with the system's recommended path, which is used to identify the implicit modification behavior of nursing staff to the system strategy.

2. The method for dynamic optimization of perioperative nursing plans driven by digital twins according to claim 1, characterized in that... The damping feedback component comprises: A physiological change monitoring interface is used to collect patients' physiological change data in real time after the execution of nursing tasks. The data includes heart rate, blood pressure, muscle tone and blood oxygen saturation. The intensity of change analysis unit is connected to the physiological change monitoring interface and is used to identify the intensity of physiological response after the nursing task based on the physiological change data, and output the response fluctuation signal. A nursing task buffer control unit, connected to the change intensity analysis unit, is used to dynamically adjust the execution rhythm of subsequent nursing tasks based on the response fluctuation signal, wherein: When the fluctuation signal is less than the first threshold, maintain the original task rhythm; When the fluctuation signal exceeds the second threshold, a task buffer delay instruction is inserted. When the fluctuation signals of two consecutive tasks both exceed the second threshold, the rhythm suppression state is activated. An interactive linkage interface, connected to the buffer control unit, is used to transmit the task adjustment information to the nursing operation interface in real time and provide interactive options for accepting or rejecting delayed execution. The twin status update interface, connected to the digital twin master system, is used to synchronize the rhythm adjustment results of nursing tasks and the results of manual selection to the twin system in order to update the patient status and acceptable care threshold.

3. The method for dynamic optimization of perioperative nursing plans driven by digital twins according to claim 2, characterized in that... The physiological change monitoring interface includes a nursing behavior recognition subunit, which is used to determine the start time and execution method of nursing behavior and synchronously collect physiological change data to align nursing behavior with physiological response. The change intensity analysis unit includes a critical point identification logic structure, which is used to dynamically adjust the threshold identification standard when a nonlinear change pattern is detected.

4. The method for dynamic optimization of perioperative nursing plans driven by digital twins according to claim 3, characterized in that... The nursing task buffer control unit has a task classification and reconstruction mechanism, which is used to break down the original high-load nursing task into multiple sub-tasks and execute them in stages under rhythm suppression. The breakdown scheme is derived from the individual patient tolerance data in twins. The interactive linkage interface is connected to the nursing staff identification module and is used to dynamically adjust the execution strategy of the task postponement suggestions given by the system according to the executor's level and experience level.

5. The method for dynamic optimization of perioperative nursing plans driven by digital twins according to claim 1, characterized in that... The first threshold in the nursing task buffer control unit is that, within a preset time period after the execution of the nursing task, the rate of change of the physiological parameters does not exceed 1.5 times the average rate of change of the parameters under the preoperative stable state, which is used to determine that the patient is in a physiologically stable state and allows the original nursing task rhythm to be maintained.

6. The method for dynamic optimization of perioperative nursing plans driven by digital twins according to claim 1, characterized in that... The second threshold is defined as the rate of change of the physiological parameter exceeding 2.5 times the average rate of change of the parameter in the preoperative stable state within a preset time period after the execution of the nursing task. This threshold is used to determine that the patient is in a state of high physiological response load, triggering a delayed task execution or rhythm suppression mode. The physiological parameters include heart rate, systolic blood pressure, diastolic blood pressure, blood oxygen saturation, or muscle tone. The rate of change is the magnitude of the numerical change per unit time divided by the length of time.

7. The method for dynamic optimization of perioperative nursing plans driven by digital twins according to claim 1, characterized in that... The methods for identifying implicit modification behaviors of nursing staff to system strategies include: The nursing path suggestions generated by the digital twin system are recorded as recommended paths. The paths include the operation type of the nursing task, the recommended execution time, the sequence and frequency information. The action information of the nursing staff when actually performing the nursing task is collected by the acquisition device set on the nursing terminal, the hospital bed or the nursing equipment, including the actual operation time, operation type, execution sequence, and records of task skipping or repeated execution. The actual execution path is compared with the recommended path to identify whether there is a continuous deviation behavior, including task advancement, delay, frequency change, substitution behavior, or sequence misalignment. After identifying the above deviation behavior, it is determined whether it constitutes a stable operation mode deviation. If the preset trend conditions are met, it is identified as an implicit correction behavior of the nursing staff to the system suggestions.

8. The method for dynamic optimization of perioperative nursing plans driven by digital twins according to claim 7, characterized in that... The steps for determining whether the implicit modification behavior constitutes the aforementioned behavior include: The identified deviation behaviors are categorized into rhythm correction, alternative actions, skipped operations, and sequence rearrangement. Identification logic and processing rules are set for different types of deviations, enabling the system to optimize paths based on the classification results. The current nursing staff's performance is compared with the operation trajectories of multiple users in similar situations in the past. When multiple users have similar deviation patterns for the same recommended task, the task is marked as a weak recommendation and its priority is reduced.

9. The method for dynamic optimization of perioperative nursing plans driven by digital twins according to claim 8, characterized in that... The identification of the implicit correction behavior includes: The system combines the patient's current physiological state data with the execution of deviation behaviors for cross-judgment. When nursing staff continuously avoid or adjust nursing tasks under a certain pathological state, the deviation behavior is attributed to patient state-driven corrective behavior and given priority for learning. After identifying three consecutive similar deviation behaviors, the system sends an active inquiry prompt to the nursing staff through the nursing terminal, requesting an explanation of the correction reason and allowing the nursing staff to choose from experience optimization, patient intolerance, or device limitation options. The feedback will be used to label the deviation behavior attributes.

Images