A personalized training system for postoperative respiratory muscle function rehabilitation

Abstract

The application discloses a postoperative respiratory muscle function rehabilitation personalized training system, which aims to solve the problems of lack of individual adaptability and inability to dynamically respond to function changes in traditional training schemes, and the core scheme comprises the following steps: a data acquisition module is integrated in a respiratory training carrier to obtain individual characteristic data and real-time respiratory physiological data; a characteristic analysis module completes clinical characteristic classification, generates a recovery rhythm curve and internally sets an abnormality early warning; a scheme generation module generates a three-stage personalized scheme of activation, promotion and consolidation in combination with an intelligent mapping library standard template; an adaptive updating module generates a special template based on a curve deviation state to incrementally update and optimize the intelligent mapping library; a communication module realizes data interaction with a medical and nursing terminal and a user mobile terminal, and an interaction execution module is responsible for training guidance, progress feedback and abnormality alarm, so that the rehabilitation process is dynamically adapted through closed-loop feedback, the effect and safety are taken into account, remote monitoring is supported, and the pertinence and efficiency of postoperative respiratory muscle rehabilitation are significantly improved.

Description

Technical Field

[0001] This invention relates to the field of medical rehabilitation technology, and in particular to a personalized training system for postoperative respiratory muscle function rehabilitation. Background Technology

[0002] Postoperative respiratory muscle dysfunction is a common postoperative complication in patients undergoing thoracic surgery, lung surgery, and some abdominal surgeries. It severely affects patients' pulmonary ventilation function, activity level, and postoperative recovery process. Therefore, postoperative respiratory muscle function rehabilitation training has become a crucial aspect of clinical rehabilitation treatment. Currently, although the number of respiratory muscle rehabilitation training devices and systems available in the clinical setting and on the market is gradually increasing, their technical solutions still have significant limitations. Existing training systems generally adopt preset static training programs, providing only fixed training parameters, such as fixed inspiratory volume target values ​​and training sets. They fail to fully consider the differences in individual clinical characteristics of patients, including surgical type, postoperative days, preoperative baseline pulmonary function values, and history of complications. These factors have a decisive impact on rehabilitation needs. In the data acquisition stage, many systems suffer from a lack of dimensionality, collecting only a limited number of respiratory parameters. Furthermore, the processing of raw data on individual characteristics and real-time respiratory physiological data is not standardized, lacking effective preprocessing steps such as filtering, outlier removal, and time axis synchronization calibration. This results in low accuracy and poor reliability of the collected respiratory physiological data, failing to truly reflect the dynamic changes in the patient's respiratory muscle function. Furthermore, the training protocol generation mechanism lacks dynamic linkage with the patient's real-time recovery status. Most systems generate one-time protocols based solely on the initial assessment, failing to dynamically adjust according to real-time changes in the recovery rhythm, such as the inspiratory volume increase slope and respiratory rhythm stability. Simultaneously, existing systems' training template libraries are mostly statically set, lacking an adaptive update mechanism. This makes it difficult to adapt to different patients' recovery paces and specific recovery types, resulting in low template matching and a disconnect between the training protocol and the patient's actual recovery rhythm. These deficiencies prevent existing postoperative respiratory muscle function rehabilitation training systems from achieving accurate clinical classification based on patients' multi-dimensional individual clinical characteristics. They also cannot generate dynamic recovery rhythm curves by combining standardized preprocessed real-time respiratory physiological data, thus failing to construct a dynamic matching mechanism for individualized training protocols. Additionally, they lack template libraries that can be iteratively optimized using patient recovery data. Ultimately, the training protocols cannot accurately adapt to individual patient differences and dynamic changes in respiratory muscle function, easily leading to problems such as excessive training intensity causing respiratory muscle fatigue or insufficient training intensity causing slow recovery progress. This severely impacts rehabilitation efficiency and safety, failing to meet the urgent clinical need for personalized and precise postoperative respiratory muscle rehabilitation training. Summary of the Invention

[0003] The purpose of this invention is to provide a personalized training system for postoperative respiratory muscle function rehabilitation, which enables accurate clinical classification based on individual patient characteristics and dynamic recovery rhythm curve generation based on real-time respiratory physiological data, thereby dynamically adjusting the personalized training program to improve rehabilitation efficiency and safety.

[0004] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a personalized training system for postoperative respiratory muscle function rehabilitation, comprising:

[0005] The data acquisition module is used to acquire users' individual characteristic data and real-time respiratory physiological data; The feature analysis module, connected to the data acquisition module, is used to determine the user's clinical feature category based on the individual feature data, and to generate a recovery rhythm curve characterizing the dynamic changes in respiratory muscle function based on the real-time respiratory physiological data. The intelligent mapping library contains pre-stored standard recovery rhythm templates and training parameter benchmarks corresponding to different clinical feature categories; The scheme generation module is connected to the feature analysis module and the intelligent mapping library respectively. It is used to match and analyze the recovery rhythm curve with the corresponding standard recovery rhythm template, and call or generate a personalized breathing training scheme based on the matching results. An interactive execution module, connected to the scheme generation module, is used to output and guide the user to execute the personalized breathing training scheme; An adaptive update module connects the feature analysis module and the intelligent mapping library to monitor the deviation of the recovery rhythm curve from the standard recovery rhythm template. When the deviation meets preset conditions, a user-specific rhythm template is generated and the intelligent mapping library is updated. The communication module is used to enable bidirectional data transmission between the system and medical terminals and user mobile terminals; A breathing training carrier is used to provide users with a physical interface for breathing training. The sensor components used to collect respiratory flow field signals in the data acquisition module are integrated into the breathing training carrier.

[0006] By adopting the above technical solutions, a complete closed-loop rehabilitation system integrating data acquisition, intelligent analysis, program generation and execution, and self-evolution was constructed. This system acquires individual user characteristics and respiratory physiological data in real time and generates dynamic recovery rhythm curves, thereby transforming abstract respiratory muscle function into an objective trajectory that can be quantified and analyzed. Utilizing a smart mapping library of pre-stored standard templates and parameter benchmarks, the system can quickly match or generate precise personalized training programs for users of different clinical categories and guide their execution through an interactive module. At the same time, the adaptive update module automatically generates user-specific templates and feeds back to optimize the knowledge base by monitoring the continuous deviation between individual recovery paths and group standards, enabling the system to have continuous learning capabilities. The communication module and dedicated respiratory training carrier respectively realize remote collaborative management and high-fidelity data acquisition. Ultimately, the system achieves a fundamental transformation from standardized rehabilitation to highly personalized rehabilitation, from static programs to dynamic adaptation, and from local operation to remote intelligent monitoring.

[0007] The present invention is further configured such that: the individual characteristic data includes the type of surgery, the number of days after surgery and the baseline value of preoperative lung function; and the real-time respiratory physiological data includes inspiratory volume, peak inspiratory flow rate and respiratory rhythm stability parameters. The data acquisition module includes a physiological parameter sensor and a data preprocessing unit. The physiological parameter sensor is integrated into the breathing interface of the breathing training carrier. The data preprocessing unit is used to filter, calibrate, and align the acquired raw individual characteristic data and real-time respiratory physiological data with timestamps. The processed raw individual characteristic data and real-time respiratory physiological data are synchronously transmitted to the feature analysis module.

[0008] By adopting the above technical solution, the key core parameters constituting individual characteristic data and real-time respiratory physiological data are clearly defined. The former focuses on three decisive dimensions: surgical type, postoperative days, and preoperative baseline lung function values, providing precise anchor points for accurate classification. The latter focuses on inspiratory volume, peak flow rate, and rhythm stability, comprehensively covering the assessment of respiratory muscle strength, speed, and coordination. At the hardware level, physiological sensors are integrated into the respiratory training carrier interface, enabling in-situ synchronous measurement of respiratory mechanics parameters. The independent data preprocessing unit effectively eliminates noise interference and systematic errors by performing filtering, calibration, and timestamp alignment, ensuring that the data flow from the original individual characteristic data and real-time respiratory physiological data collected at the source to the data transmission to the analysis module has a high degree of consistency, reliability, and timing accuracy, laying an indisputable data quality foundation for all subsequent advanced analysis and decisions.

[0009] The present invention is further configured such that: the feature analysis module calculates the fitting degree between the recovery rhythm curve and the preset trend model; when the fitting degree is lower than the preset standard, the data acquisition module is triggered to re-acquire data; and the recovery rhythm curve and the clinical feature category determination result are simultaneously sent to the protocol generation module and the adaptive update module.

[0010] By adopting the above technical solution, a key data quality self-checking and efficient distribution mechanism is introduced into the feature analysis module. By calculating the fit between the recovery rhythm curve and the preset trend model and setting a minimum standard threshold, the system can automatically identify and filter out low-quality or invalid data trajectories caused by abnormal collection or unstable user status, and immediately trigger re-collection. This closed-loop verification fundamentally prevents junk data from flowing into the downstream decision-making process, ensuring the stability and reliability of the system output. After the data quality is confirmed, the system distributes the effective recovery rhythm curve and clinical classification results synchronously and in parallel to the plan generation module and the adaptive update module. This design realizes the one-time generation of analysis results and their immediate use by multiple parties, greatly improving the internal collaborative efficiency of the system and providing synchronous data support for real-time personalized plan formulation and long-term model evolution.

[0011] The present invention is further configured such that: the intelligent mapping library stores a standard recovery rhythm template and training parameter benchmarks uniquely associated with each clinical feature category, the training parameter benchmarks including target inspiratory volume, inspiratory hold time and number of training sets per day; The intelligent mapping library supports template updates. The adaptive update module counts the frequency of successful application of user-specific rhythm templates. When the frequency exceeds a preset adoption threshold, the original standard recovery rhythm template of the corresponding category is replaced or supplemented with the user-specific rhythm template.

[0012] By adopting the above technical solution, the static structure and dynamic evolution mechanism of the intelligent mapping library are defined. Its static structure is characterized by a unique correspondence between each clinical feature category and a set of standard recovery rhythm templates and training parameter benchmarks. This strong correlation ensures the initial scientific nature and standardization of the generated plan. The defined benchmark parameters, such as the target value of inspiratory volume, the holding time, and the number of training sets, directly constitute the core skeleton of the executable training plan. Its dynamic evolution mechanism, by statistically analyzing the successful application frequency of user-specific templates and comparing it with the preset adoption threshold, enables the system to automatically identify personalized rehabilitation patterns that have been repeatedly verified in practice and perform better than the original standards, and upgrade them to new standards or supplementary templates. This makes the system's knowledge base no longer a static archive, but a living knowledge system that can continuously absorb clinical best practices and constantly improve and grow itself.

[0013] The present invention is further configured such that: the matching analysis of the scheme generation module includes calculating the Euclidean distance between the feature vectors of the recovery rhythm curve and the standard recovery rhythm template as the difference degree, and determining the specific parameters of the personalized breathing training scheme by directly calling the benchmark parameters, adjusting the benchmark parameters proportionally, or generating new parameters based on historical data according to the preset numerical range of the difference degree.

[0014] By adopting the above technical solution, the core matching decision-making process of the solution generation module is fully quantified and programmed. By calculating the Euclidean distance between the restored rhythm curve and the feature vector of the standard template, the complex and multidimensional morphological similarity comparison is transformed into a precise and calculable scalar difference degree. Then, through the preset difference degree value range, three differentiated parameter decision strategies are clearly associated: direct invocation, proportional adjustment, or new generation based on historical data. This constructs a clear, objective, and automated hierarchical decision-making logic. This mechanism perfectly balances the standardization and personalization of the solution, ensuring that the system can adopt the most reasonable strategy when facing different degrees of individual differences. It avoids the mechanical rigidity of the solution and prevents arbitrary and disorderly adjustments, achieving precise and controllable intelligent decision-making.

[0015] The present invention is further configured such that: the personalized breathing training program is configured to include a progressively advancing activation stage, enhancement stage and consolidation stage; the program generation module determines the current stage based on the number of days after surgery and the recovery rhythm curve in the individual characteristic data; and the training parameters of each stage are set according to its preset stage goal and the trend slope of the corresponding indicator in the recovery rhythm curve.

[0016] By adopting the above technical solution, the generated personalized breathing training program is designed as a phased and structured program with clear physiological basis. By dividing it into three stages—activation, enhancement, and consolidation—it accurately corresponds to the natural recovery process of the respiratory muscles after surgery, from nerve activation and muscle strength growth to functional solidification. The program generation module automatically determines the current stage based on the objective number of days after surgery, ensuring the timeliness and scientific nature of the intervention. The training parameters in each stage are not fixed values, but are dynamically set according to the preset core goals of the stage and the trend slope of the recovery rhythm curve that reflects the user's actual progress rate. This allows the training load to respond to the user's recovery rhythm in real time, achieving synchronous optimization of the rehabilitation process and the individual's physiological change curve, thereby pursuing the best rehabilitation efficiency within a safe framework.

[0017] The present invention is further configured such that: the preset conditions include a deviation threshold and the number of consecutive exceedances; when the real-time deviation between the recovery rhythm curve and the standard recovery rhythm template reaches the deviation threshold and the number of consecutive monitoring cycles reaches the number of consecutive exceedances, the adaptive update module triggers the generation of a user-specific rhythm template. The deviation threshold and the number of consecutive exceedances are set based on historical application data statistics of the training parameter benchmark.

[0018] By adopting the above technical solution, a rigorous dual criterion is set for triggering the generation of user-specific templates: a deviation threshold and the number of consecutive exceedances. The deviation threshold ensures that only differences with significant clinical significance are considered, while the requirement for the number of consecutive exceedances confirms the persistence and stability of the difference over time, rather than random fluctuations. The combination of the two forms a highly specific triggering logic, effectively avoiding erroneous actions caused by temporary data anomalies. Furthermore, the setting of these two key thresholds is not subjective, but is based on historical application data statistics of the training parameter benchmark. This makes the triggering criteria itself objective and adaptive, and can be dynamically optimized as the system accumulates clinical experience, thereby ensuring that the activation of the system's evolution mechanism is both sensitive and robust.

[0019] The present invention is further configured such that: the adaptive update module updates the intelligent mapping library in an incremental update manner, and only stores the feature parameters in the user-specific rhythm template whose difference from the existing standard template exceeds a preset difference threshold preset according to the corresponding clinical feature category as an auxiliary template associated with the corresponding clinical feature category. When the clinical feature category determination result of the user matches the category, the scheme generation module calls the standard template and the auxiliary template at the same time during the matching analysis.

[0020] By adopting the above technical solution, efficient and accurate incremental knowledge updates to the intelligent mapping library are achieved. The update process only filters and stores the feature parameters in the user-specific template that differ from the existing standard template by more than a preset threshold, forming an auxiliary template associated with the original category. This filtering mechanism ensures that the newly added knowledge has sufficient uniqueness and supplementary value, avoiding redundant expansion of the knowledge base. When users with similar clinical characteristics appear later, the solution generation module will call the standard template and the relevant auxiliary templates in the matching analysis. This is equivalent to equipping the system with an experience toolbox for handling atypical situations or special cases, significantly enhancing the system's ability to accurately adapt to complex individual differences, especially in handling cases that deviate from the mainstream recovery pattern, and improving decision-making flexibility.

[0021] The present invention is further configured such that: after each update of the recovery rhythm curve, the feature analysis module sends a trigger signal containing new curve data to the scheme generation module, and the scheme generation module then recalculates the matching difference based on the new curve. If the change in training parameters indicated by the old and new matching results exceeds a preset adjustment threshold, a scheme update instruction is generated and sent to the interactive execution module, which then notifies the user to adjust the scheme. The updated scheme data is synchronously fed back to the adaptive update module.

[0022] By adopting the above technical solution, a high-frequency, automatic, real-time program optimization closed loop is established within the system. Each time the feature analysis module updates the recovery rhythm curve, it sends a trigger signal to the program generation module, driving it to recalculate the matching difference based on the latest data. By comparing whether the parameter changes indicated by the old and new results exceed the adjustment threshold, it determines whether the program needs to be adjusted immediately. Once needed, the update instruction is automatically generated and notified to the user through the interaction module for execution, while feedback is also recorded in the update module. This process enables the training program to closely follow the latest subtle changes in the user's recovery status, achieving near real-time dynamic calibration and personalized fine-tuning. This ensures that the user is always under the most suitable training load, greatly improving the precision and adaptability of the rehabilitation process.

[0023] The present invention is further configured such that: the feature analysis module is equipped with a respiratory function abnormality warning threshold preset based on clinical safety data; when any real-time respiratory physiological data continuously exceeds the warning threshold for a preset duration, the feature analysis module triggers a warning signal; the warning signal synchronously controls the interactive execution module to issue an audio-visual alarm and suspend training, and controls the adaptive update module to suspend the update process and send the abnormal data packet to the medical terminal through the communication module. The interactive execution module establishes a connection with the user's mobile terminal through the communication module, and is used to send training reminders and recovery progress reports to the mobile terminal, and receive user subjective status scores from the mobile terminal. The scheme generation module is configured with intensity adjustment rules, which are used to dynamically adjust the intensity parameters of the personalized breathing training scheme within a preset adjustment range based on the received user subjective state score.

[0024] By adopting the above technical solutions, an enhanced system integrating proactive safety protection and deep human-computer interaction was constructed. At the safety level, based on pre-set warning thresholds and duration criteria using clinical data, it can reliably identify respiratory function abnormalities. Triggered warning signals can simultaneously drive the interaction module to issue an alarm and pause training, while also suspending the update module and instantly sending abnormal data to the medical staff. This achieves a complete safety closed loop from local emergency intervention to remote professional alerts. At the interaction level, training reminders, report pushes, and subjective status score collection are achieved through mobile terminals, quantifying and introducing user subjective feelings into the system. The plan generation module dynamically adjusts the plan intensity based on preset rules and this score. This allows the rehabilitation plan to not only respond to objective physiological signals but also integrate user subjective experiences, achieving more humane and tailored rehabilitation management while ensuring safety standards.

[0025] This invention, by employing the above technical solutions, achieves significant technical effects: through the synergistic interaction of various technical features, it constitutes a highly intelligent, adaptive, safe, and reliable overall respiratory rehabilitation solution. Based on a closed-loop architecture, the system ensures the reliability of decision-making sources through high-fidelity data acquisition and quality verification; it utilizes intelligent mapping and quantitative matching to achieve precise transformation from group knowledge to individual plans; it ensures synchronization between training load and user recovery rhythm through phased dynamic design and real-time fine-tuning of the closed loop; and it continuously optimizes the system's knowledge base through a self-evolutionary mechanism. Ultimately, with the guarantee of proactive safety protection and humanized interaction, this system effectively addresses the core pain points of traditional rehabilitation, such as insufficient personalization, disconnect between static plans and dynamic recovery, lack of safety monitoring, and low compliance, achieving a systematic improvement in rehabilitation effectiveness, efficiency, and safety. Attached Figure Description

[0026] Figure 1 This is a flowchart of a personalized training system for postoperative respiratory muscle function rehabilitation. Detailed Implementation

[0027] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.

[0028] Example: Traditional postoperative respiratory muscle function rehabilitation training systems mostly use pre-set static training programs, which lack precise differentiation of individual clinical characteristics of patients. The data collection dimensions are single or the preprocessing is not standardized, resulting in insufficient data accuracy. The training program generation mechanism lacks dynamic correlation with the patient's real-time recovery status, and the training template library is mostly fixed, which is difficult to cover the needs of patients with different recovery rhythms. This leads to the training program being out of sync with the patient's actual recovery rhythm, affecting rehabilitation efficiency and safety.

[0029] This application proposes a personalized training system for postoperative respiratory muscle function rehabilitation, including: The data acquisition module is used to acquire users' individual characteristic data and real-time respiratory physiological data; The feature analysis module, connected to the data acquisition module, is used to determine the user's clinical feature category based on the individual feature data, and to generate a recovery rhythm curve that characterizes the dynamic changes in respiratory muscle function based on the real-time respiratory physiological data. The intelligent mapping library contains pre-stored standard recovery rhythm templates and training parameter benchmarks corresponding to different clinical feature categories; The scheme generation module is connected to the feature analysis module and the intelligent mapping library respectively. It is used to match and analyze the recovery rhythm curve with the corresponding standard recovery rhythm template, and call or generate a personalized breathing training scheme based on the matching results. The interactive execution module, connected to the scheme generation module, is used to output and guide the user to execute the personalized breathing training scheme; The adaptive update module connects the feature analysis module and the intelligent mapping library to monitor the deviation of the recovery rhythm curve from the standard recovery rhythm template. When the deviation meets the preset conditions, it generates a user-specific rhythm template and updates the intelligent mapping library. The communication module is used to enable bidirectional data transmission between the system and medical terminals and user mobile terminals; A breathing training carrier is used to provide users with a physical interface for breathing training. The sensor components used to collect respiratory flow field signals in the data acquisition module are integrated into the breathing training carrier.

[0030] The core of the personalized training system for postoperative respiratory muscle function rehabilitation in this embodiment lies in achieving a precise response to the individualized rehabilitation needs of users through modular design.

[0031] The data acquisition module is responsible for acquiring the user's individual characteristic data and real-time respiratory physiological data. Specifically, individual characteristic data refers to data describing the user's individual characteristics, such as the user's medical history, physical condition, or demographic information. Individual characteristic data can be obtained through manual entry, import from existing medical information systems, or by the user filling out electronic questionnaires. Real-time respiratory physiological data refers to the dynamic measurement data of the user's respiratory function continuously acquired during training or monitoring. Real-time respiratory physiological data can be acquired by connecting external sensor devices, such as placing a flow sensor or pressure sensor independently on the user's breathing path, or by acquiring it through wearable devices. The acquired raw individual characteristic data and raw real-time respiratory physiological data are directly transmitted to the system.

[0032] The feature analysis module connects to the data acquisition module and processes the received data. This module can classify users into different clinical feature categories based on individual feature data, such as through a preset rule set or a simple classification algorithm. Clinical feature categories refer to dividing users into different groups based on their individual clinical characteristics to assist in the development of personalized treatment plans. This module performs time series analysis on real-time respiratory physiological data, such as calculating the average value or trend over a period of time, thereby generating a recovery rhythm curve that characterizes the dynamic changes in respiratory muscle function. The recovery rhythm curve is a curve that graphically presents the dynamic changes in the user's respiratory muscle function over time, reflecting their recovery progress. This curve can be presented in numerical sequence or graphical form.

[0033] The intelligent mapping library pre-stores standard recovery rhythm templates and training parameter benchmarks corresponding to different clinical feature categories. The standard recovery rhythm template refers to a preset respiratory muscle function recovery pattern or model associated with a specific clinical feature category. As a rehabilitation benchmark, the library can be a structured database that stores typical rehabilitation pattern data for different clinical feature groups, as well as corresponding training guidance parameters, such as training duration, number of repetitions, or target intensity range.

[0034] The protocol generation module connects to the feature analysis module and the intelligent mapping library. This module matches the recovery rhythm curve generated by the feature analysis module with the standard recovery rhythm templates corresponding to the clinical feature categories in the intelligent mapping library. For example, by comparing the shape similarity or key indicator differences of the two curves, based on the results of the matching analysis, this module can select the most matching personalized breathing training protocol from the preset protocol library, or modify the baseline training parameters according to the preset adjustment rules to generate a new training protocol. A personalized breathing training protocol refers to a set of breathing exercises and parameters tailored to a single user's unique rehabilitation status and clinical needs.

[0035] The interactive execution module connects to the scheme generation module, which is used to output the generated personalized breathing training scheme to the user and guide the user to execute it. This module can provide training instructions to the user through a display screen, voice prompts or vibration feedback, such as prompting the user to inhale, exhale or hold the action, and provide real-time training progress feedback.

[0036] The adaptive update module connects the feature analysis module and the intelligent mapping library. This module continuously monitors the deviation of the user's recovery rhythm curve from the standard recovery rhythm template. For example, by calculating the difference between the real-time curve and the standard template, when this deviation reaches a preset condition, the module saves the current user's recovery rhythm curve as a user-specific rhythm template and adds it to the intelligent mapping library to achieve dynamic updates to the library content. The user-specific rhythm template refers to a unique recovery pattern generated for the user when their actual recovery situation deviates significantly from the standard template, reflecting their specific recovery trajectory.

[0037] The communication module is used to realize bidirectional data transmission between the system and medical terminals and user mobile terminals. This module can use wired or wireless communication technologies, such as local area network or Internet protocols, to transmit user training data, system status information, early warning notifications, and solution updates.

[0038] A breathing training carrier provides a physical interface for users to train their breathing. The carrier can be a simple breathing tube or mouthpiece, which integrates sensor components for collecting respiratory flow field signals in a data acquisition module. For example, an airflow sensor or pressure sensor can be directly embedded into the structure of the carrier so that physiological data can be collected synchronously when the user is training their breathing.

[0039] The data acquisition module acquires multi-dimensional individual characteristic data and real-time respiratory physiological data, and the feature analysis module performs precise clinical classification and generates dynamic recovery rhythm curves. Combined with standard templates in the intelligent mapping library, the plan generation module can dynamically match and generate personalized respiratory training plans. The adaptive update module can iteratively update the template library according to the user's actual recovery situation. This effectively solves the problem that existing systems cannot accurately adapt to individual patient differences and dynamic changes in respiratory muscle function, avoids rehabilitation efficiency and safety issues caused by inappropriate training intensity, and meets the clinical needs for personalized and precise postoperative respiratory muscle rehabilitation training.

[0040] A personalized training system for postoperative respiratory muscle function rehabilitation, while its data acquisition module can obtain users' individual characteristic data and real-time respiratory physiological data, may fail if the data type is not specific enough or the data quality is poor. This could lead to the feature analysis module being unable to accurately determine the user's clinical characteristic category or accurately generate a recovery rhythm curve that represents the dynamic changes in respiratory muscle function. Consequently, it may affect the accuracy of subsequent personalized training program generation and adaptive updates, resulting in a significant reduction in training effectiveness.

[0041] This application further proposes a more specific data collection and preprocessing scheme. In some implementations of the above system, individual characteristic data is explicitly defined as including surgical type, postoperative days, and preoperative baseline lung function values. Among them, surgical type refers to the specific type of surgery the user has undergone, such as thoracic surgery, abdominal surgery, etc. Different types of surgery have significant differences in the degree of impact on respiratory muscle function and recovery pattern. Using it as individual characteristic data can provide a key basis for subsequent clinical characteristic classification. Postoperative days refer to the time elapsed since the user completed the surgery. Postoperative recovery is a dynamic process, and the physiological state and rehabilitation needs of different postoperative stages are different. Taking postoperative days into consideration helps the system accurately determine the user's recovery stage. Preoperative baseline lung function values ​​refer to lung function-related indicators measured by the user before surgery, such as forced vital capacity (FVC) and forced expiratory volume in one second (FEV1). These baseline values ​​reflect the user's respiratory system health status before surgery and are an important baseline for assessing the degree of postoperative respiratory muscle function recovery.

[0042] Real-time respiratory physiological data is defined as including inspiratory volume, peak inspiratory flow rate, and respiratory rhythm stability parameters. Inspiratory volume refers to the maximum amount of gas a user can inhale during a single inhalation; it is a key indicator for assessing respiratory muscle strength and lung expansion capacity. Peak inspiratory flow rate refers to the maximum airflow velocity a user achieves during inhalation; this parameter reflects the explosive power and coordination of the respiratory muscles. Respiratory rhythm stability parameters refer to the degree of fluctuation or regularity of parameters such as respiratory rate and tidal volume over a period of time. For example, it can be quantified by calculating the coefficient of variation of the respiratory cycle and the standard deviation of respiratory rate. A stable respiratory rhythm is a sign of good respiratory muscle function. These specific and crucial individual characteristic data and real-time respiratory physiological data provide rich and targeted information for the system to conduct precise personalized analysis.

[0043] To ensure data quality and accuracy, the data acquisition module further includes physiological parameter sensors and a data preprocessing unit. The physiological parameter sensors are responsible for converting the user's respiratory physiological signals, such as airflow and pressure, into electrical signals. These sensors can be differential pressure sensors, thermistor flow sensors, or ultrasonic flow sensors, and are integrated into the breathing interface of the breathing training device. Directly integrating the sensors here minimizes signal transmission loss and interference, ensuring that the acquired respiratory flow field signal most closely approximates the user's actual physiological state, thus improving the accuracy and real-time performance of data acquisition. The data preprocessing unit receives the raw electrical signals from the physiological parameter sensors and performs preliminary processing, specifically for processing the acquired data. The raw individual characteristic data and real-time respiratory physiological data undergo filtering, calibration, and timestamp alignment. Filtering aims to remove noise and interference from the raw signal, such as environmental noise and motion artifacts, to obtain a purer and more reliable physiological signal. Calibration corrects measurement errors or drift in the sensor itself, ensuring the accuracy of the measurement results. Timestamp alignment marks each acquired data point with precise time information and ensures time synchronization between different physiological parameters. After these processes, the processed individual characteristic data and real-time respiratory physiological data are synchronously transmitted to the feature analysis module. Synchronous transmission ensures that when the feature analysis module receives data, the individual characteristic data and the corresponding real-time respiratory physiological data are in sync. The system is well-matched, ensuring that the feature analysis module can determine clinical feature categories and generate recovery rhythm curves based on the latest, most accurate, and pre-processed data. This allows the system to acquire more comprehensive and high-quality user data. By collecting individual characteristic data such as surgery type, postoperative days, and preoperative baseline lung function values, the system can more accurately assess the user's clinical background and recovery stage, providing a refined basis for subsequent clinical feature classification. Real-time acquisition of key respiratory physiological data such as inspiratory volume, peak inspiratory flow rate, and respiratory rhythm stability parameters, along with integration with physiological parameter sensors at the respiratory interface of the respiratory training vehicle, ensures the real-time performance and accuracy of the respiratory flow field signal. The data preprocessing unit preprocesses the raw data of individual characteristics... By filtering, calibrating, and aligning the real-time respiratory physiological data with the raw data, noise and errors are effectively eliminated, ensuring data purity and synchronization. This refined individual characteristic data is synchronously transmitted to the feature analysis module along with the real-time respiratory physiological data. This allows the feature analysis module to perform analysis based on more reliable and representative data, thereby more accurately determining the user's clinical characteristic category and generating a more precise and dynamically responsive recovery rhythm curve. This significantly improves the accuracy of the system's assessment of the user's respiratory muscle function status, laying a solid foundation for the protocol generation module to call or generate truly personalized and effective respiratory training protocols, and thus optimizing the training effect of postoperative respiratory muscle function rehabilitation.

[0044] In some of the embodiments described above in this application, a personalized training system for postoperative respiratory muscle function rehabilitation is proposed. This system can acquire the user's individual characteristic data and real-time respiratory physiological data, and generate a recovery rhythm curve based on this data. Then, it matches a standard template to generate a personalized training plan. However, in practical applications, how to classify the user's clinical characteristics more precisely, and how to ensure the accuracy and reliability of the generated recovery rhythm curve, are key issues affecting the effectiveness of subsequent training plans. If the clinical characteristic classification is not accurate enough, or if the recovery rhythm curve is affected by data quality issues, the personalized training plan provided by the system may not meet the user's actual rehabilitation needs, and may even delay the rehabilitation process.

[0045] This application further proposes optimizations to the aforementioned personalized training system. The feature analysis module performs clustering based on three dimensions of individual feature data: surgical type, postoperative days, and preoperative baseline lung function, resulting in at least three clinical feature categories. This clustering process can be implemented using various unsupervised learning algorithms, such as K-means clustering, hierarchical clustering, or Gaussian mixture models. Before clustering, to eliminate the influence of differences in the dimensions of data on the clustering results, the surgical type, postoperative days, and preoperative baseline lung function data can be standardized. This multi-dimensional clustering method allows for a more objective and detailed grouping of users with similar rehabilitation characteristics into one category, making the subsequent preset standard recovery rhythm templates and training parameter benchmarks for that category of users more representative and instructive.

[0046] This application defines the recovery rhythm curve as a two-dimensional curve constructed with time as the horizontal axis and the rate of change of inspiratory volume and the respiratory rhythm stability index as the vertical axes. The rate of change of inspiratory volume can be expressed as the relative change in inspiratory volume at adjacent time points, for example, by calculating the slope of linear regression of inspiratory volume within a time window. The respiratory rhythm stability index can be quantified by analyzing the coefficient of variation of the respiratory cycle, such as inspiratory and expiratory time, or by using nonlinear dynamic indicators such as entropy. This two-dimensional curve construction method allows the recovery rhythm curve to not only reflect the recovery trend of respiratory muscle strength and endurance, but also simultaneously reflect the coordination and efficiency of respiratory control, thus providing a more comprehensive and refined assessment of the dynamic changes in respiratory muscle function.

[0047] To further ensure the quality of the recovery rhythm curve, the feature analysis module also calculates the goodness of fit between the recovery rhythm curve and the preset trend model. This preset trend model can be a typical recovery pattern obtained based on a large amount of clinical rehabilitation data, such as an exponential growth model or an S-curve model, or it can be an average recovery trend model for a specific clinical feature category. The goodness of fit can be quantified by statistical methods such as mean squared error (MSE), R-squared value, or Pearson correlation coefficient. When the goodness of fit is lower than the preset standard, for example, if the R-squared value is lower than a certain threshold, the data acquisition module will be triggered to re-acquire data. This mechanism can effectively identify and avoid low-quality or abnormal data caused by sensor failure, improper user operation, or physiological abnormalities, ensuring that the recovery rhythm curve used for subsequent analysis and protocol generation is reliable and effective.

[0048] After the above processing is completed, the generated recovery rhythm curve and clinical feature category determination results are simultaneously sent to the protocol generation module and the adaptive update module. This synchronous transmission mechanism ensures that the protocol generation module can obtain the latest and most accurate recovery rhythm curve and user classification information in a timely manner, so as to perform accurate matching analysis and generate personalized training protocols. The adaptive update module can also receive these key data in real time, providing necessary data support for subsequent monitoring of the deviation of the recovery rhythm curve from the standard recovery rhythm template, and for generating user-specific rhythm templates and updating the intelligent mapping library when preset conditions are met. The system can analyze user data based on multi-dimensional individual feature data. More refined clinical feature classification makes rehabilitation guidance more targeted. By constructing a two-dimensional recovery rhythm curve that includes the rate of change in inspiratory volume and the respiratory rhythm stability index, the system can more comprehensively and accurately assess the dynamic changes in the user's respiratory muscle function. More importantly, the introduction of a fitting degree calculation mechanism between the recovery rhythm curve and the preset trend model, as well as a data re-acquisition mechanism, effectively improves the reliability of the recovery rhythm curve and the data quality, avoiding the risk of generating training programs based on inaccurate data. These improvements work together to enable the system to provide more accurate, effective, and safe personalized respiratory training programs, thereby significantly improving the efficiency and effectiveness of postoperative respiratory muscle function rehabilitation.

[0049] In some embodiments described above in this application, a feature analysis module is proposed to determine the clinical feature category based on the user's individual feature data and generate a recovery rhythm curve based on real-time respiratory physiological data. Subsequently, the scheme generation module matches and analyzes the recovery rhythm curve with the pre-stored standard recovery rhythm template in the intelligent mapping library to generate a personalized breathing training scheme. However, in practical applications, the preset standard recovery rhythm template may not be able to fully cover the individual differences of all users, or with the progress of medical research and the accumulation of a large amount of user data, the original standard template may have room for optimization, resulting in the generated training scheme failing to achieve the best personalized effect.

[0050] This application further proposes that the intelligent mapping library is configured to store a standard recovery rhythm template and training parameter benchmark uniquely associated with each clinical feature category. Here, the clinical feature categories are formed by clustering and classifying user individual feature data using a feature analysis module, ensuring the specificity of the initial plan. Individual feature data includes surgical type, postoperative days, and preoperative baseline lung function values. The standard recovery rhythm template can be a mathematical model or a sequence of feature points representing a two-dimensional curve constructed with time as the horizontal axis and the rate of change in inspiratory volume and respiratory rhythm stability index as the vertical axis, used to characterize the typical recovery trend of patients in different clinical feature categories. The training parameter benchmark is a set of parameters associated with this standard template. The corresponding initial recommended training parameters include the target inspiratory volume, inspiratory hold time, and daily training sets. The target inspiratory volume refers to the target inspiratory volume that the user should achieve with each inspiratory breath. The inspiratory hold time refers to the duration that the user needs to maintain the inspiratory state after reaching peak inspiratory capacity. The daily training sets refer to the number of training cycles or sets that the user needs to complete in a day. These parameters together constitute the core elements of a personalized breathing training program, directly guiding the user's training behavior. The intelligent mapping library can be implemented using a relational database or a non-relational database. Its data structure can clearly map clinical feature categories to corresponding standard templates and training parameter benchmarks.

[0051] To ensure continuous optimization and adaptability of the system, the intelligent mapping library is designed to support template updates. This means that the intelligent mapping library is not static but has the ability to dynamically adjust and optimize. It allows the system to modify, replace or supplement the stored standard recovery rhythm templates based on actual operating results and user feedback. This update mechanism is the foundation for the system to achieve adaptive learning and evolution.

[0052] The adaptive update module is configured to count the frequency of successful application of user-specific rhythm templates. When the adaptive update module detects that the deviation of the recovery rhythm curve from the standard recovery rhythm template meets preset conditions, it generates a user-specific rhythm template. Successful application here means that the user-specific rhythm template is adopted by the protocol generation module and used to guide actual training, and achieves positive recovery results in actual execution. The adaptive update module maintains a counter for each user-specific rhythm template. Each time the template is successfully applied, the counter is incremented. When the successful application frequency of a user-specific rhythm template exceeds a preset adoption threshold, the adaptive update module will trigger an update operation of the intelligent mapping library. At this time, the user-specific rhythm template will be used to replace or supplement the original standard recovery rhythm template in the corresponding category. The replacement operation is applicable when the user-specific rhythm template shows significantly better performance than the original standard template in a certain clinical feature category; the supplementation operation... When a user-specific rhythm template represents a specific subgroup or recovery path that the original standard template failed to cover, the intelligent mapping library can continuously learn and evolve from a large amount of user practice data through this mechanism, making its stored standard templates more closely aligned with actual rehabilitation needs. This application implements a dynamic update mechanism for the standard recovery rhythm templates and training parameter benchmarks in the intelligent mapping library, overcoming the limitations that traditional fixed templates may have in personalized training. When a user-specific rhythm template generated by the adaptive update module proves to have high effectiveness and universality in practical applications, the system can incorporate it into the intelligent mapping library, thereby continuously optimizing and enriching the standard template library. This enables the system to learn from a large amount of user practice data, continuously improving the accuracy and adaptability of personalized breathing training programs, ensuring that the training program is always highly matched with the user's actual recovery process, and thus significantly improving the overall effect of postoperative respiratory muscle function rehabilitation.

[0053] In some of the embodiments described above in this application, a method is proposed to match and analyze the user's recovery rhythm curve with a pre-stored standard recovery rhythm template in an intelligent mapping library, and to call or generate a personalized breathing training program based on the matching results. However, in practical applications, the user's recovery rhythm curve may differ from the standard template to varying degrees. Simply calling preset benchmark parameters or generating a rough template may not be able to adequately adapt to individual differences, resulting in insufficient accuracy and personalization of the training program, thereby affecting the rehabilitation effect.

[0054] This application further proposes a matching analysis for the scheme generation module, which includes calculating the Euclidean distance between the feature vectors of the recovery rhythm curve and the standard recovery rhythm template as the difference degree. Based on the preset numerical range of the difference degree, the specific parameters of the personalized breathing training scheme are determined by directly calling the benchmark parameters, adjusting the benchmark parameters proportionally, or generating new parameters based on historical data.

[0055] To quantitatively compare the recovery rhythm curve with the standard recovery rhythm template, it is necessary to transform them into a computable mathematical form. This can be achieved by extracting the rate of change of inspiratory volume and the respiratory rhythm stability index at key time points in both the recovery rhythm curve and the standard recovery rhythm template, forming a multidimensional vector. For example, several fixed time points in the rehabilitation process can be selected, such as day 3, day 7, and day 14 post-surgery, with the curve values ​​corresponding to these time points serving as the various dimensions of the vector. Alternatively, statistical features of the curve can be extracted, or frequency domain features can be extracted using methods such as Fourier transform and wavelet analysis to construct a more representative feature vector. Statistical features can include average value, peak value, slope, area under the curve, etc. Euclidean distance, a commonly used distance metric, is used to calculate the straight-line distance between two feature vectors in multidimensional space; the result is the degree of difference, which quantifies the deviation between the user's current recovery state and the ideal standard template.

[0056] The system predefines multiple difference intervals, each corresponding to a parameter determination strategy. For example, a very low difference interval (e.g., 0-0.1), a medium difference interval (e.g., 0.1-0.5), and a high difference interval (e.g., greater than 0.5) can be set. These intervals can be optimized based on clinical experience, expert knowledge, or historical training data. When the calculated difference is in the very low difference interval, it indicates that the user's recovery rhythm curve closely matches the standard recovery rhythm template. In this case, the protocol generation module directly retrieves the training parameter benchmark associated with the standard template from the intelligent mapping library as a personalized parameter. The specific parameters of the breathing training program include benchmark parameters such as target inspiratory volume, inspiratory hold time, and number of training sets per day. When the difference is in the moderate range, it indicates that the user's recovery has deviated to some extent, but adjustments can still be made based on the standard template. The program generation module will adjust the benchmark parameters proportionally according to the magnitude of the difference. For example, if the difference shows that the user's recovery progress is slightly slower than the standard, the target inspiratory volume can be appropriately reduced or the inspiratory hold time extended to reduce the training intensity; conversely, if the recovery progress is ahead of schedule, the training intensity can be appropriately increased. The adjustment ratio can be linear with the difference. The system employs a non-linear relationship to ensure the rationality of adjustments. When the difference is in the high difference range, it indicates a significant deviation between the user's recovery rhythm and the standard template. In this case, simply adjusting the baseline parameters may no longer be applicable. The program generation module utilizes historical training data stored in the intelligent mapping library, which includes training programs and rehabilitation effects of other users with similar high differences. Combined with machine learning algorithms such as regression analysis and reinforcement learning, it generates a new set of training parameters that better match the user's unique recovery state. This application can accurately quantify the degree of matching between the user's recovery rhythm curve and the standard recovery rhythm template, avoiding the problem of coarse programs that may result from simple matching. Based on the difference calculated using Euclidean distance, the system can identify subtle deviations in the user's recovery state and intelligently select the most suitable parameters to determine the strategy according to the degree of deviation. Whether it is a high degree of matching, slight deviation, or significant deviation, the system can provide accurate and personalized training parameters, thereby ensuring that the personalized breathing training program can more closely match the user's actual rehabilitation process. This refined parameter generation mechanism significantly improves the adaptability and effectiveness of the training program, helps optimize rehabilitation effects, and reduces the risks caused by inappropriate training programs.

[0057] The system can generate personalized breathing training plans based on the user's real-time physiological data and clinical characteristics, and adjust the training parameters. However, in the actual rehabilitation process, the user's physiological state and rehabilitation needs will change in stages over time. If the training plan fails to effectively match these stage changes, it may lead to poor training results or hinder the rehabilitation process.

[0058] This application further proposes an optimized configuration of a personalized breathing training program, which is configured to include a progressive activation phase, an enhancement phase, and a consolidation phase. This phased training design aims to systematize and structure the entire postoperative respiratory muscle function rehabilitation process in order to better adapt to the physiological characteristics and needs of patients at different rehabilitation stages.

[0059] The activation phase typically corresponds to the early postoperative period. Its main goal is to gently activate respiratory muscle function, prevent postoperative complications, and help patients restore basic breathing patterns. During this phase, the training intensity is usually low, focusing more on respiratory control and activation of shallow respiratory muscles. As the patient's recovery progresses and physical tolerance increases, the training program will enter the enhancement phase. The core goal of this phase is to gradually enhance the strength and endurance of the respiratory muscles. Therefore, the training intensity and duration will increase accordingly, and resistance training or more complex breathing exercises may be introduced to promote a comprehensive improvement in respiratory muscle function. Finally, in the later stages of rehabilitation, the training program will enter the consolidation phase. This phase aims to maintain and further optimize respiratory muscle function, prevent functional degeneration, and help patients integrate their restored breathing ability into daily activities, ultimately achieving the goal of the patient returning to a normal life.

[0060] To achieve dynamic management of this phased training, the program generation module determines the current phase based on the number of days after surgery and the recovery rhythm curve in the individual's characteristic data. Specifically, the program generation module can have a set of phase division rules built in. For example, days 1-7 after surgery can be defined as the activation phase, days 8-28 after surgery can be defined as the enhancement phase, and days 29 and beyond after surgery can be defined as the consolidation phase. By comparing the user's current number of days after surgery with these preset time ranges, the system can accurately determine the user's rehabilitation phase. These phase division rules can be set and optimized based on a large amount of clinical experience data, rehabilitation guidelines for different types of surgery, or expert knowledge.

[0061] Building upon this foundation, the training parameters for each stage are no longer solely determined by a single matching result. Instead, they are comprehensively set based on the pre-defined stage goals and the trend slope of the corresponding indicators in the recovery rhythm curve. Each stage has a clear rehabilitation objective. For example, the activation stage may focus on restoring baseline inspiratory volume, the enhancement stage on increasing peak inspiratory flow, and the consolidation stage on improving respiratory rhythm stability. These pre-defined stage goals provide directional guidance for the initial setting of training parameters, such as the slope of the rate of change of inspiratory volume or the slope of the respiratory rhythm stability index in the recovery rhythm curve. This system can reflect the speed and direction of a user's rehabilitation progress in real time. The program generation module analyzes these trend slopes. For example, if the trend slope of the inspiratory volume change rate remains positive and is above a certain threshold, it indicates that the user's rehabilitation progress is good, and the target inspiratory volume value can be appropriately increased. Conversely, if the slope tends to flatten or decrease, it may be necessary to adjust the inspiratory hold time or reduce the training intensity. This parameter setting method, which combines stage goals and dynamic trend slopes, ensures that the training program not only conforms to the macroscopic stage rehabilitation pattern but also finely adapts to the dynamic changes in individual rehabilitation. This application overcomes the limitations of traditional personalized training programs in the rehabilitation process. The lack of systematic stage division and dynamic adaptability is addressed by clearly configuring personalized breathing training programs into progressively advancing activation, enhancement, and consolidation stages. This is achieved by utilizing postoperative days and recovery rhythm curves from individual characteristic data for stage determination. The system provides users with a structured training framework that aligns with physiological rehabilitation principles. This allows the training program to better match the patient's physiological needs and capacity changes at different stages of rehabilitation, avoiding mismatches between training intensity and the patient's current state. Furthermore, the training parameters for each stage are not only set based on preset stage goals but also dynamically adjusted according to the trend slope of corresponding indicators in the recovery rhythm curve. This approach allows the training program to provide macro-level stage guidance while also enabling fine-tuning at the micro-level based on the user's actual recovery speed and trends. For example, when the recovery rhythm curve shows rapid recovery progress, the system can appropriately increase training intensity to accelerate recovery; when progress is slow or a plateau occurs, the strategy can be adjusted promptly to avoid overtraining or undertraining. This significantly improves the effectiveness and safety of training, ensuring a smooth recovery process and optimizing the overall rehabilitation effect, making personalized training programs more adaptable and precise.

[0062] In some of the embodiments described above in this application, an adaptive update module is proposed to monitor the deviation of the recovery rhythm curve from the standard recovery rhythm template and generate a user-specific rhythm template when the deviation meets preset conditions. However, in practical applications, if the preset conditions are too simple, such as being based only on instantaneous deviation, the system may become overly sensitive to short-term physiological fluctuations, frequently triggering the generation of user-specific rhythm templates, thereby affecting the stability of the training scheme and the operating efficiency of the system.

[0063] This application further proposes the aforementioned preset conditions, including a deviation threshold and the number of consecutive exceedances. When the real-time deviation between the recovery rhythm curve and the standard recovery rhythm template reaches the deviation threshold, and the number of consecutive monitoring cycles reaches the number of consecutive exceedances, the adaptive update module triggers the generation of a user-specific rhythm template. The deviation threshold and the number of consecutive exceedances are set based on the historical application data statistics of the training parameter benchmark.

[0064] The preset conditions are the criteria for triggering the adaptive update module to generate a user-specific rhythm template. These conditions are refined into two interdependent parameters: a deviation threshold and a number of consecutive exceedances. These parameters aim to improve the accuracy and robustness of the judgment. The deviation threshold refers to the maximum allowable difference between the recovery rhythm curve and the standard recovery rhythm template. When the real-time monitored deviation exceeds this threshold, it indicates that the user's recovery may differ significantly from expectations. This threshold can be set based on clinical experience, statistical analysis, or machine learning models. For example, it can be defined as Euclidean distance, correlation coefficient, or a percentage deviation of a specific indicator, such as the rate of change in inspiratory volume or the respiratory rhythm stability index. The number of consecutive exceedances refers to the number of monitoring cycles in which the deviation consistently exceeds the deviation threshold. This parameter is introduced to avoid misjudgment due to accidental, transient physiological fluctuations. It ensures that only when the deviation persists and reaches a certain level is the user's recovery pattern considered to have a substantial difference from the standard template, thus triggering the generation of the user-specific rhythm template. The monitoring cycle can be daily, half-day, or per training session.

[0065] Furthermore, the adaptive update module continuously receives the recovery rhythm curves generated by the feature analysis module and compares them in real time with the standard recovery rhythm templates corresponding to the clinical feature categories stored in the intelligent mapping library. The comparison method can use Euclidean distance, dynamic time warping (DTW) algorithm, or other curve similarity measurement methods to calculate the real-time deviation. When the calculated real-time deviation exceeds the preset deviation threshold for the first time, the system will start a counter. In each subsequent monitoring cycle, if the deviation continues to exceed the threshold, the counter will increment; if the deviation falls back to within the threshold, the counter will be reset to zero. Only when the counter reaches the preset number of consecutive exceedances will the adaptive update module issue an instruction to start the generation process of the user-specific rhythm template.

[0066] To ensure the scientific validity and rationality of the deviation thresholds and the number of consecutive exceedances, and to better reflect the actual clinical recovery and training effects, this application proposes that these parameters be set based on historical application data of the training parameter benchmark. The system collects a large amount of historical data on users' training under different training parameter benchmarks. This historical data includes the user's previous training data and effective rehabilitation data of similar clinical feature categories in the intelligent mapping library, including their recovery rhythm curves, training effects, and whether the program needs adjustment. Statistical analysis of this historical data can be performed; for example, machine learning algorithms such as classification, regression, or anomaly detection can be used to identify the degree and duration of deviation. Deviation patterns ultimately lead to poor training effects or require manual intervention. Based on these analysis results, the deviation thresholds and the number of consecutive exceedances can be dynamically or offline optimized. For example, it can be analyzed at which deviation degrees and durations significantly deviate from the standard template in the user's final recovery path. Differentiation is used to determine appropriate thresholds. This method, based on historical data, allows the system to learn and optimize its adaptive capabilities from practical applications, avoiding the limitations of subjective judgment or empirical settings. Deviation thresholds and consecutive exceedance counts are introduced as preset conditions to trigger the generation of user-specific rhythm templates. These are statistically set based on historical application data from training parameter benchmarks, effectively solving the problems of system misjudgment and frequent generation of new templates caused by instantaneous physiological fluctuations. This dual judgment mechanism, combined with thresholds set by historical data statistics, significantly improves the accuracy and robustness of the system's judgment of the user's recovery state. It can effectively distinguish between normal physiological fluctuations and actual deviations from the recovery path, avoiding unnecessary template updates and ensuring the stability and continuity of the training program. Simultaneously, when the user's recovery path does change significantly, it can promptly and accurately generate a more suitable user-specific template, further enhancing the accuracy and effectiveness of personalized training programs.

[0067] In some of the embodiments described above in this application, the system can generate a user-specific rhythm template and update the intelligent mapping library when preset conditions are met, based on the deviation between the user's recovery rhythm curve and the standard recovery rhythm template. However, if the standard template in the intelligent mapping library is updated by completely replacing or simply supplementing the user-specific rhythm template each time it is generated, it may cause the size of the intelligent mapping library to expand rapidly, increase the storage and management burden, and may introduce too many templates with subtle differences, reducing the template's generalization ability and matching efficiency.

[0068] This application further proposes an adaptive update module that updates the intelligent mapping library in an incremental manner. Only feature parameters in the user-specific rhythm template whose difference from the existing standard template exceeds a preset difference threshold for the corresponding clinical feature category are stored as auxiliary templates associated with the corresponding clinical feature category. When the clinical feature category determination result of the user matches the category, the scheme generation module calls both the standard template and the auxiliary template during the matching analysis.

[0069] The adaptive update module updates the intelligent mapping library incrementally. This means that the update of the intelligent mapping library is not a full replacement or simple appending, but rather a selective and local process. It focuses on identifying and updating or adding only the key information that is significantly different from the existing templates, rather than the entire template structure. For example, when a user-specific rhythm template is generated, the adaptive update module compares it with the corresponding standard recovery rhythm template in the current intelligent mapping library to identify the specific parameters or feature points that differ between the two, rather than storing the entire user-specific rhythm template as a completely new template.

[0070] Only feature parameters in the user-specific rhythm template that differ from the existing standard template by a preset difference threshold according to the corresponding clinical feature category are stored as auxiliary templates associated with the corresponding clinical feature category. Here, feature parameters can be understood as key indicators that constitute the recovery rhythm curve, such as the rate of change of inspiratory volume at a specific time point of the curve, respiratory rhythm stability index, peak slope, plateau length, etc. These parameters can quantitatively characterize the recovery dynamics of respiratory muscle function. The preset difference threshold is a numerical standard used to determine whether the difference between the user-specific rhythm template and the standard template is significant enough to warrant separate recording. For example, if the numerical difference of a certain feature parameter exceeds 5%, it is considered significant. The auxiliary template is not a complete and independent standard template, but a set of supplementary and corrective feature parameters. It specifically records the unique features of the user-specific rhythm template relative to the standard template under a specific clinical feature category. This storage method avoids the storage of redundant information and enables the intelligent mapping library to capture individual differences more precisely.

[0071] When a subsequent user's clinical characteristic category is matched, the protocol generation module calls both the standard template and the auxiliary template during the matching analysis. This means that when a new user is identified as belonging to a certain clinical characteristic category, the protocol generation module no longer relies solely on the standard recovery rhythm template for that category when generating a personalized breathing training plan. Instead, it first calls the standard template for that category as a basis, and also calls all auxiliary templates associated with that category. During the matching analysis, the protocol generation module comprehensively considers the general rules represented by the standard template and the unique recovery patterns of specific individuals or subgroups reflected by the auxiliary templates, thereby generating a more accurate and personalized training plan. For example, an initial plan can be determined based on the standard template, and then the initial plan can be fine-tuned according to the difference parameters recorded in the auxiliary templates.

[0072] Through the aforementioned incremental update method, the adaptive update module can avoid excessive expansion of the intelligent mapping library due to frequent full updates, effectively controlling storage resource consumption. It only stores feature parameters that differ significantly from existing standard templates as auxiliary templates, enabling the intelligent mapping library to more precisely capture and record individual recovery patterns under specific clinical feature categories without diluting the generalization ability of the standard templates. When the protocol generation module calls both the standard template and the auxiliary template during matching analysis, it can combine general recovery patterns with the unique recovery characteristics of specific individuals or subgroups, thereby generating more accurate and personalized breathing training protocols. This mechanism not only improves the adaptability and effectiveness of training protocols but also optimizes the structure and management efficiency of the intelligent mapping library, ensuring that the system remains efficient and stable during continuous learning and adaptation.

[0073] The system can generate a recovery rhythm curve based on the user's real-time respiratory physiological data and generate a personalized training plan based on the matching of the curve with a standard template. It can even update the intelligent mapping library when the curve deviates from the preset conditions. However, in actual applications, the user's respiratory muscle function recovery status may change dynamically. If the system cannot respond to the update of the recovery rhythm curve in a timely manner and adjust the training plan accordingly, it may cause the training plan to be inconsistent with the user's current actual recovery status, affecting the training effect and user experience.

[0074] This application further proposes that after each update of the recovery rhythm curve, the feature analysis module sends a trigger signal containing new curve data to the scheme generation module. The scheme generation module then recalculates the matching difference based on the new curve. If the change in training parameters indicated by the old and new matching results exceeds the preset adjustment threshold, a scheme update instruction is generated and sent to the interactive execution module. The interactive execution module then notifies the user to adjust the scheme, and the updated scheme data is synchronously fed back to the adaptive update module.

[0075] After each successful update of the recovery rhythm curve, the feature analysis module immediately sends a trigger signal to the protocol generation module. This trigger signal can be a lightweight data packet containing the complete data of the newly generated recovery rhythm curve, or a reference to the storage location of the curve. This is to ensure that the protocol generation module can obtain the latest information on the user's respiratory muscle function recovery status in a timely manner, providing a real-time basis for subsequent protocol adjustments.

[0076] Upon receiving a trigger signal from the feature analysis module, the scheme generation module immediately initiates its internal matching analysis process. This module uses the latest recovery rhythm curves to perform matching analysis with the standard recovery rhythm templates corresponding to the clinical feature categories stored in the intelligent mapping library, and recalculates the difference between the two. For example, algorithms such as Euclidean distance can be used to quantify the difference between the curves to ensure that the adjustment of the training scheme is based on the user's current most accurate recovery state.

[0077] To avoid frequent adjustments to the training protocol due to minor fluctuations, the protocol generation module compares the changes in the training parameters calculated based on the new curve with those of the previous execution. This change can be an assessment of the absolute or relative differences in key parameters such as the target inspiratory volume, inspiratory hold time, or the number of training sets per day. Only when this change exceeds a preset adjustment threshold does the system consider the protocol adjustment necessary and generate a protocol update instruction containing the new training parameters. This preset adjustment threshold can be determined based on clinical experience, expert advice, or statistical analysis of historical training data to balance the responsiveness and stability of the protocol.

[0078] Once the update instruction is generated, it will be sent to the interactive execution module. The interactive execution module is responsible for conveying the updated training plan to the user and guiding the user to train according to the new parameters. The notification method can be diversified, such as directly displaying the new training parameters on the display screen of the breathing training carrier, or sending push notifications or voice prompts by establishing a connection with the user's mobile terminal, to ensure that the user can clearly understand and execute the adjusted plan.

[0079] To continuously optimize the system's adaptive capabilities, updated protocol data is synchronously fed back to the adaptive update module. This data includes new training parameters, protocol update timestamps, and initial user feedback after implementing the new protocol. The adaptive update module can utilize this information, combined with subsequently collected real-time respiratory physiological data, to more accurately evaluate the effectiveness of the new protocol. It also provides more comprehensive data support for generating user-specific rhythm templates or updating the intelligent mapping library, thereby further enhancing the system's intelligence. The system can achieve real-time dynamic monitoring and rapid response to the user's respiratory muscle function recovery status. The feature analysis module immediately triggers the protocol generation module for re-evaluation after the recovery rhythm curve is updated, ensuring that the training protocol remains synchronized with the user's latest physiological data. By setting adjustment thresholds, the system avoids frequent fine-tuning due to oversensitivity, while ensuring timely and effective adjustments to the training protocol when the user's recovery status changes significantly. The interactive execution module promptly notifies the user, ensuring that the user can train according to the latest personalized protocol, thus improving the accuracy and effectiveness of training. The updated protocol data fed back to the adaptive update module provides a more accurate basis for subsequent intelligent mapping library optimization and user-specific rhythm template generation, further enhancing the system's adaptive capabilities and long-term training effects.

[0080] In some of the embodiments described above in this application, although the system can provide users with personalized breathing training programs and make dynamic adjustments based on the user's individual characteristic data and real-time respiratory physiological data through modules such as feature analysis, program generation, and adaptive updates, if the user experiences abnormalities in respiratory function indicators that continuously exceed the safe range during training, the existing programs may not be able to detect and take effective intervention measures in a timely manner, thereby posing potential health risks and affecting the safety of training.

[0081] This application further proposes a personalized training system, wherein the feature analysis module is equipped with a pre-set respiratory function abnormality warning threshold based on clinical safety data. When any real-time respiratory physiological data continuously exceeds the warning threshold for a preset duration, the feature analysis module triggers a warning signal. This warning signal simultaneously controls the interactive execution module to issue an audio-visual alarm and suspend training, and controls the adaptive update module to suspend the update process and send the abnormal data packet to the medical terminal through the communication module.

[0082] The preset respiratory function abnormality warning thresholds in the feature analysis module are based on a large amount of clinical safety data, such as the physiological range of healthy people, the safety range during postoperative recovery, and the contraindications for patients with specific diseases. These thresholds are determined through statistical analysis and expert experience. They can be stored in the internal memory of the feature analysis module or loaded from an external database. For example, the lower limit threshold of inspiratory volume, the upper limit threshold of peak inspiratory flow, and the fluctuation range threshold of respiratory rhythm stability parameters can be set. These thresholds can be fixed values ​​or dynamically adjusted according to the user's individual characteristics such as age, gender, and underlying diseases to ensure the accuracy and personalization of the warnings.

[0083] When the feature analysis module continuously monitors the real-time respiratory physiological data received from the data acquisition module, it compares it with the corresponding respiratory function abnormality warning threshold. If a real-time respiratory physiological data continuously exceeds its corresponding warning threshold and the duration reaches a preset duration, such as 3, 5, or 10 seconds, the feature analysis module will determine it as a real abnormality and immediately trigger a warning signal. This preset duration can be configured according to clinical experience and data sensitivity to avoid false alarms caused by instantaneous fluctuations and ensure the reliability of the warning.

[0084] Once the warning signal is triggered, it will simultaneously control the interactive execution module and the adaptive update module. Upon receiving the warning signal, the interactive execution module will immediately perform two operations: First, it will issue visual and auditory alarms, such as flashing indicator lights, sounding buzzers, or providing voice prompts on the breathing training carrier. If the system is connected to a user's mobile terminal, it can also issue vibration, pop-up, or sound alarms through the mobile terminal. Second, it will pause the training, that is, it will send a command to the breathing training carrier to stop providing training feedback or resistance and stop guiding the user to perform breathing actions. At the same time, the training timer and progress record inside the system will also be paused.

[0085] The warning signal will also control the adaptive update module to pause the update process. Upon receiving the pause command, the adaptive update module will immediately stop any ongoing matching analysis of recovery rhythm curves and standard templates, generation of user-specific rhythm templates, or update of the intelligent mapping library. This can effectively prevent the system from updating the model based on inaccurate or dangerous data when the user is in an abnormal physiological state, which could lead to incorrect adjustments to the subsequent training scheme.

[0086] The feature analysis module or communication module also sends abnormal data packets to the medical terminal via the communication module. When an alarm signal is triggered, the system immediately packages the current abnormal real-time respiratory physiological data, the time of the alarm trigger, the duration, and the user's individual characteristic data into an abnormal data packet. The communication module encrypts this data packet via the network and sends it to a preset medical terminal, such as a doctor's workstation, a nurse's station tablet, or a mobile app. After receiving the data, the medical terminal can immediately display the alarm information and allow medical staff to view detailed abnormal data for remote diagnosis or to arrange on-site examinations. This application can promptly detect abnormalities that users may encounter during training. In case of any physiological abnormalities, if the abnormal data continues to exceed the warning threshold for a preset duration, the system can immediately trigger a warning signal, issue an audio-visual alarm through the interactive execution module, and suspend training. This effectively prevents users from continuing training in an unsafe state, maximizing user training safety. The warning signal also controls the adaptive update module to pause the update process, preventing abnormal data from contaminating the model and ensuring the accuracy of subsequent training plans. The abnormal data packets are promptly sent to the medical terminal through the communication module, enabling medical staff to remotely monitor the user's health status in real time and provide timely intervention and guidance, further improving the safety and effectiveness of postoperative respiratory muscle function rehabilitation training.

[0087] The interactive execution module establishes a connection with the user's mobile terminal via a communication module. This connection can be achieved using various wireless communication technologies, such as Bluetooth, Wi-Fi, or cellular networks, ensuring stable and reliable data transmission between the system and the user's mobile terminal. Using this connection, the interactive execution module can send training reminders and recovery progress reports to the user's mobile terminal. Training reminders can be sent to the user's mobile terminal before a preset training time via push notifications, SMS, or voice messages, prompting the user to begin training and including the main objectives or precautions for the training session. The recovery progress reports periodically update the user on training completion status and respiratory muscle function recovery. Information such as the trend of rhythm curve changes, the degree of matching with the standard recovery rhythm template, and the achievement of phased goals are sent to the user's mobile terminal in the form of intuitive charts or text, so that the user can clearly understand their own rehabilitation progress. The interactive execution module is also used to receive the user's subjective state rating from the user's mobile terminal. The user's mobile terminal can provide a user interface that allows users to quantify their fatigue level, comfort, pain or perceived training difficulty after training or at a specific time point, for example, using a Likert scale of 1-5 points or a visual analog scale (VAS). These subjective rating data are transmitted back to the system through the communication module.

[0088] Based on this, the plan generation module is configured with intensity adjustment rules. These rules can be a pre-defined lookup table, a fuzzy logic-based inference system, or a model trained through machine learning. For example, when the user's subjective state rating is "very tired" and the training difficulty is high, the rule might instruct that intensity parameters in the personalized breathing training plan, such as the target inspiratory volume and inspiratory hold time, be reduced by a certain percentage. Conversely, if the user's rating is "good" and the training difficulty is low, it might suggest appropriately increasing the intensity parameters. The plan generation module dynamically adjusts the intensity parameters of the personalized breathing training plan within a pre-defined adjustment range based on the received user subjective state rating. This pre-defined adjustment range ensures the rationality and safety of the plan adjustment. For example, the target inspiratory volume can be adjusted within ±15% of the current value, and the inspiratory hold time can be adjusted within ±20%. The dynamic adjustment of intensity parameters can take effect immediately to guide the next training session. Alternatively, it can take effect after periodic evaluation. This application incorporates the user's subjective feelings into the adjustment mechanism of the personalized breathing training program, overcoming the limitations that may be caused by relying solely on objective physiological data. The connection between the interactive execution module and the user's mobile terminal enables the system to proactively provide training information and progress feedback to the user, and conveniently collect the user's subjective state scores, greatly enhancing the user's sense of participation and control over the rehabilitation process. The program generation module dynamically adjusts the intensity parameters of the training program based on these subjective scores, making the training program not only scientific and rigorous, but also in line with the user's real-time physical feelings and psychological state. This helps to avoid overtraining due to an overly aggressive program or undertraining due to an overly conservative program, ensuring that the training intensity is always in an optimal range that can effectively stimulate muscle recovery without causing discomfort, thereby significantly improving the user's compliance with training, optimizing the training effect of respiratory muscle function rehabilitation, and ultimately improving the overall rehabilitation experience.

Claims

1. A personalized training system for postoperative respiratory muscle function rehabilitation, characterized in that, include: The data acquisition module is used to acquire users' individual characteristic data and real-time respiratory physiological data; The feature analysis module, connected to the data acquisition module, is used to determine the user's clinical feature category based on the individual feature data, and to generate a recovery rhythm curve characterizing the dynamic changes in respiratory muscle function based on the real-time respiratory physiological data. The intelligent mapping library contains pre-stored standard recovery rhythm templates and training parameter benchmarks corresponding to different clinical feature categories; The scheme generation module is connected to the feature analysis module and the intelligent mapping library respectively. It is used to match and analyze the recovery rhythm curve with the corresponding standard recovery rhythm template, and call or generate a personalized breathing training scheme based on the matching results. An interactive execution module, connected to the scheme generation module, is used to output and guide the user to execute the personalized breathing training scheme; An adaptive update module connects the feature analysis module and the intelligent mapping library to monitor the deviation of the recovery rhythm curve from the standard recovery rhythm template. When the deviation meets preset conditions, a user-specific rhythm template is generated and the intelligent mapping library is updated. The communication module is used to enable bidirectional data transmission between the system and medical terminals and user mobile terminals; A breathing training carrier is used to provide users with a physical interface for breathing training. The sensor components used to collect respiratory flow field signals in the data acquisition module are integrated into the breathing training carrier.

2. The personalized training system for postoperative respiratory muscle function rehabilitation according to claim 1, characterized in that, The individual characteristic data includes the type of surgery, postoperative days and preoperative baseline lung function values, and the real-time respiratory physiological data includes inspiratory volume, peak inspiratory flow rate and respiratory rhythm stability parameters. The data acquisition module includes a physiological parameter sensor and a data preprocessing unit. The physiological parameter sensor is integrated into the breathing interface of the breathing training carrier. The data preprocessing unit is used to filter, calibrate, and align the acquired raw individual characteristic data and real-time respiratory physiological data with timestamps. The processed raw individual characteristic data and real-time respiratory physiological data are synchronously transmitted to the feature analysis module.

3. The personalized training system for postoperative respiratory muscle function rehabilitation according to claim 2, characterized in that, The feature analysis module calculates the fit between the recovery rhythm curve and the preset trend model. When the fit is lower than the preset standard, the data acquisition module is triggered to re-acquire data. The recovery rhythm curve and the clinical feature category determination results are simultaneously sent to the protocol generation module and the adaptive update module.

4. A personalized training system for postoperative respiratory muscle function rehabilitation according to claim 1 or 3, characterized in that, The intelligent mapping library stores a standard recovery rhythm template and training parameter benchmarks that are uniquely associated with each clinical feature category. The training parameter benchmarks include target inspiratory volume, inspiratory hold time, and number of training sets per day. The intelligent mapping library supports template updates. The adaptive update module counts the frequency of successful application of user-specific rhythm templates. When the frequency exceeds a preset adoption threshold, the original standard recovery rhythm template of the corresponding category is replaced or supplemented with the user-specific rhythm template.

5. The personalized training system for postoperative respiratory muscle function rehabilitation according to claim 4, characterized in that, The matching analysis of the scheme generation module includes calculating the Euclidean distance between the feature vectors of the recovery rhythm curve and the standard recovery rhythm template as the difference degree. Based on the preset numerical range of the difference degree, the specific parameters of the personalized breathing training scheme are determined by directly calling the benchmark parameters, adjusting the benchmark parameters proportionally, or generating new parameters based on historical data.

6. The personalized training system for postoperative respiratory muscle function rehabilitation according to claim 5, characterized in that, The personalized breathing training program is configured to include a progressively advancing activation phase, enhancement phase, and consolidation phase. The program generation module determines the current phase based on the number of days after surgery and the recovery rhythm curve in the individual characteristic data. The training parameters for each phase are set according to their preset phase goals and the trend slope of the corresponding indicators in the recovery rhythm curve.

7. The personalized training system for postoperative respiratory muscle function rehabilitation according to claim 6, characterized in that, The preset conditions include a deviation threshold and the number of consecutive exceedances. When the real-time deviation between the recovery rhythm curve and the standard recovery rhythm template reaches the deviation threshold and the number of consecutive monitoring cycles reaches the number of consecutive exceedances, the adaptive update module triggers the generation of a user-specific rhythm template. The deviation threshold and the number of consecutive exceedances are set based on historical application data statistics of the training parameter benchmark.

8. The personalized training system for postoperative respiratory muscle function rehabilitation according to claim 7, characterized in that, The adaptive update module updates the intelligent mapping library in an incremental manner, storing only the feature parameters in the user-specific rhythm template whose difference from the existing standard template exceeds a preset difference threshold for the corresponding clinical feature category as an auxiliary template associated with the corresponding clinical feature category. When the clinical feature category determination result of the user matches the category, the scheme generation module calls both the standard template and the auxiliary template during the matching analysis.

9. A personalized training system for postoperative respiratory muscle function rehabilitation according to claim 8, characterized in that, After each update of the recovery rhythm curve, the feature analysis module sends a trigger signal containing new curve data to the scheme generation module. The scheme generation module then recalculates the matching difference based on the new curve. If the change in training parameters indicated by the old and new matching results exceeds a preset adjustment threshold, a scheme update instruction is generated and sent to the interactive execution module. The interactive execution module then notifies the user to adjust the scheme, and the updated scheme data is synchronously fed back to the adaptive update module.

10. A personalized training system for postoperative respiratory muscle function rehabilitation according to claim 9, characterized in that, The feature analysis module is equipped with a respiratory function abnormality warning threshold preset based on clinical safety data. When any real-time respiratory physiological data continuously exceeds the warning threshold for a preset duration, the feature analysis module triggers a warning signal. The warning signal synchronously controls the interactive execution module to issue an audio-visual alarm and suspend training, and controls the adaptive update module to suspend the update process and send the abnormal data packet to the medical terminal through the communication module. The interactive execution module establishes a connection with the user's mobile terminal through the communication module, and is used to send training reminders and recovery progress reports to the mobile terminal, and receive user subjective status scores from the mobile terminal. The scheme generation module is configured with intensity adjustment rules, which are used to dynamically adjust the intensity parameters of the personalized breathing training scheme within a preset adjustment range based on the received user subjective state score.

Images