Respirator parameter analysis method and device based on multi-dimensional treatment data

Abstract

The invention discloses a breathing machine parameter analysis method and device based on multi-dimensional treatment data, and relates to the technical field of medical instruments. The method comprises the following steps: acquiring real-time treatment data of the breathing machine; comparing the real-time treatment data with each preset rule and each incidence relation, determining a set diagnosis result of the breathing machine according to a rule matching result and a historical mode matching result, and if the set diagnosis result is unreasonable, executing a self-test process on the breathing machine to obtain a self-test result; if the self-test result is unreasonable, performing multi-round iterative adjustment on key parameters of the breathing machine based on a preset numerical value adjustment strategy, and determining an accumulated reward corresponding to each iteration; and taking the parameter scheme of the breathing machine when the accumulated award is maximum as an optimal parameter adjustment scheme corresponding to the breathing machine, and pushing the optimal parameter adjustment scheme to the user. According to the method, the unreasonable parameter setting can be identified, the adjustment suggestion can be generated, the problem of unreasonable parameter setting can be found and solved in time, and the use effect of the breathing machine is improved.

Description

Technical Field

[0001] This application relates to the field of medical device technology, and in particular to a method and apparatus for analyzing ventilator parameters based on multi-dimensional treatment data. Background Technology

[0002] With the development of technology, ventilators now have a variety of treatment parameter settings, covering therapeutic parameters such as ventilation mode settings, treatment pressure, comfort parameters such as humidity, preheating, and delayed pressure increase, as well as general parameters such as alarms for excessive air leakage and low respiratory rate. However, in actual use, these settings are often misused or forgotten to be activated, resulting in a significant reduction in the effectiveness of the ventilator.

[0003] Currently, ventilator parameter settings primarily rely on physician guidance. Physicians, using their professional knowledge and clinical experience, adjust the parameters according to general clinical procedures based on the patient's condition. While some ventilators have built-in preset parameters, these are generic and do not adequately consider individual patient differences. Although physicians provide guidance during ventilator use, they sometimes operate based on general settings, making it difficult to provide appropriate parameter settings tailored to each patient's historical and current condition, and lacking comprehensive, dynamic data analysis support. Summary of the Invention

[0004] The purpose of this application is to address at least one of the aforementioned technical deficiencies.

[0005] On the one hand, embodiments of this application provide a method for analyzing ventilator parameters based on multi-dimensional treatment data, the method comprising: Acquire real-time treatment data from the ventilator, including treatment curve data, parameter setting data, alarm event records, and basic patient information; Real-time treatment data is compared with each preset rule and each correlation to obtain rule matching results and historical pattern matching results. Each preset rule consists of parameter setting range and adjustment suggestions, and each correlation consists of curve data, parameter setting range and problem type. The ventilator's setting diagnosis result is determined based on the rule matching result and the historical pattern matching result. If the setting diagnosis result is unreasonable, the ventilator will perform a self-test process to obtain the self-test result. If the self-test results are unreasonable, the key parameters of the ventilator will be adjusted in multiple rounds based on the preset numerical adjustment strategy, and the cumulative reward corresponding to each iteration will be determined. The parameter scheme of the ventilator when the cumulative reward is maximized is taken as the optimal parameter adjustment scheme for the ventilator, and the optimal parameter adjustment scheme is pushed to the user through the human-computer interaction interface.

[0006] Optionally, each association is determined in the following ways: Collect historical multi-dimensional treatment data during ventilator operation. The historical multi-dimensional treatment data includes historical treatment curve data, historical parameter setting records, and historical alarm event logs. Calculate the similarity between different historical treatment curve data, and cluster the parameter setting problem based on the similarity between different historical treatment curve data to obtain each association.

[0007] Optionally, real-time treatment data can be compared with each preset rule and each association relationship to obtain rule matching results and historical pattern matching results, including: The parameter setting data is compared with the parameter setting range in each preset rule to obtain the rule matching result; Calculate the DTW similarity between the treatment curve data and the curve data in each association to obtain the DTW similarity between the treatment curve data and each curve data. Based on the preset similarity threshold and the DTW similarity between the treatment curve data and each curve data, determine the target curve data from each association. The parameter setting data is compared with the parameter setting range corresponding to the target curve data to obtain the historical pattern matching results.

[0008] Optionally, the diagnostic results can be configured to include diagnostic questions. The ventilator will then perform a self-test procedure and obtain the self-test results in the following manner: Based on the diagnostic results, the preset rules, and the correlations, determine the baseline parameter adjustment plan; The parameters of the ventilator are adjusted according to the baseline parameter adjustment scheme to obtain the adjusted ventilator. The adjusted ventilator is then run for the set duration to obtain the self-test results.

[0009] Optionally, run the adjusted ventilator for the set duration and obtain self-test results, including: Collect the test treatment data corresponding to the operation of the adjusted ventilator, and clean and extract features from the test treatment data to obtain the processed test treatment data; Identify the key assessment indicators corresponding to the diagnostic problem, and determine the first target data corresponding to the key assessment indicators in the processed test and treatment data, and the second target data corresponding to the key assessment indicators in the patient's physiological data; Obtain the target improvement requirements corresponding to the target data, and determine the self-test results based on the first target data, the second target data, and the target improvement requirements.

[0010] Optionally, the key parameters of the ventilator may be iteratively adjusted at least once based on a preset numerical adjustment strategy, including: Determine key parameters from the ventilator monitoring parameters based on the diagnostic problem; The key parameters of the ventilator are adjusted at least once according to the numerical adjustment strategy to obtain at least one adjusted key parameter. The ventilator is then run based on at least one adjusted key parameter. The numerical adjustment strategy is a step-by-step adjustment strategy, with each adjustment step being smaller than the parameter adjustment range during the first self-test.

[0011] Optionally, determine the cumulative reward for each iteration, including: Data on key assessment indicators corresponding to patients were collected when the ventilator was running based on key parameters after each adjustment, and at least one test physiological indicator data was obtained. A reward model is constructed based on at least one test physiological indicator data, and the cumulative reward for each iteration is determined based on the reward model.

[0012] Optionally, a reward model is constructed based on at least one test physiological indicator data, and the cumulative reward for each iteration is determined based on the reward model, including: For each iteration, a positive reward value is determined based on the degree of improvement in the physiological indicator data after each iteration, and a negative penalty value is determined based on the amount of deterioration in safety indicators and the number of alarm events triggered. The cumulative reward for each iteration is obtained by summing the reward and penalty values ​​determined after each iteration.

[0013] Optionally, the method further includes: The optimal parameter adjustment scheme is updated as a new case into the preset rules and relationships.

[0014] Furthermore, embodiments of this application provide a ventilator parameter analysis device based on multi-dimensional treatment data, comprising: The data aggregation module is used to acquire real-time treatment data from the ventilator, including treatment curve data, parameter setting data, alarm event records, and basic patient information. The rule base construction module is used to compare real-time treatment data with each preset rule and each correlation to obtain rule matching results and historical pattern matching results. Each preset rule consists of parameter setting range and adjustment suggestions, and each correlation consists of curve data, parameter setting range and problem type to form a mapping relationship. The machine learning model module is used to determine the ventilator setting diagnosis result based on the rule matching result and the historical pattern matching result. If the setting diagnosis result is unreasonable, the ventilator will perform a self-test process to obtain the self-test result. If the self-test result is unreasonable, the key parameters of the ventilator will be iteratively adjusted in multiple rounds based on the preset numerical adjustment strategy, and the cumulative reward corresponding to each iteration will be determined. The feedback module is used to take the parameter scheme of the ventilator when the cumulative reward is maximized as the optimal parameter adjustment scheme for the ventilator, and push the optimal parameter adjustment scheme to the user through the human-computer interaction interface.

[0015] In another aspect, embodiments of this application provide an electronic device, including a processor and a memory: The memory is configured to store machine-readable instructions that, when executed by the processor, cause the processor to perform any of the methods in a ventilator parameter analysis method based on multidimensional treatment data.

[0016] The beneficial effects of the technical solutions provided in this application include at least the following: This application, by collecting real-time treatment data from multi-dimensional ventilators, can fully consider individual patient differences and provide personalized parameter adjustment suggestions for each patient, thus solving the problem of lack of personalization in parameter settings.

[0017] Furthermore, this application utilizes intelligent analysis of multi-dimensional ventilator data to identify unreasonable parameter settings and generate adjustment suggestions. This allows for timely detection and resolution of parameter setting issues, preventing parameters from being forgotten and improving ventilator effectiveness. In addition, it dynamically correlates multi-dimensional ventilator data with ventilator parameters, constructs a parameter rationality rule base based on clinical guidelines, and combines this with machine learning models to comprehensively and deeply analyze the impact of parameter settings on sleep apnea events. This provides physicians with more scientific decision-making support, addressing the problem of insufficient comprehensive data support for physician decision-making. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 A flowchart illustrating a ventilator parameter analysis method based on multi-dimensional treatment data provided in this application embodiment; Figure 2A schematic diagram illustrating another method for analyzing ventilator parameters based on multi-dimensional treatment data provided in this application embodiment; Figure 3 A schematic diagram of a ventilator parameter analysis device based on multi-dimensional treatment data provided in this application embodiment; Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0020] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain this application, and should not be construed as limiting the invention.

[0021] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in this application means the presence of the stated features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. It should be understood that when we say an element is “connected” or “coupled” to another element, it can be directly connected or coupled to the other element, or there may be intermediate elements. Furthermore, “connected” or “coupled” as used herein can include wireless connections or wireless coupling. The term “and / or” as used herein includes all or any units and all combinations of one or more associated listed items.

[0022] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0023] Ventilators: As an effective means of artificially replacing spontaneous ventilation, ventilators are widely used in respiratory failure caused by various reasons, anesthetic respiratory management during major surgery, respiratory support therapy, and emergency resuscitation, occupying a very important position in the field of modern medicine. Ventilators are crucial medical devices that can prevent and treat respiratory failure, reduce complications, and save and prolong patients' lives. Currently, the setting of ventilator parameters mainly relies on the guidance of doctors. However, in practical applications, existing methods for adjusting ventilator parameters have the following shortcomings: 1. Lack of personalized parameter settings: In existing technical solutions, doctors mostly set ventilator parameters according to general settings or universal preset modes, making it difficult to provide appropriate parameter settings based on each patient's historical and current specific conditions. Different patients have different physical conditions, disease severity, physiological indicators, etc., and universal settings cannot meet personalized treatment needs, thus affecting treatment outcomes.

[0024] 2. Inappropriate or forgotten parameter settings: In actual use, many treatment parameters of the ventilator are often set inappropriately or forgotten to be turned on, and the existing technical solutions lack effective means to detect and correct these problems in a timely manner.

[0025] 3. Lack of comprehensive data support for doctors' decision-making: When guiding the setting of ventilator parameters, doctors often can only make judgments based on the patient's general condition and simple follow-up information. They lack comprehensive, dynamic, and multi-dimensional data support, making it difficult to accurately judge the impact of parameter settings on the patient's sleep breathing events.

[0026] 4. Lack of intuitive analysis and feedback: The existing parameter adjustment process lacks an intuitive feedback mechanism, making it difficult for doctors and patients to intuitively understand the parameter settings and adjustment suggestions.

[0027] Based on this, this application provides a method for analyzing ventilator parameters based on multi-dimensional treatment data, aiming to solve the above-mentioned technical problems of the prior art.

[0028] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0029] Specifically, such as Figure 1 As shown, the method may include: Step S101: Obtain real-time treatment data from the ventilator. The real-time treatment data includes treatment curve data, parameter setting data, alarm event records, and basic patient information.

[0030] Optionally, real-time treatment data from the ventilator can be collected. This real-time treatment data includes treatment curve data, parameter setting data, and alarm event records. Treatment curve data can be pressure-time curves, flow-time curves, blood oxygen saturation-time curves, etc. Parameter setting data can include data on all adjustable parameters such as pressure value, ventilation mode, IPAP (Inspiratory Positive Airway Pressure), EPAP (Expiratory Positive Airway Pressure), humidifier temperature, and delayed pressure increase. Alarm event records can include the number of "excessive leakage" alarms that occurred in the past hour. Basic patient information can include the patient's age, weight, and historical diagnostic results, such as obstructive sleep apnea, chronic obstructive pulmonary disease, etc.

[0031] This application, by collecting real-time treatment data from multi-dimensional ventilators, can fully consider individual patient differences and provide personalized parameter adjustment suggestions for each patient, thus solving the problem of lack of personalization in parameter settings.

[0032] Step S102: The real-time treatment data is compared with each preset rule and each correlation to obtain the rule matching result and the historical pattern matching result. Each preset rule consists of a parameter setting range and adjustment suggestions, and each correlation consists of a mapping relationship composed of curve data, parameter setting range and problem type.

[0033] Optionally, a rule base can be built based on medical guidelines and clinical expert experience. This rule base includes various preset rules, each of which includes range rules and logical rules. Range rules define the allowed value range for parameters. For example, the pressure range for CPAP (Continuous Positive Airway Pressure) mode is 4-20 cmH2O, while in BiPAP mode, IPAP should generally be greater than EPAP, with a difference of at least 4 cmH2O, and the humidification temperature range is 30-37°C. Logical rules define the constraints between parameters, such as: "If the patient is using a nasal mask and the leakage rate is consistently >30 L / min, the mask fitting should be checked or a mouth-nose mask should be considered," "If the oxygen saturation is consistently below 90%, the pressure setting may be insufficient," and "If the respiratory rate is consistently below 8 breaths / min, the trigger sensitivity may need to be adjusted," etc.

[0034] In optional embodiments of this application, each association is determined in the following manner: Collect historical multi-dimensional treatment data during ventilator operation. The historical multi-dimensional treatment data includes historical treatment curve data, historical parameter setting records, and historical alarm event logs. Calculate the similarity between different historical treatment curve data, and cluster the parameter setting problem based on the similarity between different historical treatment curve data to obtain each association.

[0035] Optionally, historical data of the ventilator can also be collected, specifically including historical treatment curve data, historical parameter setting data, historical alarm events, and historical labeled problem types. This historical data is then cleaned to obtain cleaned data. This data cleaning process may include removing outliers, handling missing values, and time alignment. Then, the mean, standard deviation, maximum, minimum, rising slope, falling slope, and area within a specific time period of each cleaned historical treatment curve are extracted as features. Further, the similarity between historical treatment curves is calculated based on these extracted features, and then clustering is performed based on this similarity. The similarity calculation can employ the DTW (Dynamic Time Warping) algorithm and the cosine similarity calculation method. The DTW algorithm is suitable for situations where the extracted feature lengths of historical treatment curves are inconsistent; it can calculate the optimal matching path between the extracted features of two historical treatment curves and obtain a distance value. The smaller the distance, the higher the similarity. The cosine similarity calculation method involves extracting features from historical treatment curve data and converting them into vectors, then calculating the cosine value of the angle between the two vectors. The closer the cosine value of the angle is to 1, the more similar they are.

[0036] Optionally, when clustering based on similarity, if the historical data has already been labeled with problem types, supervised learning methods can be directly used to assign the treatment curve data to the known problem types. For example, the K-nearest neighbors algorithm can be used to find the K closest historical curves based on similarity, and then the corresponding problem type can be determined by voting based on the problem types of these historical curves. If the problem types have not been labeled, unsupervised clustering can be used to divide the curves into several classes, and then experts in the field can label each class with a problem type. Subsequently, new curves can be assigned to the labeled classes.

[0037] Furthermore, for each category, i.e., problem type, the common parameter setting characteristics of all treatment curves within that category are analyzed. For example, the average value and range of pressure parameters, the average value and range of blood oxygen saturation, etc., are calculated for that category. Then, the typical shapes of the curves for that category are summarized, such as a pressure curve remaining at a consistently high level and a blood oxygen curve remaining at a consistently low level. This forms a characteristic description of that type of problem, and further establishes a mapping relationship between curve data, parameter setting ranges, and problem types, resulting in a relational database composed of these relationships.

[0038] Correspondingly, after obtaining real-time treatment data, the real-time treatment data can be compared with each preset rule and each correlation to obtain rule matching results and historical pattern matching results.

[0039] In optional embodiments of this application, real-time treatment data is compared with each preset rule and each association relationship to obtain rule matching results and historical pattern matching results, including: The parameter setting data is compared with the parameter setting range in each preset rule to obtain the rule matching result; Calculate the DTW similarity between the treatment curve data and the curve data in each association to obtain the DTW similarity between the treatment curve data and each curve data. Based on the preset similarity threshold and the DTW similarity between the treatment curve data and each curve data, determine the target curve data from each association. The parameter setting data is compared with the parameter setting range corresponding to the target curve data to obtain the historical pattern matching results.

[0040] Optionally, the obtained parameter setting data can be compared with the range rules one by one to obtain the rule matching result. For example, if the parameter setting data is IPAP=12 cmH2O, and the preset rule is "IPAP in BiPAP mode usually does not exceed 15 cmH2O", then the parameter setting data meets the range rule, and the rule matching result is passed; or if the parameter setting data is humidification temperature=28°C, and the preset rule is "humidification temperature must be ≥30°C", then the parameter setting data does not meet the range rule, and the rule matching result is failed.

[0041] Additionally, the DTW similarity between the treatment curve data and the curve data in each associated relationship can be calculated to obtain the DTW similarity between the treatment curve data and each curve data. Then, a preset similarity threshold is obtained, and the DTW similarity between the treatment curve data and each curve data is compared with the similarity threshold. Curve data with a DTW similarity greater than the preset threshold is identified as target curve data. Correspondingly, the acquired parameter setting data is compared with the parameter setting range corresponding to the target curve data to obtain historical pattern matching results. If the acquired parameter setting data and the parameter setting range corresponding to the target curve data are similar in value and combination, it indicates that the ventilator is currently reproducing the historical problem corresponding to the target curve data. In this case, the obtained historical pattern matching result is unqualified and includes the corresponding historical problem.

[0042] When calculating the DTW similarity between treatment curve data and curve data, the Euclidean distance is calculated pairwise between each data point on the treatment curve data and each data point on the curve data. Simply put, this calculates the difference between the value of the treatment curve data at time i and the value of the curve data at time j. After calculating the pairwise Euclidean distances between the data points, all distance values ​​form a two-dimensional matrix. Each cell (i, j) in the matrix stores the distance between treatment curve data point i and curve data point j. The smaller the distance, the closer the two points are numerically.

[0043] Furthermore, a path with the minimum cumulative distance is found from the top left corner to the bottom right corner of the two-dimensional matrix. This path represents the optimal matching scheme for the two curves, indicating that the i-th point on the treatment curve data matches the j-th point on the curve data. Then, the distance values ​​of all the cells traversed along this path are summed to obtain the total normalized distance, which directly measures the overall difference between the two curves in their optimal alignment state.

[0044] Furthermore, since the total normalized distance is an absolute value and not easily understood directly, it needs to be normalized to obtain the similarity score. F = 1 / (1 + D) Where F is the similarity and D is the total normalized distance. The value of the similarity is limited to between 0 and 1. The closer the value is to 1, the more similar the two curves are after alignment. The closer the value is to 0, the greater the difference between the two curves.

[0045] Step S103: Determine the ventilator setting diagnosis result based on the rule matching result and the historical pattern matching result. If the setting diagnosis result is unreasonable, execute the self-test process on the ventilator to obtain the self-test result.

[0046] Optionally, when obtaining rule matching results and historical pattern matching results, if the rule matching result is unreasonable but the historical pattern matching result does not find similar historical problems, the diagnosis result is still unreasonable; conversely, if the rule matching result is reasonable but the historical pattern matching result finds similar historical problems, the diagnosis result is also still unreasonable. Similarly, if both the rule matching result and the historical pattern matching result are marked as unreasonable, the diagnosis result is set to unreasonable and strongly marked as "unreasonable," indicating a high confidence level for the problem.

[0047] Furthermore, the questions marked as "unreasonable," their types, confidence levels, and related evidence are sent to the machine learning model module to trigger the ventilator's self-testing and optimization process, thereby obtaining the self-testing results.

[0048] In an optional embodiment of this application, the diagnostic results are set to include diagnostic questions, and the ventilator performs a self-test process in the following manner to obtain the self-test results: Based on the diagnostic results, the preset rules, and the correlations, determine the baseline parameter adjustment plan; The parameters of the ventilator are adjusted according to the baseline parameter adjustment scheme to obtain the adjusted ventilator. The adjusted ventilator is then run for the set duration to obtain the self-test results.

[0049] Optionally, after receiving the self-test optimization process trigger instruction, the machine learning model module determines the type of problem to be optimized, such as the problem type being that the pressure setting is too high, determines the safe adjustment range for the problem type, such as the IPAP allowable adjustment range of 8-16 cmH2O, with a minimum step unit of 0.5 cmH2O, determines the monitoring indicators corresponding to the problem type, such as blood oxygen saturation (SpO2), tidal volume, and leakage volume, and determines the duration of the self-test.

[0050] Furthermore, for the identified diagnostic problem, a baseline parameter adjustment scheme is determined from various preset rules and correlations. For example, for "pressure setting too high," the baseline parameter adjustment scheme might be to reduce IPAP from the current 12 cmH2O by 2 cmH2O to 10 cmH2O, while keeping other parameters unchanged. The generated baseline parameter adjustment scheme is then sent to the ventilator via a communication interface, such as Wi-Fi or Bluetooth. The ventilator parameters are adjusted according to the baseline parameter adjustment scheme to obtain the adjusted ventilator. The adjusted ventilator is then run for the set duration to obtain the self-test results.

[0051] In an optional embodiment of this application, the adjusted ventilator is run for a set duration to obtain self-test results, including: Collect the test treatment data corresponding to the operation of the adjusted ventilator, and clean and extract features from the test treatment data to obtain the processed test treatment data; Identify the key assessment indicators corresponding to the diagnostic problem, and determine the first target data corresponding to the key assessment indicators in the processed test and treatment data, and the second target data corresponding to the key assessment indicators in the patient's physiological data; Obtain the target improvement requirements corresponding to the target data, and determine the self-test results based on the first target data, the second target data, and the target improvement requirements.

[0052] Optionally, after the adjusted ventilator is in operation, corresponding test and treatment data are collected during the ventilator's operation. This test and treatment data includes treatment parameter data, physiological indicators, event data, and patient feedback. Among them, treatment parameters include actual operating pressure, flow rate, etc.; physiological indicators include blood oxygen saturation (SpO2), heart rate, respiratory rate, tidal volume, etc.; event data includes whether there are alarm events, such as excessive air leakage or apnea; and patient feedback is recorded as the patient's subjective comfort score.

[0053] Furthermore, key assessment indicators are selected based on the problem type. For example, for the problem of "pressure setting too high," key assessment indicators include blood oxygen saturation and tidal volume. It is necessary to determine whether blood oxygen saturation has increased while ensuring that tidal volume has not decreased significantly. Correspondingly, the first target data corresponding to the key assessment indicators is determined from the processed test and treatment data, and the second target data corresponding to the key assessment indicators is determined from the patient's physiological data obtained before the self-test. Then, the self-test result is determined based on the first target data, the second target data, and the corresponding improvement requirements. For example, if the key indicators improve, such as the average SpO2 increasing from 88% before the self-test to 92% after the self-test without adverse reactions, the self-test is considered successful; if the key indicators do not improve or new problems arise, such as low tidal volume, the self-test is considered unsuccessful.

[0054] In step S104, if the self-test result is unreasonable, the key parameters of the ventilator are adjusted in multiple rounds based on the preset numerical adjustment strategy, and the cumulative reward corresponding to each iteration is determined.

[0055] Optionally, if the self-test results are unreasonable, it means that the current ventilator parameters are not very suitable for the patient. In this case, the key parameters of the ventilator can be adjusted in multiple rounds based on the preset numerical adjustment strategy, and the cumulative reward corresponding to each iteration can be determined.

[0056] In optional embodiments of this application, the key parameters of the ventilator are adjusted iteratively for at least one round based on a preset numerical adjustment strategy, including: Determine key parameters from the ventilator monitoring parameters based on the diagnostic problem; The key parameters of the ventilator are adjusted at least once according to the numerical adjustment strategy to obtain at least one adjusted key parameter. The ventilator is then run based on at least one adjusted key parameter. The numerical adjustment strategy is a step-by-step adjustment strategy, with each adjustment step being smaller than the parameter adjustment range during the first self-test.

[0057] Optionally, to avoid too many interfering variables and focus on the core issue, key parameters can be determined from the ventilator's monitoring parameters based on the diagnostic problem. For example, select 1-2 key variables and 1-2 adjustment parameters that are most relevant to the diagnostic problem from all monitoring indicators as key parameters. Then, within a safe range, adjust the data of the key parameters of the ventilator at least once according to the numerical adjustment strategy to obtain at least one adjusted key parameter. Run the ventilator sequentially based on at least one adjusted key parameter and determine the improvement effect of each run to obtain the cumulative reward corresponding to each iteration.

[0058] In optional embodiments of this application, determining the cumulative reward corresponding to each iteration includes: Data on key assessment indicators corresponding to patients were collected when the ventilator was running based on key parameters after each adjustment, and at least one test physiological indicator data was obtained. A reward model is constructed based on at least one test physiological indicator data, and the cumulative reward for each iteration is determined based on the reward model.

[0059] Optionally, after running the ventilator based on the adjusted key parameters each time, data corresponding to key assessment indicators of the patient can be collected during ventilator operation to obtain the number of tested physiological indicators. Then, a reward model can be constructed based on the collected data of at least one tested physiological indicator, and the cumulative reward corresponding to each iteration can be determined based on the reward model. The reward model can be expressed by the following formula: in, To accumulate reward value, No. i One positive reward value, correspond The weighting coefficients, No. j The negative penalty value for each safety indicator or alarm event. correspond The penalty coefficient.

[0060] In an optional embodiment of this application, a reward model is constructed based on at least one number of test physiological indicators, and the cumulative reward corresponding to each iteration is determined based on the reward model, including: For each iteration, a positive reward value is determined based on the degree of improvement in the physiological indicator data after each iteration, and a negative penalty value is determined based on the amount of deterioration in safety indicators and the number of alarm events triggered. The cumulative reward for each iteration is obtained by summing the reward and penalty values ​​determined after each iteration.

[0061] Optionally, after each iteration, the degree of improvement of each key physiological indicator can be determined based on the change in the test physiological indicator data, and quantified into a corresponding positive reward value. The amount of deterioration of safety indicators and the amount of alarm events triggered can be used to determine negative penalty values. Then, the reward value and penalty value determined after each iteration are added into the formula of the reward model to obtain the cumulative reward corresponding to each iteration.

[0062] Step S105: The parameter scheme of the ventilator when the cumulative reward is maximized is taken as the optimal parameter adjustment scheme for the ventilator, and the optimal parameter adjustment scheme is pushed to the user through the human-machine interface.

[0063] Optionally, after determining the cumulative reward for each iteration, the parameter scheme of the ventilator when the cumulative reward is maximized can be used as the optimal parameter adjustment scheme for the ventilator, and the optimal parameter adjustment scheme can be pushed to the user through the human-computer interaction interface.

[0064] When the optimal parameter adjustment plan is pushed to the user through the human-computer interaction interface, the optimal parameter adjustment plan can be dynamically linked with the original data charts, such as respiratory waveforms and AHI (Apnea-Hypopnea Index) curves. It also supports clicking on the markers to view details and adjustment suggestions, realizing a closed loop of "data → analysis → feedback". This allows doctors and patients to understand the parameter settings more intuitively, making it easier to adjust and optimize them.

[0065] In optional embodiments of this application, the method further includes: The optimal parameter adjustment scheme is updated as a new case into the preset rules and relationships.

[0066] Optionally, after the parameter adjustment for the call is completed, the determined optimal parameter adjustment scheme can be updated into the preset rules and relationships as a new case for future matching and learning.

[0067] This application utilizes intelligent analysis of multi-dimensional ventilator data to identify unreasonable parameter settings and generate adjustment suggestions. This allows for timely detection and resolution of parameter setting issues, preventing parameters from being forgotten and improving ventilator effectiveness. Furthermore, it dynamically correlates multi-dimensional ventilator data with ventilator parameters, constructs a parameter rationality rule base based on clinical guidelines, and combines this with machine learning models to comprehensively and deeply analyze the impact of parameter settings on sleep apnea events. This provides physicians with more scientific decision-making support, addressing the problem of insufficient comprehensive data support for physician decision-making.

[0068] To better understand the methods provided in the embodiments of this application, the solutions provided in this application will be explained below. Figure 2The flowchart shown illustrates this process. Specifically, it may include: knowledge base construction, real-time data acquisition, intelligent analysis and detection, root cause analysis, focusing and strategy generation, simulation evaluation and learning, strategy optimization and decision-making, and output and execution.

[0069] The goal of the knowledge base construction is to establish the foundation for system analysis. It not only aggregates curves related to common respiratory problems but also integrates medical rules to form a "composite rule base," containing various preset rules and their relationships. This enables the system to both identify abnormal patterns and determine the clinical rationality of parameter settings. Real-time data acquisition involves the system continuously collecting respiratory monitoring data and patient vital signs data as real-time input for analysis. Intelligent analysis and detection inputs real-time data into the "composite rule base" to determine the rationality of current parameter settings. This is a prerequisite and extension for determining "whether a fault exists," first judging whether the "settings are optimal" before investigating "whether a fault exists." Root cause analysis involves initiating a parameter self-optimization process if the parameter settings are determined to be unreasonable. Focusing and strategy generation narrows the analysis scope for parameter problems and generates multiple feasible parameter adjustment strategies. Simulation evaluation and learning evaluate the potential effects of each strategy in a safe and controllable environment and establish a learning reward model to quantify "what kind of problem brings how much improvement." Strategy optimization and decision-making calculate and compare the cumulative rewards of adjustment strategies based on the reward model, automatically selecting the optimal personalized parameter adjustment scheme. Output and execution involve outputting the optimal treatment plan in the form of clear treatment recommendations for medical staff to review and implement.

[0070] Optionally, the system for implementing the method provided in the embodiments of this application may specifically include a data aggregation module, a rule base construction module, a machine learning model module, and a feedback module. The functions of each module will be described in detail below.

[0071] (1) Data aggregation module: Multi-dimensional treatment data is collected during ventilator operation, including treatment curve data, parameter setting data, alarm event records, and user feedback information. The similarity between different treatment curve data is calculated. For example, the similarity is determined based on the Dynamic Time Warping (DTW) algorithm or cosine similarity. Common parameter setting problems are clustered based on similarity. For example, the curve pattern of "high pressure accompanied by low blood oxygen saturation" is classified as "pressure setting too high". At the same time, a relationship library of "curve data-parameter setting-problem type" is constructed.

[0072] (2) Rule base construction module: Based on medical guidelines and clinical expert experience, reasonable ranges for ventilator parameters and configuration logic rules are predefined, such as "the pressure range for adult CPAP mode is usually 4-20 cmH2O" and "humidification temperature needs to be dynamically adjusted to 30-37℃ according to the ambient temperature." These rules are combined with curve-problem relationships generated by the data aggregation module to form a composite rule base containing standardized rules and historical problem cases. (3) Machine Learning Model Module: When the rule base detects that the current parameter settings are unreasonable, such as a parameter exceeding the predefined range or matching a problem pattern corresponding to a historical similar curve, a self-testing process is triggered—an optimized parameter setting combination is sent to the ventilator, such as adjusting the pressure value by ±2cmH2O and switching the ventilation mode, and treatment data during the self-testing period is collected, such as changes in blood oxygen saturation and patient respiratory mechanics parameters; if the self-testing results still show that the parameters are unreasonable, such as blood oxygen not improving at the target pressure, the monitored data category is narrowed down, and a numerical adjustment strategy is designed for the narrowed data category, such as gradually increasing or decreasing the pressure value by 0.5cmH2O; the improvement effect of each numerical adjustment sequence on the narrowed data category is further analyzed, such as a tidal volume increase of 15% after a pressure increase, and a reward model for spliced ​​data is generated based on the improvement effect.

[0073] (4) Feedback module: Construct a set of states corresponding to narrowed data categories and treatment data, such as "low pressure - high leakage rate state" and "high humidity - improved comfort state". Calculate the cumulative reward corresponding to different numerical adjustment sequences, and finally output the optimal parameter adjustment scheme, such as "adjust the current pressure from 8cmH2O to 10cmH2O and activate the delayed pressure boost function". Push feedback suggestions to users or medical staff through human-computer interaction interfaces, such as ventilator displays and mobile applications.

[0074] Accordingly, in order to make the execution process of the method provided in the embodiments of this application clearer, the method in this application will be described in detail below with reference to specific embodiments: Example 1: Automatic identification and clustering of problems caused by unreasonable parameter settings During ventilator operation, the data aggregation module collects treatment curve data in real time, including pressure-time curve, flow-time curve, and blood oxygen saturation-time curve; parameter setting records, such as the current ventilation mode being "Bilevel Positive Airway Pressure (BiPAP)", IPAP=12cmH2O, EPAP=4cmH2O, and humidity set to 50%; alarm event records, such as "excessive air leakage alarm triggered 3 times / hour", etc.

[0075] Furthermore, by calculating the similarity between the treatment curve data and 1000 curves in the historical database, it was found that the current curve had a similarity of 0.92 with the historical curve of "high IPAP pressure accompanied by low blood oxygen saturation," thus clustering this scenario into "pressure setting too high problem." At the same time, the rule base module retrieved predefined rules: "In BiPAP mode, IPAP is usually no more than 15 cmH2O, but it needs to be adjusted according to the patient's lung compliance; if blood oxygen saturation is consistently <90%, it is recommended to reduce IPAP or check the mask seal."

[0076] Example 2: Dynamic optimization process of machine learning module Upon identifying an "overly high pressure setting" issue, the machine learning module triggers a self-testing process: it issues temporary parameter adjustment instructions to the ventilator, such as reducing IPAP from 12 cmH2O to 10 cmH2O while maintaining EPAP at 4 cmH2O. Data is collected over 10 minutes during the self-test, such as an increase in blood oxygen saturation from 88% to 92% and a decrease in tidal volume from 450 ml to 400 ml, but still within a safe range. If blood oxygen saturation still does not meet the target after the self-test, such as only increasing to 91%, the monitoring data category is narrowed down to the key variable "IPAP-blood oxygen saturation," eliminating secondary variables such as ambient humidity. A numerical adjustment strategy is then designed: IPAP is gradually increased in increments of 0.5 cmH2O, successively trying 10.5 cmH2O, 11 cmH2O, and 11.5 cmH2O, recording the blood oxygen changes after each adjustment; for example, blood oxygen reaches 94% when IPAP=11 cmH2O. The improvement effect of each adjustment step was analyzed. For example, when IPAP increased from 10cmH2O to 11cmH2O, blood oxygen saturation increased by 3%, and the reward value increased by 1; when IPAP increased from 11cmH2O to 11.5cmH2O, blood oxygen saturation only increased by 1%, and the reward value increased by 0.2. A reward model was generated, and the total reward = 1 + 0.2 = 1.2. Finally, the feedback module constructed a state set of "moderate pressure (11cmH2O) - high blood oxygen (94%)", calculated the parameter combination with the highest cumulative reward, namely IPAP = 11cmH2O and EPAP = 4cmH2O, and pushed feedback to the user: "It is recommended to adjust IPAP from 12cmH2O to 11cmH2O. The current setting can significantly improve blood oxygen saturation and maintain stable tidal volume."

[0077] This application provides a ventilator parameter analysis device based on multi-dimensional treatment data, such as... Figure 3 As shown, the device may include a data aggregation module 301, a rule base construction module 302, a machine learning model module 303, and a feedback module 304, wherein... The data aggregation module is used to acquire real-time treatment data from the ventilator, including treatment curve data, parameter setting data, alarm event records, and basic patient information. The rule base construction module is used to compare real-time treatment data with each preset rule and each correlation to obtain rule matching results and historical pattern matching results. Each preset rule consists of parameter setting range and adjustment suggestions, and each correlation consists of curve data, parameter setting range and problem type to form a mapping relationship. The machine learning model module is used to determine the ventilator setting diagnosis result based on the rule matching result and the historical pattern matching result. If the setting diagnosis result is unreasonable, the ventilator will perform a self-test process to obtain the self-test result. If the self-test result is unreasonable, the key parameters of the ventilator will be iteratively adjusted in multiple rounds based on the preset numerical adjustment strategy, and the cumulative reward corresponding to each iteration will be determined. The feedback module is used to take the parameter scheme of the ventilator when the cumulative reward is maximized as the optimal parameter adjustment scheme for the ventilator, and push the optimal parameter adjustment scheme to the user through the human-computer interaction interface.

[0078] Optionally, each association is determined in the following ways: Collect historical multi-dimensional treatment data during ventilator operation. The historical multi-dimensional treatment data includes historical treatment curve data, historical parameter setting records, and historical alarm event logs. Calculate the similarity between different historical treatment curve data, and cluster the parameter setting problem based on the similarity between different historical treatment curve data to obtain each association.

[0079] Optionally, the rule base construction module, when comparing real-time treatment data with each preset rule and each association to obtain rule matching results and historical pattern matching results, is specifically used for: The parameter setting data is compared with the parameter setting range in each preset rule to obtain the rule matching result; Calculate the DTW similarity between the treatment curve data and the curve data in each association to obtain the DTW similarity between the treatment curve data and each curve data. Based on the preset similarity threshold and the DTW similarity between the treatment curve data and each curve data, determine the target curve data from each association. The parameter setting data is compared with the parameter setting range corresponding to the target curve data to obtain the historical pattern matching results.

[0080] Optionally, the diagnostic results can be configured to include diagnostic questions. The ventilator will then perform a self-test procedure and obtain the self-test results in the following manner: Based on the diagnostic results, the preset rules, and the correlations, determine the baseline parameter adjustment plan; The parameters of the ventilator are adjusted according to the baseline parameter adjustment scheme to obtain the adjusted ventilator. The adjusted ventilator is then run for the set duration to obtain the self-test results.

[0081] Optionally, when the machine learning model module runs the adjusted ventilator for a set duration and obtains self-test results, it is specifically used for: Collect the test treatment data corresponding to the operation of the adjusted ventilator, and clean and extract features from the test treatment data to obtain the processed test treatment data; Identify the key assessment indicators corresponding to the diagnostic problem, and determine the first target data corresponding to the key assessment indicators in the processed test and treatment data, and the second target data corresponding to the key assessment indicators in the patient's physiological data; Obtain the target improvement requirements corresponding to the target data, and determine the self-test results based on the first target data, the second target data, and the target improvement requirements.

[0082] Optionally, when the machine learning model module performs at least one round of iterative adjustments to the key parameters of the ventilator based on a preset numerical adjustment strategy, it is specifically used for: Determine key parameters from the ventilator monitoring parameters based on the diagnostic problem; The key parameters of the ventilator are adjusted at least once according to the numerical adjustment strategy to obtain at least one adjusted key parameter. The ventilator is then run based on at least one adjusted key parameter. The numerical adjustment strategy is a step-by-step adjustment strategy, with each adjustment step being smaller than the parameter adjustment range during the first self-test.

[0083] Optionally, the machine learning model module is specifically used to determine the cumulative reward for each iteration in the following ways: Data on key assessment indicators corresponding to patients were collected when the ventilator was running based on key parameters after each adjustment, and at least one test physiological indicator data was obtained. A reward model is constructed based on at least one test physiological indicator data, and the cumulative reward for each iteration is determined based on the reward model.

[0084] Optionally, when constructing a reward model based on at least one test physiological indicator and determining the cumulative reward for each iteration based on the reward model, the machine learning model module is specifically used for: For each iteration, a positive reward value is determined based on the degree of improvement in the physiological indicator data after each iteration, and a negative penalty value is determined based on the amount of deterioration in safety indicators and the number of alarm events triggered. The cumulative reward for each iteration is obtained by summing the reward and penalty values ​​determined after each iteration.

[0085] Optionally, the rule base building module is also used for: The optimal parameter adjustment scheme is updated as a new case into the preset rules and relationships.

[0086] The ventilator parameter analysis device based on multi-dimensional treatment data in this embodiment can execute the ventilator parameter analysis method based on multi-dimensional treatment data shown in the embodiment of this application. The implementation principle is similar and will not be described again here.

[0087] This application provides an electronic device, which includes a processor and a memory configured to store machine-readable instructions that, when executed by the processor, cause the processor to perform a ventilator parameter analysis method based on multi-dimensional treatment data.

[0088] This application provides an electronic device, such as... Figure 4 As shown, Figure 4 The illustrated electronic device includes a processor 2001 and a memory 2003. The processor 2001 and the memory 2003 are connected, for example, via a bus 2002. Optionally, the electronic device 2000 may further include a transceiver 2004. It should be noted that in practical applications, the transceiver 2004 is not limited to one type, and the structure of this electronic device 2000 does not constitute a limitation on the embodiments of this application.

[0089] Processor 2001 may be a CPU, a general-purpose processor, a DSP, an ASIC, an FPGA, or other programmable logic device, transistor logic device, hardware component, or any combination thereof. It may implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 2001 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0090] Bus 2002 may include a pathway for transmitting information between the aforementioned components. Bus 2002 may be a PCI bus or an EISA bus, etc. Bus 2002 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 4 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0091] The memory 2003 may be ROM or other type of static storage device capable of storing static information and instructions, RAM or other type of dynamic storage device capable of storing information and instructions, or EEPROM, CD-ROM or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto.

[0092] The memory 2003 stores the application code that executes the scheme of this application, and its execution is controlled by the processor 2001. The processor 2001 executes the application code stored in the memory 2003 to implement... Figure 3 The embodiment shown illustrates the operation of a ventilator parameter analysis device based on multi-dimensional treatment data.

[0093] It should be understood that although the steps in the flowcharts of the accompanying figures are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the accompanying figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0094] The above description is only a partial embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for analyzing ventilator parameters based on multi-dimensional therapy data, the method comprising: The application relates to a method for setting diagnosis of a ventilator, comprising the following steps: acquiring real-time treatment data of the ventilator, wherein the real-time treatment data comprises treatment curve data, parameter setting data, alarm event records and patient basic information; comparing the real-time treatment data with each preset rule and each association relationship respectively to obtain rule matching results and historical mode matching results, wherein each preset rule comprises a parameter setting range and an adjustment suggestion, and each association relationship comprises a mapping relationship between curve data, a parameter setting range and a problem type; determining a setting diagnosis result of the ventilator according to the rule matching results and the historical mode matching results, and executing a self-test process on the ventilator if the setting diagnosis result is unreasonable to obtain a self-test result; if the self-test result is unreasonable, iteratively adjusting key parameters of the ventilator based on a preset numerical adjustment strategy, and determining a cumulative reward corresponding to each iteration; taking a parameter scheme of the ventilator when the cumulative reward is maximum as an optimal parameter adjustment scheme corresponding to the ventilator, and pushing the optimal parameter adjustment scheme to a user through a man-machine interactive interface.

2. The method of claim 1, wherein, Each association relationship is determined by the following method: collecting historical multidimensional treatment data in the running process of the ventilator, wherein the historical multidimensional treatment data comprises historical treatment curve data, historical parameter setting records and historical alarm event logs; calculating the similarity between different historical treatment curve data, and clustering parameter setting problems based on the similarity between different historical treatment curve data to obtain each association relationship.

3. The method of claim 1, wherein, The comparison of the real-time treatment data with each preset rule and each association relationship to obtain rule matching results and historical mode matching results comprises the following steps: comparing the parameter setting data with the parameter setting range in each preset rule to obtain the rule matching results; calculating the DTW similarity between the treatment curve data and the curve data in each association relationship to obtain the DTW similarity between the treatment curve data and each curve data, and determining target curve data from each association relationship according to a preset similarity threshold and the DTW similarity between the treatment curve data and each curve data; comparing the parameter setting data with the parameter setting range corresponding to the target curve data to obtain the historical mode matching results.

4. The method of claim 1, wherein, The setting diagnosis result comprises a diagnosis problem, and the ventilator executes the self-test process to obtain the self-test result by the following method: determining a reference parameter adjustment scheme according to the diagnosis result, each preset rule and each association relationship; adjusting the parameters of the ventilator according to the reference parameter adjustment scheme to obtain an adjusted ventilator, and running the adjusted ventilator for a set time length to obtain the self-test result.

5. The method of claim 4, wherein, The running of the adjusted ventilator for a set time length to obtain the self-test result comprises the following steps: collecting test treatment data corresponding to the adjusted ventilator during running, and performing cleaning and feature extraction processing on the test treatment data to obtain processed test treatment data; determining a key evaluation index corresponding to the diagnosed problem, and determining first target data corresponding to the key evaluation index in the processed test treatment data and second target data corresponding to the key evaluation index in the patient physiological data; obtaining a target improvement requirement corresponding to the target data, and determining the self-test result based on the first target data, the second target data and the target improvement requirement.

6. The method of claim 4, wherein, The method further comprises: determining a key parameter of the ventilator from the monitoring parameters of the ventilator according to the diagnosed problem; adjusting the data of the key parameter of the ventilator at least once according to the numerical adjustment strategy to obtain at least one adjusted key parameter, and running the ventilator based on the at least one adjusted key parameter, wherein the numerical adjustment strategy is a step-by-step adjustment strategy, and the adjustment step length is smaller than the parameter adjustment amplitude in the first self-test.

7. The method of claim 6, wherein, The method further comprises: collecting data corresponding to the key evaluation index of the patient when the ventilator is running based on the adjusted key parameter to obtain at least one test physiological index data; constructing a reward model based on the at least one test physiological index data, and determining the cumulative reward corresponding to each iteration based on the reward model.

8. The method of claim 7, wherein, The method further comprises: for each iteration, determining a positive reward value according to the improvement degree of the test physiological index data after each iteration, and determining a negative penalty value according to the deterioration of the safety index and the triggering of the alarm event; accumulating the reward value and the penalty value determined after each iteration to obtain the cumulative reward corresponding to each iteration.

9. The method of claim 1, wherein, The method further comprises: updating the optimal parameter adjustment scheme as a new case to the preset rules and the association relationships.

10. A ventilator parameter analysis device based on multi-dimensional therapy data, characterized by, The method further comprises: a data aggregation module configured to obtain real-time treatment data of the ventilator, wherein the real-time treatment data comprises treatment curve data, parameter setting data, alarm event records and patient basic information; a rule library construction module configured to compare the real-time treatment data with each preset rule and each association relationship respectively to obtain rule matching results and historical pattern matching results, wherein each preset rule comprises parameter setting range and adjustment suggestion, and each association relationship comprises a mapping relationship composed of curve data, parameter setting range and problem type; a machine learning model module configured to determine a setting diagnosis result of the ventilator according to the rule matching results and the historical pattern matching results, and if the setting diagnosis result is unreasonable, performing a self-test process on the ventilator to obtain a self-test result; and if the self-test result is unreasonable, performing multiple rounds of iterative adjustment on the key parameter of the ventilator based on a preset numerical adjustment strategy, and determining the cumulative reward corresponding to each iteration. The feedback module is configured to determine a parameter scheme of the ventilator corresponding to an optimal parameter adjustment scheme of the ventilator when the cumulative reward is the largest, and push the optimal parameter adjustment scheme to a user through a human-computer interaction interface.

Images