Inflation control method for replaceable air bag type nasal cavity compressor

Abstract

The invention relates to the technical field of medical engineering, and discloses an inflation control method for a replaceable air bag type nasal cavity compressor, which comprises the following steps of: determining an inflation pressure strategy of a bleeding rate based on diagnosis and treatment data and a nasal cavity environment of an inflation channel; the inflation rate is determined based on the traversal result of the bleeding rate in the historical data set, when the inflation rate is determined as a model control strategy, an initial inflation control model is constructed based on the data sample set, the data sample set is changed according to the evaluation performance of the initial inflation control model, a target inflation control model is determined, and the inflation rate is output; when the characteristic data set and the data sample set are determined to be a comprehensive control strategy, the characteristic data set and the data sample set are combined, a comprehensive inflation control model is determined according to the combination result, and the inflation rate is output, and according to the dynamic determination of the inflation pressure strategy and the targeted training of the model, the stability and reliability of inflation rate control are ensured.

Description

Technical Field

[0001] This invention relates to the field of medical engineering technology, and more specifically, to an inflation control method for a replaceable cuff-type nasal compressor. Background Technology

[0002] In the process of hemostasis after nasal surgery, replaceable cuffed nasal compressors are widely used due to their advantages such as ease of operation and reusability. Replaceable cuffed nasal compressors work by inflating a cuff, using the pressure generated by the inflated cuff to compress the internal tissues of the nasal cavity, thereby achieving hemostasis.

[0003] However, the inflation operation of replaceable cuff nasal compressors currently still relies on human experience for control. Different users have individual differences in their nasal physiological structure, such as nasal cavity size and mucosal thickness. If the inflation rate is insufficient, the pressure generated by the cuff will be too small to achieve effective hemostasis or fixation, and may even require secondary treatment. If the inflation rate is excessive and the pressure is too high, it may cause excessive compression of the nasal mucosa, leading to adverse consequences such as mucosal damage. In addition, different people have different operating habits and judgment standards, resulting in inconsistent control of the inflation rate in different operating scenarios. Relying solely on experience to judge the inflation rate and adjust the inflation pressure is difficult to accurately match the actual compression needs.

[0004] Therefore, it is necessary to design an inflation control method for replaceable airbag nasal compressors to address the problems existing in the current technology. Summary of the Invention

[0005] In view of this, the present invention proposes an inflation control method for a replaceable airbag nasal compressor, which aims to solve the problem that individual differences in the nasal physiological structure of different users, as well as differences in the operating habits and judgment standards of different personnel, lead to inconsistent control of inflation rate in different operating scenarios, and that it is difficult to accurately match the actual compression needs by relying solely on experience to judge the inflation rate and change the inflation pressure.

[0006] This invention proposes an inflation control method for a replaceable cuff-type nasal compressor, comprising:

[0007] The bleeding volume per unit time is obtained and the bleeding rate is determined. An inflation pressure strategy for determining the bleeding rate is determined based on the diagnostic and treatment data and the nasal cavity environment of the inflation channel. The inflation pressure strategy includes a historical control strategy, a model control strategy, and a comprehensive control strategy.

[0008] When the historical control strategy is determined, the inflation rate is determined based on the bleeding rate in the historical dataset.

[0009] When the model control strategy is determined, a data sample set is acquired, an initial inflation control model is constructed based on the data sample set, and the data sample set is changed according to the evaluation performance of the initial inflation control model to determine the target inflation control model and output the inflation rate.

[0010] When a comprehensive control strategy is determined, a feature dataset is acquired, and the feature dataset and the data sample set are merged. Based on the merging result, a comprehensive inflation control model is determined and the inflation rate is output.

[0011] Furthermore, when the historical control strategy is determined, the inflation rate is determined based on the bleeding rate's traversal of the historical dataset, including:

[0012] When historical medical data that is consistent with the aforementioned medical data exists, it is determined to be the historical control strategy;

[0013] When no historical treatment data consistent with the treatment data is available, the model control strategy or comprehensive control strategy is determined based on the nasal cavity environment.

[0014] When the nasal resistance of the nasal cavity environment is within the standard nasal resistance range, it is determined to be the model control strategy.

[0015] When the nasal resistance of the nasal cavity environment is not within the standard nasal resistance range, it is determined to be the comprehensive control strategy.

[0016] Furthermore, when determining the inflation rate based on the bleeding rate through a historical dataset, the process includes:

[0017] The bleeding rate and nasal resistance are used as a diagnostic set. The historical dataset includes several historical diagnostic sets and historical inflation rates. The historical diagnostic sets include historical bleeding rates and historical nasal resistance, and each historical diagnostic set corresponds to a historical inflation rate.

[0018] When there is a historical treatment set in the historical dataset that is the same as the treatment set, the inflation rate is determined by the historical inflation rate corresponding to the historical treatment set.

[0019] When there is no historical treatment set in the historical dataset that is identical to the treatment set, the inflation rate is determined based on similarity.

[0020] Furthermore, when determining the inflation rate based on similarity, the following includes:

[0021] Obtain the similarity between the treatment set and each historical treatment set;

[0022] When there is no historical treatment set with a similarity greater than the similarity threshold, the inflation rate is determined based on the clustering results;

[0023] When there is a historical treatment set with a similarity greater than the similarity threshold, the inflation rate is determined based on the historical treatment set.

[0024] When the historical diagnosis set with a similarity greater than the similarity threshold is unique, the average of the historical inflation rates corresponding to several historical diagnosis sets is determined as the inflation rate.

[0025] When the historical diagnosis set with a similarity greater than the similarity threshold is not unique, the inflation rate is determined by the historical inflation rate corresponding to the historical diagnosis set.

[0026] Furthermore, when there is no historical treatment set with a similarity greater than the similarity threshold, determining the inflation rate based on the clustering results includes:

[0027] The diagnosis set and the historical dataset are used as the dataset to be clustered, and the historical inflation rate corresponding to each historical diagnosis set in the dataset to be clustered is extracted. The expected number of clusters k is determined to be 2, and the parameters of the Gaussian distribution are initialized.

[0028] The responsibility value is determined by calculating the probability that each data point in the dataset to be clustered belongs to each Gaussian distribution. Based on the responsibility value, the clustered dataset corresponding to the diagnosis set is obtained, and the mean of the historical inflation rate in the clustered dataset is determined as the inflation rate.

[0029] Furthermore, when constructing the initial inflation control model based on the data sample set, the following steps are included:

[0030] The data sample set is used to identify extreme sample data based on interquartile range, and the identified extreme sample data is deleted to determine the effective samples. The continuous data of the effective samples are normalized by min-max, and the classification feature data of the effective samples are one-hot encoded to construct a 16-dimensional nasal cavity feature vector. The inflation time series in the effective samples is used to derive features to construct an 8-dimensional nasal cavity temporal feature vector.

[0031] Based on the Pearson correlation coefficient and variance inflation factor, 20-dimensional sample feature vectors are selected from the 16-dimensional nasal cavity feature vector and the 8-dimensional nasal cavity temporal feature vector, and a sample feature matrix is ​​constructed. All sample feature matrices are divided into training set and test set according to the sampling ratio, and the initial inflation control model is constructed based on the training set and test set.

[0032] Furthermore, when constructing the initial inflation control model based on the training set and test set, the following steps are included:

[0033] A weight mapping table is constructed based on the sample features of the training set, and weights are assigned to the training set based on the weight mapping table. Bootstrap resampling is then performed on the training set after weight assignment to determine the target sample set.

[0034] A random forest model is pre-selected, and the target sample set is used as input to train the random forest model. When splitting nodes, high-weight features are split first to construct an initial decision tree. The initial decision tree is validated based on the test set to determine the model residual.

[0035] S1: Mark the target samples in the target sample set that are greater than the model residual threshold, and increase the splitting probability of the marked target samples when training the next decision tree;

[0036] S2: Record the current splitting features of the current node split and the historical splitting features of the previous node split, and determine the overlap between the current splitting features and the historical splitting features. Based on the overlap, switch the training feature type, and complete the training of the next decision tree according to the switching result.

[0037] S3: Repeat steps S1 to S3 until a preset number of decision trees are trained to form the initial inflation control model.

[0038] Furthermore, when modifying the data sample set based on the evaluation performance of the initial inflation control model, determining the target inflation control model, and outputting the inflation rate, the process includes:

[0039] The initial inflation control model was validated and the overall model deviation was determined based on the test set.

[0040] When the deviation of the integrated model is less than or equal to the deviation threshold of the integrated model, the performance of the initial inflation control model is determined to meet the standard, and the initial inflation control model is determined as the target inflation control model.

[0041] When the deviation of the integrated model is greater than the deviation threshold of the integrated model, it is determined that the performance of the initial inflation control model is not up to standard, and the training set is adjusted. The target inflation control model is determined based on the adjusted training set.

[0042] The inflation rate is determined based on the bleeding rate, nasal resistance, and the target inflation control model.

[0043] Furthermore, when adjusting the training set and determining the target inflation control model based on the adjusted training set, the process includes:

[0044] The training set is divided into several training subsets, and samples are drawn from the test set and several training subsets. The data features of the test set and each training subset are determined according to the drawn samples. The feature distribution is determined by probability density estimation based on the data features.

[0045] A loss function is constructed based on the feature distribution of the test set and each of the training subsets. The sampling probability of each of the training subsets is determined based on the loss function. Samples are extracted from each of the training subsets based on the sampling probability to determine a reconstructed training set. The target inflation control model is determined based on the reconstructed training set and the test set.

[0046] Furthermore, when merging the feature dataset and the data sample set, determining the integrated inflation control model based on the merging result, and outputting the inflation rate, the process includes:

[0047] When the target inflation control model is determined based on the training set and the test set, the feature dataset is supplemented to the training set and the test set according to the sampling ratio, and the target inflation control model is trained to determine the comprehensive inflation control model.

[0048] When the target inflation control model is determined based on the reconstructed training set and test set, the feature dataset is supplemented to the reconstructed training set and test set according to the sampling ratio, and the target inflation control model is trained to determine the comprehensive inflation control model.

[0049] The inflation rate is determined based on the bleeding rate, nasal resistance, and the integrated inflation control model.

[0050] Compared with existing technologies, the beneficial effects of this invention are as follows: It determines the corresponding inflation pressure strategy based on diagnostic data and the nasal cavity environment of the inflation channel; through a historical control strategy, it determines the inflation rate based on the results of traversing historical datasets using the bleeding rate; it can directly reuse effective parameters from historical experience, reducing inflation rate deviations caused by human operational differences, thus ensuring the stability and consistency of the inflation rate; the model control strategy is constructed and optimized through a data sample set to determine a universal target inflation control model and output the inflation rate, dynamically adapting to different nasal physiological structures, so that the inflation rate can meet the pressure requirements of the nasal cavity environment, avoiding the uncertainty and blindness of human experience, and achieving targeted and flexible adjustment of the inflation rate. The integrated control strategy combines feature datasets and data sample sets, taking into account both common patterns and individual differences, to construct a comprehensive inflation control model and output the inflation rate. This further enhances the dynamic adjustment capability of the inflation rate, thereby solving the problem of insufficient adaptation of general models to individual differences. It retains the group's patterns while improving the prediction accuracy of individual differences through feature datasets. The hierarchical control strategy realizes intelligent and personalized control of the inflation rate, ensuring the stability and reliability of the control. Attached Figure Description

[0051] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0052] Figure 1 This is a flowchart of an inflation control method for a replaceable airbag nasal compress provided in an embodiment of the present invention.

[0053] Figure 2 This is a schematic diagram of the replaceable airbag nasal compress provided in an embodiment of the present invention.

[0054] In the diagram: 1. Airbag; 2. Air tube; 3. Air delivery tube; 4. Buckle; 5. Air cylinder; 6. Adjusting block; 7. Connecting rod; 8. Air valve; 9. Air delivery tube; 10. Air pump. Detailed Implementation

[0055] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the disclosure to those skilled in the art. It should be noted that, unless otherwise specified, embodiments and features in the embodiments of the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0056] In some embodiments of this application, see Figure 1 As shown, an inflation control method for a replaceable cuff-type nasal compress includes:

[0057] S100: Obtain the amount of bleeding per unit time and determine the bleeding rate. The inflation pressure strategy for determining the bleeding rate is based on the diagnostic data and the nasal cavity environment of the inflation channel. The inflation pressure strategy includes historical control strategy, model control strategy and comprehensive control strategy.

[0058] S200: When the historical control strategy is determined, the inflation rate is determined based on the bleeding rate traversal results in the historical dataset.

[0059] S300: When the model control strategy is determined, a data sample set is acquired, an initial inflation control model is constructed based on the data sample set, and the data sample set is changed according to the evaluation performance of the initial inflation control model to determine the target inflation control model and output the inflation rate.

[0060] S400: When the integrated control strategy is determined, the feature dataset is acquired, and the feature dataset and the data sample set are merged. Based on the merging result, the integrated inflation control model is determined and the inflation rate is output.

[0061] Specifically, during compression, the bleeding rate is determined by measuring the amount of blood loss per unit time within the nasal cavity. The amount of blood loss per unit time is a fundamental indicator for quantifying the severity of bleeding, obtained through imaging techniques (such as ultrasound or CT scans) or miniature sensor devices. The bleeding rate, calculated from the amount of blood loss per unit time (e.g., ml / min), reflects the urgency of the bleeding. Inflation pressure strategies are selected based on clinical data, including reports on the user's nasal structure, nasal environment, and the use of a replaceable cuff-type nasal compressor. The inflation channel is used to place a replaceable cuff-type nasal compressor. When using a cuffed nasal compressor, the nasal cavity access is used to determine the inflation pressure strategy for bleeding rate based on clinical data and the nasal environment of the inflation channel. For example, for common bleeding rates and users who have used replaceable cuffed nasal compressors, a historical control strategy is selected; for rare bleeding rates or new users, a model control strategy is selected, using big data analysis to determine the appropriate inflation rate; and for users with special circumstances, such as young age or fragile mucosa, the control situation is relatively complex, and a comprehensive control strategy is adopted to accurately adapt to various situations and avoid the uncertainty and blindness of human experience.

[0062] Specifically, when a historical control strategy is adopted, cases with similar bleeding rates to the current rate are traversed in the historical dataset. This historical dataset contains relevant data such as bleeding rates and inflation rates from previous users. Relying on the direct reuse of historical experience ensures the consistency and stability of inflation rate control. When a model-based control strategy is adopted, a data sample set is acquired, and an initial inflation control model is constructed based on this set. The data sample set covers known data from multiple users, including their personal characteristics, nasal cavity structural parameters, and inflation process data. Personal characteristics include gender, age, and history of rhinitis / sinusitis. Nasal cavity structural parameters include the degree of nasal septum deviation and nasal mucosal thickness. The inflation process includes data on initial cuff pressure, the time series from inflation to compression pressure, inflation rate, and settling time. Using this data to construct the initial inflation control model ensures the model's generalization ability and data capture capability. The data sample set is modified based on the evaluation performance of the initial inflation control model. During model training, when the data sample set is sufficiently large, the data distribution needs to remain consistent to ensure better performance in real-world applications. Otherwise, it may introduce biases into the actual training results. Modifying the data sample set based on the initial inflation control model's evaluation performance, through a data-driven approach, eliminates biases during data testing. This allows for repeated training to determine a relatively stable target inflation control model that ultimately outputs the inflation rate, ensuring the stability of the model's predictions and the accuracy of inflation control. For users with unique circumstances, a feature dataset is acquired. This feature dataset reflects the user's personal data, such as weight, nasal cavity structure development, left and right nasal cavity volumes, nasal mucosal blood vessel distribution density, wound area and depth, etc. This data is unique to each user and generally lacks universality. Therefore, for users with special circumstances, the feature dataset and the data sample set are merged to ensure independent predictions for that specific user. This achieves a comprehensive inflation control model that combines common and individual characteristics for prediction, ensuring a precise match between the inflation rate and actual compression requirements.

[0063] Understandably, dynamically determining the inflation rate based on historical control strategies, model control strategies, and integrated control strategies reduces operational deviations and reliance on human experience and uncertainty in different scenarios. By combining data quantification, strategy adaptation, and data models, and using a data-driven approach to precisely control the inflation rate, the stability and reliability of the inflation rate are ensured.

[0064] In some embodiments of this application, when a historical control strategy is determined, the inflation rate is determined based on the bleeding rate traversal results in the historical dataset, including: when there is historical medical data consistent with the medical data, it is determined to be a historical control strategy; when there is no historical medical data consistent with the medical data, it is determined to be a model control strategy or a comprehensive control strategy based on the nasal environment; when the nasal resistance of the nasal environment is within the standard nasal resistance range, it is determined to be a model control strategy; when the nasal resistance of the nasal environment is not within the standard nasal resistance range, it is determined to be a comprehensive control strategy.

[0065] Specifically, the system prioritizes determining if there is historical treatment data consistent with the current treatment data (user's nasal structure report, nasal environment, and data on the use of a replaceable cuffed nasal compress, etc.). If such data exists, it indicates a high degree of similarity between the current bleeding situation and previous bleeding situations using a replaceable cuffed nasal compress. In this case, the historical control strategy is directly selected, and the inflation rate is quickly controlled by reusing the historical inflation rate. If no consistent historical treatment data exists, it indicates that no valid information exists for the user. In this case, nasal resistance is used as the basis for judging the complexity of the nasal environment. Nasal resistance reflects the resistance to nasal ventilation. The standard nasal resistance range refers to the normal resistance range of a healthy nasal cavity, which is 0.1 to 0.3 kPa·s / L. Nasal resistance is measured using equipment such as a nasal resistance meter. When the nasal resistance of the nasal environment is within the standard nasal resistance range, it indicates that the user's nasal cavity has a certain degree of universality, and a model control strategy is adopted. The inflation rate is effectively controlled through a big data model. When the nasal resistance of the nasal environment is not within the standard nasal resistance range, it indicates that the user's nasal environment is more complex and will produce individual differences due to the structure of the individual's nasal cavity. In this case, a comprehensive control strategy is adopted to avoid the error of a one-size-fits-all control approach.

[0066] Understandably, by dynamically selecting historical control strategies, model control strategies, and integrated control strategies, and prioritizing the reuse of historical data, complex calculations are avoided, and the response time for controlling the inflation rate is shortened. Nasal cavity complexity is categorized based on nasal resistance. Slightly simpler nasal environments utilize a universally applicable model to control the inflation rate, reducing resource consumption. Conversely, a comprehensive control strategy is employed for relatively complex nasal environments, improving the stability and adaptability of personalized inflation rate control.

[0067] In some embodiments of this application, when determining the inflation rate based on the bleeding rate through the historical dataset, the method includes: using the bleeding rate and nasal resistance as a diagnostic set; the historical dataset includes several historical diagnostic sets and historical inflation rates; the historical diagnostic sets include historical bleeding rates and historical nasal resistance, and each historical diagnostic set corresponds to a historical inflation rate; when there is a historical diagnostic set in the historical dataset that is the same as the diagnostic set, the inflation rate is determined based on the historical inflation rate corresponding to that historical diagnostic set; when there is no historical diagnostic set in the historical dataset that is the same as the diagnostic set, the inflation rate is determined based on similarity.

[0068] In some embodiments of this application, determining the inflation rate based on similarity includes: obtaining the similarity between the treatment set and each historical treatment set; when there is no historical treatment set with a similarity greater than a similarity threshold, determining the inflation rate based on the clustering result; when there is a historical treatment set with a similarity greater than a similarity threshold, determining the inflation rate based on the historical treatment set; when the historical treatment set with a similarity greater than the similarity threshold is unique, determining the average of the historical inflation rates corresponding to several historical treatment sets as the inflation rate; when the historical treatment set with a similarity greater than the similarity threshold is not unique, determining the inflation rate based on the historical inflation rate corresponding to that historical treatment set.

[0069] Specifically, the historical treatment sets in the historical dataset also consist of historical bleeding rates and historical nasal resistance, and each historical treatment set corresponds to a historical inflation rate. If the current treatment set is completely identical to a historical treatment set (its bleeding rate and nasal resistance match those of the historical bleeding rate and historical nasal resistance), the corresponding historical inflation rate is directly reused to ensure consistency and stability in inflation rate control. If no completely matching historical treatment set exists, the difference between the current treatment set and each historical treatment set is calculated using similarity. Similarity can be determined based on Euclidean distance and cosine similarity, etc. The historical treatment set with the highest similarity is selected to derive the inflation rate, avoiding... To mitigate the uncertainty and blind spots inherent in human experience, a similarity threshold of 0.8 is preferred. When the calculated similarities do not exceed the threshold, it indicates the absence of similar historical cases. In this case, clustering is used to determine the historical case set closest to the current case set. If a historical case set with a similarity exceeding the threshold exists, the number of historical case sets is further differentiated. If it is unique, its corresponding historical inflation rate is directly reused; otherwise, the average of these historical inflation rates is taken to balance the differences between different historical data, reduce random errors in similarity cases, ensure the accuracy of inflation rate control, and ultimately achieve a precise match between compression pressure and actual needs.

[0070] In some embodiments of this application, when there is no historical treatment set with a similarity greater than the similarity threshold, the method for determining the inflation rate based on the clustering results includes: using the treatment set and the historical dataset as the dataset to be clustered, extracting the historical inflation rate corresponding to each historical treatment set in the dataset to be clustered, determining the expected number of clusters k as 2, initializing the parameters of the Gaussian distribution, calculating the probability that each data in the dataset to be clustered belongs to each Gaussian distribution to determine the responsibility value, obtaining the clustering dataset corresponding to the treatment set based on the responsibility value, and determining the mean of the historical inflation rates in the clustering dataset as the inflation rate.

[0071] Specifically, the dataset to be clustered includes the current treatment set and all historical treatment sets in the historical dataset, and is associated with the corresponding historical inflation rates. The expected number of clusters is k=2, and Gaussian distribution parameters (mean, variance, weight) are initialized to describe the probability distribution characteristics of each data set. The responsibility value refers to the probability that a single data point belongs to a certain Gaussian distribution. Finally, based on the responsibility value of the current treatment set, the cluster dataset to which it belongs is determined, and the mean of the historical inflation rates within that cluster is taken as the inflation rate. This reduces the reliance on human experience and judgment, and reduces the human error in controlling the inflation rate. Even if the historical dataset has limited data, the mean within the cluster can be used to smooth individual differences. By analyzing the historical data through the clustering algorithm, the cluster dataset that is closest to the current conditions is found, thus improving the accuracy of the inflation rate determination.

[0072] In some embodiments of this application, the process of constructing an initial inflation control model based on a data sample set includes: identifying extreme sample data in the data sample set according to the interquartile range, deleting the identified extreme sample data to determine valid samples, applying min-max normalization to the continuous data of the valid samples, performing one-hot encoding on the classification feature data of the valid samples, constructing a 16-dimensional nasal cavity feature vector, performing feature derivation on the inflation time series in the valid samples to construct an 8-dimensional nasal cavity temporal feature vector, selecting 20-dimensional sample feature vectors from the 16-dimensional nasal cavity feature vector and the 8-dimensional nasal cavity temporal feature vector based on the Pearson correlation coefficient and variance inflation factor, constructing a sample feature matrix, dividing all sample feature matrices into a training set and a test set according to the sampling ratio, and constructing an initial inflation control model based on the training set and the test set.

[0073] Specifically, the data sample set encompasses known data from multiple users, including their personal characteristics, nasal cavity structural parameters, and inflation process data. Personal characteristics include gender, age, and history of rhinitis / sinusitis. Nasal cavity structural parameters include the degree of nasal septum deviation and nasal mucosal thickness. The inflation process includes data on initial balloon pressure, the time series from inflation to compression pressure, inflation rate, and settling time. First, the interquartile range (IQR) is used to identify extreme sample data, avoiding interference from extreme data values. Extremely abnormal samples, accounting for less than 5%, are then removed to reduce the number of individuals. To mitigate the impact of fluctuations, an effective sample of feature data that aligns with the characteristics of most users is constructed. Then, min-max normalization is applied to continuous data such as age and nasal mucosal thickness to map them to the 0-1 range, thus eliminating dimensional differences. For example, unprocessed data for age (range 18-80 years) and mucosal thickness (range 1-5 mm) exhibit significant numerical variations, potentially causing the model to overemphasize features with larger values. After min-max normalization, both age and nasal mucosal thickness fall within the 0-1 range, ensuring the model can fairly learn the relationship between different continuous features and inflation rate. To avoid the interference of dimensions on feature importance, one-hot encoding is used for categorical feature data, including the degree of nasal septum deviation, gender, history of rhinitis / sinusitis, and incision type. One-hot encoding transforms categorical feature data into a numerical form that the model can recognize. The encoded data exists in a 0 / 1 format, allowing the model to resolve differences between categories. If the degree of deviation is represented by numbers 1-4 (1 = none, 4 = severe), the model might misjudge 4 as 4 times 1. One-hot encoding, however, uses independent dimensions, ensuring that each category feature has equal status, thus accurately reflecting qualitative differences between categories and ensuring that these features can be... The model learns effectively, thus constructing a 16-dimensional nasal cavity feature vector. This 16-dimensional nasal cavity feature vector is a digital representation of the user's nasal cavity structure and basic information. At the same time, the inflation time series in the effective samples are used to derive features, including the first derivative (to calculate the average inflation rate under the same time interval), the pressure curve inflection point (to identify the nodes of inflation rate change by the point where the second derivative is zero), and the fluctuation coefficient (the standard deviation / mean of the inflation pressure). Finally, an 8-dimensional nasal cavity temporal feature vector is generated, which is a quantitative description of the dynamic process of airbag inflation.

[0074] Understandably, while the matrix formed by combining the 16-dimensional nasal cavity feature vector and the 8-dimensional nasal cavity temporal feature vector can effectively train the model, the high dimensionality of the vectors and the presence of redundant features can interfere with the model. Therefore, a dual screening process using Pearson correlation coefficient and variance inflation factor (VIF) is employed to eliminate 4-dimensional feature vectors, resulting in a 20-dimensional sample feature vector. This 20-dimensional sample feature vector is the optimal feature vector obtained from the 16-dimensional nasal cavity feature vector and the 8-dimensional nasal cavity temporal feature vector. It retains valuable information while avoiding the interference of redundant features on the model, providing effective input for subsequent model training. Based on the 20-dimensional sample feature vector, a sample feature matrix is ​​constructed, determining a parameter set for training and testing the model. These sample feature matrices are then divided into training and testing sets according to a sampling ratio, typically 3:2, to ensure the model's generalization and data capture capabilities, laying a data foundation for precise control of the inflation rate.

[0075] In some embodiments of this application, the process of constructing an initial inflation control model based on a training set and a test set includes: constructing a weight mapping table based on the sample features of the training set, assigning weights to the training set based on the weight mapping table, performing bootstrap resampling on the weighted training set to determine the target sample set, pre-selecting a random forest model, using the target sample set as input to train the random forest model, and prioritizing splitting high-weight features to construct an initial decision tree when splitting nodes, and validating the initial decision tree based on the test set to determine the model residuals. S1: Marking target samples in the target sample set that are greater than the model residual threshold, and increasing the splitting probability of the marked target samples when training the next decision tree. S2: Recording the current splitting feature of the current node split and the historical splitting feature of the previous node split, and determining the overlap between the current splitting feature and the historical splitting feature, switching the training feature type based on the overlap, and completing the training of the next decision tree based on the switching result. S3: Repeating steps S1 to S3 until a preset number of decision trees are trained to form the initial inflation control model.

[0076] Specifically, sample features closely related to the risk level of controlling inflation rate, such as "degree of deviation," are selected from the sample feature matrix to construct a weight mapping table. For "degree of deviation," the weight of "no deviation" is set to 1.0, "mild deviation" to 1.05, and "moderate deviation" to 1.1 (moderate deviation alters nasal airflow dynamics, affecting inflation). The weight of "severe deviation" is set to 1.15 (severe deviation alters nasal cavity shape, increasing inflation difficulty and risk). Similarly, weights are dynamically assigned based on the different degrees of sample features, and bootstrap resampling is performed on the weighted training set. The bootstrap resampling is based on the weighted training set as the overall... The random forest model employs repeated sampling with replacement. Each sampling iteration constructs a new bootstrap sample set, the target sample set. During each sampling, a probability of selection is assigned to each sample based on a weighted mapping table, making it more likely to be included in the new bootstrap sample set across multiple sampling iterations. Multiple target sample sets are generated during the repeated sampling process, each containing a different proportion of sample features with higher weights. Therefore, the frequency of sample features with higher weights is higher than the random distribution of sample features in the unweighted training set. The target sample sets are used as input to train the random forest model. During the growth (training) of the decision tree, node splitting prioritizes higher-weight sample features based on the mean squared error (MSE). For example, when splitting a node, if it contains many highly skewed, high-weight sample features, the model training will tend to prioritize closely related features, such as mucosal thickness and fluctuation coefficient, as the splitting criterion, rather than lower-weight general features, such as gender, thus constructing the initial decision tree.

[0077] Understandably, the initial decision tree is validated using the test set to determine the model residual. The model residual is the difference between the initial decision tree's prediction on the test set and the actual inflation rate on the test set. Target samples in the test set that exceed the model residual threshold are selected, i.e., samples with higher weights. These high-weighted features are then labeled. When training the next decision tree, the splitting probability of these high-weighted features is increased, causing the next decision tree to pay more attention to these features when splitting nodes, thereby specifically optimizing prediction bias. The current splitting feature of the current node and the historical splitting features of the previous node are recorded. For example, time-series features with higher weights, such as the pressure inflection point time, are typically considered. Generally, if the current splitting feature overlaps with the historical splitting features of the previous node... If the degree exceeds approximately 40%, the training feature type is forcibly switched, such as from time-series features with higher weights to skewness features with higher weights. This prevents the tree group from over-relying on a certain type of feature during training, thereby improving the model's generalization ability. This continues until a preset number of decision trees are trained to form the initial inflation control model. The preset number of decision trees is determined by the number of target sample sets. If the number of target sample sets is large, the preset number of decision trees will be larger. Since each target sample set has different proportions of sample features with higher weights, the trained initial inflation control model will have differences in feature selection and node splitting, thus forming a universal and diverse model. This improves the accuracy and reliability of the model in predicting inflation rate and ensures the stability of inflation rate control.

[0078] In some embodiments of this application, when modifying the data sample set based on the evaluation performance of the initial inflation control model, determining the target inflation control model, and outputting the inflation rate, the process includes: verifying the initial inflation control model based on the test set to determine the comprehensive model deviation; when the comprehensive model deviation is less than or equal to the comprehensive model deviation threshold, the performance of the initial inflation control model is determined to be up to standard, and the initial inflation control model is determined as the target inflation control model; when the comprehensive model deviation is greater than the comprehensive model deviation threshold, the performance of the initial inflation control model is determined to be down to standard, and the training set is adjusted; the target inflation control model is determined based on the adjusted training set; and the inflation rate is determined based on the bleeding rate, nasal resistance, and the target inflation control model.

[0079] Specifically, the overall model deviation refers to the total residual of the initial inflation control model on the test set, reflecting the degree of deviation between the model's predicted inflation rate and the actual inflation rate. The overall model deviation threshold is the critical value for judging whether the model is qualified. Typically, the overall model deviation threshold is set to around 5-10%, but the specific value needs to be dynamically selected based on the deviation. If the overall model deviation is less than or equal to the overall model deviation threshold, it indicates that the model's prediction accuracy on the test set meets the standard, and the initial inflation control model is directly determined as the target inflation control model. If the overall model deviation is greater than the overall model deviation threshold, it indicates... The model's performance was insufficient, failing to adequately learn and capture the relationships between data on the target sample set during training. Therefore, the training set needed to be adjusted to redetermine the target sample set and ultimately determine the target inflation control model. This allowed for real-time input of bleeding rate and nasal resistance to determine the appropriate inflation rate. By rigorously screening models through a comprehensive model deviation threshold, inadequate models were avoided from making predictions, reducing control failures or insufficient control caused by inaccurate predictions. For cases where the deviation exceeded the standard, the training set was adjusted to supplement the corresponding data, improving the model's prediction accuracy for complex situations and achieving precise control of the inflation rate, thus ensuring the reliability of the inflation rate.

[0080] In some embodiments of this application, when adjusting the training set and determining the target inflation control model based on the adjusted training set, the process includes: dividing the training set into several training subsets, extracting samples from the test set and the several training subsets, determining the data features of the test set and each training subset based on the extracted samples, determining the feature distribution by performing probability density estimation based on the data features, constructing a loss function based on the feature distribution of the test set and each training subset, determining the extraction probability of each training subset based on the loss function, extracting samples from each training subset based on the extraction probability to determine the reconstructed training set, and determining the target inflation control model based on the reconstructed training set and the test set.

[0081] Specifically, to ensure the model performs well on the test set, the data in the training and test sets need to be consistent. Conversely, for the model to perform well in real-world applications, the distribution of the training and actual data needs to be consistent. Therefore, the selection of the training set has a significant impact on the learning and training effectiveness of the random forest model. If the training set already contains a large amount of data, simply adding more data can improve the model's performance, but each additional amount of data will increase the training time and resource consumption, and may also lead to data redundancy. Therefore, it is necessary to construct a completely new training set based on the existing training set, i.e., reconstruct the training set, and bring the feature distribution of the reconstructed training set as close as possible to that of the test set. The training set is divided into several training subsets using the K-means algorithm to ensure that the samples drawn from the training and test sets are as similar as possible in the feature space. Data features include the mean bleeding rate, variance of nasal resistance, and mucosal thickness distribution, which describe the statistical properties of the data. The samples drawn from the training subsets and the corresponding data features for each sample are random variables conforming to a certain probability distribution. Similarly, the samples drawn from the test set also conform to a certain probability distribution. Therefore, probability density estimation is performed based on the data features to determine the feature distribution. Gaussian functions and other methods are used to estimate the probability density distributions of the test set and each training subset. To make the feature distributions of the reconstructed training set and the test set as close as possible, a loss function is constructed as the squared modulus of the difference between the feature distributions of the reconstructed training set and the test set. The loss function is solved to determine the sampling probability of each training subset. The solution of the loss function is a quadratic function of the sampling probability, which can calculate the minimum solution. Therefore, there is an optimal estimate of the sampling probability. That is, samples are drawn from each training subset according to the solved sampling probability to determine the reconstructed training set. This solves the distribution offset between the training set and the test set. For example, rare data in the test set does not account for enough of the training set, thus strengthening the scarce but important data, thereby improving training efficiency and enhancing the generalization ability of the model.

[0082] Understandably, after determining the reconstructed training set, weights are still assigned based on the weight mapping table to ultimately determine the target inflation control model. The overall training process is consistent with the process of forming the initial inflation control model, and will not be repeated here. The target inflation control model is determined based on the reconstructed training and test sets. The reconstructed training set takes into account the diversity and specificity of the data, reducing overfitting and ensuring the accuracy of the output inflation rate under different nasal cavity environments. This allows for precise control of the inflation rate, avoiding the uncertainty and blind spots of human experience.

[0083] In some embodiments of this application, when merging the feature dataset and the data sample set, determining the comprehensive inflation control model based on the merging result, and outputting the inflation rate, the process includes: when the target inflation control model is determined based on the training set and the test set, the feature dataset is supplemented to the training set and the test set according to the sampling ratio, and the target inflation control model is trained to determine the comprehensive inflation control model; when the target inflation control model is determined based on the reconstructed training set and the test set, the feature dataset is supplemented to the reconstructed training set and the test set according to the sampling ratio, and the target inflation control model is trained to determine the comprehensive inflation control model, and the inflation rate is determined based on the bleeding rate, nasal resistance, and the comprehensive inflation control model.

[0084] Specifically, the feature dataset reflects the user's personal data, such as user weight, nasal cavity structure development, left and right nasal cavity volume, nasal mucosal blood vessel distribution density, wound area and depth, etc. This data is unique to each user and usually lacks universality. While established target inflation control models have universality and generality, the addition of the feature dataset allows the model to incorporate the user's personal data, thus addressing the problem of insufficient adaptation of general models to individual differences. This preserves the group patterns observed during training (ensuring the reliability of the model's foundation) while improving the prediction accuracy of individual differences through the feature dataset. During training, the target inflation control model is further trained based on the supplemented test set and the supplemented reconstructed training set or training set to determine the comprehensive inflation control model. This training aims to improve the overall performance of the model by simply using the target inflation control model to fit the supplemented feature dataset. This allows the comprehensive inflation control model to respond to both the common patterns of the population and the specific needs of individuals. The bleeding rate and nasal resistance are substituted into the comprehensive inflation control model to determine the inflation rate, thereby achieving precise control of the inflation rate and improving the adaptability and flexibility of the inflation rate to different nasal environments.

[0085] See Figure 2As shown, a replaceable airbag nasal compressor includes: an airbag 1 and an air delivery tube 9. An air inlet tube 2 and an air guide tube 3 are disposed within the inner cavity of the airbag 1. The air inlet tube 2 passes through the airbag 1 and ensures gas flow within the airbag 1. A buckle 4 is provided on the right side of the airbag 1 to secure the airbag 1, air inlet tube 2, and air guide tube 3. A sealing ring is provided on the left side of the buckle 4 to prevent gas leakage from the airbag 1. The air guide tube 3 is connected to an air delivery cylinder 5, which supplies gas to the air delivery tube 3. Connecting rods 7 are provided on the upper and lower sides of the air delivery cylinder 5. Adjusting blocks 6 are fixedly connected to the connecting rods 7. Pulling the connecting blocks moves the air guide tube 3, causing the buckle 4 to move the airbag 1. The movement is synchronized to pull out the airbag 1. The airbag 1 can be replaced by the buckle 4, so that it can still maintain a good working condition when the airbag 1 is used for nasal compression next time. The upper side of the air cylinder 5 is also equipped with an air valve 8, and the right side of the air cylinder 5 is connected to the air supply pipe 9. The end of the air supply pipe 9 away from the air cylinder 5 is connected to the air pump 10. The air pump 10 is adjusted accordingly by a certain inflation rate so that the air pump 10 delivers gas through the air supply pipe 9. During the delivery process, the air valve 8 is opened in advance so that the gas in the air supply pipe 9 can be delivered smoothly to the air cylinder 5, and then delivered to the airbag 1 along the air guide tube 3 to ensure the compression needs of different users.

[0086] In summary, the beneficial effects of this invention are as follows: It determines the corresponding inflation pressure strategy based on diagnostic data and the nasal cavity environment of the inflation channel; through a historical control strategy, it determines the inflation rate based on the bleeding rate traversal results of historical datasets, directly reusing effective parameters from historical experience, reducing inflation rate deviations caused by human operational differences, and ensuring the stability and consistency of the inflation rate; the model control strategy is constructed and optimized through a data sample set to determine a universal target inflation control model and output the inflation rate, dynamically adapting to different nasal physiological structures, so that the inflation rate can conform to the pressure requirements of the nasal cavity environment, avoiding the uncertainty and blindness of human experience, and achieving targeted and flexible adjustment of the inflation rate. The integrated control strategy combines feature datasets and data sample sets, taking into account both common patterns and individual differences, to construct a comprehensive inflation control model and output the inflation rate. This further enhances the dynamic adjustment capability of the inflation rate, thereby solving the problem of insufficient adaptation of general models to individual differences. It retains the group's patterns while improving the prediction accuracy of individual differences through feature datasets. The hierarchical control strategy realizes intelligent and personalized control of the inflation rate, ensuring the stability and reliability of the control.

[0087] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program goods. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program goods embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0088] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program goods according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0089] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0090] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0091] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for controlling the inflation of a replaceable cuff-type nasal compressor, characterized in that, include: The bleeding volume per unit time is obtained and the bleeding rate is determined. An inflation pressure strategy for determining the bleeding rate is determined based on the diagnostic and treatment data and the nasal cavity environment of the inflation channel. The inflation pressure strategy includes a historical control strategy, a model control strategy, and a comprehensive control strategy. When the historical control strategy is determined, the inflation rate is determined based on the bleeding rate in the historical dataset. When the model control strategy is determined, a data sample set is acquired, an initial inflation control model is constructed based on the data sample set, and the data sample set is changed according to the evaluation performance of the initial inflation control model to determine the target inflation control model and output the inflation rate. When a comprehensive control strategy is determined, a feature dataset is acquired, and the feature dataset and the data sample set are merged. Based on the merging result, a comprehensive inflation control model is determined and the inflation rate is output.

2. The inflation control method for a replaceable air-filled nasal compress according to claim 1, characterized in that, When the historical control strategy is determined, the inflation rate is determined based on the bleeding rate's traversal of the historical dataset, including: When historical medical data that is consistent with the aforementioned medical data exists, it is determined to be the historical control strategy; When no historical treatment data consistent with the treatment data is available, the model control strategy or comprehensive control strategy is determined based on the nasal cavity environment. When the nasal resistance of the nasal cavity environment is within the standard nasal resistance range, it is determined to be the model control strategy. When the nasal resistance of the nasal cavity environment is not within the standard nasal resistance range, it is determined to be the comprehensive control strategy.

3. The inflation control method for a replaceable cuff-type nasal compress according to claim 2, characterized in that, When determining the inflation rate based on the bleeding rate through a historical dataset, the following steps are included: The bleeding rate and nasal resistance are used as a diagnostic set. The historical dataset includes several historical diagnostic sets and historical inflation rates. The historical diagnostic sets include historical bleeding rates and historical nasal resistance, and each historical diagnostic set corresponds to a historical inflation rate. When there is a historical treatment set in the historical dataset that is the same as the treatment set, the inflation rate is determined by the historical inflation rate corresponding to the historical treatment set. When there is no historical treatment set in the historical dataset that is identical to the treatment set, the inflation rate is determined based on similarity.

4. The inflation control method for a replaceable cuff-type nasal compress according to claim 3, characterized in that, When determining inflation rate based on similarity, the following are included: Obtain the similarity between the treatment set and each historical treatment set; When there is no historical treatment set with a similarity greater than the similarity threshold, the inflation rate is determined based on the clustering results; When there is a historical treatment set with a similarity greater than the similarity threshold, the inflation rate is determined based on the historical treatment set. When the historical diagnosis set with a similarity greater than the similarity threshold is unique, the average of the historical inflation rates corresponding to several historical diagnosis sets is determined as the inflation rate. When the historical diagnosis set with a similarity greater than the similarity threshold is not unique, the inflation rate is determined by the historical inflation rate corresponding to the historical diagnosis set.

5. The inflation control method for a replaceable cuff-type nasal compress according to claim 4, characterized in that, When there is no historical treatment set with a similarity greater than the similarity threshold, the determination of the inflation rate based on the clustering results includes: The diagnosis set and the historical dataset are used as the dataset to be clustered, and the historical inflation rate corresponding to each historical diagnosis set in the dataset to be clustered is extracted. The expected number of clusters k is determined to be 2, and the parameters of the Gaussian distribution are initialized. The responsibility value is determined by calculating the probability that each data point in the dataset to be clustered belongs to each Gaussian distribution. Based on the responsibility value, the clustered dataset corresponding to the diagnosis set is obtained, and the mean of the historical inflation rate in the clustered dataset is determined as the inflation rate.

6. The inflation control method for a replaceable cuff-type nasal compress according to claim 5, characterized in that, When constructing the initial inflation control model based on the data sample set, the following are included: The data sample set is used to identify extreme sample data based on interquartile range, and the identified extreme sample data is deleted to determine the effective samples. The continuous data of the effective samples are normalized by min-max, and the classification feature data of the effective samples are one-hot encoded to construct a 16-dimensional nasal cavity feature vector. The inflation time series in the effective samples is used to derive features to construct an 8-dimensional nasal cavity temporal feature vector. Based on the Pearson correlation coefficient and variance inflation factor, 20-dimensional sample feature vectors are selected from the 16-dimensional nasal cavity feature vector and the 8-dimensional nasal cavity temporal feature vector, and a sample feature matrix is ​​constructed. All sample feature matrices are divided into training set and test set according to the sampling ratio, and the initial inflation control model is constructed based on the training set and test set.

7. The inflation control method for a replaceable cuff-type nasal compress according to claim 6, characterized in that, When constructing the initial inflation control model based on the training set and the test set, the following steps are included: A weight mapping table is constructed based on the sample features of the training set, and weights are assigned to the training set based on the weight mapping table. Bootstrap resampling is then performed on the training set after weight assignment to determine the target sample set. A random forest model is pre-selected, and the target sample set is used as input to train the random forest model. When splitting nodes, high-weight features are split first to construct an initial decision tree. The initial decision tree is validated based on the test set to determine the model residual. S1: Mark the target samples in the target sample set that are greater than the model residual threshold, and increase the splitting probability of the marked target samples when training the next decision tree; S2: Record the current splitting features of the current node split and the historical splitting features of the previous node split, and determine the overlap between the current splitting features and the historical splitting features. Based on the overlap, switch the training feature type, and complete the training of the next decision tree according to the switching result. S3: Repeat steps S1 to S3 until a preset number of decision trees are trained to form the initial inflation control model.

8. The inflation control method for a replaceable cuff-type nasal compress according to claim 7, characterized in that, When modifying the data sample set based on the evaluation performance of the initial inflation control model, determining the target inflation control model, and outputting the inflation rate, the process includes: The initial inflation control model was validated and the overall model deviation was determined based on the test set. When the deviation of the integrated model is less than or equal to the deviation threshold of the integrated model, the performance of the initial inflation control model is determined to meet the standard, and the initial inflation control model is determined as the target inflation control model. When the deviation of the integrated model is greater than the deviation threshold of the integrated model, it is determined that the performance of the initial inflation control model is not up to standard, and the training set is adjusted. The target inflation control model is determined based on the adjusted training set. The inflation rate is determined based on the bleeding rate, nasal resistance, and the target inflation control model.

9. The inflation control method for a replaceable cuff-type nasal compress according to claim 8, characterized in that, When adjusting the training set and determining the target inflation control model based on the adjusted training set, the process includes: The training set is divided into several training subsets, and samples are drawn from the test set and several training subsets. The data features of the test set and each training subset are determined according to the drawn samples. The feature distribution is determined by probability density estimation based on the data features. A loss function is constructed based on the feature distribution of the test set and each of the training subsets. The sampling probability of each of the training subsets is determined based on the loss function. Samples are extracted from each of the training subsets based on the sampling probability to determine a reconstructed training set. The target inflation control model is determined based on the reconstructed training set and the test set.

10. The inflation control method for a replaceable cuff-type nasal compress according to claim 9, characterized in that, When merging the feature dataset and data sample set, determining the integrated inflation control model based on the merging result, and outputting the inflation rate, the process includes: When the target inflation control model is determined based on the training set and the test set, the feature dataset is supplemented to the training set and the test set according to the sampling ratio, and the target inflation control model is trained to determine the comprehensive inflation control model. When the target inflation control model is determined based on the reconstructed training set and test set, the feature dataset is supplemented to the reconstructed training set and test set according to the sampling ratio, and the target inflation control model is trained to determine the comprehensive inflation control model. The inflation rate is determined based on the bleeding rate, nasal resistance, and the integrated inflation control model.

Images