CN114191665A

CN114191665A - Method and device for classifying man-machine asynchronous phenomena in mechanical ventilation process

Info

Publication number: CN114191665A
Application number: CN202111452509.6A
Authority: CN
Inventors: 马良; 仲为; 熊富海; 廖天正; 颜延; 李慧慧; 王磊
Original assignee: Shenzhen Institute of Advanced Technology of CAS
Current assignee: Shenzhen Institute of Advanced Technology of CAS
Priority date: 2021-12-01
Filing date: 2021-12-01
Publication date: 2022-03-18
Also published as: WO2023097780A1

Abstract

The invention discloses a classification method and a classification device for man-machine asynchronous phenomena in a mechanical ventilation process. The classification method comprises the following steps: acquiring real-time respiratory waveform data of an object to be detected in a mechanical ventilation process; extracting Poincare map features of the real-time respiratory waveform data; inputting the Poincare image characteristics into a pre-trained classification model, and outputting the human-computer asynchronous type corresponding to the real-time respiratory waveform data by the classification model. The method extracts Poincare graph characteristics from an original waveform, does not depend on other factors except the waveform to better reflect information such as form change of the waveform, obtains a classification model for various man-machine asynchronous classification tasks in various mechanical ventilation modes from training, and finally realizes accurate classification of real-time respiratory waveform data.

Description

Method and device for classifying man-machine asynchronous phenomena in mechanical ventilation process

Technical Field

The invention belongs to the technical field of electrophysiological detection monitoring, and particularly relates to a classification method, a classification device, a computer readable storage medium and computer equipment for man-machine asynchronous phenomena in a mechanical ventilation process.

Background

Ventilators play an extremely important role in life support systems as important equipment in Intensive Care Units (ICUs) of various hospitals. The interactive process of a ventilator with a patient suffering from respiratory disease is called Mechanical Ventilation (MV), and the process has two main modes: 1) a control mode, in which a patient is often sedated or anesthetized, in which the breathing rhythm of the patient is controlled by a ventilator, such as Pressure Controlled Ventilation (PCV) and Volume Controlled Ventilation (VCV); 2) an assist mode, such as Pressure Support Ventilation (PSV), in which the patient has a certain level of respiratory effort and the ventilator functions to assist the patient in breathing, thus reducing the patient's work of breathing. In either case, however, the interaction between the two is not so smooth, and some degree of incompatibility occurs, which is commonly referred to as human-machine asynchrony (PVA).

Since different patients have different conditions, even the same patient may have different conditions with the passage of time during the mechanical ventilation process, in the decades of the development of mechanical ventilation, various types of man-machine asynchrony have been discovered up to now, such as common Double Triggering (DT), Ineffective Effort (IE), advanced switching (PC), delayed switching (DC), and Reversed Triggering (RT), etc., and fig. 1 shows the four typical waveforms of the common asynchrony. With years of observation and research in the medical field, it is found that human-computer asynchronism often brings about increasing probability of lung injury induced by a breathing machine, prolonging hospitalization time and even increasing death rate in ICU of patients. Therefore, the asynchronous identification and management of man-machine is continuously attracting research interest from clinicians to related professionals.

The most common method in clinic is an observation method, namely a bedside doctor observes ventilation waveforms on a respirator screen, including airway pressure time waveforms, flow rate time waveforms and tidal volume time waveforms, to judge which kind of asynchrony occurs at the end, and then makes corresponding coping actions according to professional training of the doctor. However, there are also studies to use rule-based algorithms, i.e. combine clinical expertise and metadata in mechanical ventilation to formulate relevant detection rules, such as calculating the ratio of tidal volume during expiration to tidal volume during expiration, the ratio of inspiration time to expiration time, etc.

Since observation is often time consuming and requires a high level of expertise on the part of the caregiver, a trained person is required to quickly and accurately identify whether the currently-facing respiratory cycle contains asynchronous waveforms. However, the rule-based method often needs to set a threshold according to the professional knowledge of clinical staff, and the selection of the threshold can only be used as a decision basis for the current sample, and cannot flexibly adapt to more scenes.

At present, the academic community focuses on the study of man-machine asynchrony in algorithms based on machine learning and deep learning, namely, a large amount of respiratory waveform data are divided according to respiratory cycles, then characteristics are extracted from the respiratory waveform data, the characteristics are used as the input of an algorithm model, and finally the model learns the capability of identifying various man-machine asynchrony by using the advantages of the existing computing resources and supervised learning.

The following is a current investigation of relevant research in this area:

1. makowa et al, in the patent (CN201910149798.9) A mechanical ventilation man-machine asynchronous detection method based on a recurrent neural network, mention that the algorithm of the recurrent neural network is used to detect man-machine asynchrony, GRUs (gated cyclic units) of two channels respectively extract pressure waveform characteristics and flow velocity time waveform characteristics, then after the two characteristics are fused, BGRUs (bidirectional gated cyclic neural units) are used to extract higher dimensional characteristics, and finally softmax full-connection layers are used to obtain the classification result of man-machine asynchrony types. In this patent, their dataset labels are labeled by a professional physician, with four major classes of human-machine asynchrony detected: flow rate, trigger, cycle, and others.

2. Li-ice et al, in the patent CN202011431650.3, man-machine asynchronous recognition method based on one-dimensional interpretable convolutional neural network, propose to input a data set labeled by a doctor into a one-dimensional convolutional neural network after preprocessing for learning and training to obtain a prediction model based on the neural network. During the prediction process, a visual explanation of the model classification decision can be obtained by a gradient weighting class activation mapping mode.

3. Panqing et al, in the patent (CN202110208054.7) Man-machine asynchronous recognition method based on small data set and convolutional neural network, convert the collected original respiratory signal into two-dimensional image, firstly train the multi-classification model of the two-dimensional image by using the public image data set ImageNet, then input the two-dimensional image formed by respiratory waveform into the model in a transfer learning mode and finely adjust the weight of the layer above the last layer of the full connection layer, so as to obtain the convolutional neural network which can be used for respiratory waveform classification.

4. Kudzuvine, et al, in a patent (CN202010474275.4) DBA-DTW-KNN-based mechanical ventilation man-machine asynchronous fast recognition method, reads respiratory waveform data in real time to form a test sequence, calculates DTW distances between the test sequence and all sequences in a training set after standardization, calculates similarity distances by using DTW, and classifies the test sequence by combining a KNN clustering idea. The invention is used for judging invalid inspiration efforts in the man-machine asynchronous phenomenon.

5. CaoRui et al, in a patent (CN201610289472.2) epilepsia electroencephalogram signal classification method based on fuzzy entropy, use fuzzy entropy to analyze electroencephalogram signals, then select and extract fuzzy entropy under corresponding electrodes reflecting electroencephalogram signal characteristics through characteristics as input characteristics, and finally use the characteristics for classification.

6. In journal article, applied wavelet multi-scale feature detection mechanical ventilation man-machine asynchrony, Luyunfei and the like, wavelet scale transformation is firstly adopted to carry out one-time transformation on an original respiratory waveform, on the basis, various entropy features are used for extracting nonlinear features, an optimal feature combination is selected by using a previous item selection algorithm and then is used as input of a support vector machine classification algorithm for classification, and the invention only classifies the man-machine asynchrony phenomenon of invalid inspiration effort, and belongs to a two-classification task.

Research finds that the machine learning and the deep learning have similar performance on the classification effect, the difference is that the performance of the machine learning algorithm is influenced by feature selection, and most of the current feature selection is formed by combining statistical features and some clinical data; and the deep learning algorithm is more complicated to model than the former. Importantly, there is essentially no research available on the multi-classification task of various modes of ventilation with respect to the phenomenon of man-machine asynchrony.

Disclosure of Invention

(I) technical problems to be solved by the invention

The technical problem solved by the invention is as follows: how to realize that a plurality of man-machine asynchronous types under a plurality of ventilation modes can be classified simultaneously.

(II) the technical scheme adopted by the invention

A method of classifying an asynchronous phenomenon in a mechanical ventilation process, the method comprising:

acquiring real-time respiratory waveform data of an object to be detected in a mechanical ventilation process;

extracting Poincare map features of the real-time respiratory waveform data;

inputting the Poincare image characteristics into a pre-trained classification model, and outputting the human-computer asynchronous type corresponding to the real-time respiratory waveform data by the classification model.

Preferably, the classification method further comprises:

acquiring historical respiratory waveform data in the mechanical ventilation process;

performing waveform segmentation and data annotation on the historical respiratory waveform data to obtain respiratory waveform data of multiple periods, wherein the respiratory waveform data of the multiple periods correspond to multiple different man-machine asynchronous types;

sequentially extracting Poincare graph characteristics of the respiratory waveform data of each period to form a training sample;

and training the classification model to be trained by utilizing the training sample to obtain the trained classification model.

Preferably, the poincare graph feature extraction operation is sequentially performed on the waveform data of the three channels, namely the pressure time waveform data, the flow time waveform data and the capacity time waveform data, so as to obtain the poincare graph feature corresponding to the waveform data of each channel, so as to form the training sample of the current period.

Preferably, the poincare map feature extraction operation performed on the waveform data of each channel includes the following steps:

dividing waveform data of a current channel into n sections of data, wherein n is more than or equal to 2;

and calculating the short-time standard deviation, the long-time standard deviation and the ratio of the short-time standard deviation to the long-time standard deviation of each section of data in sequence to obtain n x 3-dimensional feature data which are used as Poincare diagram features corresponding to the waveform data of the current channel.

Preferably, the short-time standard difference SD1 is calculated as follows:

where, x (i) represents the signal of the ith sampling point of the current segment, and N represents the length of the data of the current segment.

Preferably, the long-term standard deviation is calculated as follows:

wherein X (i) represents the signal of the ith sampling point in the current segment,

represents the mean value of the current segment signal and N represents the length of the current segment data.

Preferably, the plurality of different man-machine-asynchronous types comprises at least a dual trigger type, an ineffective effort type, an early handover type, a late handover type.

The application also discloses sorter of man-machine asynchronous phenomenon among mechanical ventilation process, sorter includes:

the waveform acquisition unit is used for acquiring real-time respiratory waveform data of the object to be detected in the mechanical ventilation process;

the characteristic extraction unit is used for extracting Poincare graph characteristics of the real-time respiratory waveform data;

and the type prediction unit is used for predicting and obtaining the man-machine asynchronous type corresponding to the real-time respiratory waveform data according to the input Poincare diagram characteristics.

The application also discloses a computer readable storage medium, wherein the computer readable storage medium stores a classification program of the human-machine asynchronous phenomenon in the mechanical ventilation process, and the classification program of the human-machine asynchronous phenomenon in the mechanical ventilation process is executed by a processor to realize the classification method of the human-machine asynchronous phenomenon in the mechanical ventilation process.

The application also discloses a computer device, which comprises a computer readable storage medium, a processor and a classification program of the man-machine asynchronous phenomenon in the mechanical ventilation process, wherein the classification program of the man-machine asynchronous phenomenon in the mechanical ventilation process is stored in the computer readable storage medium, and when being executed by the processor, the classification program of the man-machine asynchronous phenomenon in the mechanical ventilation process realizes the classification method of the man-machine asynchronous phenomenon in the mechanical ventilation process.

(III) advantageous effects

The invention discloses a classification method and a classification device for man-machine asynchronous phenomena in a mechanical ventilation process, which have the following technical effects compared with the prior art:

1. the method uses the Poincare diagram to extract the characteristics from the original waveform, does not depend on other factors except the waveform, is only related to the form of the waveform, and the extracted characteristics can reflect the information such as the form change of the waveform;

2. compared with a deep learning model, the machine learning model built by the method is simpler and easier to deploy;

3. the method puts forward that the established machine learning model is directly used for various human-machine asynchronous classification tasks in various mechanical ventilation modes, and avoids the repeated modeling process.

Drawings

FIG. 1 is a diagram of four common asynchronous typical waveforms;

FIG. 2 is a flowchart illustrating a method for classifying the human-machine asynchronous phenomenon in the mechanical ventilation process according to a first embodiment of the present invention;

FIG. 3 is a diagram illustrating a training process of a classification model of the human-machine asynchronous phenomenon in the mechanical ventilation process according to a first embodiment of the present invention;

FIG. 4 is a schematic diagram illustrating a Poincare feature map extraction process of respiratory waveform data according to an embodiment of the present invention;

FIG. 5 is a schematic structural diagram of an apparatus for classifying the asynchronous phenomenon in the mechanical ventilation process according to a second embodiment of the present invention;

fig. 6 is a schematic diagram of a computer device according to a fourth embodiment of the present invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

Before describing in detail the various embodiments of the present application, the inventive concepts of the present application are first briefly described: the existing method for judging the man-machine asynchronous phenomenon comprises an observation method and a method based on machine learning, wherein the former is generally only suitable for current individuals, does not have applicability and is highly dependent on the experience of doctors, the latter is dependent on extracted features, most of the features adopted at present are statistical features and clinical data combinations, the accuracy of a model obtained by training is not high, most of the features are suitable for a two-classification task, and the multi-classification task is not realized. Therefore, the method for classifying the human-computer asynchronous phenomena in the mechanical ventilation process is provided, the Poincare image features of original respiratory waveform data corresponding to various human-computer asynchronizations are extracted to form a training sample, an existing machine learning model is trained to obtain a classification model capable of classifying the various human-computer asynchronous phenomena simultaneously, and then the classification model is used for predicting real-time respiratory waveform data in the mechanical ventilation process to obtain corresponding human-computer asynchronous types.

As shown in fig. 2, the first embodiment discloses a method for classifying the human-machine asynchronous phenomenon in the mechanical ventilation process, which includes the following steps:

step S11: acquiring real-time respiratory waveform data of an object to be detected in a mechanical ventilation process;

step S12: extracting Poincare map features of the real-time respiratory waveform data;

step S13: inputting the Poincare image characteristics into a pre-trained classification model, and outputting the human-computer asynchronous type corresponding to the real-time respiratory waveform data by the classification model.

Further, as shown in fig. 3, the first embodiment discloses that the method for classifying the human-machine asynchronous phenomenon in the mechanical ventilation process further includes the following steps:

step S21: acquiring historical respiratory waveform data in the mechanical ventilation process;

step S22: performing waveform segmentation and data annotation on the historical respiratory waveform data to obtain respiratory waveform data of multiple periods, wherein the respiratory waveform data of the multiple periods at least correspond to four different man-machine asynchronous types;

step S23: sequentially extracting Poincare graph characteristics of the respiratory waveform data of each period to form a training sample;

step S24: and training the classification model to be trained by utilizing the training sample to obtain the trained classification model.

Specifically, the first embodiment discloses a method for classifying the human-machine asynchronous phenomenon in the mechanical ventilation process, which includes two parts, i.e., step S21 to step S24, and model prediction step S11 to step S13, and the following description focuses on the model training part.

In step S21, a patient with respiratory disease is first simulated using TestLung (a simulated lung device), and by adjusting different parameters, patients with different respiratory frequencies, different respiratory intensities, and different respiratory diseases can be simulated. And then, mechanically ventilating by using a respirator and the TestLung, and adjusting the ventilation mode configuration of the respirator to obtain different man-machine asynchronous data under different ventilation modes. The respiratory waveform data of TestLung can be derived by software. In this embodiment, the historical respiratory waveform data used in the present embodiment are data of three channels, i.e., a pressure time waveform, a flow rate time waveform, and a volume time waveform.

The method for waveform segmentation of the historical respiratory waveform data in step S22 is as follows: the data obtained in step S21 is divided according to the breathing cycle. In the first embodiment, the acquired historical respiratory waveform data includes four common asynchronous types, namely a dual trigger type (DT), an invalid effort type (IE), an early switching type (PC) and a delayed switching type (DC), the four asynchronous types are respectively observed and then set up a segmentation rule, corresponding codes are written for segmentation, some of the segmented data cannot be distinguished from the asynchronous types, the data is usually discarded, a respiratory cycle is finally obtained by segmentation, the data of each respiratory cycle is respectively stored in a csv file according to the appearance sequence, and the file names are named from 1 according to the appearance sequence.

The method for performing data annotation on the historical respiratory waveform data in step S22 is as follows: because the nature of the problem determines that a supervised learning mode needs to be used for training the model, labeling is also needed for the segmented data, namely, a label is added to the waveform of each period. The process of adding the labels is completed by three persons of related professionals who are trained through man-machine asynchronous recognition, the labeling of each label is completed through the process of labeling one person and another person for examination, and if the labeling opinions are inconsistent, the periodic data is discarded. Wherein, the labels of the five types of the four asynchronous plus Other types mentioned above are digitalized to 0-4, wherein the Other type is composed of any one of DT, IE, PC and DC.

In the labeling process, firstly, the data of each period is read by using a code, secondly, the data of the three channels are drawn in a graph according to the sequence of pressure, flow and capacity, then, a labeling person observes the form of the graph and judges which type of asynchronization the period belongs to. And after the marking is finished, generating a csv file by the code, wherein the file is composed of two columns, one column is the file name of each respiratory cycle, and the other column corresponds to the label of the cycle.

In step S23, the poincare chart features of the respiratory waveform data of each cycle are sequentially extracted, and the method for constructing the training sample includes: and sequentially carrying out poincare graph feature extraction on the waveform data of the three channels of the pressure time waveform data, the flow time waveform data and the capacity time waveform data to obtain poincare graph features corresponding to the waveform data of each channel so as to form a training sample of the current period.

Specifically, there are two reasons why the original waveform is not directly used as the input of the model, one is that the dimensionality of the original data is too large, and the training difficulty is increased by directly inputting the model; the second is that it is generally not allowed for the model because the length of each data cycle is usually not uniform. Therefore, key features are extracted from the original data to reduce the data dimension, and finally, a consistent dimension can be obtained to be used as the input of the model. Features are extracted from the labeled data of each breathing cycle by utilizing a Poincare graph algorithm, wherein 3 features are extracted from each channel respectively and then are flattened to be used as integral features of three general waveforms of the breathing cycle, and finally the features are used as training samples of a machine learning model.

In step S23, as shown in fig. 4, the poincare graph feature extraction operation includes the following steps:

and calculating the short-time standard deviation, the long-time standard deviation and the ratio of the short-time standard deviation to the long-time standard deviation of each section of data in sequence to obtain n x 3-dimensional feature data which are used as Poincare diagram features corresponding to the waveform data of the current channel. Finally, Poincare graph features corresponding to the waveform data of the three channels in the current period, namely n x 3 dimensional feature data, are obtained and used as training samples in the current period.

The calculation formula of the short-time standard difference SD1 is as follows:

the long-term standard deviation is calculated as follows:

representing the mean of the current segment signal and N represents the length of the current segment.

The Ratio of the short time standard deviation to the long time standard deviation is SD1/SD 2.

In step S24, the specific process of training the classification model to be trained by using the training samples is as follows: a plurality of machine learning models are built by utilizing the existing machine learning framework, wherein the models comprise models of random forests, support vector machines, logistic regression and the like. And (5) dividing 80% of the training samples obtained in the step (S23) into a training set and the rest 20% into a test set, training the model by using the training set, and evaluating the model by using the test set to obtain a classification model with optimal performance.

Through experimental verification, the results of training and testing of a Logistic Regression model (Logistic Regression), a Support Vector Machine (SVM), a Random Forest model (Random Forest), a Voting model and an XGboost model show that the accuracy of a general model can reach more than 80%, for most models, the extracted features of the scheme can enable the overall accuracy of human-computer asynchronous classification to be more than 95%, and the results are shown in Table 1.

Table 1 test set of results for each model

Serial number	Model (model)	Accuracy	Cohenkappa
				1	Logisticregression	0.8294	0.7868
2	Randomforest	0.9538	0.9422
				3	SVM	0.9076	0.8845
4	Voting	0.9562	0.9452
				5	XGBoost	0.9550	0.9437

Further, after the classification model is obtained by training through the methods of steps S21 to S24, the method of steps S11 to S13 is used to predict the human-machine asynchronous type corresponding to the real-time respiratory waveform data of the subject during the mechanical ventilation.

As shown in fig. 5, the second embodiment further discloses a classification device for the human-machine asynchronous phenomenon in the mechanical ventilation process, which includes a waveform obtaining unit 100, a feature extracting unit 200, and a type predicting unit 300. The waveform acquisition unit 100 is used for acquiring real-time respiratory waveform data of a to-be-detected object in a mechanical ventilation process; the feature extraction unit 200 is configured to extract poincare map features of the real-time respiratory waveform data; the type prediction unit 300 is configured to predict a human-machine asynchronous type corresponding to the real-time respiratory waveform data according to the inputted poincare graph characteristics.

The type prediction unit 300 is a classification model trained in advance, and the training process of the classification model is obtained by referring to the method from step S21 to step S24 in the first embodiment, which is not described herein again.

A third embodiment further discloses a computer-readable storage medium, where a program for classifying the asynchronous phenomenon during the mechanical ventilation process is stored in the computer-readable storage medium, and when the program for classifying the asynchronous phenomenon during the mechanical ventilation process is executed by a processor, the method for classifying the asynchronous phenomenon during the mechanical ventilation process according to the first embodiment is implemented.

In the fourth embodiment, a computer device is further disclosed, and in a hardware level, as shown in fig. 6, the terminal includes a processor 12, an internal bus 13, a network interface 14, and a computer-readable storage medium 11. The processor 12 reads a corresponding computer program from the computer-readable storage medium and then runs, forming a request processing apparatus on a logical level. Of course, besides software implementation, the one or more embodiments in this specification do not exclude other implementations, such as logic devices or combinations of software and hardware, and so on, that is, the execution subject of the following processing flow is not limited to each logic unit, and may also be hardware or logic devices. The computer-readable storage medium 11 stores a program for classifying the asynchronous phenomenon during mechanical ventilation, and the program for classifying the asynchronous phenomenon during mechanical ventilation, when executed by a processor, implements the method for classifying the asynchronous phenomenon during mechanical ventilation according to the first embodiment.

Computer-readable storage media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of computer-readable storage media include, but are not limited to, phase change memory (PRAM), Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), Read Only Memory (ROM), Electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), Digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic disk storage, quantum memory, graphene-based storage media or other magnetic storage devices, or any other non-transmission medium that can be used to store information that can be accessed by a computing device.

Although a few embodiments of the present invention have been shown and described, it would be appreciated by those skilled in the art that changes may be made in these embodiments without departing from the principles and spirit of the invention, the scope of which is defined in the claims and their equivalents, and that such changes and modifications are intended to be within the scope of the invention.

Claims

1. A method for classifying human-machine asynchrony phenomena in a mechanical ventilation process, the method comprising:

extracting Poincare map features of the real-time respiratory waveform data;

2. The method for classifying the phenomenon of man-machine asynchrony during mechanical ventilation according to claim 1, further comprising:

3. The method of classifying asynchrony phenomena during mechanical ventilation as claimed in claim 2, wherein the method of extracting Poincare map features of respiratory waveform data for each cycle comprises:

acquiring pressure time waveform data, flow time waveform data and capacity time waveform data of each period;

and sequentially carrying out poincare graph feature extraction on the waveform data of the three channels of the pressure time waveform data, the flow time waveform data and the capacity time waveform data to obtain poincare graph features corresponding to the waveform data of each channel so as to form a training sample of the current period.

4. The method for classifying the human-machine asynchronous phenomenon in the mechanical ventilation process as claimed in claim 3, wherein the Poincare map feature extraction operation performed on the waveform data of each channel comprises the following steps:

5. The method for classifying the phenomenon of man-machine asynchrony during mechanical ventilation according to claim 3, wherein the short time standard difference SD1 is calculated as follows:

6. The method for classifying the phenomenon of man-machine asynchrony during mechanical ventilation according to claim 3, wherein the long-term standard deviation is calculated as follows:

7. The method of classifying asynchrony phenomena during mechanical ventilation according to claim 4, wherein the plurality of different asynchrony types includes at least a dual trigger type, an ineffective effort type, an early switch type, and a late switch type.

8. A device for classifying the phenomenon of man-machine asynchrony during mechanical ventilation, said device comprising:

9. A computer-readable storage medium, wherein the computer-readable storage medium stores a program for classifying an asynchronous phenomenon during mechanical ventilation, and the program for classifying an asynchronous phenomenon during mechanical ventilation, when executed by a processor, implements the method for classifying an asynchronous phenomenon during mechanical ventilation according to any one of claims 1 to 7.

10. A computer device comprising a computer-readable storage medium, a processor, and a program for classifying an asynchrony phenomenon during mechanical ventilation stored in the computer-readable storage medium, wherein the program for classifying an asynchrony phenomenon during mechanical ventilation, when executed by the processor, implements the method for classifying an asynchrony phenomenon during mechanical ventilation according to any one of claims 1 to 7.