Prediction method and system for withdrawal behavior of medical breathing machine

Abstract

The invention discloses a method and a system for predicting a machine withdrawal behavior of a medical respirator, and relates to the technical field of artificial intelligence, the method for predicting the machine withdrawal behavior of the medical respirator comprises the following steps: acquiring first respiration data generated in the process that a current patient uses the medical respirator in real time; inputting the first respiration data into a trained neural network model; based on the first breathing data, the neural network model generates second breathing data generated after the current patient uses the medical breathing machine in a first time period in the future; whether machine withdrawal can be carried out currently is predicted in real time based on the second respiration data and the neural network model, intelligent prediction is carried out through the neural network model, automatic machine withdrawal decision support is provided by the neural network model based on multiple key parameters such as the respiration frequency, the tidal volume, the oxygenation index and the lung compliance index, the burden of manual judgment is reduced, and the working efficiency is improved. The prediction accuracy and efficiency are improved, and the risk of machine withdrawal failure is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of medical device use. Specifically, the present invention provides a method, system, medium, and device for predicting the weaning behavior of a medical ventilator. Background Art

[0002] A medical ventilator is a device that can replace, control, or change a person's normal physiological breathing to increase pulmonary ventilation volume, improve respiratory function, reduce respiratory power consumption, and save cardiac reserve capacity. During use, the medical ventilator needs to judge whether it is possible to perform the key action of ventilator weaning based on the current usage situation of the patient, so as to gradually reduce the patient's dependence on the ventilator and finally remove mechanical ventilation. For patients in the intensive care unit, the timing of medical ventilator weaning is crucial. Inaccurate judgment of the weaning timing may lead to weaning failure, re-intubation, or prolonged mechanical ventilation time, increasing the risk of complications. Currently, the criteria used clinically to judge whether a patient is suitable for weaning are mainly based on some key physiological parameters and empirical decisions, but these methods have the following defects:

[0003] (1) The weaning criteria largely rely on the experience of medical staff. The weaning criteria of different doctors are not unified, with a large degree of subjectivity, time-consuming and laborious, and low precision, resulting in a relatively high risk of weaning failure;

[0004] (2) The physiological conditions of different patients vary greatly. The weaning criteria are usually based on widely applicable statistical rules and cannot dynamically evaluate the individualized disease fluctuations, with weak applicability, reduced precision, and a relatively high risk of weaning failure;

[0005] (3) The current weaning criteria are based on the patient's current instantaneous respiratory data and cannot predict the patient's respiratory data in the future for a period of time. The patient may be re-intubated due to unstable subsequent respiratory function, further reducing the precision. Summary of the Invention

[0006] The purpose of this application is to overcome the defects existing in the prior art.

[0007] To achieve the above purpose, this application proposes a method, system, medium, and device for predicting the weaning behavior of a medical ventilator.

[0008] In the first aspect, the method for predicting the weaning behavior of a medical ventilator proposed in this application includes:

[0009] S101, obtaining first respiratory data generated during the use of a medical ventilator by the current patient in real time;

[0010] S102, inputting the first respiratory data into a trained neural network model;

[0011] S103. Based on the first respiratory data, the neural network model generates second respiratory data generated by the current patient after using a medical ventilator within a future first time period, where the first respiratory data and the second respiratory data respectively include at least respiratory rate, tidal volume, oxygenation index, pulmonary compliance index, probability of successful weaning for similar cases, and parameter fluctuation index;

[0012] S104. Based on the second respiratory data, the neural network model predicts in real time whether weaning is currently possible.

[0013] In some examples, based on the second respiratory data, the neural network model predicting in real time whether a medical ventilator can be weaned currently includes:

[0014] According to the formula P = S1W1 + S2W2 + S3W3 + S4W4 + S5W5 + S6W6, the neural network model calculates the probability of successful weaning P, where S1, S2, S3, S4, S5, and S6 are the scores corresponding to the respiratory rate, tidal volume, oxygenation index, pulmonary compliance index, similarity of similar cases, and parameter fluctuation index respectively, and P1, P2, P3, P4, P5, and P6 are the preset weights corresponding to the respiratory rate, tidal volume, oxygenation index, pulmonary compliance index, probability of successful weaning for similar cases, and parameter fluctuation index respectively;

[0015] According to the probability of successful weaning P, the neural network model determines whether weaning is currently possible.

[0016] In some examples, according to the probability of successful weaning P, the neural network model determining whether weaning is currently possible includes:

[0017] If the probability of successful weaning P is greater than a set first threshold, the neural network model determines that weaning is currently possible;

[0018] If the probability of successful weaning P is not less than a set second threshold and not greater than a set third threshold, the neural network model generates a first adjustment suggestion;

[0019] If the probability of successful weaning P is less than a set fourth threshold, the neural network model determines that weaning is not currently possible and generates a second adjustment suggestion.

[0020] In some examples, after the neural network model determines that weaning is not currently possible, the method further includes:

[0021] Repeatedly execute steps S101 - S102, and the neural network model generates third respiratory data generated by the current patient after using a medical ventilator within a future second time period, where the second time period is greater than the first time period;

[0022] Based on the third breathing data, the neural network model predicts in real time whether extubation can be performed currently.

[0023] In some examples, the training process of the trained neural network model includes:

[0024] Obtain the fourth breathing data generated by multiple historical patients after using a medical ventilator respectively, and generate a first training dataset;

[0025] Input the first training dataset into the neural network model to train the neural network model, and obtain an initial trained neural network model;

[0026] Obtain the fifth breathing data generated by the current patient after using a medical ventilator within a set historical time period, and generate a second training dataset;

[0027] Input the second training dataset into the initial neural network model to optimize the initial neural network model, and obtain a trained neural network model.

[0028] In some examples, before inputting the first breathing data into the trained neural network model, the method further includes:

[0029] Obtain the abnormal parameters in the first breathing data, and eliminate or correct the abnormal parameters;

[0030] Perform normalization processing on each parameter in the first breathing data after eliminating or correcting the abnormal parameters;

[0031] Perform fusion processing on the first breathing data after normalization processing.

[0032] In some examples, the neural network model includes a first model and a second model different from the first model, where the first model is an LSTM model and the second model is a Transformer model.

[0033] In a second aspect, the prediction system for the extubation behavior of a medical ventilator provided by an embodiment of the present invention includes:

[0034] An acquisition module, configured to acquire in real time the first breathing data generated during the process of the current patient using a medical ventilator, where the at least includes respiratory rate, tidal volume, oxygenation index, and lung compliance index;

[0035] An input module, configured to input the first breathing data into the trained neural network model;

[0036] A generation module, configured to generate, based on the first breathing data, the second breathing data generated by the current patient after using a medical ventilator within a first future time period;

[0037] A prediction module, configured to predict in real time whether extubation can be performed currently based on the second respiratory data.

[0038] In a third aspect, the present application further provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the method described in the first aspect above is implemented.

[0039] In a fourth aspect, the present application further provides a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, the processor is caused to execute the method described in the first aspect above.

[0040] Compared with the prior art, the prediction method, system, medium, and device for the extubation behavior of a medical ventilator provided in the embodiments of the present invention have the following beneficial effects:

[0041] (1) By using neural network models (LSTM model and Transformer model) for intelligent prediction, based on multiple key parameters such as respiratory rate, tidal volume, oxygenation index, lung compliance index, etc., the neural network models provide automated extubation decision support, reducing the burden of manual judgment, improving the accuracy and efficiency of prediction, and reducing the risk of extubation failure;

[0042] (2) By collecting the individual data of the current patient in real time and using these data to perform personalized fine-tuning on the neural network model, the prediction model can dynamically adapt to the physiological characteristics of different patients, with strong applicability. This personalized fine-tuning method can improve the accuracy of prediction and reduce the risk of extubation failure;

[0043] (3) By generating adjustment suggestions, it can help the patient maintain stable respiratory support and further reduce the risk of extubation failure.

[0044] Compared with the prior art, the advantages of the present application are:

[0045] (1) By using deep learning models (LSTM model and Transformer model) for intelligent prediction, based on multiple key parameters such as the change rate, peak value, and valley value corresponding to airway pressure waveform data, the accuracy of prediction is improved;

[0046] (2) By collecting the individual data of the current patient in real time and using these data to perform personalized fine-tuning on the deep learning model, the prediction model can dynamically adapt to the physiological characteristics of different patients, with strong applicability. This personalized fine-tuning method can further improve the accuracy of prediction;

[0047] (3) Automatically select the LSTM model and the Transformer model for prediction according to actual needs, and support prediction at a time point or within a time period, with high flexibility and strong applicability. Description of the Drawings

[0048] Figure 1 Schematic flow chart of the prediction method for the ventilator weaning behavior provided by the embodiment of the present invention.

[0049] Figure 2 Block diagram of the structure of the dynamic prediction system for the ventilator weaning behavior provided by the embodiment of the present invention.

[0050] Figure 3 Principle block diagram of an electronic device as a classical computing device according to an embodiment of the present invention. Detailed Embodiments

[0051] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the accompanying drawings required for use in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0052] Embodiment 1

[0053] As Figure 1 shown, the prediction method for the ventilator weaning behavior provided by the embodiment of the present invention includes the following steps:

[0054] S101, obtain the first respiratory data generated during the current patient's use of the medical ventilator in real time.

[0055] Among them, the first respiratory data includes respiratory rate, tidal volume, oxygenation index, lung compliance index, weaning success probability of similar cases, and parameter fluctuation index.

[0056] Specifically, the first respiratory data may also include airway pressure, airflow rate, carbon dioxide exhalation concentration, and respiratory mode.

[0057] Among them, the respiratory rate is the number of breaths per unit time, which is used to reflect the patient's respiratory condition; the tidal volume is the volume of gas inhaled or exhaled per breath, which is used to reflect the capacity and function of the patient's lungs; the oxygenation index is the partial pressure of arterial oxygen and the oxygen concentration in the inhaled gas, which is used to evaluate the patient's blood oxygen condition; the lung compliance index is the inflation ability of the lungs under a certain pressure, which is used to evaluate the support effect of the medical ventilator on the patient; the parameter fluctuation index is used to reflect the overall current change of each parameter relative to the previous time period; the airway pressure is the pressure change of the airflow in the respiratory tract, which is used to evaluate the working state of the medical ventilator and the patient's breathing ability; the airflow rate is the rate of gas entering and leaving the lungs, which is an important parameter for predicting the respiratory rate and tidal volume and helps to judge the patient's breathing pattern; the carbon dioxide exhalation concentration is the carbon dioxide exhalation concentration at the end of each exhalation, which can reflect the ventilation energy efficiency of the patient; the breathing patterns include deep breathing, shallow breathing, rapid breathing, and periodic breathing, etc.

[0058] S102, input the first respiratory data into the trained neural network model.

[0059] In some examples, the training process of the trained neural network model includes:

[0060] Obtain the fourth respiratory data generated by multiple historical patients after using the medical ventilator respectively, and generate the first training data set;

[0061] Input the first training data set into the neural network model to train the neural network model, and obtain the initially trained neural network model;

[0062] Obtain the fifth respiratory data generated by the current patient after using the medical ventilator within the historical set time period, and generate the second training data set;

[0063] Input the second training data set into the initial neural network model to optimize the initial neural network model, and obtain the trained neural network model.

[0064] In some examples, before inputting the first respiratory data into the trained neural network model, the method further includes:

[0065] Obtain the abnormal parameters in the first respiratory data, and eliminate or correct the abnormal parameters;

[0066] Normalize each parameter in the first respiratory data after eliminating or correcting the abnormal parameters;

[0067] Perform fusion processing on the first respiratory data after normalization processing.

[0068] Specifically, after the first respiratory data is collected, the following processing is performed on the first respiratory data:

[0069] Outlier detection and processing: Identify and process outliers in the first respiratory data, such as unreasonable data caused by sensor failures or other reasons. Statistical methods or machine learning algorithms can be used to detect outliers and perform operations such as removal, correction, or interpolation;

[0070] Data standardization: Normalize each parameter of different types to give them a unified scale for subsequent processing;

[0071] Data fusion: Fuse the parameters in the first respiratory data after normalization. There are correlations and influences among different parameters, and data fusion can improve the accuracy and reliability of prediction. For example, fuse parameters such as respiratory rate, tidal volume, oxygenation index, and lung compliance index with respiratory waveform data, and use the fusion layer of a deep learning model or other fusion methods to integrate multi-source data into a unified representation.

[0072] After the first respiratory data is processed, the dimensional difference between different respiratory parameters can be eliminated, and the calculation efficiency can be improved.

[0073] Specifically, the present invention collects the individual data of the current patient in real time and uses this data to perform personalized fine-tuning on the neural network model, enabling the prediction model to dynamically adapt to the physiological characteristics of different patients. This personalized fine-tuning method can improve the accuracy of prediction and reduce the risk of weaning failure.

[0074] In some examples, the neural network model includes a first model and a second model different from the first model. Among them, the first model is an LSTM model, and the second model is a Transformer model.

[0075] Specifically, the LSTM model is good at processing parts with strong dependencies in time series data and is mainly used for predicting respiratory data within a short time period (such as within 30 seconds). The Transformer model uses its self-attention mechanism to capture more extensive time series information and is mainly used for predicting respiratory data within a long time period (such as within 5 minutes). This combination of short-term and long-term prediction helps medical staff formulate more flexible and comprehensive weaning plans and improve the success rate of weaning.

[0076] S103, Based on the first respiratory data, the neural network model generates second respiratory data generated by the current patient after using a medical ventilator within a future first time period. The second respiratory data includes respiratory rate, tidal volume, oxygenation index, lung compliance index, probability of successful weaning of similar cases, and parameter fluctuation index.

[0077] Specifically, the second respiratory data may further include airway pressure, airflow rate, exhaled carbon dioxide concentration, and respiratory pattern.

[0078] S104. Based on the second respiratory data, the trained neural network model predicts in real time whether extubation is possible currently.

[0079] In the present invention, through the use of a neural network model (a combination of an LSTM model and a Transformer model) for intelligent prediction, based on multiple key parameters such as respiratory rate, tidal volume, oxygenation index, and lung compliance index, the neural network model provides automated decision support for extubation, reducing the burden of manual judgment, improving the accuracy and efficiency of prediction, and reducing the risk of extubation failure.

[0080] In some examples, this step specifically includes:

[0081] According to the formula P = S1W1 + S2W2 + S3W3 + S4W4 + S5W5 + S6W6, the neural network model calculates the extubation success probability P, where S1, S2, S3, S4, S5, and S6 are the scores corresponding to the respiratory rate, tidal volume, oxygenation index, lung compliance index, similarity of similar cases, and parameter fluctuation index respectively, and P1, P2, P3, P4, P5, and P6 are the preset weights corresponding to the respiratory rate, tidal volume, oxygenation index, lung compliance index, extubation success probability of similar cases, and parameter fluctuation index respectively;

[0082] Based on the extubation success probability P, the neural network model determines whether extubation is possible currently.

[0083] Specifically, according to the numerical values of each parameter in the second respiratory data, the score of each parameter is determined. That is, the closer the numerical value of each parameter is to the normal value, the higher its score. According to the importance of each parameter, the weight of each parameter is set. That is, the higher the importance, the greater its corresponding weight. For example, according to the importance, the weights of the respiratory rate, tidal volume, oxygenation index, lung compliance index, extubation success probability of similar cases, and parameter fluctuation index can be set to: 0.25, 0.3, 0.25, 0.1, 0.05, 0.05 respectively. Among them, the calculation methods of weights mainly include the following:

[0084] (1) Expert evaluation method: Invite respiratory physicians, intensive care experts, etc. to conduct multiple rounds of discussions to determine the subjective importance of each parameter for extubation success, score (for example, 1 - 10 points), and then calculate the weighted average.

[0085] (2) Data - driven method: Collect historical data, conduct regression analysis on each parameter, train a model using machine learning algorithms (such as linear regression, logistic regression), and determine the relative importance of each parameter according to the coefficients of the model.

[0086] (3) Analytic Hierarchy Process: The Analytic Hierarchy Process is used to construct a weight matrix. Pairwise comparisons are made for each parameter, and the final weight value is obtained through consistency testing.

[0087] (4) Principal Component Analysis: Principal Component Analysis is used to extract the correlations between parameters. Multiple parameters are synthesized into several variables to reduce dimensions and determine weights.

[0088] (5) Fuzzy Comprehensive Evaluation Method: In view of the uncertainty of parameters, fuzzy mathematics is used to fuzzify each parameter, generate a comprehensive evaluation score, and convert it into a weight.

[0089] In practical applications, these weights need to be verified and adjusted regularly. Based on new respiratory data and more clinical feedback, a machine learning model is used for self-learning and optimization to improve the accuracy and reliability of prediction.

[0090] In some examples, according to the probability P of successful weaning, whether the neural network model can determine to wean the patient currently specifically includes the following steps:

[0091] If the probability of successful weaning is greater than a set first threshold, the neural network model determines that the patient can be weaned currently.

[0092] Specifically, the first threshold is 80%. If the weaning success rate is greater than 80% and the parameters in the patient's respiratory data are stable (such as respiratory rate, tidal volume, oxygenation index, etc.), the neural network model recommends that the patient can be weaned currently. During the weaning process, continuously monitor the patient's airway pressure and oxygenation index, and adjust the parameters of the medical ventilator according to the patient's real-time respiratory data if necessary.

[0093] If the probability of successful weaning is not less than a set second threshold and not greater than a set third threshold, the neural network model generates a first adjustment recommendation.

[0094] Specifically, the second threshold is 60% and the third threshold is 80%. If the weaning success rate is not less than 60% and not greater than 80%, the neural network model does not give a weaning or non-weaning recommendation, but will prompt the medical staff to adjust the parameters of the medical ventilator according to the values of each parameter in the second respiratory data, such as: moderately increasing the carbon dioxide exhalation concentration, increasing the airway pressure, adjusting the breathing mode, etc.

[0095] If the probability of successful weaning is less than a set fourth threshold, the neural network model determines that the patient cannot be weaned currently and generates a second adjustment recommendation.

[0096] Specifically, the fourth threshold is 60%. If the weaning success rate is less than 60% and multiple parameters in the current respiratory data of the patient are abnormal (such as unstable respiratory rate, insufficient tidal volume, low oxygenation index, poor lung compliance, etc.), the neural network model indicates that weaning is not possible at present and recommends adjusting these parameters of the medical ventilator: increasing the tidal volume and respiratory rate, increasing the carbon dioxide exhalation concentration, increasing the oxygen concentration and monitoring the oxygenation index, and adopting the pressure support mode to optimize the airway pressure.

[0097] Specifically, by generating adjustment suggestions, it can help the patient maintain stable respiratory support and further reduce the risk of weaning failure.

[0098] In some examples, after the neural network model determines that weaning is not possible at present, the method further includes:

[0099] Repeatedly execute the above steps S101 - S102, and the neural network model generates the third respiratory data generated by the current patient after using the medical ventilator within the next second time period, where the second time period is greater than the first time period;

[0100] Based on the third respiratory data, the neural network model predicts in real time whether weaning is possible at present.

[0101] Specifically, the neural network model uses the LSTM model to predict the respiratory data of the patient within the next first time period (1 minute). When it is determined that weaning is not possible based on this respiratory data, the neural network model uses the Transformer model to predict the respiratory data of the patient within the next second time period (5 minutes), and the prediction is more accurate.

[0102] Embodiment 2

[0103] As Figure 2 shown, the prediction system for the weaning behavior of the medical ventilator provided by the embodiment of the present invention includes:

[0104] An acquisition module, configured to acquire in real time the first respiratory data generated by the current patient during the use of the medical ventilator;

[0105] An input module, configured to input the first respiratory data into a trained neural network model;

[0106] A generation module, configured to generate, based on the first respiratory data, the second respiratory data generated by the current patient after using the medical ventilator within the next first time period, where the first respiratory data and the second respiratory data respectively include at least the respiratory rate, tidal volume, oxygenation index, lung compliance index, weaning success probability of similar cases, and parameter fluctuation index;

[0107] A prediction module, configured to predict in real time whether extubation can be performed currently based on the second respiratory data.

[0108] Embodiment 3

[0109] Figure 3 Fig. 7 shows a block diagram of the hardware structure principle of an embodiment of the electronic device provided by the present invention. The electronic device includes a processor 601 and a memory 602 storing computer program instructions. When the processor executes the computer program instructions, the methods disclosed in the above embodiments of the present invention are implemented.

[0110] Specifically, the above-mentioned processor 601 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured as one or more integrated circuits implementing the embodiments of the present invention.

[0111] The memory 602 may include a mass storage for data or instructions. By way of example and not limitation, the memory 602 may include a hard disk drive (HDD), a floppy disk drive, a flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a universal serial bus (USB) drive or a combination of two or more of these. In a suitable case, the memory 602 may include a removable or non-removable (or fixed) medium. In a suitable case, the memory 602 may be inside or outside the integrated gateway disaster recovery device. In a specific embodiment, the memory 602 is a non-volatile solid-state memory.

[0112] The memory may include a read-only memory (ROM), a random access memory (RAM), a magnetic disk storage media device, an optical storage media device, a flash memory device, an electrical, optical, or other physical / tangible memory storage device. Thus, in general, the memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described in reference to the methods according to one aspect of the present disclosure.

[0113] In some examples, as Figure 3 shown, the electronic device may further include a communication interface 603 and a bus 610. Among them, as Figure 3 shown, the processor 601, the memory 602, and the communication interface 603 are connected through the bus 610 and complete communication with each other. The electronic device in the embodiments of the present invention may be a local server or other computing devices, or may also be a cloud server.

[0114] The communication interface 603 is mainly used to implement the communication between each module, device, unit, and / or equipment in the embodiments of the present invention.

[0115] The bus 610 includes hardware, software, or both, and couples the components of the online data flow meter charging device to each other. By way of example and not limitation, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a HyperTransport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an InfiniBand interconnect, a Low Pin Count (LPC) bus, a memory bus, a MicroChannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-E) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local (VLB) bus, or other suitable buses or a combination of two or more of these. In suitable cases, the bus 610 may include one or more buses. Although the embodiments of the present invention describe and illustrate specific buses, the present invention contemplates any suitable bus or interconnect.

[0116] The present invention also provides a computer program product, which includes computer program instructions. When the computer program instructions are executed by a processor, the method provided by the embodiments of the present invention is implemented. Such computer program product is, for example, a software installation package, a plug-in compatible with a related software system, etc.

[0117] It should be clear that the present invention is not limited to the specific configurations and processes described above and illustrated in the figures. For the sake of brevity, the detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and illustrated as examples. However, the method process of the present invention is not limited to the specific steps described and illustrated. Those skilled in the art can make various changes, modifications, and additions, or change the order between steps after understanding the spirit of the present invention.

[0118] In addition, the "first", "second", and similar terms used in the present disclosure do not denote any order, quantity, or importance, but are only used to distinguish different parts.

[0119] As described above, only the specific embodiments of the present invention are provided. Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the systems, modules, and units described above can refer to the processes in the foregoing method embodiments, which will not be repeated here. It should be understood that the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of various equivalent modifications or substitutions within the technical scope disclosed by the present invention, and these modifications or substitutions should be covered within the protection scope of the present invention.

Claims

1. A prediction method for the behavior of weaning a medical ventilator, comprising: S101, obtaining first respiratory data generated during the process of a current patient using a medical ventilator in real time; S102, inputting the first respiratory data into a trained neural network model; S103, based on the first respiratory data, the neural network model generates second respiratory data generated by the current patient after using the medical ventilator within a first future time period, wherein the first respiratory data and the second respiratory data respectively at least include respiratory rate, tidal volume, oxygenation index, pulmonary compliance index, weaning success probability of similar cases, and parameter fluctuation index; S104, based on the second respiratory data, the neural network model predicts in real time whether weaning can be performed currently.

2. The prediction method according to claim 1, characterized in that, Based on the second respiratory data, the neural network model predicting in real time whether the medical ventilator can be weaned currently includes: According to the formula P = S1W1 + S2W2 + S3W3 + S4W4 + S5W5 + S6W6, the neural network model calculates the weaning success probability P, where S1, S2, S3, S4, S5, and S6 are the scores corresponding to the respiratory rate, tidal volume, oxygenation index, pulmonary compliance index, similarity of similar cases, and parameter fluctuation index respectively, and P1, P2, P3, P4, P5, and P6 are the preset weights corresponding to the respiratory rate, tidal volume, oxygenation index, pulmonary compliance index, weaning success probability of similar cases, and parameter fluctuation index respectively; According to the weaning success probability P, the neural network model determines whether weaning can be performed currently.

3. The prediction method according to claim 2, wherein According to the weaning success probability P, the neural network model determining whether weaning can be performed currently includes: If the weaning success probability P is greater than a set first threshold, the neural network model determines that weaning can be performed currently; If the weaning success probability P is not less than a set second threshold and not greater than a set third threshold, the neural network model generates a first adjustment suggestion; If the weaning success probability P is less than a set fourth threshold, the neural network model determines that weaning cannot be performed currently and generates a second adjustment suggestion.

4. The prediction method according to claim 3, wherein After the neural network model determines that weaning cannot be performed currently, it further includes: Repeating steps S101 - S102, the neural network model generates third respiratory data generated by the current patient after using the medical ventilator within a second future time period, wherein the second time period is greater than the first time period; Based on the third respiratory data, the neural network model predicts in real time whether weaning can be performed currently.

5. The prediction method according to claim 1, wherein The training process of the trained neural network model includes: Respectively obtaining fourth respiratory data generated by multiple historical patients after using the medical ventilator to generate a first training data set; Inputting the first training data set into the neural network model to train the neural network model to obtain a trained initial neural network model; Obtaining fifth respiratory data generated by the current patient after using the medical ventilator within a set historical time period to generate a second training data set; Inputting the second training data set into the initial neural network model to optimize the initial neural network model to obtain a trained neural network model.

6. The prediction method according to claim 1, wherein Before inputting the first respiratory data into the trained neural network model, the method further includes: Obtaining abnormal parameters in the first respiratory data, and removing or correcting the abnormal parameters; Normalizing each parameter in the first respiratory data after removing or correcting the abnormal parameters; Performing fusion processing on the first respiratory data after the normalization processing.

7. The prediction method according to any one of claims 1-6, characterized in that, The neural network model includes a first model and a second model different from the first model. Among them, the first model is an LSTM model, and the second model is a Transformer model.

8. A prediction system for the weaning behavior of a medical ventilator, comprising: An acquisition module, configured to acquire in real time first respiratory data generated during the use of a medical ventilator by a current patient, where the at least includes respiratory rate, tidal volume, oxygenation index, and lung compliance index; An input module, configured to input the first respiratory data into a trained neural network model; A generation module, configured to generate, based on the first respiratory data, second respiratory data generated after the current patient uses the medical ventilator within a first future time period; A prediction module, configured to predict in real time whether weaning can be performed currently based on the second respiratory data.

9. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, and the computer program is used to execute the method according to any one of claims 1-7 above.

10. An electronic device, characterized in that, The electronic device includes: A processor; A memory for storing executable instructions that can be executed by the processor; The processor is used to read the executable instructions from the memory and execute the instructions to implement the method according to any one of claims 1-7 above.

Images