Mechanical ventilation data processing method based on AI large model and related equipment

Abstract

The embodiment of the invention provides a mechanical ventilation data processing method based on an AI large model and related equipment. The method comprises the steps that object attribute data and sleep respiration detection data of a target object are acquired; calling a preset neural network model to perform mechanical ventilation parameter generation on the basis of the object attribute data and the sleep respiration detection data to generate mechanical ventilation data; obtaining object feedback data; calling a neural network model to generate mechanical ventilation effect evaluation data based on the object attribute data, the sleep respiration detection data, the mechanical ventilation data and the object feedback data; when the mechanical ventilation effect does not meet the preset mechanical ventilation effect condition, a neural network model is called to adjust the mechanical ventilation data based on the object attribute data, the sleep respiration detection data, the mechanical ventilation data, the object feedback data and the mechanical ventilation effect evaluation result. The mechanical ventilation parameter setting accuracy can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent control of medical equipment, and particularly relates to a mechanical ventilation data processing method based on an AI large model and related equipment. BACKGROUND

[0002] Sleep breathing disorder, especially obstructive sleep apnea-hypopnea syndrome (OSAHS), is a common chronic sleep disease. Its main feature is that the upper airway collapses repeatedly during sleep, causing apnea or hypopnea, and causing intermittent hypoxia, hypercapnia and sleep structure disorder. Long-term untreated OSAHS can cause hypertension, coronary heart disease, diabetes, stroke and other complications, seriously affecting the quality of life and life of patients.

[0003] Continuous positive airway pressure (CPAP) respirators, auto-CPAP respirators and bilevel positive airway pressure (BiPAP) respirators are the main technical means for treating OSAHS at present, and the accurate setting of the mechanical ventilation parameters of the respirator is the key to ensuring the treatment effect.

[0004] However, the setting method of the mechanical ventilation parameters in the related art has the technical problem of low accuracy of the mechanical ventilation parameter setting, thereby resulting in poor treatment effect of OSAHS. SUMMARY

[0005] The main purpose of the embodiments of the present application is to propose a mechanical ventilation data processing method based on an AI large model and related equipment, aiming to improve the accuracy of the mechanical ventilation parameter setting.

[0006] To achieve the above-mentioned purpose, a first aspect of the embodiments of the present application proposes a mechanical ventilation data processing method based on an AI large model, the method comprising: obtaining object attribute data and sleep breathing detection data of a target object; calling a preset neural network model to generate mechanical ventilation parameters based on the object attribute data and the sleep breathing detection data, to obtain mechanical ventilation data corresponding to the target object; obtaining object feedback data after mechanical ventilation processing of the target object based on the mechanical ventilation data; The neural network model is called to perform mechanical ventilation effect evaluation based on the object attribute data, the sleep respiration detection data, the mechanical ventilation data, and the object feedback data, to obtain mechanical ventilation effect evaluation data. When the mechanical ventilation effect evaluation data indicates that the mechanical ventilation effect does not meet a preset mechanical ventilation effect condition, the neural network model is called to adjust the mechanical ventilation data based on the object attribute data, the sleep respiration detection data, the mechanical ventilation data, the object feedback data, and the mechanical ventilation effect evaluation result, to obtain updated mechanical ventilation data.

[0007] To achieve the above object, a second aspect of the embodiment of the present application proposes a mechanical ventilation data processing device based on an AI large model, which comprises: A first obtaining unit is configured to obtain object attribute data and sleep respiration detection data of a target object. A generating unit is configured to call a preset neural network model to perform mechanical ventilation parameter generation based on the object attribute data and the sleep respiration detection data, to obtain mechanical ventilation data corresponding to the target object. A second obtaining unit is configured to obtain object feedback data after mechanical ventilation processing of the target object based on the mechanical ventilation data. An evaluation unit is configured to call the neural network model to perform mechanical ventilation effect evaluation based on the object attribute data, the sleep respiration detection data, the mechanical ventilation data, and the object feedback data, to obtain mechanical ventilation effect evaluation data. An updating unit is configured to call the neural network model to adjust the mechanical ventilation data based on the object attribute data, the sleep respiration detection data, the mechanical ventilation data, the object feedback data, and the mechanical ventilation effect evaluation result when the mechanical ventilation effect evaluation data indicates that the mechanical ventilation effect does not meet a preset mechanical ventilation effect condition, to obtain updated mechanical ventilation data.

[0008] To achieve the above object, a third aspect of the embodiment of the present application proposes an electronic device, which comprises a memory and a processor, the memory stores a computer program, and the processor implements the method of the first aspect when executing the computer program.

[0009] To achieve the above object, a fourth aspect of the embodiment of the present application proposes a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the method of the first aspect.

[0010] The AI large model-based mechanical ventilation data processing method and related equipment provided by the present application can improve the accuracy and individualization of initial mechanical ventilation parameter setting by obtaining object attribute data and sleep respiration detection data of a target object, calling a preset neural network model to generate initial mechanical ventilation data based on the multi-dimensional data, and then obtaining object feedback data after treatment based on the mechanical ventilation data, and calling the same neural network model again to evaluate the mechanical ventilation effect in combination with the initial data, setting parameters and feedback effect, and forming a quantitative judgment on the effectiveness of the parameters. When the evaluation result does not meet the preset effect condition, the model is further called to adaptively adjust the model parameters based on all the data obtained in the above steps to generate updated mechanical ventilation data. In this way, the mechanical ventilation parameter setting method provided by the present application can dynamically respond to the specific treatment response of the target object, thereby systematically improving the accuracy of mechanical ventilation parameter setting. BRIEF DESCRIPTION OF DRAWINGS

[0011] Figure 1 is a flowchart of the AI large model-based mechanical ventilation data processing method provided by the embodiments of the present application; Figure 2 is a flowchart of step S102 in Figure 1 provided by the embodiments of the present application; Figure 3 is a flowchart of step S204 in Figure 2 provided by the embodiments of the present application; Figure 4 is a schematic diagram of a mechanical ventilation data setting page provided by the embodiments of the present application; Figure 5 is a schematic diagram of a model processing flow provided by the embodiments of the present application; Figure 6 is a structural schematic diagram of an AI large model-based mechanical ventilation data processing device provided by the embodiments of the present application; Figure 7 is a hardware structural schematic diagram of an electronic device provided by the embodiments of the present application. DETAILED DESCRIPTION

[0012] In order to make the purpose, technical scheme and advantages of the present application clearer, the present application will be further described in detail below in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application.

[0013] It should be noted that although the functional modules are divided in the device schematic diagram, the logical order is shown in the flowchart, but in some cases, the steps shown or described can be performed in a manner different from the module division in the device or the order in the flowchart. The terms "first", "second", etc. in the specification and claims and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily describe a specific order or sequence.

[0014] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in the specification herein is for the purpose of describing the embodiments of the present application only and is not intended to be limiting of the present application.

[0015] First, the meanings of several terms involved in the present application are analyzed: Artificial Intelligence (AI) large model: AI large model, usually refers to deep learning model with huge number of parameters, massive training data and huge consumption of computing resources. At present, the most common large model is Large Language Model (LLM). In addition, there are also multi-modal large models focusing on image generation, such as Stable Diffusion (SD) model, etc.

[0016] Sleep breathing disorder, especially Obstructive Sleep Apnea-Hypopnea Syndrome (OSAHS), is a common chronic sleep disease, its main feature is that the upper airway collapses repeatedly during sleep, leading to apnea or hypopnea, causing intermittent hypoxia, hypercapnia and sleep structure disorder. Long-term untreated OSAHS can lead to hypertension, coronary heart disease, diabetes, stroke and other complications, seriously affecting the quality of life and life of patients.

[0017] Continuous Positive Airway Pressure (CPAP) ventilator, Auto-CPAP ventilator and Bilevel Positive Airway Pressure (BiPAP) ventilator are the main methods for treating OSAHS at present, and the accurate setting of mechanical ventilation parameters is the key to ensure the treatment effect.

[0018] The mechanical ventilation parameter method in the related art mainly relies on artificial experience setting, that is, a doctor gives initial parameters of mechanical ventilation according to patient conditions (such as an apnea-hypopnea index (AHI) and blood oxygen conditions) and self experience, and then adjusts the parameters according to patient feedback and subsequent review data. This method is highly subjective and has low accuracy, and it is difficult to realize personalization of mechanical ventilation parameter setting. The method in the related art also includes polysomnography (PSG) pressure titration, which needs to be performed in a sleep laboratory. The whole process is complex, time-consuming and costly, and the difference between the laboratory environment and the natural sleep environment may cause deviation between the titration result and the actual home treatment requirement.

[0019] In addition, the related art also sets a mechanical ventilation scheme by embedding a simple algorithm in an automatic continuous positive airway pressure (Auto-CPAP) device and adjusting pressure according to real-time respiratory events. However, the algorithm model of this method is usually simple, and feedback adjustment is based on a single or a small number of physiological signals (such as airflow and pressure), which fails to fully integrate multi-dimensional data such as patient's basic information and detailed sleep monitoring reports for comprehensive prediction, resulting in low personalization of mechanical ventilation setting. Therefore, the related art has the technical problem of low accuracy of mechanical ventilation parameter setting, which leads to poor treatment effect.

[0020] Therefore, the embodiments of the present application provide a mechanical ventilation data processing method based on an AI large model and related equipment, which aims to improve the accuracy of mechanical ventilation parameter setting.

[0021] The mechanical ventilation data processing method based on an AI large model and related equipment provided by the embodiments of the present application are specifically explained by the following embodiments. First, the mechanical ventilation data processing method based on an AI large model in the embodiments of the present application is described.

[0022] The AI large model-based mechanical ventilation data processing method provided by the embodiments of the present application relates to the technical field of artificial intelligence. The AI large model-based mechanical ventilation data processing method provided by the embodiments of the present application can be applied to a terminal, can also be applied to a server end, and can also be software running in the terminal or the server end. In some embodiments, the terminal can be a smart phone, a tablet computer, a notebook computer, a desktop computer, or the like; the server end can be configured as a stand-alone physical server, can also be configured as a server cluster or a distributed system formed by multiple physical servers, and can also be configured as a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDNs, and big data and artificial intelligence platforms; and the software can be an application for implementing the AI large model-based mechanical ventilation data processing method, but is not limited to the above forms.

[0023] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld devices or portable devices, tablet devices, multi-processor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, and the like. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, and the like that perform specific tasks or implement specific abstract data types. The present application can also be practiced in a distributed computing environment in which tasks are performed by remote processing devices connected by a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media, including storage devices.

[0024] Figure 1 is an optional flowchart of the AI large model-based mechanical ventilation data processing method provided by the embodiments of the present application, Figure 1 The method in the above can include but is not limited to including steps S101 to S105.

[0025] Step S101, obtaining object attribute data and sleep respiration detection data of a target object; Step S102, calling a preset neural network model to generate mechanical ventilation parameters based on the object attribute data and the sleep respiration detection data, to obtain mechanical ventilation data corresponding to the target object; Step S103, obtaining object feedback data after mechanical ventilation processing of the target object based on the mechanical ventilation data; In step S104, the neural network model is called to perform mechanical ventilation effect evaluation based on the object attribute data, sleep respiration detection data, mechanical ventilation data, and object feedback data, to obtain mechanical ventilation effect evaluation data. In step S105, when the mechanical ventilation effect evaluation data indicates that the mechanical ventilation effect does not meet the preset mechanical ventilation effect condition, the neural network model is called to adjust the mechanical ventilation data based on the object attribute data, sleep respiration detection data, mechanical ventilation data, object feedback data, and mechanical ventilation effect evaluation result, to obtain updated mechanical ventilation data.

[0026] The steps S101 to S105 shown in the embodiments of the present application, by obtaining the object attribute data and sleep respiration detection data of the target object, and calling the preset neural network model to generate initial mechanical ventilation data based on the above multi-dimensional data, improve the accuracy and individualization of the initial mechanical ventilation parameter setting; then by obtaining the object feedback data after treatment based on the mechanical ventilation data, and calling the same neural network model again, the mechanical ventilation effect is evaluated in combination with the initial data, the set parameters and the feedback effect, to form a quantitative judgment on the effectiveness of the parameters; when the evaluation result does not reach the preset effect condition, the model is further called to adaptively adjust the model parameters based on all the data obtained in the above steps, to generate updated mechanical ventilation data. In this way, the mechanical ventilation parameter setting method provided by the present application can dynamically respond to the specific treatment response of the target object, thereby systematically improving the accuracy of the mechanical ventilation parameter setting.

[0027] In step S101 of some embodiments, the object attribute data and sleep respiration detection data of the target object can be obtained first, wherein the object attribute data represents the basic physical data of the sleep apnea patient, for example, including but not limited to height, weight (or BMI), neck circumference, medical history (such as hypertension, diabetes, cardiovascular and cerebrovascular diseases, etc.), smoking history, drinking history, and Epworth Sleepiness Scale (ESS) score, etc. The sleep respiration detection data refers to the sleep respiration related parameters obtained by sleep respiration detection of the target object through polysomnography (PSG) or portable sleep monitoring device, such as sleep AHI value, respiratory event frequency and respiratory duration, total number of apnea, blood oxygen saturation data (such as minimum blood oxygen, average blood oxygen, Oxygen Desaturation Index (ODI), cumulative time of blood oxygen saturation in different intervals (such as 90%-100%, 80%-89%, etc.), original respiratory waveform, chest and abdominal movement, and snoring sound, etc. The multi-dimensional data of the patient can be used to quantify the severity of respiratory disorders, and by comprehensively collecting and integrating these multi-source data, sufficient data support can be provided for intelligent parameter generation, so as to facilitate subsequent comprehensive analysis and personalized mechanical ventilation parameter recommendation.

[0028] In step S102 of some embodiments, a preset neural network model can be called to generate mechanical ventilation parameters based on the object attribute data and the sleep respiration detection data obtained in step S101, to obtain mechanical ventilation data corresponding to the target object. The preset neural network model can be an artificial intelligence large model (AI large model) trained by a large number of training samples, which can mine the complex correlation between multi-dimensional features and mechanical ventilation parameters. For example, the preset neural network model can be a large language model (LLM). The model outputs the mechanical ventilation data corresponding to the target object by performing nonlinear transformation and feature fusion on the input data. The mechanical ventilation data can include the optimal mechanical ventilation parameter combination under different ventilator modes and the corresponding respiratory health recommendations. For example, for the CPAP ventilator mode, the mechanical ventilation data output by the model can include the optimal fixed treatment pressure value; for the Auto-CPAP mode, the model can recommend the optimal pressure range (minimum pressure and maximum pressure) and related response parameters (such as pressure rise rate); for the BiPAP mode, the model can recommend the inspiratory positive airway pressure (IPAP), expiratory positive airway pressure (EPAP), and standby breathing frequency, etc. The respiratory health recommendations can include the analysis of the sleep condition of the target object, the reason analysis of the recommended current mechanical ventilation parameter combination, and the respiratory health precautions for the target object. The mechanical ventilation data can also include the adjustment range of the provided mechanical ventilation parameter combination, so as to provide adjustment suggestions for the mechanical ventilation parameters when the treatment effect is not good.

[0029] In some embodiments, please refer to Figure 2 The neural network model includes a fully connected network, a time sequence feature extraction network, and a cross-attention network. The preset neural network model is called to generate mechanical ventilation parameters based on the object attribute data and the sleep respiration detection data, to obtain mechanical ventilation data corresponding to the target object, including the following steps S201 to S204: Step S201, calling the fully connected network to extract features from the object attribute data to obtain object attribute features; Step S202, performing sliding time window segmentation on the sleep respiration detection data to obtain sleep respiration detection data sequences, and calling the time sequence feature extraction network to extract features from the sleep respiration detection data sequences to obtain sleep respiration detection data features; Step S203, calling the cross-attention network to perform cross-attention processing on the object attribute features and the sleep respiration detection data features to obtain corresponding structured data. Step S204, generating mechanical ventilation parameters based on the structured data to obtain mechanical ventilation data corresponding to the target object.

[0030] In step S201 of some embodiments, the preset neural network model can include a fully connected network, a time sequence feature extraction network, and a cross-attention network. Specifically, data features of the object attribute data can be extracted to obtain object attribute features, which can be implemented by a fully connected network (FCN). The fully connected network is a feedforward neural network structure that can perform nonlinear combination and transformation on the input features, thereby learning the deep patterns in the data. After the object attribute data is standardized or encoded, it is input into the fully connected network, which abstracts the object attribute data through its multi-layer connection structure and outputs a high-dimensional feature vector, i.e., the object attribute features, which can more effectively represent the individual physiological state of the target object and the risk factors of sleep respiration.

[0031] In step S202 of some embodiments, the sleep respiration detection data can be segmented by a sliding time window to obtain a sleep respiration detection data sequence, and a time sequence feature extraction network is called to extract features from the sequence to obtain sleep respiration detection data features. The sleep respiration detection data is time series data, and the sliding time window segmentation is to slice the continuous time series according to a fixed length (such as 30 seconds per window) and a step to form a series of short sequences, so as to facilitate the model to capture local time sequence patterns, for example, the PatchTST architecture can divide the sleep respiration detection data into subsequence blocks (Patches). The sleep respiration detection data is segmented by a sliding time window to obtain a sleep respiration detection data sequence. The time sequence feature extraction network can be a long short-term memory (LSTM) network or a Transformer encoder, which can capture long-term dependencies and dynamic change laws in the data. The time sequence feature extraction network is called to extract features from the sleep respiration detection data sequence to obtain sleep respiration detection data features.

[0032] In step S203 of some embodiments, a cross-attention network can be invoked to perform cross-attention processing on the object attribute features and the sleep respiration detection data features to obtain corresponding structured data. The cross-attention network is a neural network mechanism that allows feature vectors from different modalities or sources to pay attention to important information of each other. In this embodiment, the object attribute features (representing the static features of the target object) and the sleep respiration detection data features (representing the dynamic physiological change features of the target object) are input into the network as two information sources. The cross-attention mechanism calculates attention weights to dynamically modulate the weights of the time series features (sleep respiration detection data features) by the static features (object attribute features) (for example, a diabetes history can enhance the sensitivity of blood oxygen fluctuation events). In this way, it can dynamically determine how much attention the object attribute features of the target object should pay to its sleep respiration detection data features when generating the final representation, thereby realizing deep interaction and information complementation of the two features. The output of the cross-attention processing is a structured data that integrates static attributes and dynamic time series information.

[0033] In step S204 of some embodiments, a preset regression layer or small prediction network can be invoked to generate mechanical ventilation parameters based on the structured data generated in the above steps to obtain mechanical ventilation data corresponding to the target object. Specifically, the fused structured data can be input into the prediction network. The network outputs the recommended mechanical ventilation mode and its corresponding mechanical ventilation parameter value or parameter combination, such as the optimal treatment pressure value for the CPAP mode, or the inspiratory positive pressure and expiratory positive pressure parameter combination for the BiPAP mode, through the learned complex mapping relationship.

[0034] The embodiments of the present application can effectively extract the static attribute features and dynamic physiological time series features of the target object, and perform deep information fusion on the two features. The structured data obtained by fusion can more comprehensively and accurately reflect the individualized condition of the target object, thereby improving the individualization degree of generating mechanical ventilation parameters and improving the accuracy of setting mechanical ventilation parameters.

[0035] In some embodiments, please refer to Figure 3 , generating mechanical ventilation parameters based on the structured data to obtain mechanical ventilation data corresponding to the target object, including steps S301 to S303: Step S301, generating mechanical ventilation parameters based on the structured data to obtain a mechanical ventilation mode set for the target object and mechanical ventilation parameters corresponding to the mechanical ventilation mode; Step S302, invoking a preset large language model to generate a corresponding reason analysis text based on the mechanical ventilation mode and the mechanical ventilation parameters; Step S303, determining the mechanical ventilation data corresponding to the target object according to the mechanical ventilation mode, the mechanical ventilation parameters, and the cause analysis text.

[0036] In step S301 of some embodiments, mechanical ventilation parameter generation can be performed based on the structured data generated in the foregoing steps, to obtain the mechanical ventilation mode set for the target object and the mechanical ventilation parameters corresponding to the mechanical ventilation mode. The structured data refers to a high-level feature vector that can comprehensively represent the static attributes and dynamic physiological state of the target object after processing by the foregoing feature extraction and fusion module (such as a cross-attention network). Mechanical ventilation data generation according to the structured data can be implemented through a fully connected layer or a small neural network. These networks output mechanical ventilation data through the mapping relationship learned by them.

[0037] In step S302 of some embodiments, as previously described, the preset neural network model can be a large language model. Specifically, the large language model can be called to generate a corresponding cause analysis text based on the mechanical ventilation mode and the mechanical ventilation parameters generated in the foregoing steps. Specifically, the mechanical ventilation mode, the mechanical ventilation parameters, and possibly associated object attribute data (such as AHI value and BMI, etc.) are input as prompt information into the large language model. Then, the large language model automatically generates a natural language description of the cause analysis text based on its embedded medical knowledge and logical reasoning ability, to explain the basis for recommending the specific mode and parameter combination. For example, the model can generate the following text: recommend using Auto-CPAP mode, pressure range set to 7-15 cmH2O, recommended reason is that the AHI value of the object is 38.2 times / hour, which belongs to severe sleep apnea syndrome, and is accompanied by obesity (BMI = 31.2), a higher upper limit of pressure is needed to effectively cope with obstruction events, and setting the pressure range can improve comfort.

[0038] In step S303 of some embodiments, the mechanical ventilation data corresponding to the target object can be determined according to the mechanical ventilation mode, the mechanical ventilation parameters, and the corresponding cause analysis text. The mechanical ventilation data is a comprehensive data output structure, which is not a single parameter value, but a data set encapsulating the treatment mode, specific parameter settings, and explanatory notes. The above data is encapsulated and then output as a personalized treatment plan that can be used in clinical practice and contains the basis for decision-making.

[0039] Embodiments of the present application provide an intelligent parameter generation process from data to decision-making with explainability. The model trained by specific field training data is used for mechanical ventilation parameter recommendation, which improves the accuracy of mechanical ventilation parameters, thereby improving the clinical credibility of recommended mechanical ventilation parameters, and can help doctors quickly understand the decision-making logic of the model, and improve the reliability of the intelligent recommendation plan.

[0040] In some examples, after determining the mechanical ventilation data corresponding to the target object according to the mechanical ventilation mode, the mechanical ventilation parameter, and the cause analysis text, the method provided by the present application further includes the following steps: displaying the mechanical ventilation data in the mechanical ventilation data setting page; displaying a voice broadcast control in the mechanical ventilation data setting page, and in response to a triggering operation on the voice broadcast control, broadcasting the mechanical ventilation data.

[0041] In the embodiments of the present application, the mechanical ventilation data can be displayed in the mechanical ventilation data setting page, wherein the mechanical ventilation data setting page can be used to clearly present the individualized treatment plan generated by the model, specifically including the mechanical ventilation data determined in the foregoing steps, to ensure that the object can quickly and accurately obtain all the recommended content.

[0042] The voice broadcast control can also be displayed in the mechanical ventilation data setting page. When the object (especially the object who is inconvenient to operate or has a visual assistance demand) triggers the control, the object using the terminal can call the built-in text-to-speech engine to convert the mechanical ventilation data (including the mode, parameter value, and cause analysis text, etc.) displayed in the mechanical ventilation data setting page into a voice signal and play it, for example, the voice outputs the following content: the recommended mode for you is Auto-CPAP, the minimum pressure is 6 cmH2O, and the maximum pressure is 14 cmH2O. The recommended reason is that the target object belongs to severe sleep apnea and is obese, and this pressure range can effectively eliminate most respiratory events while taking into account comfort. For example, please refer to Figure 4 , Figure 4 The schematic diagram of the mechanical ventilation data setting page provided by the embodiments of the present application is shown in FIG. 4, wherein the mechanical ventilation data setting page 400 includes a voice broadcast control 410, in response to a triggering operation on the voice broadcast control 410, the mechanical ventilation parameter content and health analysis and safety suggestions shown in FIG. 5 can be broadcasted by voice. In this way, the generated mechanical ventilation parameters are output to the doctor or the target object in a clear and easy-to-understand manner, improving the usability of the intelligent recommended scheme of the mechanical ventilation parameters. Figure 4 Figure 4

[0043] ​​The embodiments of the present application realize the bimodal output of key information, i.e., visual display and voice broadcast, at the user interface interaction level. Therefore, the present application not only guarantees the intuitiveness and accuracy of information presentation, facilitates the object to carefully read and confirm the complex treatment parameters, but also provides convenience for the object with visual impairment or who prefers to receive auditory information, reduces the risk of operation errors caused by misreading or missing parameters. At the same time, the voice broadcast reason analysis text helps doctors to more naturally understand the decision logic of the AI model, thereby improving the clinical acceptance of the intelligent scheme of mechanical ventilation parameters.

[0044] In some embodiments, the training process of the neural network model includes the following steps: obtaining training sample data; updating the model parameters of the neural network model by taking the sample object attribute data and the sample sleep respiration detection data as the input of the preset neural network model, and taking the sample mechanical ventilation data as the output label of the neural network model.

[0045] In the embodiments of the present application, the training sample data can be obtained first, wherein each training sample data includes sample object attribute data, sample sleep respiration detection data and sample mechanical ventilation data of a sample target object, and can also include the clinically verified effective mechanical ventilation parameters corresponding to each sample target object, such as CPAP pressure and Auto-CPAP pressure range, and can also include corresponding treatment effect feedback data, which can specifically include indicators after treatment, such as AHI value after treatment, blood oxygen improvement (such as blood oxygen saturation change) and satisfaction of the sample target object, and each training sample data can correspond to a complete record of a historical sleep respiration treatment process. The sample object attribute data includes the basic information of the sample target object, and the sample sleep respiration detection data is derived from the sleep condition detection result of the sample target object; and the sample mechanical ventilation data can include the treatment scheme verified by the clinic as effective for the sample target object, which can specifically include the mechanical ventilation model and the corresponding mechanical ventilation parameters, for example, the optimal CPAP pressure value or Auto-CPAP pressure range determined by the doctor through PSG titration method can be taken as the standard answer required by the model to learn, i.e., the output label. Before training the neural network model according to the training sample data, the training sample data can be cleaned and standardized to obtain a feature vector in a unified format, so as to ensure the data quality and meet the input requirements of model training.

[0046] Further, the sample object attribute data and the sample sleep respiration detection data are taken as inputs of a preset neural network model, and the sample mechanical ventilation data is taken as an output label of the neural network model, and model parameters of the neural network model are updated. Specifically, the neural network model learns a mapping relationship from object features to optimal mechanical ventilation parameters by receiving a large amount of input data and output label pairs. In the training iteration process, the model makes a prediction according to the input data, compares the prediction result with the real sample mechanical ventilation data (i.e., the output label, such as CPAP pressure, Auto-CPAP pressure range, IPAP / EPAP of BiPAP, and standby respiratory frequency, etc.), calculates the value of a loss function (such as mean square error), and then adjusts and updates the model parameters (i.e., weights and biases) according to the loss value gradient by using an optimization method such as a back propagation algorithm, so that the prediction output of the model is continuously close to the real clinically effective parameters, and the prediction error is minimized. It should be noted that the treatment effect feedback data can be used for supervised learning in model training to optimize the prediction accuracy, but not as a direct output label.

[0047] The embodiments of the present application can obtain high-quality training sample data, perform supervised parameter learning on the neural network model, and ensure that the model deeply mines the internal correlation between the features of the target object and the mechanical ventilation parameters. In this way, the finally trained model can have good generalization ability and can accurately generate individualized mechanical ventilation data for unobserved target objects, thereby ensuring the clinical effectiveness and reliability of the intelligent mechanical ventilation parameter recommendation scheme.

[0048] In some embodiments, the sample object attribute data and the sample sleep respiration detection data are taken as inputs of a preset neural network model, and the sample mechanical ventilation data is taken as an output label of the neural network model, and model parameters of the neural network model are updated, including the following steps: The sample object attribute data and the sample sleep respiration detection data are taken as inputs of a preset neural network model to perform mechanical ventilation parameter prediction, and predicted mechanical ventilation data is obtained. The mechanical ventilation parameter loss is calculated according to the sample mechanical ventilation data and the predicted mechanical ventilation data, and the model parameters of the neural network model are updated according to the mechanical ventilation parameter loss.

[0049] In the embodiment of the present application, the sample object attribute data and the sample sleep respiration detection data are used as the input of the preset neural network model to predict the mechanical ventilation parameter, and the predicted mechanical ventilation data is obtained. This step is the forward propagation process in the model training process. The sample object attribute data and the sample sleep respiration detection data contained in the training sample data in a batch are used as the input of the neural network model. The model encodes, fuses and reasons the input information through the internal multi-layer nonlinear transformation and feature interaction processing, and finally outputs the predicted mechanical ventilation data. For example, the model may predict that the Auto-CPAP pressure range suitable for the sample target object is a minimum pressure of 7 cmH2O and a maximum pressure of 13 cmH2O.

[0050] It should be noted that the present application adopts the Cosine Annealing Warm Restarts strategy to periodically restart the learning rate in the later training period to avoid local optimum, and uses FP16 and FP32 mixed precision to accelerate the calculation, which greatly reduces the memory occupation.

[0051] Then, the mechanical ventilation parameter loss can be calculated according to the sample mechanical ventilation data and the predicted mechanical ventilation data, and the model parameters of the neural network model can be updated according to the mechanical ventilation parameter loss. The loss function needs to optimize the mechanical ventilation parameter prediction (regression task) and the corresponding treatment effect evaluation (classification and regression joint task) at the same time. After the loss value is calculated, the gradient of the loss value to the model learnable parameters (i.e. model parameters, including weights and biases) can be calculated through the back propagation mechanism (such as the gradient descent optimizer), and the model parameters are updated according to the optimization rule. In this way, the loss of the next prediction is continuously reduced, so that the prediction output of the model gradually approaches the true clinical parameters.

[0052] The training method provided in the embodiment of the present application can enable the neural network model to continuously learn from a large number of labeled clinical samples and gradually optimize its internal representation and mapping ability. This supervised end-to-end training method ensures that the model can accurately capture the deep nonlinear relationship between the complex multi-modal object features and the individualized and accurate mechanical ventilation parameters, thereby improving the accuracy of setting the mechanical ventilation parameters.

[0053] In some embodiments, the training sample data further includes sample object feedback data obtained after the sample target object is subjected to mechanical ventilation treatment based on the predicted mechanical ventilation data, and the model parameters of the neural network model are updated according to the mechanical ventilation parameter loss, including the following steps: The neural network model is called to predict the mechanical ventilation effect based on the sample object attribute data, the sample sleep respiration detection data and the predicted mechanical ventilation data, and the predicted object feedback data is obtained. The mechanical ventilation effect loss is calculated according to the predicted object feedback data and the sample object feedback data, and the model parameters of the neural network model are updated based on the mechanical ventilation effect loss and the mechanical ventilation parameter loss.

[0054] In the embodiments of the present application, the neural network model can be called to predict the mechanical ventilation effect based on the sample object attribute data, the sample sleep respiration detection data and the predicted mechanical ventilation data to obtain the predicted object feedback data. In this way, the learning goal of the neural network model can be expanded, which requires it not only to predict the mechanical ventilation parameters, but also to predict the treatment results. The model outputs the predicted object feedback data based on the comprehensive reasoning of the above information, that is, estimates the effect that the sample target object may produce after treatment with the predicted mechanical ventilation parameters, such as the predicted AHI value after treatment, the average blood oxygen saturation, the air leakage index or the object comfort score, etc.

[0055] Further, the mechanical ventilation effect loss can be calculated according to the predicted object feedback data and the sample object feedback data, and the model parameters of the neural network model are updated based on the mechanical ventilation effect loss and the mechanical ventilation parameter loss. Specifically, the expression of the mechanical ventilation parameter loss constructed by the present application is as follows:

[0056] Among them, represents the total loss function, represents the mechanical ventilation parameter loss, which is a weighted mean absolute error (MAE) loss, and gives a higher weight to the key parameter (CPAP pressure). The mechanical ventilation parameter loss is used to quantify the error between the predicted mechanical ventilation data and the sample mechanical ventilation data. For example, the minimum pressure in the true label is 6.5 cmH2O, and the model predicts 7 cmH2O, then the mean square error loss will calculate the square of this difference. represents the mechanical ventilation effect loss, that is, the loss function representing the treatment effect, which is calculated by combining the classification loss (patient satisfaction) and the regression loss (AHI / blood oxygen improvement rate). and are weight parameters corresponding to the mechanical ventilation parameter loss and the treatment effect loss, respectively, is a regularization term used to control the responsibility of the model, Typically, it is 0.01. The mechanical ventilation effectiveness loss is used to quantify the difference between the predicted subject feedback data and the sample subject feedback data. Subsequently, the training process combines the mechanical ventilation effectiveness loss with the mechanical ventilation parameter loss (a loss that measures the prediction accuracy of the parameters) calculated in the previous step, typically by weighted summation to form a joint loss. The optimizer of the model aims to minimize this joint loss, so that during backpropagation, the update direction of the model parameters of the neural network model is guided by both the parameter prediction accuracy and the effectiveness prediction rationality. The calculation expression of the joint loss provided by the present application is as follows:

[0057] including binary cross-entropy (BCE) and mean squared error (MSE), represents the satisfaction of the sample target object predicted by the model, represents the sample satisfaction of the sample target object in the training sample data, i.e., the sample object feedback data, represents the AHI value after treatment of the mechanical ventilation data predicted by the model, represents the sample AHI value of the sample target object in the training sample data, represents a weight parameter for controlling the attention degree of the above two losses.

[0058] The multi-task learning and optimization paradigm of the embodiments of the present application enables the model to not only learn to accurately predict the mechanical ventilation parameters during the training process, but also learn to predict the treatment effect brought by the mechanical ventilation parameters, so that the optimization target of the model is improved from simple parameter matching to curative effect guidance. In this way, the model can be forced to internalize medical knowledge, understand the complex relationship between object characteristics, mechanical ventilation parameters and clinical treatment results, and provide personalized mechanical ventilation parameter setting schemes with explainability.

[0059] In step S103 of some embodiments, the system obtains object feedback data after the target object is treated by mechanical ventilation based on the mechanical ventilation data. The object feedback data includes objective treatment data and subjective patient feedback. The objective data is derived from the actual treatment parameters recorded by the built-in sensors of the ventilator, such as AHI after treatment, air leakage, pressure curve, blood oxygen change, etc.; the subjective feedback includes the patient's evaluation of treatment comfort, mask adaptability, sleep quality, etc. These feedback data reflect the effect of the recommended parameters in actual application, providing an important basis for subsequent evaluation and adjustment.

[0060] ​​In step S104 of some embodiments, the neural network model can be invoked to evaluate the mechanical ventilation effect based on the subject attribute data, the sleep respiration detection data, the mechanical ventilation data, and the subject feedback data, to obtain mechanical ventilation effect evaluation data. Specifically, the model evaluates whether the mechanical ventilation parameters achieve the preset treatment effect by comprehensively analyzing the historical data and the feedback data of the current treatment, such as whether the AHI value of the target subject is reduced to the normal range (for example, the AHI value is less than 5 times / hour), whether the blood oxygen is improved, and whether the target subject is uncomfortable, and then outputs the mechanical ventilation effect evaluation data. The model can quantify the generated mechanical ventilation effect evaluation data into a score or a classification label, such as a label that the mechanical ventilation parameters need to be adjusted or the label that the mechanical ventilation parameters are effective, thereby providing decision support for whether the mechanical ventilation parameter optimization is needed.

[0061] In step S105 of some embodiments, when the mechanical ventilation effect evaluation data indicates that the mechanical ventilation effect does not meet the preset mechanical ventilation effect condition, the neural network model can be invoked again to adjust the mechanical ventilation data based on the subject attribute data, the sleep respiration detection data, the mechanical ventilation data, the subject feedback data, and the mechanical ventilation effect evaluation result, to obtain updated mechanical ventilation data. Further, the model performs inference again by fusing multiple sources of information, such as reducing the pressure P max or adjusting the pressure range or the response parameter when the AHI value after treatment does not meet the standard, to achieve dynamic optimization of the parameters. In this way, the embodiments of the present application can ensure that the treatment parameters can be adaptively adjusted according to the state change of the target subject, and further improve the accuracy of the mechanical ventilation parameter setting.

[0062] In some embodiments, please refer to Figure 5, time series data such as respiratory waveform, blood oxygen saturation curve, chest and abdominal movement, etc. from polysomnography or portable monitoring equipment, non-time series data indicating static basic information of the target object. Treatment feedback data refers to data generated after the target object is treated using the initial recommended parameters. The data preprocessing engine is used to clean, standardize, normalize, and feature extract the above multi-source heterogeneous data, and convert the original data into a unified format suitable for model processing. The model uses a double-branch feature extraction architecture to process different types of data, where the time series feature branch is used to extract long-term dependencies and dynamic patterns in time series data, and the static feature branch is used to learn deep features in non-time series data. The dynamic feature interaction layer includes a cross-attention mechanism that fuses time series data and non-time series data through the cross-attention mechanism. The high-dimensional features after fusion are sent to multiple parallel task heads, which simultaneously complete different tasks and achieve collaborative optimization. Specifically, the parameter prediction head (regression) regresses to predict specific mechanical ventilation parameters, the treatment effect evaluation head (classification / regression) predicts the treatment effect that may be brought about by the parameter setting, such as API target achievement probability (e.g. probability of AHI <5) and blood oxygen improvement rate, etc., and the root cause analysis head (multi-label classification) provides interpretability. If the prediction effect is not good, this module will analyze the possible failure root causes, such as the main central respiratory events not being solved or possibly caused by mask leakage, to provide direction for subsequent adjustment. Through clinical verification, the model-recommended parameters are applied to clinical practice, and real treatment feedback data is collected. If the effect evaluation does not meet expectations, new feedback data can be collected, combined with root cause analysis results, to automatically generate parameter adjustment suggestions. New data can be used to retrain the model, enabling the entire system to continuously learn and evolve, and become more accurate.

[0063] See Figure 6 The embodiment of the present application also provides an AI large model-based mechanical ventilation data processing device 600, which can implement the AI large model-based mechanical ventilation data processing method described above. The device comprises: A first acquisition unit 610 is configured to acquire object attribute data and sleep respiration detection data of a target object. A generation unit 620 is configured to call a preset neural network model to generate mechanical ventilation parameters based on the object attribute data and the sleep respiration detection data, and obtain mechanical ventilation data corresponding to the target object. A second acquisition unit 630 is configured to acquire object feedback data of the target object after mechanical ventilation treatment based on the mechanical ventilation data. An evaluation unit 640 is configured to call a neural network model to evaluate the effect of mechanical ventilation based on the object attribute data, the sleep respiration detection data, the mechanical ventilation data, and the object feedback data, and obtain mechanical ventilation effect evaluation data. The updating unit 650 is configured to, when the mechanical ventilation effect evaluation data indicates that the mechanical ventilation effect does not meet the preset mechanical ventilation effect condition, call the neural network model to adjust the mechanical ventilation data based on the object attribute data, the sleep respiration detection data, the mechanical ventilation data, the object feedback data, and the mechanical ventilation effect evaluation result, to obtain updated mechanical ventilation data.

[0064] The specific implementation of the AI large model-based mechanical ventilation data processing apparatus is basically the same as that of the above-described AI large model-based mechanical ventilation data processing method, and will not be described here again.

[0065] The embodiments of the present application further provide an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor implements the above-described AI large model-based mechanical ventilation data processing method when executing the computer program. The electronic device can be any intelligent terminal, such as a tablet computer or a vehicle-mounted computer.

[0066] Please refer to Figure 7 , Figure 7 The hardware structure of the electronic device of another embodiment is illustrated, which includes: The processor 701 can be implemented in a general-purpose central processing unit (CPU), a microprocessor, an application specific integrated circuit (ASIC), or one or more integrated circuits, and is configured to execute related programs to implement the technical solutions provided by the embodiments of the present application. The memory 702 can be implemented in a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 702 can store an operating system and other application programs. When the technical solutions provided by the embodiments of the present application are implemented by software or firmware, the related program codes are stored in the memory 702 and called and executed by the processor 701 to implement the AI large model-based mechanical ventilation data processing method of the embodiments of the present application. The input / output interface 703 is configured to realize information input and output. The communication interface 704 is configured to realize the communication interaction between the device and other devices. The communication can be realized by a wired manner (for example, a USB, a network cable, etc.) or a wireless manner (for example, a mobile network, WIFI, Bluetooth, etc.). The bus 705 transmits information between various components (for example, the processor 701, the memory 702, the input / output interface 703, and the communication interface 704) of the device. The processor 701, the memory 702, the input / output interface 703, and the communication interface 704 are communicatively connected with each other through the bus 705.

[0067] The application also provides a computer readable storage medium, which stores a computer program. The computer program is executed by a processor to implement the mechanical ventilation data processing method based on an AI large model.

[0068] The memory, as a non-transitory computer readable storage medium, can be used to store non-transitory software programs and non-transitory computer executable programs. In addition, the memory can include a high-speed random access memory and can also include a non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state memory device. In some embodiments, the memory can optionally include a memory remotely arranged relative to the processor, and these remote memories can be connected to the processor through a network. Examples of the above network include but are not limited to the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof.

[0069] The embodiments described in the application embodiments are used to more clearly illustrate the technical solutions of the application embodiments, and do not constitute a limitation on the technical solutions provided by the application embodiments. Those skilled in the art can know that, with the evolution of technology and the appearance of new application scenarios, the technical solutions provided by the application embodiments are also applicable to similar technical problems.

[0070] Those skilled in the art can understand that the technical solutions shown in the figures do not constitute a limitation on the application embodiments, and can include more or fewer steps than the figures, or combine certain steps or different steps.

[0071] The device embodiments described above are only schematic, and the units described as separate components can or can not be physically separated, that is, they can be located in one place or distributed on multiple network units. According to actual needs, part or all of the modules can be selected to achieve the purpose of the embodiments.

[0072] Those skilled in the art can understand that all or some steps in the above disclosed method, the function modules / units in the system and the device can be implemented as software, firmware, hardware and their appropriate combinations.

[0073] The terms "first", "second", "third", "fourth", and the like in the description and in the claims of this application, if any, are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the use of the terms so termed is interchangeable under appropriate circumstances such that the embodiments of the application described herein are, for example, capable of orderly or chronological mundane operation, reverse order operation, based on circuitry availability, based on stated preference or the like, and that "default" or other orderings are thus permissible. Further, the terms "comprise", "comprising", "include", "including", and the like, are specifically intended to be open-ended. That is, references to individual steps and the like do not suhstantially exclude the presence of two or more of a recited step or its integral sub-steps or additional steps whether or not readily ascertainable from the description or the like. Further, the words "a" or "an", as used herein in the disclosure and elsewhere, are used indiscriminately and are to be interpreted in the same way, i.e. as meaning "one or more".

[0074] It should be understood that, in the application, "at least one" refers to one or more, and "multiple" refers to two or more. "And / or" is used to describe the relationship between the associated objects, which means that there can be three relationships, for example, "A and / or B" can represent three cases: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally represents an "or" relationship between the front and rear associated objects. "At least one of the following" or similar expressions means any combination of these items, including any combination of single or multiple items. For example, at least one of a, b or c can mean a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0075] In several embodiments provided in the application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the above-mentioned units is only a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be omitted or not executed. In addition, the coupling or direct coupling or communication connection between the displayed or discussed objects can be indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0076] The units described above as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, i.e. they can be located in one place or distributed on multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiment of the application.

[0077] In addition, each function unit in each embodiment of the present application can be integrated in one processing unit, or each unit can be physically present separately, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software function unit.

[0078] If the integrated unit is realized in the form of a software function unit and sold or used as an independent product, it can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application, essentially or in part, or all or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes multiple instructions used to cause a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods in the embodiments of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various other media that can store programs.

[0079] The preferred embodiments of the embodiments of the present application are described above with reference to the accompanying drawings, and are not limited to the scope of the embodiments of the present application. Any modifications, equivalent replacements and improvements made by those skilled in the art without departing from the scope and essence of the embodiments of the present application shall be within the scope of the embodiments of the present application.

Claims

1. An AI large model-based mechanical ventilation data processing method, characterized in that, The method comprises: obtaining object attribute data and sleep respiration detection data of a target object; calling a preset neural network model to generate mechanical ventilation parameters based on the object attribute data and the sleep respiration detection data, to obtain mechanical ventilation data corresponding to the target object; obtaining object feedback data after mechanical ventilation processing of the target object based on the mechanical ventilation data; calling the neural network model to evaluate mechanical ventilation effect based on the object attribute data, the sleep respiration detection data, the mechanical ventilation data and the object feedback data, to obtain mechanical ventilation effect evaluation data; when the mechanical ventilation effect evaluation data indicates that the mechanical ventilation effect does not meet a preset mechanical ventilation effect condition, calling the neural network model to adjust the mechanical ventilation data based on the object attribute data, the sleep respiration detection data, the mechanical ventilation data, the object feedback data and the mechanical ventilation effect evaluation result, to obtain updated mechanical ventilation data.

2. The method of claim 1, wherein, The training process of the neural network model comprises the following steps: obtaining training sample data, each of the training sample data comprising sample object attribute data, sample sleep respiration detection data and sample mechanical ventilation data of a sample target object; updating model parameters of a preset neural network model by taking the sample object attribute data and the sample sleep respiration detection data as inputs of the neural network model and taking the sample mechanical ventilation data as output labels of the neural network model.

3. The method of claim 2, wherein, The updating of the model parameters of the neural network model by taking the sample object attribute data and the sample sleep respiration detection data as inputs of the neural network model and taking the sample mechanical ventilation data as output labels of the neural network model comprises: performing mechanical ventilation parameter prediction by taking the sample object attribute data and the sample sleep respiration detection data as inputs of the neural network model, to obtain predicted mechanical ventilation data; calculating a mechanical ventilation parameter loss according to the sample mechanical ventilation data and the predicted mechanical ventilation data, and updating the model parameters of the neural network model according to the mechanical ventilation parameter loss.

4. The method of claim 3, wherein, The training sample data further comprises sample object feedback data after mechanical ventilation processing of the sample target object based on the predicted mechanical ventilation data, and the updating of the model parameters of the neural network model according to the mechanical ventilation parameter loss comprises: calling the neural network model to perform mechanical ventilation effect prediction based on the sample object attribute data, the sample sleep respiration detection data and the predicted mechanical ventilation data, to obtain predicted object feedback data; calculating a mechanical ventilation effect loss according to the predicted object feedback data and the sample object feedback data, and updating the model parameters of the neural network model based on the mechanical ventilation effect loss and the mechanical ventilation parameter loss.

5. The method of claim 1, wherein, The neural network model comprises a full connection network, a time sequence feature extraction network, and a cross attention network, the preset neural network model is called to generate mechanical ventilation parameters based on the object attribute data and the sleep respiration detection data, and mechanical ventilation data corresponding to the target object is obtained, which comprises: The full connection network is called to extract features of the object attribute data, and object attribute features are obtained; The sleep respiration detection data is segmented by a sliding time window, sleep respiration detection data sequences are obtained, and the time sequence feature extraction network is called to extract features of the sleep respiration detection data sequences, and sleep respiration detection data features are obtained; The cross attention network is called to perform cross attention processing on the object attribute features and the sleep respiration detection data features, and corresponding structured data is obtained; The structured data is used to generate mechanical ventilation parameters, and mechanical ventilation data corresponding to the target object is obtained.

6. The method of claim 5, wherein, The structured data is used to generate mechanical ventilation parameters, and mechanical ventilation data corresponding to the target object is obtained, which comprises: The structured data is used to generate mechanical ventilation parameters, and a mechanical ventilation mode set for the target object and mechanical ventilation parameters corresponding to the mechanical ventilation mode are obtained; A preset large language model is called to generate corresponding reason analysis text based on the mechanical ventilation mode and the mechanical ventilation parameters; The mechanical ventilation mode, the mechanical ventilation parameters, and the reason analysis text are used to determine mechanical ventilation data corresponding to the target object.

7. The method of claim 6, wherein, After the mechanical ventilation mode, the mechanical ventilation parameters, and the reason analysis text are used to determine mechanical ventilation data corresponding to the target object, the method further comprises: The mechanical ventilation data is displayed in a mechanical ventilation data setting page; A voice broadcast control is displayed in the mechanical ventilation data setting page, and in response to a triggering operation on the voice broadcast control, the mechanical ventilation data is broadcasted.

8. An AI large model-based mechanical ventilation data processing apparatus, characterized by, The device comprises: A first acquisition unit is configured to acquire object attribute data and sleep respiration detection data of a target object; A generation unit is configured to call a preset neural network model to generate mechanical ventilation parameters based on the object attribute data and the sleep respiration detection data, and obtain mechanical ventilation data corresponding to the target object; A second acquisition unit is configured to acquire object feedback data after mechanical ventilation processing of the target object based on the mechanical ventilation data; An evaluation unit is configured to call the neural network model to evaluate mechanical ventilation effect based on the object attribute data, the sleep respiration detection data, the mechanical ventilation data, and the object feedback data, and obtain mechanical ventilation effect evaluation data; An updating unit is configured to, when the mechanical ventilation effect evaluation data indicates that the mechanical ventilation effect does not meet a preset mechanical ventilation effect condition, call the neural network model to adjust the mechanical ventilation data based on the object attribute data, the sleep respiration detection data, the mechanical ventilation data, the object feedback data, and the mechanical ventilation effect evaluation result, to obtain updated mechanical ventilation data.

9. An electronic device, comprising: The electronic device comprises a memory and a processor, the memory stores a computer program, and the processor implements the AI large model-based mechanical ventilation data processing method in any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium storing a computer program, the computer program comprising instructions that, when executed by a computer, cause the computer to perform the method of any one of claims 1 to 9. The computer program is executed by the processor to implement the AI large model-based mechanical ventilation data processing method in any one of claims 1 to 7.

Images