Breathing machine control strategy debugging method and system based on EIT image

Abstract

The invention relates to the technical field of precise monitoring medical treatment, and discloses a breathing machine control strategy debugging method and system based on an EIT image, and the method comprises the steps: constructing a breathing machine control algorithm model based on deep learning through a case library, and representing a mapping relation between a case EIT image and a key index thereof and a breathing machine ventilation control parameter; acquiring current EIT images of a plurality of respiratory cycles of the patient, inputting the current EIT images into the respirator control algorithm model, matching the most similar case from the case library, and performing respirator ventilation control by using target respirator ventilation control parameters of the most similar case; operation is controlled to start, adjustment EIT images and key indexes of a plurality of respiratory cycles of the patient are obtained, and the adjustment EIT images and the key indexes are compared with the EIT images and the key indexes of the most similar cases for judgment; if the requirements are met, continuously controlling and updating the case library, and optimizing the model; otherwise, switching the control mode. Determination of breathing machine control parameters participated by a patient is reduced, trial and error times of clinical parameter adjustment are reduced, and patient safety and control accuracy are both considered.

Description

Technical Field

[0001] The present invention relates to the field of precision monitoring medical technology, and in particular to a ventilator control strategy debugging method and system based on EIT images. Background Art

[0002] The research field of combining electrical impedance tomography (EIT) technology with mechanical ventilation equipment control has made significant progress in recent years. The adaptive adjustment technology of ventilators based on EIT provides an innovative solution for mechanical ventilation. The practical significance of this technology is that it can provide regional physiological information of the lungs through EIT images, allowing doctors to more accurately monitor and evaluate the patient's lung function status, and adjust mechanical ventilation based on key indicators in EIT images to achieve personalized and optimized treatment.

[0003] When used, the EIT device is wrapped around the patient's chest, and multiple electrodes are used to send weak currents and measure voltage changes at different locations. These data reflect the conductivity distribution of the internal tissues of the chest cavity, and can generate images reflecting the ventilation status of the lungs. The acquired data is transmitted to the connected computer system in real time, and the data is analyzed and processed by a pre-built algorithm model. The algorithm model is designed to identify and interpret the features in the EIT images, such as key indicators such as alveolar expansion and gas exchange efficiency, and dynamically adjust the working parameters of the ventilator accordingly, such as tidal volume, inspiratory pressure, positive end-expiratory pressure (PEEP), etc., to achieve the best mechanical ventilation effect.

[0004] However, although this technology provides personalized treatment plans in theory, there are still some significant flaws in practical applications. One of the main problems is that it is highly dependent on real-time data streams for calculations and adjustments, which means that physiological data must be collected from patients continuously for a long time, and each adjustment may require additional time for effect feedback and optimization, which not only increases the workload of medical staff, but more importantly, it may increase the risk to patients due to the uncertainty caused by frequent adjustments. In addition, since each patient's physiological responses are different, finding the best settings that best suit the individual is often a time-consuming process, which may lead to certain risks due to the length of adjustment time, affecting the patient's immediate safety and treatment effect. Summary of the invention

[0005] The present invention intends to provide a ventilator control strategy debugging method and system based on EIT images, aiming to explore a more intelligent and faster response and clinical guidance based on EIT and ventilator joint control algorithm, strive to improve the timeliness and reliability of ventilator control, reduce unnecessary risks for patients, and ensure the immediate safety of patients.

[0006] The basic solution provided by the present invention is: a method for debugging a ventilator control strategy based on EIT images, the method comprising:

[0007] Constructing a ventilator control algorithm model based on deep learning using a case library, for characterizing the mapping relationship between the EIT images of cases, their key indicators, and the ventilator ventilation control parameters;

[0008] Obtaining the current EIT images of a patient for a number of respiratory cycles, and inputting them into the ventilator control algorithm model, using the current EIT images to match the most similar case from the case library, and retrieving the corresponding target EIT image, its key indicators, and the target ventilator ventilation control parameters of the most similar case according to the mapping relationship;

[0009] Controlling the ventilator ventilation using the target ventilator ventilation control parameters; starting from the beginning of the ventilator ventilation control operation, obtaining the adjusted EIT images and key indicators of the patient for a number of respiratory cycles, and comparing and judging them with the EIT images and key indicators of the most similar case;

[0010] If the requirements are met, control the ventilator ventilation according to the target ventilator ventilation control parameters, and at the same time, at least aggregate the EIT images, key indicators, and ventilator ventilation control parameters of this patient obtained during the process to form a new case and add it to the case library to optimize the ventilator control algorithm model; otherwise, switch to other control modes.

[0011] The present invention also provides a system for debugging a ventilator control strategy based on EIT images for implementing a method for debugging a ventilator control strategy based on EIT images. The system includes a ventilator, and further includes:

[0012] An acquisition unit, configured to acquire the EIT images of the patient as required and send them to the model unit, where the EIT images include the current EIT images of the patient for a number of respiratory cycles before the ventilator ventilation control operation, and the adjusted EIT images of the patient for a number of respiratory cycles starting from the beginning of the ventilator ventilation control operation;

[0013] A model unit is used to construct a case library; it is also used to construct a deep - learning - based ventilator control algorithm model using several cases in the case library to represent the mapping relationship between EIT images, their key indicators, and ventilator ventilation control parameters; it is further used to input the current EIT image into the ventilator ventilation control algorithm model, match the most similar case from the case library using the EIT image, and call the corresponding target EIT image, its key indicators, and target ventilator ventilation control parameters of the most similar case according to the mapping relationship; it is also used to send the target ventilator ventilation control parameters to the ventilator for ventilator ventilation control, and send the target EIT image and key indicators to the evaluation unit; it is further used to input the received adjusted EIT image into the ventilator ventilation control algorithm model, extract the corresponding key indicators, and send the adjusted EIT image and its key indicators to the evaluation unit; it is also used to at least aggregate the EIT image, key indicators, and ventilator ventilation control parameters of this patient obtained during the process to form a new case and add it to the case library to optimize the ventilator control algorithm model;

[0014] An evaluation unit is used to receive the adjusted EIT image and its key indicators, compare them with the target EIT image and its key indicators of the most similar case, and make a judgment. If the requirements are met, the current ventilator ventilation control is performed according to the target ventilator ventilation control parameters; otherwise, other control modes are switched.

[0015] The working principle and advantages of the present invention are as follows:

[0016] In order to overcome the defects of the prior art, this solution is based on the joint control application of electrical impedance tomography (EIT) and ventilators, and proposes a deep - learning - driven electrical impedance tomography method to achieve fast and effective guidance for clinical ventilator control. A simulation control model is constructed in advance to complete the optimization of ventilator ventilation control parameters. During clinical application, patients do not need to participate in the training process of the ventilator control algorithm itself, greatly reducing the patients' participation in determining ventilator control parameters and reducing the number of trial - and - error times for clinical parameter adjustment, taking into account both patient safety and the accuracy of control parameters.

[0017] Based on a large number of existing cases and combined with deep learning technology, this solution constructs a ventilator control algorithm model that characterizes the mapping relationship between EIT images, their key indicators, and ventilator ventilation control parameters, improving the model adaptability and prediction accuracy. It provides a rapid feedback mechanism for clinical applications. With just a few respiratory cycles of the patient (such as 2 - 5 respiratory cycles), the most matching ventilator control parameters can be directly retrieved from the ventilator control algorithm model through EIT images for ventilator control operation, quickly completing the ventilator control decision-making judgment. At the same time, it runs several respiratory cycles for a quick assessment of the ventilation effect and, in case of non-compliance, quickly and reasonably makes switching measures to reduce patient mechanical ventilation injury and maximize patient safety and treatment effect. The electrical impedance tomography method driven by deep learning enhances the ability of EIT technology to accurately reflect lung function, can reveal the deep mechanism of lung ventilation status, and then improve the accuracy and efficiency of lung electrical impedance tomography, enabling more accurate adjustment parameters to be provided for ventilator ventilation and enhancing the accuracy and safety of clinical applications. Additionally, through the continuous addition of new cases during clinical application, the ventilator control algorithm model is continuously and effectively optimized to improve the matching accuracy of ventilator control parameters and further optimize the co-control effect.

[0018] This solution theoretically promotes the development of EIT imaging technology and deep learning in the field of biomedical imaging. In practical applications, it provides a more efficient and accurate method for the diagnosis and treatment of lung diseases, with significant clinical application value and social significance. Brief Description of the Drawings

[0019] Figure 1 It is a schematic flowchart of a method for debugging a ventilator control strategy based on EIT images provided by an embodiment of the present invention;

[0020] Figure 2 It is a schematic diagram of the process for verifying the effectiveness of a lung perfusion / ventilation simulation device provided by an embodiment of the present invention;

[0021] Figure 3 It is a schematic diagram of the training process of a ventilator control algorithm model provided by an embodiment of the present invention;

[0022] Figure 4 It is a schematic structure diagram of a system for debugging a ventilator control strategy based on EIT images provided by an embodiment of the present invention Figure 1 ;

[0023] Figure 5 It is a schematic structure diagram of a system for debugging a ventilator control strategy based on EIT images provided by an embodiment of the present invention Figure 2 。 Detailed Description of the Embodiment

[0024] The following is a further detailed description through specific embodiments:

[0025] Embodiment 1

[0026] Basically as shown in the appendix Figure 1 A method for debugging a ventilator control strategy based on EIT images, the method comprising:

[0027] Construct a ventilator control algorithm model based on deep learning using a case library to characterize the mapping relationship between the EIT images of cases, their key indicators, and ventilator ventilation control parameters.

[0028] Specifically, obtain a number of actual real clinical cases to form a case library. Each case in the case library is not limited to including real lung EIT images, their key indicators, and real ventilator ventilation control parameters, covering from basic personal information such as age, gender, height, and weight to more complex physiological state parameters. Physiological state parameters include the patient's respiratory physiological data and blood gas analysis results. Respiratory physiological data such as respiratory waveforms, tidal volume, airway pressure and other respiratory mechanics indicators, lung EIT images and key indicators can reflect the ventilation distribution in the lung area; blood gas analysis results, such as arterial oxygen saturation, end-tidal carbon dioxide partial pressure (ETCO 2 ) etc., can evaluate the gas exchange efficiency; it also involves clinical records, such as the patient's medical history, current diagnosis, treatment response, and complication situation. These comprehensive information is crucial for optimizing mechanical ventilation settings, monitoring the progress of the disease, and adjusting treatment plans. By integrating the multi-dimensional data of cases, it can help deep learning improve the accuracy of the ventilator control algorithm model of this solution.

[0029] This solution combines the electrical impedance tomography method and deep learning technology to explore new algorithms to improve the speed and accuracy of imaging. The deep learning method can learn the complex relationships between impedance images and in-vivo impedance distributions, and between impedance images and ventilator ventilation control parameters from a large amount of training data, so as to achieve faster and more accurate image reconstruction and optimization of ventilator ventilation control parameters.

[0030] First, a lung perfusion / ventilation simulation device is constructed, which can simulate the respiration of a living lung with high precision. Among them, by deeply studying the key mechanical factors in the breathing process, such as pulmonary gas exchange, respiratory tract mechanical properties, and lung tissue elastic properties, based on detailed biomechanical principles, the development of the lung perfusion / ventilation simulation device is carried out. This device is not limited to simulating the dynamic behavior of the lungs and chest cavity, including the simulation of lung inflation and contraction, as well as chest wall movements related to breathing, and can accurately reproduce the biophysical characteristics during human breathing, improving the accuracy of the model and the accuracy of imaging feedback. The kinetic parameters of the lung perfusion / ventilation simulation device are adjustable, and it can simulate the respiratory systems corresponding to various different types, different age groups, different genders, etc. The main set parameters include airway resistance R (adjustable range 5 - 500 cmH 2 O·L －1 ·min －1 ), airway compliance C (adjustable range 10 - 150 mL / cm H 2 O), spontaneous breathing frequency (adjustable range 1 - 60 bpm), and spontaneous breathing tidal volume (adjustable range 50 - 2000 mL).

[0031] As Figure 2 shown, before model training, the effectiveness of the lung perfusion / ventilation simulation device is verified, that is, the performance of the device in capturing the dynamic changes of the lung ventilation state is verified, improving the accuracy and efficiency of pulmonary electrical impedance tomography. Specifically, call the existing cases in the case library, adjust the parameters of the lung perfusion / ventilation simulation device and run it, obtain the verification EIT imaging based on the breathing state simulated by the simulation device, and compare it with the real EIT imaging of this case to judge the effectiveness of the lung perfusion / ventilation simulation device by the imaging difference.

[0032] During the EIT experiment, assuming that the excitation current is constant, the inverse problem of EIT (i.e., image reconstruction) reconstructs the conductivity distribution σ of the measured area through the boundary measurement voltage U and the given current density J. In pulmonary EIT, regional changes in the electro-biological impedance of lung tissue can be observed in dynamic imaging. During human inspiration, the alveolar wall is stretched, resulting in a local increase in tissue impedance, thus causing an increase in the measured voltage value in the corresponding area. The Tikhonov regularization algorithm can be used for imaging to obtain an image simulation of the change in lung area.

[0033] Select the key index PT that characterizes the change in lung morphology during breathing. PT can be the ratio of the lung cross-sectional area at the current moment to the lung cross-sectional area at the end of inspiration, or other indexes that can characterize the change in lung morphology.

[0034] As Figure 3 shown, after the lung perfusion / ventilation simulation device is effective, it enters the adaptive algorithm training stage to construct a deep learning-based ventilator control algorithm model, including:

[0035] Take the key index PT, which characterizes the morphological changes of the lungs during the breathing process, as the target input and set the initial value.

[0036] Input the case parameters into the lung perfusion / ventilation simulation device (ventilation platform), collect EIT data, reconstruct the PC image, extract the key indexes of the image, and simulate to form a real-time response curve. Compare it with the ideal response curve, adjust the ventilation control parameters of the ventilator, so that the real-time response curve formed by the simulation conforms to the ideal response curve. The ventilator control algorithm model learns the adjustment process of the ventilation control parameters of the ventilator. When the average error of the key indexes is within 0.6%, the simulation results show that using the key indexes as input parameters can better control the ventilation system, and thus complete the construction of the ventilator control algorithm model.

[0037] The deep learning model is completed by using the existing model, which is not limited here, as long as the functions of this solution can be realized.

[0038] Conduct a mechanical ventilation test, and the test platform is as Figure 4 shown:

[0039] 1) Set the ventilator parameters. Set the ideal body weight to 60 kg, insert an endotracheal tube, the inner diameter of the artificial airway is 8.0 mm, the positive end-expiratory pressure (PEEP) is set to 2 cmH2O, the fraction of inspired oxygen (FiO2) is set to 21%, the expiratory sensitivity (Esens) is default 3%, the flow trigger (V’SENS) is set to 3.0 L / min, and each alarm parameter is set to the maximum value.

[0040] 2) Randomly select all different ventilation modes, PAV (the proportionality coefficients are PA35%, PA45%, PA65%, PA75%, PA89% respectively), PSV (the support pressure range is 12 - 20 cmH 2 O), BiLevel (the high-pressure phase pressure range is 12 - 20 cmH 2 O). After selecting the mode, run for about 5 minutes and record each respiratory mechanics measurement parameter of the ventilator.

[0041] 3) Set the R value (the setting range is 0 - 20 cmH 2 O·L －1 ·min －1 ) and the C value (the setting range is 19 - 90 mL / cm H 2 O) of the simulated lung according to the case parameters, and repeat step 2.

[0042] A total of 189 ventilation tests were simulated in this study. Table 1 shows the effects of PAV ventilation with different assist ratios on other respiratory mechanics parameters. The peak airway pressure (Ppeak), mean airway pressure (Pmean), exhaled tidal volume (Vte), minute ventilation volume (Vetot), and inspiratory time (Ti) were positively correlated with the proportionality coefficient PA, and the correlation coefficients r were all positively correlated, with a significance level of P < 0.01. It can be seen that the PAV mode achieved real-time and effective estimation of respiratory system mechanics parameters, provided ventilation support proportional to the degree of the patient's inspiratory effort, realized the control of the ventilator to deliver gas by the simulation device throughout the inspiratory phase, and had good synchronization performance.

[0043] Table 1 One-way ANOVA of the effects of PAV ventilation with different assist ratios on respiratory mechanics parameters

[0044]

[0045]

[0046] In actual application, replace the lung simulation device with the patient's actual real lungs, obtain the current EIT images of several respiratory cycles of the patient, and input them into the ventilator ventilation control algorithm model. Use the current EIT images to match the most similar cases from the case library, and retrieve the corresponding target EIT images, their key indicators, and target ventilator ventilation control parameters of the most similar cases according to the mapping relationship.

[0047] Specifically, to match the most similar case from the case library using the current EIT image, calculate the similarity between the current EIT image and the case EIT image, and determine the case corresponding to the case EIT image with the highest similarity as the most similar case. Specifically, it can be determined by using a similarity metric to evaluate the correspondence between the two images. In addition, the changes in key indicators can also be compared.

[0048] Use the target ventilator ventilation control parameters to control the ventilator ventilation; starting from the operation of the ventilator ventilation control, obtain the adjusted EIT images and key indicators of several respiratory cycles of the patient, and compare and judge them with the EIT images and key indicators of the most similar cases.

[0049] Specifically, if the requirements are met, control the ventilator ventilation according to the target ventilator ventilation control parameters, and at the same time, at least aggregate the EIT images, key indicators, and ventilator ventilation control parameters of this patient obtained during the process to form a new case and add it to the case library to optimize the ventilator control algorithm model; otherwise, switch to other control modes.

[0050] Specifically, meeting the requirements means the values and the time of stable maintenance that the oxygenation parameters represented by the blood oxygen saturation and the ventilation parameters represented by the carbon dioxide partial pressure need to reach under the treatment of the ventilator ventilation control operation for the patient.

[0051] Switching to other control modes includes switching to positive end-expiratory pressure (PEEP) control mode. When the control effect of positive end-expiratory pressure does not meet the requirements, switch to manual control mode, which is controlled by medical staff. Thus, an orderly switching of the three control modes can be formed to balance the timeliness and accuracy of control and the degree of medical staff's control intervention.

[0052] As Figure 5 shown, the present invention executes a debugging method for a ventilator control strategy based on EIT images, and also provides a debugging system for a ventilator control strategy based on EIT images. The system includes a ventilator and further includes:

[0053] An acquisition unit, configured to acquire EIT images of a patient as required and send them to the model unit. The EIT images include the current EIT images of several respiratory cycles of the patient before the ventilator ventilation control runs, and the adjusted EIT images of several respiratory cycles of the patient starting from the beginning of the ventilator ventilation control run;

[0054] A model unit, configured to build a case library; also configured to build a ventilator control algorithm model based on deep learning by using several cases in the case library to represent the mapping relationship between EIT images and their key indicators and ventilator ventilation control parameters; also configured to input the current EIT image into the ventilator ventilation control algorithm model, match the most similar case from the case library by using the EIT image, and call the corresponding target EIT image, its key indicators and target ventilator ventilation control parameters of the most similar case according to the mapping relationship; also configured to send the target ventilator ventilation control parameters to the ventilator for ventilator ventilation control, and send the target EIT image and key indicators to the evaluation unit; also configured to input the received adjusted EIT image into the ventilator ventilation control algorithm model, obtain the corresponding key indicators, and send the adjusted EIT image and its key indicators to the evaluation unit; also configured to at least aggregate the EIT images, key indicators and ventilator ventilation control parameters of this patient obtained during the process to form a new case and add it to the case library to optimize the ventilator control algorithm model;

[0055] An evaluation unit, configured to receive the adjusted EIT image and its key indicators, compare them with the target EIT image and its key indicators of the most similar case, and make a comparison and judgment. If it meets the requirements, perform the current ventilator ventilation control according to the target ventilator ventilation control parameters; otherwise, switch to other control modes.

[0056] The acquisition unit includes an EIT imaging device;

[0057] The model unit includes a lung perfusion / ventilation simulation device and a host computer, where the host computer stores a case library, a ventilator ventilation control algorithm model, and a model training program.

[0058] The lung perfusion / ventilation simulation device is used to provide simulated lung perfusion and lung ventilation parameters during model training, and the parameters include respiratory rate, expiratory volume, inspiratory volume, expiratory pressure change and inspiratory pressure change.

[0059] It can be understood that the present system is able to fully execute the above method, and the specific process will not be described in detail.

[0060] The present embodiment is a method and system for debugging a ventilator control strategy based on EIT images. Based on a large number of existing cases and combined with deep learning technology, a ventilator control algorithm model is constructed to characterize the mapping relationship between EIT images and their key indicators and ventilator ventilation control parameters, thereby improving the model adaptability and prediction accuracy. A rapid feedback mechanism is provided for clinical applications. Only a few breathing cycles of the patient are required to directly retrieve the most matching ventilator control parameters from the ventilator control algorithm model through the EIT image to perform ventilator control operation, quickly complete ventilator control decision judgment, and test several breathing cycles to quickly evaluate the ventilation effect. When the requirements are not met, switching measures are quickly and reasonably made to reduce the patient's mechanical ventilation injury, maximize the protection of patient safety and improve the treatment effect. The electrical impedance imaging method driven by deep learning improves the ability of EIT technology in accurately reflecting lung function, can reveal the deep mechanism of lung ventilation status, and thus improve the accuracy and efficiency of lung electrical impedance imaging, can provide more accurate adjustment parameters for ventilator ventilation, and improve the accuracy and safety of clinical applications. In addition, through the continuous addition of new cases during clinical application, the ventilator control algorithm model can be continuously and effectively optimized, the matching accuracy of the ventilator control parameters can be improved, and the joint control effect can be further optimized.

[0061] Embodiment 2

[0062] Taking an adult patient case as an example, a 50-year-old male patient was admitted to the ICU and received mechanical ventilation treatment due to respiratory failure caused by severe pneumonia.

[0063] The patient's lung ventilation distribution was obtained through EIT images of three respiratory cycles, and it was found that there was obvious uneven ventilation in the lungs, and the ventilation in the consolidation area of ​​the right lower lung was significantly reduced.

[0064] Based on the EIT image results and the algorithm model of this scheme, the tidal volume was 6 ml / kg ideal body weight, the respiratory rate was 12 times / minute, the inspiratory-expiratory ratio was 1:2, and the patient's uneven lung ventilation was improved. After several days of adjustment and treatment, the patient's lung ventilation gradually returned to normal and was successfully taken off the ventilator.

[0065] Embodiment 3

[0066] Taking a child patient case as an example, an 8-year-old child suffered from acute respiratory distress syndrome due to drowning and required mechanical ventilation support.

[0067] The EIT images over three respiratory cycles clearly show the damaged areas and ventilation abnormal regions in children's lungs, and diffuse exudation in both lungs and severe uneven ventilation distribution are found.

[0068] Based on the EIT image results and the algorithm model of this solution, a tidal volume of 5 ml / kg ideal body weight, a respiratory rate of 20 breaths per minute, and an inhalation-to-exhalation ratio of 1:1.5 are obtained. At the initial stage of treatment, it is found that some alveoli have a risk of collapse. Under the guidance of EIT images, the positive end-expiratory pressure (PEEP) is switched from 5 cmH 2 O and gradually increased to 8 cmH 2 O to improve the alveolar collapse situation. After active treatment and parameter adjustment, the pulmonary ventilation of the child gradually improves, and finally the child successfully weans from mechanical ventilation.

[0069] Example 4

[0070] Taking an elderly patient case as an example, a 75-year-old elderly person, due to acute exacerbation of chronic obstructive pulmonary disease complicated with respiratory failure, received mechanical ventilation.

[0071] Through the EIT images over two respiratory cycles, it is found that there are emphysematous regions in the elderly person's lungs, the ventilation distribution is extremely uneven, and there is gas trapping phenomenon.

[0072] Based on the EIT image results and the algorithm model of this solution, a tidal volume of 5 ml / kg ideal body weight, a respiratory rate of 12 breaths per minute, and an inhalation-to-exhalation ratio of 1:2.5 are obtained to extend the expiratory time and reduce gas trapping. After a period of fine adjustment and treatment, the pulmonary ventilation function of the elderly person gradually stabilizes, and finally the elderly person successfully weans from the ventilator.

[0073] The above are only examples of the present invention. Specific structures and common knowledge such as characteristics well known in the solution are not described in detail here. Those of ordinary skill in the art know all the common general technical knowledge in the technical field to which the invention belongs before the application date or the priority date, can know all the existing technologies in this field, and have the ability to apply the conventional experimental means before this date. Those of ordinary skill in the art can, under the inspiration given by this application, combine their own abilities to complete and implement this solution. Some typical well-known structures or well-known methods should not become an obstacle for those of ordinary skill in the art to implement this application. It should be noted that for those skilled in the art, without departing from the structure of the present invention, several deformations and improvements can still be made, and these should also be regarded as the protection scope of the present invention, and these will not affect the implementation effect of the present invention and the practicality of the patent.

Claims

1. A ventilator control strategy debugging method based on EIT image, characterized in that: The method comprises: A deep learning-based ventilator control algorithm model is constructed using the case database to characterize the mapping relationship between the EIT image of the case and its key indicators and ventilator ventilation control parameters; The current EIT images of the patient for several respiratory cycles are obtained and input into the ventilator control algorithm model. The most similar case is matched from the case library using the current EIT image, and the corresponding target EIT image of the most similar case and its key indicators and target ventilator ventilation control parameters are retrieved according to the mapping relationship. Use the target ventilator ventilation control parameters to control the ventilator ventilation; from the start of the ventilator ventilation control operation, obtain the adjusted EIT images and key indicators of the patient's several respiratory cycles, and compare and judge with the EIT images and key indicators of the most similar case; If the requirements are met, ventilator ventilation control is performed according to the target ventilator ventilation control parameters. At the same time, at least the EIT image and key indicators and ventilator ventilation control parameters of the patient obtained in the process are aggregated to form a new case and added to the case library to optimize the ventilator control algorithm model; otherwise, switch to other control modes.

2. A ventilator control strategy debugging method based on EIT image according to claim 1, characterized in that: Building a deep learning-based ventilator control algorithm model includes: The key indicators that characterize the changes in lung morphology during breathing are used as target inputs and initial values ​​are set; Case parameters are input into the lung perfusion / ventilation simulation device, EIT data is collected, PC images are reconstructed, key image indicators are extracted, and a real-time response curve is simulated to be compared with the ideal response curve. The ventilation control parameters of the ventilator are adjusted so that the real-time response curve formed by the simulation conforms to the ideal response curve. The ventilator control algorithm model learns the adjustment process of the ventilation control parameters of the ventilator.

3. A ventilator control strategy debugging method based on EIT image according to claim 2, characterized in that: The effectiveness of the lung perfusion / ventilation simulation device was verified before model training.

4. A ventilator control strategy debugging method based on EIT image according to claim 3, characterized in that: The validity verification includes: calling existing cases in the case library, adjusting the parameters of the lung perfusion / ventilation simulation device, obtaining and verifying EIT imaging based on the respiratory state simulated by the device, comparing the actual EIT imaging of the case, and judging the validity of the lung perfusion / ventilation simulation device based on the imaging differences.

5. The method for debugging a ventilator control strategy based on EIT images according to claim 1, characterized in that: The current EIT image is used to match the most similar case from the case database, the similarity between the current EIT image and the case EIT image is calculated, and the case corresponding to the case EIT image with the highest similarity is determined as the most similar case.

6. A method for debugging a ventilator control strategy based on EIT images according to claim 1, characterized in that: The requirements are that, under the ventilation control treatment of the ventilator, the oxygenation parameters represented by blood oxygen saturation and the ventilation parameters represented by carbon dioxide partial pressure need to reach certain values ​​and be maintained stably for a certain period of time.

7. A method for debugging a ventilator control strategy based on EIT images according to claim 1, characterized in that: Switching to other control modes includes switching to the positive end-expiratory pressure control mode, and switching to the manual control mode when the positive end-expiratory pressure control effect does not meet the requirements.

8. A ventilator control strategy debugging system based on EIT images, characterized in that: Execute a ventilator control strategy debugging method based on EIT images as described in any one of claims 1 to 7; the system includes a ventilator, and also includes: An acquisition unit, used for acquiring an EIT image of the patient as required and sending it to the model unit, wherein the EIT image includes a current EIT image of a number of respiratory cycles of the patient before the ventilation control of the ventilator is operated, and an adjusted EIT image of a number of respiratory cycles of the patient from the start of the ventilation control of the ventilator; A model unit is used to construct a case library; it is also used to use the case library to construct a deep learning-based ventilator control algorithm model to characterize the mapping relationship between the EIT image and its key indicators and the ventilator ventilation control parameters; it is also used to input the current EIT image into the ventilator ventilation control algorithm model, use the EIT image to match the most similar case from the case library, and call the target EIT image and its key indicators and target ventilator ventilation control parameters corresponding to the most similar case according to the mapping relationship; it is also used to send the target ventilator ventilation control parameters to the ventilator for ventilator ventilation control, and send the target EIT image and key indicators to the evaluation unit; it is also used to input the received adjusted EIT image into the ventilator ventilation control algorithm model, extract the corresponding key indicators, and send the adjusted EIT image and its key indicators to the evaluation unit; it is also used to aggregate at least the EIT image and key indicators and ventilator ventilation control parameters of the patient obtained in the process to form a new case and add it to the case library to optimize the ventilator control algorithm model; The evaluation unit is used to receive the adjusted EIT image and its key indicators, and the target EIT image and its key indicators of the most similar case, and compare and judge. If the requirements are met, the current ventilator ventilation control is performed according to the target ventilator ventilation control parameters, otherwise, other control modes are switched.

9. A ventilator control strategy debugging system based on EIT images according to claim 8, characterized in that: The acquisition unit includes an EIT imaging device; The model unit includes a lung perfusion / ventilation simulation device and a host computer, wherein the host computer stores a case library, a ventilator ventilation control algorithm model and a model training program.

10. A ventilator control strategy debugging system based on EIT images according to claim 8, characterized in that: The lung perfusion / ventilation simulation device is used to provide simulated lung perfusion and lung ventilation parameters during model training, and the parameters include respiratory rate, expiratory volume, inspiratory volume, expiratory pressure change and inspiratory pressure change.

Images