A ventilator offline auxiliary judgment method based on artificial intelligence

Abstract

The invention discloses an artificial intelligence-based ventilator offline auxiliary judgment method, which relates to the technical field of artificial intelligence. The method comprises the following steps: obtaining physiological data and magnetic resonance image data of a patient, obtaining a time series characteristic curve according to the physiological data, and obtaining characteristic data of the patient according to the time series characteristic curve; constructing a standard neural network model, and performing a primary judgment according to the standard neural network model; preprocessing the magnetic resonance image data according to the primary judgment result, and performing grayscale co-occurrence matrix analysis on the preprocessed magnetic resonance image data to obtain a texture feature set of a region corresponding to the magnetic resonance image data; performing similarity analysis on the texture feature set to obtain a texture feature similarity value of a region corresponding to the magnetic resonance image data, and performing a secondary judgment according to the texture feature similarity value; and improving the accuracy of judging the patient's respiratory state through the two judgments.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an artificial intelligence-based ventilator offline auxiliary judgment method. Background Art

[0002] Invasive mechanical ventilation is an important means of life support for respiratory failure caused by various causes in the ICU, and weaning from the ventilator (offline ventilation) is an important part of it.

[0003] At present, long-term use of mechanical ventilation will increase complications for patients. Some patients may experience loss of their own respiratory function, respiratory failure, heart failure and other complications. Therefore, after the patient's vital signs stabilize, mechanically ventilated patients need to be taken off the machine in time. However, premature weaning will cause insufficient oxygen supply to the patient when the respiratory function has not yet recovered, resulting in weaning failure.

[0004] How to judge whether the patient can stop using the ventilator and ensure the effectiveness and efficiency of the patient's respiratory function recovery is a problem we need to consider. Therefore, before the patient is disconnected from the ventilator, an effective offline auxiliary judgment tool is needed to conduct a comprehensive assessment of the patient, so as to accurately grasp the best time to go offline and reduce the risk of offline failure. In response to the above needs, the present invention provides a ventilator offline auxiliary judgment method based on artificial intelligence. Summary of the invention

[0005] In order to solve the above technical problems, the purpose of the present invention is to provide a ventilator offline auxiliary judgment method based on artificial intelligence.

[0006] In order to achieve the above object, the present invention provides the following technical solution: a ventilator offline auxiliary judgment method based on artificial intelligence, comprising the following steps:

[0007] Step S1: Acquire physiological data and magnetic resonance image data of a patient, obtain a time series characteristic curve according to the physiological data, and obtain characteristic data of the patient according to the time series characteristic curve;

[0008] Step S2: construct a standard neural network model, and make a judgment based on the standard neural network model;

[0009] Step S3: preprocessing the magnetic resonance image data according to the first judgment result, and performing gray-level co-occurrence matrix analysis on the preprocessed magnetic resonance image data to obtain a texture feature set of a region corresponding to the magnetic resonance image data;

[0010] Step S4: performing similarity analysis on the texture feature set to obtain texture feature similarity values ​​of the corresponding regions of the magnetic resonance image data, and performing secondary judgment based on the texture feature similarity values.

[0011] Furthermore, the physiological data of the patient is obtained, and the collection time of the physiological data is marked, and a collection cycle is set, and a time series characteristic curve that changes with time is obtained according to the physiological data, the collection time and the collection cycle.

[0012] Furthermore, according to the time series characteristic curve, the process of obtaining the patient's characteristic data includes:

[0013] Preset time series characteristic reference curve;

[0014] According to the comparison result between the time series characteristic curve and the time series characteristic reference curve, the patient's respiratory state is obtained. If the patient's respiratory state is a secondary respiratory pending state, characteristic data indicating that the patient's respiratory state is a secondary respiratory pending state is obtained, and a judgment is made.

[0015] Furthermore, the process of building a standard neural network model includes:

[0016] Obtain physiological sample data from several groups of patients and non-patients;

[0017] Several groups of physiological sample data are used as training sets and test sets, and are input into the neural network model to train the standard neural network model to obtain a standard neural network model after training.

[0018] Furthermore, according to the standard neural network model, a judgment process includes:

[0019] The obtained characteristic data is input into a standard neural network model for verification to obtain predicted respiratory rate data, predicted respiratory pressure data, and predicted oxygen saturation data;

[0020] Constructing a time series characteristic prediction curve according to the predicted respiratory rate data, the predicted respiratory pressure data, and the predicted oxygen saturation data;

[0021] The time series characteristic prediction curve is compared with the time series characteristic reference curve to obtain the predicted breathing state in the remaining time period within the acquisition cycle. If the breathing state is a fourth-level breathing pending state, a secondary judgment is performed.

[0022] Furthermore, the process of obtaining a texture feature set of a region corresponding to the magnetic resonance image data includes:

[0023] Acquiring magnetic resonance image data of a corresponding area of ​​the patient;

[0024] Performing denoising and grayscale preprocessing on the magnetic resonance image data to generate a grayscale image;

[0025] Setting standard magnetic resonance image data of a non-patient corresponding area, and performing denoising and grayscale preprocessing on the standard magnetic resonance image data to generate a standard grayscale image;

[0026] Subtracting the grayscale values ​​of the pixels at the corresponding positions of the standard grayscale image from the grayscale image to obtain a difference value, constructing a binary image according to the difference value, obtaining the difference value of each pixel in the binary image, setting a difference value threshold, screening out a region of pixels in the binary image whose difference value is greater than or equal to the difference value threshold, and marking the region as an abnormal region;

[0027] A gray level co-occurrence matrix of the abnormal area is obtained, and a texture feature set of the abnormal area is obtained according to the gray level co-occurrence matrix.

[0028] Furthermore, the process of obtaining the similarity value of the corresponding region of the magnetic resonance image data includes:

[0029] Preset a texture feature set of non-patient corresponding areas;

[0030] According to the texture feature set of the non-patient corresponding area, the contrast, energy, entropy rate and dispersion of the non-patient corresponding area are obtained;

[0031] Obtain the contrast, energy, entropy rate and dispersion of abnormal areas;

[0032] According to the contrast, energy, entropy rate and dispersion of the corresponding area of ​​the patient with abnormal breathing state and the contrast, energy, entropy rate and dispersion of the abnormal area, the texture feature similarity value of the abnormal area is obtained.

[0033] Furthermore, the process of performing a secondary judgment based on the texture feature similarity value includes:

[0034] Set standard texture feature threshold;

[0035] Preset the error range of texture feature similarity analysis;

[0036] The sum of the texture feature similarity value and the texture feature similarity analysis error range is compared with the standard texture feature threshold to obtain the patient's respiratory state, and determine whether the patient should be taken off the ventilator according to the patient's respiratory state.

[0037] Compared with the prior art, the beneficial effects of the present invention are: obtaining the physiological data and magnetic resonance image data of the patient, obtaining a time series characteristic curve based on the physiological data, and obtaining the characteristic data of the patient based on the time series characteristic curve; constructing a standard neural network model, and performing a primary judgment based on the standard neural network model and the characteristic data; preprocessing the magnetic resonance image data, and performing grayscale co-occurrence matrix analysis on the preprocessed magnetic resonance image data to obtain a texture feature set of a region corresponding to the magnetic resonance image data; performing a similarity analysis on the texture feature set to obtain a similarity value of a region corresponding to the magnetic resonance image data, setting a standard texture feature threshold, and performing a secondary judgment on the similarity value and the standard texture feature threshold; and improving the accuracy of judging the patient's respiratory state based on the primary judgment and the secondary judgment. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0039] Figure 1 It is a schematic diagram of the present invention. DETAILED DESCRIPTION

[0040] To make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be described in detail below. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other implementation methods obtained by ordinary technicians in this field without creative work belong to the scope of protection of the present invention.

[0041] like Figure 1 As shown, a ventilator offline auxiliary judgment method based on artificial intelligence includes the following steps:

[0042] Step S1: Acquire physiological data and magnetic resonance image data of a patient, obtain a time series characteristic curve according to the physiological data, and obtain characteristic data of the patient according to the time series characteristic curve;

[0043] Step S2: construct a standard neural network model, and make a judgment on the feature data according to the standard neural network model;

[0044] Step S3: preprocessing the magnetic resonance image data according to the first judgment result, and performing gray-level co-occurrence matrix analysis on the preprocessed magnetic resonance image data to obtain a texture feature set of a region corresponding to the magnetic resonance image data;

[0045] Step S4: performing similarity analysis on the texture feature set to obtain texture feature similarity values ​​of the corresponding regions of the magnetic resonance image data, and performing secondary judgment based on the texture feature similarity values;

[0046] It should be further explained that, in the specific implementation process, the specific process of extracting features from the physiological data and obtaining a time series characteristic curve includes:

[0047] Obtain the patient's physiological data, mark the collection time of the physiological data, and set the collection cycle;

[0048] The physiological data includes respiratory rate data, respiratory pressure data and oxygen saturation data;

[0049] Obtaining a time series characteristic curve that changes with time according to the respiratory rate data, respiratory pressure data, oxygen saturation data, acquisition time, and acquisition cycle;

[0050] The time series characteristic curves include a respiratory frequency time series characteristic curve, a respiratory pressure time series characteristic curve and an oxygen saturation time series characteristic curve;

[0051] The preset time series characteristic reference curves include a respiratory frequency time series characteristic reference curve, a respiratory pressure time series characteristic reference curve, and an oxygen saturation time series characteristic reference curve;

[0052] It should be further explained that, in the specific implementation process, the specific process of obtaining the patient's characteristic data includes:

[0053] Acquire the respiratory rate data, respiratory pressure data and oxygen saturation data at the same time as the time series characteristic curve and the time series characteristic reference curve, and judge the patient's respiratory state according to the respiratory rate data, respiratory pressure data and oxygen saturation data at the same time as the time series characteristic curve and the time series characteristic reference curve;

[0054] The breathing state includes a first-level abnormal breathing state, a second-level abnormal breathing state, a third-level abnormal breathing state, a first-level normal breathing state, a second-level normal breathing state, a third-level normal breathing state, a first-level breathing pending state, a second-level breathing pending state, a third-level breathing pending state and a first-level breathing pending state;

[0055] When the respiratory rate data, respiratory pressure data and oxygen saturation data of the time series characteristic curve at the same time are greater than the respiratory rate data, respiratory pressure data and oxygen saturation data of the time series characteristic reference curve at the same time, the patient's respiratory state is a first-level abnormal respiratory state and cannot be separated from the ventilator;

[0056] When the respiratory rate data, respiratory pressure data or oxygen saturation data of the time series characteristic curve at the same time is less than or equal to the respiratory rate data, respiratory pressure data and oxygen saturation data of the time series characteristic reference curve at the same time, the patient's respiratory state is a first-level respiratory pending state. According to the time series characteristic curve, the actual comprehensive fluctuation coefficient is obtained. , and recorded as abnormal respiratory rate data, abnormal respiratory pressure data and abnormal oxygen saturation data;

[0057] Setting the actual fluctuation

[0058] When the actual comprehensive volatility coefficient Greater than the actual fluctuation threshold When the patient's breathing state is in the secondary breathing pending state, a judgment is required;

[0059] When the actual comprehensive volatility coefficient Less than or equal to the verification fluctuation threshold When the patient's breathing state is normal at the first level, he or she can be separated from the ventilator;

[0060] Marking abnormal respiratory rate data, abnormal respiratory pressure data, and abnormal oxygen saturation data at the same time as characteristic data;

[0061] It should be further explained that the characteristic data is the total set of abnormal respiratory frequency data, abnormal respiratory pressure data and abnormal oxygen saturation data at the same time;

[0062] It should be further explained that in the specific implementation process, the actual fluctuation coefficient is obtained The specific process includes:

[0063] Obtain the respiratory rate data, respiratory pressure data, and oxygen saturation data of the time series characteristic reference curve and record them as , t=1,2,3,...,n, n is the total number of natural moments, represents the respiratory rate data at time t, represents the respiratory pressure data at the tth moment, represents the oxygen saturation data at time t;

[0064] Obtain abnormal respiratory rate data, abnormal respiratory pressure data, and abnormal oxygen saturation data, and record them as , t=1,2,3,...,n, n is the total number of natural moments, represents the abnormal respiratory rate data at time t, represents the abnormal respiratory pressure data at the tth moment, represents the abnormal oxygen saturation data at time t;

[0065] According to the respiratory rate data, respiratory pressure data and oxygen saturation data of the time series characteristic reference curve and the abnormal respiratory rate data, abnormal respiratory pressure data and abnormal oxygen saturation data, the actual fluctuation coefficient is obtained. ,in ;in is the weight coefficient of abnormal respiratory rate, is the weight coefficient of abnormal respiratory pressure and is the weight coefficient of abnormal oxygen saturation.

[0066] It should be further explained that, in the specific implementation process, according to the patient's respiratory state being in the respiratory pending state, the specific process of making a judgment includes:

[0067] A standard neural network model is constructed based on physiological sample data of several groups of patients and non-patients;

[0068] Make a judgment on the feature data according to the standard neural network model;

[0069] It should be further explained that, in the specific implementation process, the specific process of building a standard neural network model includes:

[0070] Obtain physiological sample data from several groups of patients and non-patients;

[0071] Several groups of physiological sample data are used as training sets and test sets, and the training sets and test sets are input into a standard neural network model, the standard neural network model is trained, and a standard neural network model after training is obtained.

[0072] It should be further explained that, in the specific implementation process, the specific process of judging the feature data according to the standard neural network model includes:

[0073] The abnormal respiratory rate data, abnormal respiratory pressure data and abnormal oxygen saturation data obtained at the same time are input into the standard neural network model for verification, so as to obtain the predicted respiratory rate data, predicted respiratory pressure data and predicted oxygen saturation data;

[0074] Constructing a time series characteristic prediction curve according to the predicted respiratory rate data, the predicted respiratory pressure data, and the predicted oxygen saturation data;

[0075] The time series characteristic prediction curve is compared with the time series characteristic reference curve;

[0076] When the predicted respiratory rate data, predicted respiratory pressure data and predicted oxygen saturation data of the time series characteristic prediction curve at the same time are greater than the respiratory rate data, respiratory pressure data and oxygen saturation data of the time series characteristic reference curve at the same time, the patient's respiratory state is a secondary respiratory abnormal state and cannot be separated from the ventilator;

[0077] When the respiratory rate data, predicted respiratory pressure data or predicted oxygen saturation data of the time series characteristic prediction curve at the same time is less than or equal to the respiratory rate data, respiratory pressure data and oxygen saturation data of the time series characteristic reference curve at the same time, the patient's respiratory state is the third-level respiratory pending state. According to the time series characteristic prediction curve, the predicted comprehensive trend fluctuation coefficient is obtained. ;

[0078] Setting verification fluctuations

[0079] When predicting the comprehensive trend volatility coefficient Greater than the verification fluctuation threshold When the patient's breathing state is in the fourth level of breathing pending state, a second judgment is required;

[0080] When predicting the comprehensive trend volatility coefficient Less than or equal to the verification fluctuation threshold When the patient's breathing state is in the secondary normal breathing state, he or she can be separated from the ventilator.

[0081] It should be further explained that, in the specific implementation process, the predicted comprehensive trend fluctuation coefficient is obtained according to the time series characteristic prediction curve. The specific process includes:

[0082] The predicted respiratory rate data, predicted respiratory pressure data, and predicted oxygen saturation data are recorded as , t=1,2,3,...,n, n is the total number of natural moments, represents the predicted respiratory rate data at time t, represents the predicted respiratory pressure data at time t, Represents the predicted oxygen saturation data at time t;

[0083] According to the respiratory rate data, respiratory pressure data and oxygen saturation data of the time series characteristic reference curve and the predicted respiratory rate data, predicted respiratory pressure data and predicted oxygen saturation data, the predicted comprehensive trend fluctuation coefficient is obtained. ,in ; Where A is the weight coefficient for predicting respiratory rate, B is the weight coefficient for predicting respiratory pressure, and C is the weight coefficient for predicting oxygen saturation;

[0084] It should be further explained that, in the specific implementation process, according to the first judgment result that the patient's respiratory state is in the fourth level respiratory pending state, the specific process of the second judgment includes:

[0085] According to the magnetic resonance image data, a texture feature set of a region corresponding to the magnetic resonance image data is obtained;

[0086] According to the similarity analysis of the texture feature set, a similarity value K of the corresponding area of ​​the magnetic resonance image data is obtained;

[0087] Perform secondary judgment according to the texture feature similarity value K;

[0088] It should be further explained that, in the specific implementation process, the specific process of obtaining the texture feature set of the area corresponding to the magnetic resonance image data includes:

[0089] Acquiring magnetic resonance image data of a corresponding area of ​​the patient;

[0090] Performing denoising and grayscale preprocessing on the magnetic resonance image data to generate a grayscale image;

[0091] Setting standard magnetic resonance image data of a non-patient corresponding area, and performing denoising and grayscale preprocessing on the standard magnetic resonance image data to generate a standard grayscale image;

[0092] Subtracting the grayscale values ​​of the pixels at the corresponding positions of the standard grayscale image from the grayscale image to obtain a difference value, constructing a binary image according to the difference value, obtaining the difference value of each pixel in the binary image, setting a difference value threshold, screening out a region of pixels in the binary image whose difference value is greater than or equal to the difference value threshold, and marking the region as an abnormal region;

[0093] Quantize the grayscale value of the abnormal area into discrete grayscale levels, preset fixed parameters of the grayscale co-occurrence matrix, including the relative position of pixel pairs and the interval distance between adjacent pixel pairs, obtain the grayscale co-occurrence matrix of the abnormal area, and obtain the texture feature set of the abnormal area according to the grayscale co-occurrence matrix;

[0094] It should be further explained that, in the specific implementation process, the gray level co-occurrence matrix of the abnormal area is calculated. The specific details include quantizing the gray level of the abnormal area into discrete gray levels, dividing the gray value range into several levels, for example, an 8-bit image can be divided into 16, 32, and 64 levels, and defining the parameters required for the gray level co-occurrence matrix, including distance (d) and direction (θ), which are determined according to actual application requirements. The specific steps will not be repeated. Subsequently, based on the calculated gray level co-occurrence matrix, a series of texture feature values ​​are extracted to construct a texture feature set. The texture features in the texture feature set include but are not limited to:

[0095] Contrast: It is a statistical feature that describes the contrast of pixels at different gray levels in an image. It is a feature that measures the roughness of the image texture and reflects the clarity of the image and the depth of the texture grooves. The deeper the texture grooves, the greater the contrast and the clearer the visual effect. On the contrary, when the contrast is small, the grooves are shallow and the effect is blurred.

[0096] Energy: describes the uniformity of the grayscale distribution of pixels in an image, measures the randomness contained in the image, and shows the complexity of the image. When all values ​​of the co-occurrence matrix are equal or the pixel values ​​show the greatest randomness, the entropy is the largest;

[0097] Entropy rate: describes the uncertainty of image texture, measures the randomness contained in the image, and shows the complexity of the image. When all values ​​of the co-occurrence matrix are equal or the pixel values ​​show the greatest randomness, the entropy rate is the largest;

[0098] Dispersion: describes the degree of difference between different gray value pairs in the gray level co-occurrence matrix;

[0099] It should be further explained that, in the specific implementation process, the specific process of obtaining the similarity value of the corresponding area of ​​the magnetic resonance image data by performing similarity analysis on the texture feature set includes:

[0100] Preset a texture feature set of non-patient corresponding areas;

[0101] It should be further explained that, in the specific implementation process, the process of performing similarity analysis on the texture feature set of the abnormal area includes:

[0102] According to the texture feature set of the non-patient corresponding area, the contrast, energy, entropy rate and dispersion of the non-patient corresponding area are obtained and recorded as , , as well as ;

[0103] The contrast, energy, entropy rate and dispersion of the abnormal region are respectively denoted as , , as well as ;

[0104] According to the contrast of non-patient corresponding area ,energy , Entropy rate And dispersion The contrast of the area corresponding to the abnormal area ,energy , Entropy rate And dispersion , the texture feature similarity value K is obtained;

[0105] It should be further explained that, in the specific implementation process, the calculation formula for obtaining the texture feature similarity value K between the abnormal area and the non-patient corresponding area is:

[0106] ;

[0107] in, , , as well as Respectively represent the contrast of the corresponding area of ​​the abnormal area ,energy , Entropy rate And dispersion The weight factor of .

[0108] It should be further explained that, in the specific implementation process, the specific process of performing secondary judgment according to the texture feature similarity value K includes:

[0109] Set standard texture feature threshold;

[0110] Preset texture feature similarity analysis error range ;

[0111] When the texture feature similarity value K is within the texture feature similarity analysis error range When the sum is greater than the standard texture feature threshold, the patient's respiratory state is level 3 abnormal respiratory state and cannot be separated from the ventilator;

[0112] When the texture feature similarity value K is within the texture feature similarity analysis error range When the sum is less than or equal to the standard texture feature threshold, the patient's breathing state is the third level normal breathing state and can be separated from the ventilator;

[0113] The accuracy of judging the patient's respiratory state is improved by using the primary judgment and the secondary judgment as auxiliary judgment methods.

[0114] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A ventilator offline auxiliary judgment method based on artificial intelligence, characterized in that: The following steps are involved: Step S1: Acquire physiological data and magnetic resonance image data of a patient, obtain a time series characteristic curve according to the physiological data, and obtain characteristic data of the patient according to the time series characteristic curve; Step S2: constructing a standard neural network model: obtaining physiological sample data of several groups of patients and non-patients; Inputting several groups of physiological sample data as training sets and test sets into a neural network model, training the standard neural network model, and obtaining a trained standard neural network model; According to the standard neural network model, a judgment is made: The abnormal respiratory rate data, abnormal respiratory pressure data and abnormal oxygen saturation data obtained at the same time are input into the standard neural network model for verification, so as to obtain the predicted respiratory rate data, predicted respiratory pressure data and predicted oxygen saturation data; Constructing a time series characteristic prediction curve according to the predicted respiratory rate data, the predicted respiratory pressure data, and the predicted oxygen saturation data; When the predicted respiratory rate data, predicted respiratory pressure data or predicted oxygen saturation data of the time series characteristic prediction curve at the same time is less than or equal to the respiratory rate data, respiratory pressure data and oxygen saturation data of the time series characteristic reference curve at the same time, the patient's respiratory state is the third-level respiratory pending state. According to the time series characteristic prediction curve, the predicted comprehensive trend fluctuation coefficient is obtained. ; The predicted respiratory rate data, predicted respiratory pressure data, and predicted oxygen saturation data are recorded as , t=1,2,3,...,n, n is the total number of natural moments, represents the predicted respiratory rate data at time t, represents the predicted respiratory pressure data at time t, represents the predicted oxygen saturation data at time t; Obtain the respiratory rate data, respiratory pressure data, and oxygen saturation data of the time series characteristic reference curve and record them as , t=1,2,3,...,n, n is the total number of natural moments, represents the respiratory rate data at time t, represents the respiratory pressure data at the tth moment, represents the oxygen saturation data at time t; in ; Where A is the weight coefficient for predicting respiratory rate, B is the weight coefficient for predicting respiratory pressure, and C is the weight coefficient for predicting oxygen saturation; Setting the verification fluctuation threshold ; When predicting the comprehensive trend volatility coefficient Greater than the verification fluctuation threshold When the patient's breathing state is in the fourth level of breathing pending state, a second judgment is required; When predicting the comprehensive trend volatility coefficient Less than or equal to the verification fluctuation threshold When the patient's breathing state is in the secondary normal breathing state, he / she will be taken off the ventilator; Step S3: According to the first judgment result that the patient's respiratory state is in the fourth level respiratory pending state, a second judgment is performed, the magnetic resonance image data is preprocessed, and the preprocessed magnetic resonance image data is subjected to gray level co-occurrence matrix analysis to obtain a texture feature set of the corresponding area of ​​the magnetic resonance image data; Step S4: Perform similarity analysis on the texture feature set to obtain the texture feature similarity value of the corresponding area of ​​the magnetic resonance image data: Preset a texture feature set of non-patient corresponding areas; According to the texture feature set of the non-patient corresponding area, the contrast, energy, entropy rate and dispersion of the non-patient corresponding area are obtained; Obtain the contrast, energy, entropy rate and dispersion of abnormal areas; According to the contrast, energy, entropy rate and dispersion of the non-patient corresponding area and the contrast, energy, entropy rate and dispersion of the abnormal area, a texture feature similarity value between the abnormal area and the non-patient corresponding area is obtained; A secondary judgment is performed based on the texture feature similarity value: Set standard texture feature threshold; Preset the error range of texture feature similarity analysis; The sum of the texture feature similarity value and the texture feature similarity analysis error range is compared with the standard texture feature threshold to obtain the patient's respiratory state, and determine whether the patient should be taken off the ventilator according to the patient's respiratory state.

2. According to the artificial intelligence-based ventilator offline auxiliary judgment method of claim 1, it is characterized in that: Acquire the patient's physiological data, mark the collection time of the physiological data, set the collection cycle, and obtain a time series characteristic curve that changes with time based on the physiological data, the collection time and the collection cycle.

3. The method for assisting in judging the ventilator offline based on artificial intelligence according to claim 2, characterized in that: According to the time series characteristic curve, the process of obtaining the patient's characteristic data includes: Preset time series characteristic reference curve; According to the comparison result between the time series characteristic curve and the time series characteristic reference curve, the patient's respiratory state is obtained. If the patient's respiratory state is a secondary respiratory pending state, characteristic data indicating that the patient's respiratory state is a secondary respiratory pending state is obtained, and a judgment is made.

4. The method for assisting in judging the ventilator offline based on artificial intelligence according to claim 3, characterized in that: The process of obtaining a texture feature set of a region corresponding to the magnetic resonance image data includes: Acquiring magnetic resonance image data of a corresponding area of ​​the patient; Performing denoising and grayscale preprocessing on the magnetic resonance image data to generate a grayscale image; Setting standard magnetic resonance image data of a non-patient corresponding area, and performing denoising and grayscale preprocessing on the standard magnetic resonance image data to generate a standard grayscale image; Subtracting the grayscale values ​​of the pixels at the corresponding positions of the standard grayscale image from the grayscale image to obtain a difference value, constructing a binary image according to the difference value, obtaining the difference value of each pixel in the binary image, setting a difference value threshold, screening out a region of pixels in the binary image whose difference value is greater than or equal to the difference value threshold, and marking the region as an abnormal region; A gray level co-occurrence matrix of the abnormal area is obtained, and a texture feature set of the abnormal area is obtained according to the gray level co-occurrence matrix.

Images