Modeling Method for Identifying the Degree of Ischemic Stroke Based on Evidential Reasoning Rules

Abstract

The present invention discloses a modeling method for identifying the degree of ischemic stroke based on the evidential reasoning rule. First, the present invention constructs an evidential reasoning model, taking the proportion of the ischemic area, the proportion of the total vascular area, and the proportion of vascular types as input features and the degree of stroke as the output. Then, the input feature data are obtained by measuring the real-time OCT images of the cerebral cortex vessels of rats using ImageJ. The k-means clustering algorithm is used to determine the reference levels of the input features in combination with the minimum and maximum values of the input feature data. Then, an initial evidence confidence distribution table is constructed to establish a non-linear mapping relationship between the input features and the output. The final fusion result is obtained by fusing all the evidence activated by the input features through the evidential reasoning rule. The degree of stroke is identified through the fusion result. The genetic algorithm (GA) is used to further optimize the model to make the result more accurate.

Description

Technical Field

[0001] The present invention relates to a method for identifying and modeling the degree of ischemic stroke based on cerebral cortex vascular data and evidence reasoning rules. Background Art

[0002] Stroke is mainly divided into two categories: ischemic stroke (i.e., cerebral infarction) and hemorrhagic stroke (i.e., cerebral hemorrhage). The incidence of ischemic stroke is higher than that of hemorrhagic stroke, accounting for 60%-70% of the total number of stroke patients. According to the research report of the World Health Organization, stroke has become the second leading cause of death worldwide after cancer and coronary heart disease.

[0003] In recent years, the number of cerebrovascular diseases in China has shown a sharp upward trend, and it ranks first among the causes of disease death with a mortality rate of 22.45%. Stroke has a high recurrence rate and disability rate. As a result, 75% of stroke patients show varying degrees of limb dysfunction, such as hemiplegia, hemisensory disturbance, abnormal posture and muscle tone, etc., which further leads to the loss of daily behavior ability, causing serious impacts on individuals, families, and the labor force, and bringing heavy burdens to life, work, and society. With the coming of the aging society in China, clinical rehabilitation of stroke in China faces huge challenges.

[0004] At present, in clinical practice, a rating scale mainly based on behavioral performance assessment is still mainly used to identify nerve function damage and evaluate the rehabilitation level of patients. This traditional stroke rehabilitation assessment method has obvious defects. Problems such as the mismatch and inaccuracy of the scale content caused by the self-level and subjectivity of rehabilitation therapists and the individual differences of patients themselves limit the accurate assessment of the rehabilitation level and the effective formulation of rehabilitation programs. If the blood flow changes and regeneration of damaged blood vessels and the damage and recovery of nerve tissues can be directly observed, the rehabilitation level of patients can be accurately evaluated. Therefore, an objective, accurate, intuitive, and scientific stroke degree identification model has become a necessary requirement at present. Summary of the Invention

[0005] In view of the deficiencies of the prior art, the present invention proposes a method for identifying and modeling the degree of ischemic stroke based on cerebral cortex vascular data and evidence reasoning rules.

[0006] The present invention includes the following steps:

[0007] Determine the input features and output framework of the stroke degree identification model. The input features are the two-dimensional image feature data of cerebral cortex blood vessels, and the output framework is the stroke degree.

[0008] Obtain a sample data set, which includes a training data set and a validation data set.

[0009] Determine the input feature reference level according to the sample data set and construct a similarity distribution table of evidence. Obtain the confidence distribution table of evidence through likelihood normalization, and determine the reliability factor and importance factor of evidence based on expert experience.

[0010] Calculate the matching degree distribution between the input feature data and its reference level, and fuse the reference level evidence activated by the input feature data to obtain feature evidence.

[0011] Fuse the input feature evidence through the recursive evidence reasoning rule to obtain the support degree of the input feature for the output.

[0012] Construct an optimization objective function according to the importance factor of the evidence and the mean square error of the output result of the stroke degree identification model, and use the optimized model as the final identification model.

[0013] The innovation of the method of the present invention lies in that the damage and recovery of cerebral cortical blood vessels are positively correlated with the damage and recovery of nerve tissues. By combining the biomedical research foundation with the method of artificial intelligence, a set of modeling methods for identifying the degree of ischemic stroke based on cerebral cortical blood vessel OCT image data and evidence reasoning rules is established.

[0014] Beneficial effects of the present invention:

[0015] First, by applying the modeling method provided by the present invention, a model for identifying the degree of ischemic stroke can be constructed, and the identification result is not affected by subjective factors, and is more objective, scientific and accurate.

[0016] Second, using the cerebral cortical damaged blood vessel data as the basis for identifying the degree of stroke, modeling the non-linear mapping relationship between the cerebral cortical damaged blood vessel data (input) of rats and the degree of stroke (output), and using the evidence reasoning rule to summarize the rules from the cerebral cortical blood vessel sample data and infer the degree of stroke of the sample data. Brief Description of the Drawings

[0017] Figure 1 is a schematic flow chart of the present invention.

[0018] Figure 2 is a comparison chart of the changing trends of the stroke diagnosis degree and the true degree of 150 rat cerebral cortical real-time OCT image blood vessel sample data. Detailed Embodiments

[0019] The following combines Figure 1 to explain the principle of the method of the present invention in detail.

[0020] A modeling method for identifying the degree of ischemic stroke based on evidence reasoning rules includes the steps:

[0021] Determine the input features and output framework of the stroke degree identification model. The input features are the two-dimensional image feature data of cerebral cortical blood vessels, and the output framework is the stroke degree;

[0022] Specifically, construct an evidential reasoning model for the degree of ischemic stroke. The model input is X = {x 1 , x 2 , x 3 , x 4 , x 5}, where x 1 is the proportion of the ischemic area, x 2 is the proportion of the blood vessel area, x 3 is the proportion of large blood vessels, x 4 is the proportion of medium blood vessels, x 5 is the proportion of small blood vessels; Divide the degree of ischemic stroke into 5 degrees: mild, moderate, severe, very severe, and critical, which constitute the output set Θ = {d 1 , d 2 , d 3 , d 4 , d 5}, and each proposition d j (j = 1, 2,..., 5) corresponds to a stroke degree, denoted as P(Θ). Represent x 1 , x 2 , x 3 , x 4 , x 5 and d j (j = 1, 2,..., 5) as a sample set S i = {[x 1 , x 2 , x 3 , x 4 , x 5 , d j | j = 1, 2,..., 5}, where [x 1 , x 2 , x 3 , x 4 , x 5 , d j is a sample vector. The sample data x j , x 1,j , x 2,j , x 3,j , x 4,j , x 5,j obtained under the state of the stroke degree d j constitute a sample set Q 1,j = {x 2,j , x 3,j , x 4,j , x 5,j}。Respectively obtain the sample data under each degree of stroke to form a sample set

[0023] Obtain a sample data set, where the sample data set includes a training data set and a validation data set;

[0024] Determine the input feature reference level according to the sample data set and construct a similarity distribution table of evidence. Obtain the confidence distribution table of evidence through likelihood normalization, and determine the reliability factor and importance factor of evidence through expert experience;

[0025] Calculate the matching degree distribution between the input feature data and its reference level, and fuse the reference level evidence activated by the input feature data to obtain feature evidence;

[0026] Fuse the input feature evidence through the recursive evidence reasoning rule to obtain the support degree of the input feature for the output;

[0027] Construct an optimization objective function according to the importance factor of the evidence and the mean square error of the output result of the stroke degree identification model, and use the optimized model as the final identification model.

[0028] In some embodiments, the step of obtaining the sample data set includes:

[0029] Through an animal optical window model that continuously observes the changes in the cerebral cortical blood vessels of rats before and after ischemia, use optical coherence tomography to record the damage and regeneration of blood vessels before and after vascular occlusion in real time, and obtain OCT images of cortical blood vessels with different degrees of ischemic stroke;

[0030] Analyze the characteristic information of the cerebral cortical blood vessels according to the OCT images of the cerebral cortical blood vessels, and obtain two-dimensional characteristic data of the cerebral cortical blood vessels from the images;

[0031] Perform image binarization processing on the OCT images of the cerebral cortical blood vessels based on global automatic thresholding by ImageJ, and binarize the OCT images of the cerebral cortical blood vessels to count the two-dimensional characteristic data of the cerebral cortical blood vessels.

[0032] Specifically, use ImageJ software to measure the blood vessel data of the OCT images of the rat cerebral cortex, and extract the two-dimensional characteristic data of the rat cerebral cortical blood vessels from the images. First, convert the OCT images of the cerebral cortex into 8-bit type images, and use ImageJ to measure the blood vessel area of the rat cerebral cortex. Manually select the ischemic area within the blood vessel area of the cerebral cortex, and the ratio of the ischemic area to the blood vessel area of the cerebral cortex is the proportion of the ischemic area; use ImageJ software to obtain the total area of the cerebral cortical blood vessels, and the ratio of the total area of the cerebral cortical blood vessels to the blood vessel area of the cerebral cortex is the proportion of the cerebral cortical blood vessel area; classify the rat cerebral cortical blood vessels into three types: large blood vessels, medium blood vessels, and capillaries, and use ImageJ to measure the proportions of the three types of cerebral cortical blood vessels.

[0033] In some embodiments, the steps of determining the reference level of the input feature data based on the two-dimensional cortical vascular feature dataset and constructing the similarity distribution table of the evidence include:

[0034] Calculate the average value of the contour values of each two-dimensional cortical vascular feature data, and obtain the clustering centers of the two-dimensional cortical vascular feature data through k-means clustering. The number of clustering centers is the average value of the input features, that is, the contour values of the two-dimensional cortical vascular feature data;

[0035] Determine the reference level of each two-dimensional feature data of the cortical blood vessels through the minimum value, maximum value of the two-dimensional cortical vascular feature dataset, and the clustering center of the two-dimensional cortical vascular feature data;

[0036] Calculate the matching degree distribution of the input feature data of the sample dataset and its reference level to construct the similarity distribution table of the evidence. The input feature data is included in all stroke degrees of the output framework.

[0037] Specifically, determine the input feature x 1 , x 2 , x 3 , x 4 , x 5 's reference value, perform k-means clustering on each input feature in the sample set U, and the clustering center is where T i is the number of clustering centers of the i-th feature, determined by the average value of the silhouette values of each feature in the sample set.

[0038] Arrange the minimum value i and the maximum value L i of the i-th input feature x in ascending order to form the reference level of this feature For example, when the five input features are divided into 2, 3, 3, 4, and 3 clusters respectively, the average value of the silhouette values corresponding to each feature is the largest. Taking the minimum value, maximum value, and clustering center of each cortical vascular input feature data as the reference level, the proportion of the ischemic area, the proportion of the total vascular area, and the proportion of vascular types each contain 4, 5, 5, 6, and 5 reference levels, which are respectively:

[0039] A V ={0.01426, 0.06588, 0.15153, 0.23961}

[0040] A P ={0.39521, 0.45214, 0.50012, 0.55014, 0.60012}

[0041] A Max = {0.30148, 0.32107, 0.35147, 0.39875, 0.44831}

[0042] A Middle = {0.29750, 0.29987, 0.30145, 0.30897, 0.31012, 0.31471}

[0043] A Min = {0.23698, 0.25961, 0.29876, 0.37412, 0.40102}。

[0044] Construct the evidence confidence distribution tables for five cerebral cortex vascular characteristics according to the above formulas, as shown in Tables 1 to 5.

[0045] Table 1 Evidence Confidence Distribution Table for Proportion of Ischemic Area

[0046]

[0047] Table 2 Evidence Confidence Distribution Table for Proportion of Total Vascular Area

[0048]

[0049] Table 3 Evidence Confidence Distribution Table for Proportion of Large Vessels

[0050]

[0051] Table 4 Evidence Confidence Distribution Table for Proportion of Medium-Sized Vessels

[0052]

[0053] Table 5 Evidence Confidence Distribution Table for Proportion of Capillaries

[0054]

[0055] Furthermore, the evidence confidence distribution table is obtained by likelihood normalization based on the similarity distribution table of the evidence.

[0056] Furthermore, calculating the matching degree distribution between the input feature data of the sample data set and its reference level, and fusing the reference level evidence activated by the input feature data to obtain the feature evidence includes the steps of:

[0057] Determine the reliability factor and importance factor of the reference level evidence according to expert experience, where the reference level evidence is the degree of support of each input feature data reference level in the evidence confidence distribution table for different outputs in the output framework;

[0058] Calculate the matching degree distribution of the input feature data of the sample data set with its reference levels, and determine the corresponding reference level evidence;

[0059] Fuse the matching degree distribution of the input feature data and the corresponding reference level evidence to obtain feature evidence.

[0060] Specifically, import the input feature vector X = {x 1 , x 2 , x 3 , x 4 , x 5} into the evidential reasoning model to output d j , and take the sample a[0.03459, 0.51743, 0.35218, 0.315, 0.26509, mild] randomly selected from 150 sample sets as an example to illustrate. The specific steps are as follows:

[0061] Calculate the matching degree between the input vector x i in the input feature X and the reference level of the input feature. The specific calculation process is as follows:

[0062] Set the matching degree of x i to the reference level as When or , = 1, and the matching degrees of the input feature x i to other reference levels are all 0; when , the matching degrees of the input feature x i to the reference levels and are calculated by the following two formulas. Similarly, the matching degrees of the input feature x i to other reference levels are all 0;

[0063]

[0064]

[0065] The proportion V of the ischemic area of sample a is 0.03459. The matching degrees of the value of V to the reference levels and are and Activate evidence and The proportion P of the total blood vessel area is 0.51743. The matching degrees of the value of P to the reference levels and are and Activate evidence and The proportion of large blood vessels Max = 0.35218, and the values of Max and the reference level and The matching degrees are respectively and Activation evidence and The proportion of middle blood vessels Middle = 0.315. The value of Middle is greater than the maximum value of the reference level, so the activation evidence of this feature And The proportion of capillaries Min = 0.26509, and the values of Min and the reference level and The matching degrees are respectively and Activation evidence and

[0066] In some embodiments, the fusion method for fusing the reference level evidence to obtain the feature evidence is:

[0067]

[0068] Wherein, and Are the matching degree distributions of the i-th two-dimensional input feature data of the cerebral cortex blood vessels with its Z-th and Z + 1-th reference levels, β Z,j and β Z+1,j Are the support degrees of the Z-th and Z + 1-th reference level evidences for the j-th output, Is the support degree of the i-th two-dimensional input feature data of the cerebral cortex blood vessels for the j-th output.

[0069] Specifically, for the input feature x i , when , x i Activates the reference levels and The corresponding two evidences and From the evidences and The evidence e i of the input feature x i Can be obtained through the following formula.

[0070]

[0071]

[0072] Wherein, Indicates that when the input feature x i Activates the reference level and the corresponding two pieces of evidence and output d j when the probability of. From this, the input feature x i all the corresponding evidence e i (i = 1, 2,..., 5). According to Tables 1 to 5 and the sample a, the finally activated evidence is:

[0073] e 1 ={(d 1 : 0.165), (d 2 : 0.045), (d 3 : 0.250), (d 4 : 0.268), (d 5 : 0.272)};

[0074] e 2 ={(d 1 : 0.421), (d 2 : 0.095), (d 3 : 0.291), (d 4 : 0.193), (d 5 : 0)};

[0075] e 3 ={(d 1 : 0.347), (d 2 : 0.028), (d 3 : 0.349), (d 4 : 0.273), (d 5 : 0.003)};

[0076] e 4 ={(d 1 : 0.978), (d 2 : 0), (d 3 : 0.022), (d 4 : 0), (d 5 : 0)};

[0077] e 5 ={(d 1 : 0.393), (d 2 : 0.200), (d 3 : 0.149), (d 4 : 0.194), (d 5 : 0.064)}.

[0078] In some embodiments, the feature evidence is fused through the recursive evidence reasoning rule to obtain the support degree of the input feature for the output, and the fusion calculation formula is:

[0079]

[0080] where m j,i and m j,(i-1) are the support degrees of the i-th and i-1-th two-dimensional feature evidences of the cerebral cortex blood vessels for the j-th output, r i is the reliability factor of the i-th feature evidence, m P(Θ),e(i-1) is the power set of the (i - 1)-th two-dimensional feature evidence of the cerebral cortex blood vessels, is the intersection of the support degrees of the i-th and i-1-th two-dimensional feature evidences of the cerebral cortex blood vessels for the j-th output, and m j,e(i,i-1) is the support degree of the evidence after fusing the i-th and i-1-th two-dimensional feature evidences of the cerebral cortex blood vessels for the j-th output.

[0081] Specifically, when the evidences of each feature in the input feature vector X = {x 1 , x 2 , x 3 , x 4 , x 5} are independent of each other, five evidences e i can be fused through the recursive evidence reasoning rule to obtain the support degrees of the feature vector X for five different levels of output as follows:

[0082]

[0083]

[0084] m P(Θ),e(i) = (1 - r i )m P(Θ),e(i-1)

[0085] where p j,e(5) is the joint confidence of evidences e 1 , e 2 , e 3 , e 4 , e 5 for the j-th output, indicating the credibility that when the cerebral cortex blood vessel feature vector is X = {x 1 , x 2 , x 3 , x 4 , x 5}, the degree of stroke is considered to be d j ; m j,(i-1) = w (i-1) p j,(i-1) represents the basic belief assignment; r iand w i represent the reliability factor and importance factor of evidence, satisfying 0 ≤ r i ≤ 1 and 0 ≤ w i ≤ 1. Since the stroke degree of the samples in the sample set is determined by the experience of domain experts, without loss of generality, assume that the initial evidence credibility of the sample set is 0.96. At the same time, considering that there will be human processing errors when the data used is processed by ImageJ, the initial evidence reliability is set to 0.94. Fusing the five pieces of evidence according to the support degree, the final output result is:

[0086] O(a) = {(d 1 : 0.687), (d 2 : 0.02), (d 3 : 0.149), (d 4 : 0.125), (d 5 : 0.02)}

[0087] Through the above, the stroke degree of the cerebral cortex feature vector X = {x 1 , x 2 , x 3 , x 4 , x 5} can be identified. The d j corresponding to the maximum value in O(a) is the stroke degree corresponding to the diagnostic sample a[0.03459, 0.51743, 0.35218, 0.315, 0.26509, mild] of the evidence reasoning model, that is, d 1 mild stroke.

[0088] In some embodiments, an optimization objective function is constructed according to the importance factor of the evidence and the mean square error of the model output result. The steps of optimizing the model include:

[0089] Construct an optimization objective function based on the minimum mean square error according to the model diagnostic accuracy rate and the initial importance factor of the evidence;

[0090] In the optimization process based on the genetic algorithm, the initial population is randomly generated between [0, 1], the initial population size is 60, and the optimization objective function is used as the fitness function;

[0091] Update the optimized evidence importance factor in the stroke degree identification model.

[0092] Specifically, in order to enable the stroke degree identification model to more accurately reflect the relationship between the input features and the output, the model parameters are optimized using the sample set. The optimization steps include:

[0093] Input the features of the sample set into the evidence reasoning model to be optimized. After calculation, the recognition accuracy UA can be obtained, and an optimization objective function based on the minimum mean square error is constructed:

[0094]

[0095] The parameters to be optimized are:

[0096]

[0097] In the optimization process based on the genetic algorithm, the initial population is randomly generated between [0, 1], and the number of the initial population is 60. The objective function of the optimization model is used as the fitness function, that is, f(x) = 1 - UA, and the optimized model is used as the final recognition model.

[0098] The sample data set is determined according to the continuously observed in-vivo OCT image data of the cerebral cortex blood vessels of rats. The training data set is the training data set of ischemic stroke, and the validation data set is the validation data set of ischemic stroke.

[0099] In the experiment of the present invention, 150 rat cerebral cortex blood vessel data samples are selected, and the recognition accuracy is 0.9133. Figure 2 It is a comparison chart of the changing trends of the stroke recognition degree and the true degree of 150 real-time OCT image vascular parameter samples of the rat cerebral cortex.

[0100] It can be concluded from the experimental results that: First, by applying the method provided by the present invention, a new method for recognizing the stroke degree of stroke patients has been successfully developed, and the recognition of the stroke degree of stroke patients has been extended from the traditional rating scale mainly based on behavioral performance evaluation to the evidence reasoning rule model recognition based on cerebral cortex blood vessel data. Second, different from the traditional rating scale for diagnosing the stroke degree, the modeling method provided by the present invention is based on cerebral cortex blood vessel data, directly hits the lesion, and is more objective, scientific and accurate. Third, in the experiment of diagnosing the stroke degree of 150 cerebral cortex blood vessel data samples of rats, the recognition accuracy of the stroke degree recognition model based on cerebral cortex blood vessel data and evidence reasoning rules is 0.9133, which fully proves the effectiveness of this method.

Claims

1. An ischemic stroke degree identification modeling method based on the evidential reasoning rule, characterized in that, it includes the steps of: Determine the input features and output framework of the stroke degree identification model. The input features are the two-dimensional feature data of cerebral cortical blood vessels, and the output framework is the stroke degree; Obtain a sample data set, which includes a training data set and a validation data set; Determine the reference levels of the input features according to the sample data set and construct a similarity distribution table of evidence. Obtain the confidence distribution table of evidence through likelihood normalization, and determine the reliability factor and importance factor of evidence through expert experience; Calculate the matching degree distribution between the input feature data and its reference levels, and fuse the reference level evidence activated by the input feature data to obtain feature evidence; Fuse the input feature evidence through the recursive evidential reasoning rule to obtain the support degree of the input features for the output; Construct an optimization objective function according to the importance factor of the evidence and the mean square error of the output result of the stroke degree identification model, and use the optimized model as the final identification model; Among them, calculating the matching degree distribution between the input feature data of the sample data set and its reference levels, and fusing the reference level evidence activated by the input feature data to obtain feature evidence includes the steps of: Determine the reliability factor and importance factor of the reference level evidence according to expert experience. The reference level evidence is the support degree of each input feature data reference level in the confidence distribution table of evidence for different outputs in the output framework; Calculate the matching degree distribution between the input feature data of the sample data set and its reference levels, and determine the corresponding reference level evidence; Fuse to obtain feature evidence according to the matching degree distribution of the input feature data and the corresponding reference level evidence; The fusion method for fusing the reference level evidence to obtain feature evidence is: Among them, and are the matching degree distributions of the two-dimensional input feature data of the i-th cerebral cortex blood vessel with its Z-th and Z+1-th reference levels, β Z,j and β Z+1,j are the degrees of support of the evidence of the Z-th and Z+1-th reference levels for the j-th output, is the degree of support of the two-dimensional input feature data of the i-th cerebral cortex blood vessel for the j-th output; Among them, fusing the feature evidence through the recursive evidential reasoning rule to obtain the support degree of the input features for the output, and the calculation formula is: where m j,i and m j,(i-1) are the degrees of support of the i-th and (i - 1)-th two-dimensional feature evidences of cerebral cortical blood vessels for the j-th output, r i is the reliability factor of the i-th feature evidence, m P(Θ),e(i-1) is the power set of the (i - 1)-th two-dimensional feature evidence of cerebral cortical blood vessels, is the intersection of the degrees of support of the i-th and (i - 1)-th two-dimensional feature evidences of cerebral cortical blood vessels for the j-th output, m j,e(i,i-1) is the degree of support of the evidence after fusing the i-th and (i - 1)-th two-dimensional feature evidences of cerebral cortical blood vessels for the j-th output.

2. The ischemic stroke degree identification modeling method based on the evidential reasoning rule according to claim 1, characterized in that, the steps for obtaining the sample data set include: Through an animal optical window model that continuously observes the changes in cerebral cortical blood vessels before and after ischemia in rats, use optical coherence tomography to record the damage and regeneration of blood vessels before and after vascular occlusion in real time, and obtain OCT images of cerebral cortical blood vessels with different ischemic stroke degrees; Analyze the cerebral cortical blood vessel feature information according to the OCT images of cerebral cortical blood vessels, and obtain the two-dimensional feature data of cerebral cortical blood vessels from the images; Perform image binarization processing on the OCT images of cerebral cortical blood vessels through global automatic thresholding based on ImageJ, and statistically analyze the two-dimensional feature data of cerebral cortical blood vessels.

3. The ischemic stroke degree identification modeling method based on the evidential reasoning rule according to claim 2, characterized in that, Based on the two-dimensional feature data set of cerebral cortical blood vessels, determining the reference levels of the input feature data and constructing the similarity distribution table of the evidence includes the steps of: Calculate the average value of the contour values of the two-dimensional feature data of each cerebral cortex blood vessel, and obtain the clustering centers of the two-dimensional feature data of the cerebral cortex blood vessel through k-means clustering. The number of clustering centers is the input feature, that is, the average value of the contour values of the two-dimensional feature data of the cerebral cortex blood vessel. Determine the reference level of each two-dimensional feature data of the cerebral cortex blood vessel based on the minimum value, maximum value of the two-dimensional feature data set of the cerebral cortex blood vessel and the clustering centers of the two-dimensional feature data of the cerebral cortex blood vessel. Calculate the matching degree distribution between the input feature data of the sample data set and its reference level to construct a similarity distribution table of evidence. The input feature data is included in all stroke degrees of the output framework.

4. The method for identifying and modeling the degree of ischemic stroke based on the evidence reasoning rule according to claim 3, characterized in that The confidence distribution table of evidence is obtained by likelihood normalization according to the similarity distribution table of the evidence.

5. The method for identifying and modeling the degree of ischemic stroke based on the evidence reasoning rule according to claim 1, characterized in that Construct an optimization objective function based on the importance factor of the evidence and the mean square error of the model output result. The steps of optimizing the model include: Construct an optimization objective function based on the minimum mean square error according to the model identification accuracy and the initial importance factor of the evidence. During the optimization process based on the genetic algorithm, the initial population is randomly generated between [0, 1], the number of the initial population is 60, and the optimization objective function is used as the fitness function. Update the optimized evidence importance factor in the stroke degree identification model.

6. The method for identifying and modeling the degree of ischemic stroke based on the evidence reasoning rule according to any one of claims 1-5, characterized in that The sample data set is determined according to the continuous observation of the in-vivo cerebral cortex blood vessel OCT image data of rats. The training data set is the training data set of ischemic stroke, and the validation data set is the validation data set of ischemic stroke.

Images