Evaluation of acute ischemic stroke CT imaging data model using computer chips

Abstract

The present invention belongs to the field of medical impact detection technology, and discloses a method for evaluating an acute ischemic stroke CT image data model using a computer chip. The acute ischemic stroke CT image data of each hospital is collected, and the data is classified according to different parts and different time periods after the onset of the disease, and a preliminary new acute ischemic stroke model is established according to the curve generated by the data; a part of the data is extracted from the classified data as a validation set and retained, and the remaining data is input into the initially established model for model training; the retained validation set data is input into the trained model; the results are observed and the accuracy of the model is evaluated; according to the evaluation results, the model is optimized, and the established model is embedded in a computer chip for use by other devices. The present invention ensures the availability and trainability of the model, retains a part of the validation set to verify the model, and ensures the accuracy of the model.

Description

Technical Field

[0001] The present invention belongs to the technical field of medical impact detection, and in particular relates to a method for evaluating an acute ischemic stroke CT image data model using a computer chip. Background Art

[0002] With the rapid aging of the Chinese population, and due to changes in living conditions and lifestyles, the incidence of stroke has increased year by year. As we all know, acute ischemic stroke (acute cerebral infarction) is the most common type of stroke in my country, with extremely high mortality and disability rates. Intravenous thrombolytic therapy with rt-PA can effectively reduce the mortality and disability rates of ischemic stroke, but because the onset time of stroke after awakening is uncertain, we need to consider how to rely on early CT images to perform thrombolytic therapy on patients with stroke after awakening without increasing the risk of hemorrhagic transformation. In recent years, a large number of clinical studies have also confirmed that endovascular treatment can bring significant clinical benefits to reasonably selected patients with large vessel occlusive stroke, but for patients with progressive stroke, how we can reduce the incidence of complications of interventional surgery in patients beyond the time window based on CT imaging data still requires further research. Therefore, the determination of early CT imaging data of stroke has also become an important challenge for neurologists.

[0003] The diagnosis of acute ischemic stroke mainly relies on clinical symptoms, signs and imaging. However, there is an ischemic penumbra in the early stage of acute ischemic stroke, so the time of cerebral infarction should be later than the time of clinical symptoms and signs. At present, CT images cannot determine the time of infarction. There is no good model to make timely judgments, which can easily misjudge the size of the new infarction core, increase the risk of complications such as ischemia-reperfusion injury and malignant middle cerebral artery syndrome, or miss the best treatment time for thrombolysis or endovascular treatment of acute ischemic stroke. Therefore, it is of great application value to use computer chip technology to evaluate the CT image data model of acute ischemic stroke and judge the time period of acute ischemic stroke.

[0004] Through the above analysis, the problems and defects of the prior art are as follows:

[0005] (1) Existing technologies do not have a good model to determine the time of infarction.

[0006] (2) Existing diagnostic technology makes it difficult to determine the optimal treatment timing for thrombolysis or endovascular therapy for stroke after awakening or progressive stroke. Summary of the invention

[0007] In view of the problems existing in the prior art, the present invention provides a method for evaluating an acute ischemic stroke CT image data model using a computer chip.

[0008] The present invention is implemented in this way: the method for evaluating the acute ischemic stroke CT image data model using a computer chip comprises:

[0009] Step 1: Collect CT imaging data of acute ischemic stroke from various hospitals, classify the data according to different parts and different time periods after onset, and establish a preliminary new acute ischemic stroke model based on the curve generated by the data;

[0010] Step 2: extract a portion of the classified data as a validation set and input the remaining data into the initially established model for model training;

[0011] Step 3: Input the reserved validation set data into the trained model; observe the results and evaluate the accuracy of the model;

[0012] Step 4: Based on the evaluation results, the model is optimized and the established model is embedded in a computer chip for use in other devices;

[0013] In step 3, the specific process of model training is as follows:

[0014] Input the training data into the model, preprocess the data, generate input images and label images, and process the corresponding images;

[0015] Analyze the Loss value of the model and use the MSE function to adjust it.

[0016] The process of processing the input image and the label image is as follows:

[0017] Preprocess the input image and label image and convert them into grayscale images

[0018] Find the area through the grayscale threshold, extract the fixed area through feature point extraction, and process the extracted area;

[0019] The input image and the label image are preprocessed including: digitization, geometric transformation, normalization, smoothing, restoration and enhancement;

[0020] The specific process of image enhancement is:

[0021] Assume that the noisy image is represented as:

[0022]

[0023] The original image f(x, y) is zero-mean and has a variance σ 2 The Gaussian white noise n(x, y) is contaminated and passed through the Wiener filter g(x, y);

[0024] According to the Wiener filtering criterion, we get:

[0025]

[0026] Assume that the size of f(x, y) is M×N, and H is used to represent the Wiener filter, which is:

[0027] H=R f (R f +σ 2 I) -1 ;

[0028] Among them, R f =E(ff T ) is the autocorrelation function, I is the unit matrix, and u i YesR f The eigenvector of , the corresponding eigenvalue is: i , let U=[u1,u2,…,u N ] is an M×N feature matrix, where Λ=diag[λ1,λ2,…,λ N ];

[0029] but:

[0030] R f Substitute the functional relationship into the H functional relationship;

[0031]

[0032]

[0033] In the actual image processing process, due to σ 2 The value of is unknown, but is generally estimated to be:

[0034]

[0035] in,

[0036]

[0037] Furthermore, the specific process of image digitization is as follows:

[0038] Determine appropriate points to sample and describe the image to be digitized;

[0039] After the sampling description is completed, a certain range of values ​​is used to represent each sampling point;

[0040] After the point representation is completed, all points are compressed and encoded.

[0041] Furthermore, the specific process of data classification is as follows:

[0042] In the data of different parts of brain tissue and different time periods after onset, each object is regarded as a class and the minimum distance between two objects is calculated;

[0043] The two classes with the smallest distance are merged into a new class, and the distance between the new class and all classes is repeatedly calculated until all data are classified.

[0044] Furthermore, the specific process of the image smoothing process is:

[0045] Determine the corresponding size of the image to be smoothed;

[0046] Set a circular buffer in the image to temporarily store the result image and initialize it;

[0047] Scan each pixel point in the circular buffer and sort the pixel values; assign the obtained intermediate value to the pixel point corresponding to the current point in the target image until all the pixels of the source image are processed.

[0048] Furthermore, the specific process of extracting the image feature points is as follows:

[0049] Determine the corresponding detection window in the image and normalize the image;

[0050] After the normalization process is completed, the grayscale value and gradient of the image in the window are calculated;

[0051] For each window, the gradient histogram is projected with a specified weight, and all histograms are connected to form the corresponding feature points.

[0052] Furthermore, the validation set is a small part of the data in the training set, which is used to verify the training results.

[0053] Furthermore, in step 4, the model is optimized using the Adam optimizer.

[0054] Furthermore, in step four, the equipment includes acute ischemic stroke detection equipment and acute ischemic stroke recurrence detection equipment.

[0055] Another object of the present invention is to provide a computer program product stored on a computer-readable medium, including a computer-readable program, which, when executed on an electronic device, provides a user input interface to apply the design method for evaluating an acute ischemic stroke imaging data model.

[0056] Another object of the present invention is to provide a computer-readable storage medium storing instructions, which, when executed on a computer, enable the computer to apply the method for designing an image data model for evaluating acute ischemic stroke.

[0057] In combination with all the above-mentioned technical solutions, the advantages and positive effects of the present invention are as follows: the present invention collects all existing CT image data of acute ischemic stroke, classifies the data and establishes a CT image data model of acute ischemic stroke, thereby ensuring the availability and trainability of the model, retaining a part of the verification set to verify the model to ensure the accuracy of the model, so that after the model is established and embedded in other devices, it can quickly make a situation judgment based on the patient's condition, evaluate the time window for the occurrence of acute ischemic stroke, and provide more accurate judgment for the subsequent selection of treatment plans for acute ischemic stroke and the evaluation of its prognosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments of the present application. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0059] Figure 1 It is a flow chart of a method for evaluating an acute ischemic stroke CT image data model using a computer chip provided in an embodiment of the present invention.

[0060] Figure 2 It is a flow chart of the model training method provided by an embodiment of the present invention.

[0061] Figure 3 It is a flow chart of the image digitization method provided by an embodiment of the present invention.

[0062] Figure 4 It is a flow chart of a data classification method provided by an embodiment of the present invention.

[0063] Figure 5 It is a flow chart of an image smoothing processing method provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0064] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0065] In view of the problems existing in the prior art, the present invention provides a method for evaluating an acute ischemic stroke CT image data model using a computer chip. The present invention is described in detail below in conjunction with the accompanying drawings.

[0066] like Figure 1 As shown, the method for designing a CT image data model for acute ischemic stroke provided by an embodiment of the present invention includes:

[0067] S101: Collect CT imaging data of acute ischemic stroke from various hospitals, classify the data according to different parts and different time periods after onset, and establish a preliminary new acute ischemic stroke model based on the curve generated by the data;

[0068] S102: extract a portion of the classified data as a validation set, and input the remaining data into the initially established model for model training;

[0069] S103: Input the reserved validation set data into the trained model; observe the results and evaluate the accuracy of the model;

[0070] S104: Optimize the model based on the evaluation results, and embed the established model into a computer chip for use by other devices.

[0071] In S102 provided in the embodiment of the present invention, the validation set is a small portion of data in the training set, which is used to verify the training results.

[0072] like Figure 2 As shown, in S103 provided in the embodiment of the present invention, the specific process of model training is:

[0073] S201: input the training data into the model, pre-process the data, generate input images and label images, and process the corresponding images;

[0074] S202: Analyze the Loss value of the model and use the MSE function to adjust it.

[0075] The process of processing the input image and the label image provided by the embodiment of the present invention is as follows:

[0076] Preprocessing the input image and the label image to convert them into grayscale images; the input image and the label image refer to imaging data of stroke;

[0077] Find the area through the grayscale threshold, extract the fixed area through feature point extraction, and process the extracted area.

[0078] The preprocessing of the input image and the label image provided by the embodiment of the present invention includes: digitization, geometric transformation, normalization, smoothing, restoration and enhancement.

[0079] Among them, the specific process of image enhancement is:

[0080] Assume that the noisy image is represented as:

[0081]

[0082] The original image f(x, y) is zero-mean and has a variance σ 2The Gaussian white noise n(x, y) is contaminated and passed through the Wiener filter g(x, y);

[0083] According to the Wiener filtering criterion, we get:

[0084]

[0085] Assume that the size of f(x, y) is M×N, and H is used to represent the Wiener filter, which is:

[0086] H=R f (R f +σ 2 I) -1 ;

[0087] Among them, R f =E(ff T ) is the autocorrelation function, I is the unit matrix, and u i YesR f The eigenvector of , the corresponding eigenvalue is: i , let U=[u1,u2,…,u N ] is an M×N feature matrix, where Λ=diag[λ1,λ2,…,λ N ];

[0088] but:

[0089] R f Substitute the functional relationship into the H functional relationship;

[0090]

[0091] In the actual image processing process, since the value of σ is unknown, it is generally estimated as:

[0092]

[0093] in,

[0094]

[0095] like Figure 3 As shown, the specific process of image digitization provided by the embodiment of the present invention is:

[0096] S301: Determine appropriate points to sample and describe the image to be digitized;

[0097] S302: After the sampling description is completed, each sampling point is represented using a certain range of values;

[0098] S303: After the point representation is completed, all points are compressed and encoded.

[0099] like Figure 4 As shown, the specific process of data classification provided by the embodiment of the present invention is:

[0100] S401: Data from different parts of brain tissue and different time periods after onset are taken as one class, and the minimum distance between each class is calculated;

[0101] S402: The two classes with the smallest distance are merged into a new class, and the distances between the new class and all classes are repeatedly calculated until all data are classified.

[0102] like Figure 5 As shown, the specific process of image smoothing provided by the embodiment of the present invention is:

[0103] S501: Determine the corresponding size of the image to be smoothed;

[0104] S502: setting a circular buffer in the image to temporarily store the result image and initializing it;

[0105] S503: Scan each pixel point in the circular buffer and sort the pixel values; assign the obtained intermediate value to the pixel point corresponding to the current point in the target image until all the pixel points of the source image are processed.

[0106] The specific process of image feature point extraction provided by the embodiment of the present invention is as follows:

[0107] Determine the corresponding detection window in the image and normalize the image;

[0108] After the normalization process is completed, the grayscale value and gradient of the image in the window are calculated;

[0109] For each window, the gradient histogram is projected with a specified weight, and all histograms are connected to form the corresponding feature points.

[0110] In S104 provided in the embodiment of the present invention, the model optimization uses the Adam optimizer.

[0111] In S104 provided in the embodiment of the present invention, the equipment includes an acute ischemic stroke detection equipment and an acute ischemic stroke recurrence detection equipment.

[0112] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any modifications, equivalent substitutions and improvements made by any technician familiar with the technical field within the technical scope disclosed by the present invention and within the spirit and principles of the present invention should be covered within the protection scope of the present invention.

Claims

1. A method for evaluating an acute ischemic stroke CT image data model using a computer chip, characterized in that: The design method of the CT image data model for evaluating acute ischemic stroke includes: Step 1: Collect CT imaging data of acute ischemic stroke from various hospitals, classify the data according to different parts and different time periods after onset, and establish a preliminary new acute ischemic stroke model based on the curve generated by the data; Step 2: extract a portion of the classified data as a validation set and input the remaining data into the initially established model for model training; Step 3: Input the reserved validation set data into the trained model; observe the results and evaluate the accuracy of the model; Step 4: Based on the evaluation results, the model is optimized and the established model is embedded in a computer chip for use in other devices; In step 3, the specific process of model training is as follows: Input the training data into the model, preprocess the data, generate input images and label images, and process the corresponding images; Analyze the Loss value of the model and use the MSE function to adjust it. The process of processing the input image and the label image is as follows: Preprocess the input image and label image and convert them into grayscale images; Find the area through the grayscale threshold, extract the fixed area through feature point extraction, and process the extracted area; The input image and the label image are preprocessed including: digitization, geometric transformation, normalization, smoothing, restoration and enhancement; The specific process of image enhancement is: Assume that the noisy image is represented as: The original image f(x, y) is replaced by zero mean and variance σ 2 The Gaussian white noise n(x, y) is contaminated and passed through the Wiener filter g(x, y); According to the Wiener filtering criterion, we get: Assume the size of f(x, y) is M×N, and use H to represent the Wiener filter, which is: H=R f (R f +σ 2 I) -1 ; Among them, R f =E(ff T ) is the autocorrelation function, I is the unit matrix, let u i YesR f The eigenvector of , the corresponding eigenvalue is: i , let U=[u1,u2,…,u N ] is an M×N feature matrix, where Λ=diag[λ1,λ2,…,λ N ]; but: R f Substitute the functional relationship into the H functional relationship; In the actual image processing process, since the value of σ is unknown, it is estimated as: in, The specific process of image digitization is as follows: Determine appropriate points to sample and describe the image to be digitized; After the sampling description is completed, a certain range of values ​​is used to represent each sampling point; After the point representation is completed, all points are compressed and encoded; The specific process of data classification is as follows: In the data of different parts of brain tissue and different time periods after onset, each object is regarded as a class and the minimum distance between two objects is calculated; Merge the two classes with the smallest distance into a new class, and repeatedly calculate the distance between the new class and all classes until all data are classified; The specific process of image smoothing is: Determine the corresponding size of the image to be smoothed; Set a circular buffer in the image to temporarily store the result image and initialize it; Scan each pixel point in the circular buffer and sort the pixel values; assign the obtained intermediate value to the pixel point corresponding to the current point in the target image until all the pixels of the source image are processed; The specific process of feature point extraction is as follows: Determine the corresponding detection window in the image and normalize the image; After the normalization process is completed, the grayscale value and gradient of the image in the window are calculated; For each window, the gradient histogram is projected with a specified weight, and all histograms are connected to form the corresponding feature points.

2. The method for evaluating the acute ischemic stroke CT image data model using a computer chip as claimed in claim 1, characterized in that: The validation set is a small part of the data in the training set, which is used to verify the training results.

3. The method for evaluating the acute ischemic stroke CT image data model using a computer chip as claimed in claim 1, characterized in that: In step 4, the model is optimized using the Adam optimizer.

4. The method for evaluating the acute ischemic stroke CT image data model using a computer chip as claimed in claim 1, characterized in that: In step 4, the equipment includes acute ischemic stroke detection equipment and acute ischemic stroke recurrence detection equipment.

5. A computer program product stored on a computer-readable medium, comprising a computer-readable program, which, when executed on an electronic device, provides a user input interface to apply the method for evaluating the acute ischemic stroke CT image data model using a computer chip as described in claim 1.

6. A computer-readable storage medium storing instructions, which, when executed on a computer, enable the computer to apply the method for evaluating the acute ischemic stroke CT image data model using a computer chip as claimed in claim 1.

Images