A brain disease recognition method and device

Abstract

This application provides a method and apparatus for identifying brain diseases. The method includes: acquiring user information of a user to be tested and a preset device identifier for capturing two-dimensional brain images; acquiring two-dimensional brain images of the user to be tested and converting the two-dimensional brain images into three-dimensional brain images through a preset image conversion model; inputting the three-dimensional brain images into a pre-trained image processing model to obtain at least one brain structural parameter for each brain segmentation region output by the model; fitting the mean and standard deviation of each brain structural parameter based on the user information, the preset device identifier, and at least one brain structural parameter for each brain segmentation region to calculate the percentile corresponding to each brain structural parameter; inputting the percentile corresponding to each brain structural parameter into multiple prediction models to obtain the disease probability of the user to be tested output by each prediction model; and generating a reference evaluation result for the user to be tested according to the disease probabilities corresponding to each prediction model.

Description

Technical Field

[0001] This application relates to the field of image recognition technology, and in particular to a method and apparatus for recognizing brain diseases. Background Technology

[0002] In clinical practice, screening for brain diseases and diagnosing major brain disorders (including mild cognitive impairment, Alzheimer's disease, Parkinson's disease, small vessel disease, and multiple sclerosis) largely relies on non-invasive magnetic resonance imaging (MRI). However, the imaging manifestations of these diseases are difficult to distinguish from normal aging, and while there are many similarities in the imaging presentations of these major brain diseases, there are also differences in their specific characteristics. For example, although all these diseases present with brain atrophy, the patterns of atrophy differ; Alzheimer's disease primarily affects the hippocampus, while multiple sclerosis is more commonly associated with thalamic atrophy. In clinical diagnosis and treatment, the main approach for physicians interpreting imaging is to first determine whether the brain imaging examination is normal, and if the brain imaging is abnormal, to determine which type of neurological disease it belongs to.

[0003] Currently, high-resolution three-dimensional structural imaging of the human brain or cutting-edge functional imaging using magnetic resonance imaging can provide doctors with brain images that are more conducive to diagnosis. However, the processing of three-dimensional images is relatively complex and time-consuming, making it difficult to widely apply in actual clinical work, especially for hospitals in remote areas, where limited medical equipment prevents them from capturing three-dimensional images of the human brain. Summary of the Invention

[0004] In view of this, the purpose of this application is to provide at least one method and device for identifying brain diseases, which solves the technical problems of difficulty in acquiring three-dimensional images and inaccuracy of manual identification in the prior art, thereby achieving the technical effect of improving the accuracy of brain disease identification.

[0005] This application mainly includes the following aspects: In a first aspect, embodiments of this application provide a method for identifying brain diseases. The method includes: acquiring user information of a user to be tested and a preset device identifier for capturing a two-dimensional brain image of the user to be tested; acquiring the two-dimensional brain image of the user to be tested and converting the two-dimensional brain image into a three-dimensional brain image through a preset image conversion model; inputting the three-dimensional brain image into a pre-trained image processing model to obtain each brain segmentation region of the three-dimensional brain image and at least one brain structural parameter of each brain segmentation region output by the model; fitting the mean and standard deviation of each brain structural parameter according to the user information, the preset device identifier, and at least one brain structural parameter of each brain segmentation region to calculate the percentile corresponding to each brain structural parameter, wherein the percentile is used to describe the deviation of the brain structural parameter from a reference standard of the same user information and the preset device identifier; inputting the percentile corresponding to each brain structural parameter into multiple prediction models to obtain the disease probability of the user to be tested output by each prediction model, wherein one prediction model corresponds to one brain disease; and generating a reference evaluation result for the user to be tested according to the disease probabilities corresponding to each prediction model.

[0006] Optionally, the preset graphics conversion model can be obtained by: collecting multiple pairs of two-dimensional and three-dimensional brain images, with one data pair corresponding to one user; using the three-dimensional brain image of each data pair as the training target and the two-dimensional brain image of each data pair as the model input to train the diffusion model, thereby obtaining the preset graphics conversion model.

[0007] Optionally, the image processing model is trained by: acquiring an image processing training set for training the image processing model, the image processing training set including three-dimensional brain images of multiple users; inputting the three-dimensional brain images of multiple users into brain magnetic resonance imaging analysis software to segment each three-dimensional brain image according to a preset brain partitioning template to obtain multiple brain segmentation regions, and obtaining at least one brain structural parameter corresponding to each brain segmentation region; using each three-dimensional brain image in the image processing training set as input to the image processing model, and using at least one brain structural parameter of each brain segmentation region corresponding to each three-dimensional brain image as training labels for training the image processing model.

[0008] Optionally, the mean and standard deviation of each brain structural parameter are fitted in the following manner: A modeling sample set corresponding to multiple reference users for constructing the reference standard is obtained, wherein the modeling sample set includes the user information of each reference user, the preset device identifier, and at least one brain structural parameter corresponding to each brain segmentation region; a generalized additive positional scale shape model is constructed based on the user information of each reference user, the preset device identifier, and at least one brain structural parameter corresponding to each brain segmentation region; the user information and the preset device identifier of the user to be tested are input into the generalized additive positional scale shape model to obtain the mean and standard deviation of each brain structural parameter output by the model.

[0009] Optionally, the percentile for each brain structural parameter can be calculated in the following way:

[0010] In the above formula, This refers to the percentile corresponding to the i-th brain structural parameter. This refers to the value of the i-th brain structural parameter. This refers to the average value of the i-th brain structure parameter under the same user information and the preset device identifier. This refers to the standard deviation of the i-th brain structural parameter under the same user information and the preset device identifier. This refers to the cumulative distribution function of the standard normal distribution.

[0011] Optionally, each prediction model is trained by: obtaining a target group diagnosed with the brain disease corresponding to the prediction model and a non-target group not diagnosed with the brain disease corresponding to the prediction model, wherein the target group and the non-target group each include multiple users; using the percentile corresponding to each brain structural parameter of each user in the target group and the percentile corresponding to each brain structural parameter of each user in the target group as training samples to train the prediction model, so that the prediction model predicts the probability that each user belongs to the target group.

[0012] Optionally, the prediction model is a support vector machine or a multilayer perceptron.

[0013] Secondly, embodiments of this application also provide a brain disease identification device, the device comprising: an acquisition module, configured to acquire user information of a user to be tested and a preset device identifier for capturing a two-dimensional brain image of the user to be tested; a collection module, configured to collect the two-dimensional brain image of the user to be tested and convert the two-dimensional brain image into a three-dimensional brain image through a preset image conversion model; an image processing module, configured to input the three-dimensional brain image into a pre-trained image processing model to obtain each brain segmentation region segmented by the model and at least one brain structural parameter of each brain segmentation region; and a deviation identification module, configured to, based on the user information, ... The preset device identifier and at least one brain structural parameter of each brain segmentation region are fitted to calculate the mean and standard deviation of each brain structural parameter to determine the percentile corresponding to each brain structural parameter. The percentile is used to describe the deviation of the brain structural parameter from a reference standard with the same user information and the preset device identifier. The prediction module is used to input the percentile corresponding to each brain structural parameter into multiple prediction models to obtain the disease probability of the test user output by each prediction model. One prediction model corresponds to one brain disease. The evaluation module is used to generate a reference evaluation result for the test user according to the disease probability corresponding to each prediction model.

[0014] Thirdly, embodiments of this application also provide an electronic device, including: a processor, a memory, and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor and the memory communicate through the bus, and the machine-readable instructions are executed by the processor to perform the steps of the brain disease identification method described in the first aspect or any possible implementation of the first aspect.

[0015] Fourthly, embodiments of this application also provide a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the steps of the brain disease identification method described in the first aspect or any possible implementation of the first aspect.

[0016] This application provides a method and apparatus for identifying brain diseases. The method includes: acquiring user information of a user to be tested and a preset device identifier for capturing a two-dimensional brain image of the user to be tested; acquiring the two-dimensional brain image of the user to be tested and converting the two-dimensional brain image into a three-dimensional brain image through a preset image conversion model; inputting the three-dimensional brain image into a pre-trained image processing model to obtain each brain segmentation region of the three-dimensional brain image and at least one brain structural parameter of each brain segmentation region output by the model; fitting the mean and standard deviation of each brain structural parameter according to the user information, the preset device identifier, and at least one brain structural parameter of each brain segmentation region to calculate the percentile corresponding to each brain structural parameter, wherein the percentile is used to describe the deviation of the brain structural parameter from the same reference standard of the user information and the preset device identifier; inputting the percentile corresponding to each brain structural parameter into multiple prediction models to obtain the disease probability of the user to be tested output by each prediction model, wherein one prediction model corresponds to one brain disease; and generating a reference evaluation result for the user to be tested according to the disease probabilities corresponding to each prediction model. By acquiring two-dimensional brain images of the test user, clarifying the user information and the preset device identifier for capturing the two-dimensional brain images, and using a preset image conversion model to convert the two-dimensional brain images into three-dimensional brain images, a pre-trained image processing model is used to segment the three-dimensional brain images to obtain each brain segment region and at least one brain structural parameter for each brain segment region. Based on the user information, preset device identifier, and at least one brain structural parameter for each brain segment region, the mean and standard deviation of each brain structural parameter are fitted to calculate the percentile corresponding to each brain structural parameter. The percentile corresponding to each brain structural parameter is then input into the prediction model corresponding to each brain disease to obtain the probability of the test user being diagnosed with each brain disease. Finally, a reference evaluation result for the test user is generated based on the probability corresponding to each brain disease. This solves the technical problems of difficulty in acquiring three-dimensional images and inaccurate manual identification in existing technologies, and achieves the technical effect of improving the accuracy of brain disease identification.

[0017] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 A flowchart of a brain disease identification method provided in an embodiment of this application is shown.

[0020] Figure 2 A flowchart illustrating the steps for fitting the mean and standard deviation of various brain structural parameters according to an embodiment of this application is shown.

[0021] Figure 3 This diagram illustrates a functional block diagram of a brain disease identification device provided in an embodiment of this application.

[0022] Figure 4 A schematic diagram of the structure of an electronic device provided in an embodiment of this application is shown. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. It should be understood that the drawings in this application are for illustrative and descriptive purposes only and are not intended to limit the scope of protection of this application. Furthermore, it should be understood that the schematic drawings are not drawn to scale. The flowcharts used in this application illustrate operations implemented according to some embodiments of this application. It should be understood that the operations in the flowcharts may not be implemented in sequence, and steps without logical contextual relationships may be reversed or implemented simultaneously. In addition, those skilled in the art, guided by the content of this application, may add one or more other operations to the flowcharts, or remove one or more operations from the flowcharts.

[0024] Furthermore, the described embodiments are merely some, not all, of the embodiments of this application. The components of the embodiments of this application described and illustrated herein can typically be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0025] In current technologies, obtaining three-dimensional images of the human brain through high-resolution three-dimensional structural imaging based on scientific research or cutting-edge functional magnetic resonance imaging (fMRI) requires a considerable processing time and is difficult to widely apply in actual clinical practice, especially for hospitals in remote areas where insufficient hardware and software resources prevent the scanning of data for these research sequences. Data acquisition and analysis of high-resolution three-dimensional structural imaging primarily rely on open-source brain fMRI analysis software (FreeSurfer) to extract morphological indicators of brain tissue. However, this software is slow; even with the latest brain fMRI analysis software, processing a single case still takes approximately two hours, and it has high hardware and software requirements, impacting the timeliness of diagnosis and treatment.

[0026] Based on this, this application provides a method and apparatus for identifying brain diseases. It acquires a two-dimensional brain image of a user and identifies the user's information and the pre-defined device identifier used to capture the image. A pre-defined image conversion model is used to convert the two-dimensional brain image into a three-dimensional brain image. A pre-trained image processing model is used to segment the three-dimensional brain image to obtain each brain segmentation region and at least one brain structural parameter for each region. The mean and standard deviation of each brain structural parameter are then fitted according to the user information, the pre-defined device identifier, and at least one brain structural parameter for each segmentation region to calculate the percentile corresponding to each parameter. This percentile is then input into a prediction model corresponding to each brain disease to obtain the probability of the user being diagnosed with each disease. Finally, a reference evaluation result for the user is generated based on the probability of each brain disease. This method solves the technical problems of difficulty in acquiring three-dimensional images and inaccurate manual identification in the prior art, thereby improving the accuracy of brain disease identification. Specifically, the method is as follows: Please see Figure 1 , Figure 1 This is a flowchart illustrating a brain disease identification method provided in an embodiment of this application. Figure 1 As shown in the embodiments of this application, the brain disease identification method includes the following steps: S101: Obtain user information of the user to be tested and a preset device identifier for capturing two-dimensional brain images of the user to be tested.

[0027] The user information refers to individual information that influences the diagnosis of brain diseases and / or changes in brain structural parameters. User information includes at least one of the following: gender and age. In other words, a user's gender and / or age can influence the incidence of brain diseases, and may also influence changes in brain structural parameters. Selecting and collecting user gender and age facilitates subsequent determination of the percentile for each brain structural parameter among users of the same gender and age.

[0028] Furthermore, the system captures two-dimensional brain images of the user and clearly identifies the device used to capture these images. This identification describes the testing institution to which the device belongs, the device model, and the scanning sequence used for image acquisition. In other words, the testing institution can refer to a hospital, health checkup center, etc., where the two-dimensional brain images are captured, and the scanning sequence can be a magnetization-prepared rapid gradient-echo (MPRAGE), a spoiled gradient-recalled echo (SPGR), or brain volume imaging (BRAVO). For example, the identification could describe that the two-dimensional brain images of the user were captured at Hospital A using device B according to the MPRAGE scanning sequence.

[0029] S102: Acquire a two-dimensional brain image of the user to be tested, and convert the two-dimensional brain image into a three-dimensional brain image through a preset graphics conversion model.

[0030] In other words, a two-dimensional brain image of the user to be tested is captured using a device corresponding to a preset device identifier. For example, the two-dimensional brain image in this application may refer to two-dimensional neural imaging data based on T1-weighted imaging sequences obtained by magnetic resonance imaging (MRI), and the device for acquiring the two-dimensional brain image may be a magnetic resonance scanner from General Electric, Philips, or Siemens.

[0031] The preset image conversion model is used to convert two-dimensional images into three-dimensional images. This allows for the conversion of two-dimensional brain images into three-dimensional brain images, eliminating the need to directly capture three-dimensional images of the user's brain, thus improving processing efficiency and reducing the need for every hospital to install equipment capable of capturing three-dimensional images. However, this requires a large amount of data to train the preset image conversion model so that it can obtain three-dimensional brain images from two-dimensional brain images of the user.

[0032] Specifically, the preset graphics conversion model is obtained by: collecting multiple pairs of two-dimensional and three-dimensional brain images, with one data pair corresponding to one user; using the three-dimensional brain image of each data pair as the training target and the two-dimensional brain image of each data pair as the model input to train the diffusion model, thereby obtaining the preset graphics conversion model.

[0033] In other words, multiple users' corresponding 2D and 3D brain images (MRI) are collected, and each user's 2D and 3D brain image are treated as a data pair. For each data pair, the 2D brain image is used as the input to a pre-defined image transformation model, and the 3D brain image is used as the training label to train the model. This process aims to make the predicted image output by the pre-defined image transformation model approximate the 3D brain image, thus obtaining a trained pre-defined image transformation model. The pre-defined image transformation model can be a diffusion model, which has a high-fidelity advantage in generating image details in the field of image generation.

[0034] S103: Input the three-dimensional brain image into a pre-trained preset image processing model to obtain the model output of each brain segmentation region of the three-dimensional brain image and at least one brain structural parameter of each brain segmentation region.

[0035] The pre-defined image processing model is used to segment a 3D brain image into multiple brain regions and extract brain structural parameters for each segmented region. These parameters include at least one of the following: brain volume, cortical thickness, and cortical surface area. Existing technologies for brain image segmentation are inefficient; for example, segmenting 3D brain images using software like FreeSurfer or a shape prior module (SPM) takes 2 to 8 hours, which does not meet the timely diagnostic needs of clinicians in practical clinical applications. Therefore, this application chooses to use a deep learning-based 3D adaptive medical image segmentation framework (No New U-Net, an adaptive segmentation network based on U-Net) to construct the pre-defined image processing model, and trains the model using the segmentation results from FreeSurfer.

[0036] Specifically, the preset image processing model is trained as follows: an image processing training set is obtained for training the preset image processing model, the image processing training set including three-dimensional brain images of multiple users; the three-dimensional brain images of multiple users are input into brain magnetic resonance imaging analysis software to segment each three-dimensional brain image according to a preset brain partitioning template to obtain multiple brain segmentation regions, and at least one brain structural parameter corresponding to each brain segmentation region is obtained; each three-dimensional brain image in the image processing training set is used as input to the preset image processing model, and each brain segmentation region corresponding to each three-dimensional brain image and its at least one brain structural parameter are used as training labels for training the preset image processing model.

[0037] The preset brain partition template describes a pre-planned reference template for brain segmentation. The preset brain partition template can be used to segment the brain into telencephalon, diencephalon, cerebellum, and brainstem, and then further segment each of these into multiple regions. Multiple preset brain partition templates can be set, with each preset brain partition template corresponding to a preset image processing model.

[0038] In other words, if different preset brain partition templates are set, the brain magnetic resonance imaging analysis software segments each three-dimensional brain image in the image processing training set according to the preset brain partition template to obtain multiple brain segmentation regions obtained from each three-dimensional brain image and at least one brain structural parameter corresponding to each brain segmentation region. In this way, each three-dimensional brain image in the segmented image processing training set is used as the model input, and the multiple brain segmentation regions corresponding to the preset brain partition template and at least one brain structural parameter corresponding to each brain segmentation region output by the brain magnetic resonance imaging analysis software are used as training labels to train the preset image processing model, so that the output result of the preset image processing model approaches the training labels. Thus, a preset image processing model that can segment three-dimensional brain images according to the preset brain partition template to obtain multiple brain segmentation regions and predict at least one brain structural parameter corresponding to each brain segmentation region is trained.

[0039] Furthermore, a preset brain partition template corresponds to a preset image processing model. Thus, in practical applications, the corresponding preset image processing model can be determined according to the selected preset brain partition template. By inputting the three-dimensional brain image of the user to be tested into the preset image processing model corresponding to the selected preset brain partition template, multiple brain segmentation regions corresponding to the three-dimensional brain image of the user to be tested can be obtained, and at least one brain structural parameter corresponding to each brain segmentation region can be determined.

[0040] Furthermore, the preset image processing model can improve the efficiency and accuracy of brain region segmentation and cortical reconstruction, reducing processing time to 30 minutes per case. The average DICE coefficient (an index used to measure the degree of overlap between two segmented regions) between the segmentation results obtained by the preset image processing model and those obtained by the brain magnetic resonance imaging (MRI) analysis software is 0.93. The average intraclass correlation coefficient (ICC) between the brain structural parameters obtained by the preset image processing model and those obtained by the MRI analysis software is 0.975. Therefore, in practical clinical applications, it is not necessary to wait for the segmentation results and identified brain structural parameters provided by the MRI analysis software; the preset image processing model can be directly selected to obtain multiple brain segmentation regions corresponding to the three-dimensional brain image of the test subject, as well as at least one brain structural parameter corresponding to each brain segmentation region.

[0041] For example, a preset graphic conversion model and a preset image processing model can be integrated into the MRI host and post-processing workstation. This allows for the direct conversion of two-dimensional brain images obtained from scanning into three-dimensional brain images, and the identification of each brain segmentation region and its corresponding at least one brain structural parameter, thereby improving the rate of generating brain health reports.

[0042] S104: Based on the user information, the preset device identifier, and at least one brain structural parameter of each brain segmentation region, fit the mean and standard deviation of each brain structural parameter to calculate the percentile corresponding to each brain structural parameter.

[0043] The percentile is used to describe the deviation of the brain structural parameters from a reference standard using the same user information and the preset device identifier. For example, the percentile for each brain structural parameter refers to the data distribution position of the test user's brain structural parameters compared to the brain structural parameters of other users of the same age, gender, testing institution, device model, and using the same scanning sequence. If the percentile for hippocampal volume is 31%, it means that among 100 healthy individuals whose hippocampal volumes were determined using the same scanning sequence as the test user (age, gender, testing institution, device model, and the same user), approximately 31 individuals have smaller hippocampal volumes than the test user, and 69 individuals have larger hippocampal volumes.

[0044] Please see Figure 2 , Figure 2 This is a flowchart illustrating the steps for fitting the mean and standard deviation of various brain structural parameters as provided in the embodiments of this application. Figure 2 As shown, the mean and standard deviation of each brain structural parameter were fitted using the following method: S201: Obtain the modeling sample sets corresponding to the multiple reference users used to build the reference standard.

[0045] The modeling sample set includes the user information of each reference user, the preset device identifier, and at least one brain structural parameter corresponding to each brain segmentation region.

[0046] S202: Construct a generalized additive positional scale shape model based on the user information of each reference user, the preset device identifier, and at least one brain structural parameter corresponding to each brain segmentation region.

[0047] S203: Input the user information of the user to be tested and the preset device identifier into the generalized additive positional scale shape model to obtain the mean and standard deviation of each brain structure parameter output by the model.

[0048] In this model, multiple reference users can be selected from among those who underwent two-dimensional brain image acquisition using the same testing institution, equipment model, and scanning sequence, and who do not have brain diseases. Clearly, these multiple reference users involve users of different ages and genders. Therefore, in constructing the Generalized Additive Positional Scale Shape Model (GAMLSS), age and gender are treated as fixed effects, while the testing institution, equipment model, and scanning sequence are integrated into a single index and treated as random effects. At least one brain structural parameter for each brain segmentation region of each reference user is used as the dependent variable, and age is fitted using a spline function, with the degrees of freedom of the spline function selected according to the optimal Bayesian information criterion. Furthermore, inputting the user information and preset equipment identifiers into the GAMLSS provides the mean and standard deviation fitted under the same user information and preset equipment identifiers.

[0049] The percentile for each brain structural parameter was calculated using the following method: (1) In formula (1), This refers to the percentile corresponding to the i-th brain structural parameter. This refers to the value of the i-th brain structural parameter. This refers to the average value of the i-th brain structure parameter under the same user information and the preset device identifier. This refers to the standard deviation of the i-th brain structural parameter under the same user information and the preset device identifier. This refers to the cumulative distribution function of the standard normal distribution.

[0050] In this way, a percentile is calculated for at least one brain structural parameter corresponding to each brain segmentation region of the test user. Thus, the percentiles of the brain structural parameters corresponding to each brain segmentation region of the test user are used to construct the deviation vector of the test user. By inputting the deviation vector into the prediction model corresponding to each brain disease, the probability of the test user being diagnosed with each brain disease can be obtained.

[0051] S105: Input the percentile corresponding to each brain structural parameter into multiple prediction models to obtain the disease probability of the test user output by each prediction model.

[0052] In this scenario, one prediction model corresponds to one type of brain disease. In other words, different prediction models need to be built in advance for different brain diseases, so as to predict the probability of a user being diagnosed with that brain disease using the prediction model corresponding to that brain disease.

[0053] Specifically, each prediction model is trained in the following way: a target group diagnosed with the brain disease corresponding to the prediction model and a non-target group not diagnosed with the brain disease corresponding to the prediction model are obtained, wherein the target group and the non-target group each include multiple users; the percentile corresponding to each brain structural parameter of each user in the target group and the percentile corresponding to each brain structural parameter of each user in the target group are used as training samples to train the prediction model, so that the prediction model predicts the probability that each user belongs to the target group.

[0054] For example, the prediction model is a support vector machine or a multilayer perceptron, and the target group and non-target group should correspond to different users. That is, based on machine learning models such as support vector machines or multilayer perceptrons, disease risk is predicted. For each brain disease, the percentile of at least one brain structural parameter corresponding to each brain segmentation region of patients diagnosed with the brain disease is constructed as the target sample, and the percentile of at least one brain structural parameter corresponding to each brain segmentation region of other users not diagnosed with the brain disease is constructed as the reference sample for a binary classification task. These other users not diagnosed with the brain disease can be either normal users without any diagnosed brain disease or users diagnosed with other brain diseases, to consider the potential mutual influence between brain diseases. In other words, the model input is the percentile of at least one brain structural parameter corresponding to each brain segmentation region, and the model output is the probability of belonging to the target group. Furthermore, a specific predictive model is constructed for each brain disease. Therefore, for each brain disease, the percentile corresponding to each brain structural parameter of the test user is input into the predictive model corresponding to the brain disease to obtain the probability that the test user is diagnosed with the brain disease.

[0055] S106: Generate reference evaluation results for the user to be tested according to the disease probabilities corresponding to each prediction model.

[0056] For example, the disease probabilities corresponding to each prediction model are sorted in descending order, and the brain diseases corresponding to the prediction models with the highest disease probabilities are selected as the reference assessment results for the test users, so that doctors can refer to the reference assessment results of the test users to assist in diagnosis and treatment.

[0057] Furthermore, regardless of whether the user is diagnosed or not, outcome analysis can be performed. Two-dimensional brain images of the user are resampled after 1, 3, and 5 years, and the percentile of at least one brain structural parameter corresponding to each brain segmentation region of the user is determined. Based on the percentile of each brain structural parameter of the user after 1, 3, and 5 years, the probability of being diagnosed with each brain disease is predicted. This clarifies the brain changes of the user, and for diagnosed diseases, it can clarify disease progression and provide auxiliary assessment of the patient's prognosis.

[0058] Furthermore, the technical solution of this application converts two-dimensional brain images taken in routine clinical settings into three-dimensional brain images, then uses deep learning algorithms to segment brain tissue and identify at least one brain structural parameter for each segmented brain region. Based on this parameter, the percentile corresponding to each brain structural parameter is calculated, and a prediction model is used to predict the disease probability of the user being tested. Thus, by using image conversion, image segmentation, and brain structural parameter identification, the time-consuming problem of existing technologies is solved. The model also has stronger generalization ability, better meets clinical needs, has strong clinical applicability and scalability, and high clinical application value, providing reliable support for the diagnosis and prognostic assessment of neurological diseases.

[0059] Based on the same application concept, this application also provides a brain disease identification device corresponding to the brain disease identification method provided in the above embodiments. Since the principle of the device in this application is similar to the brain disease identification method in the above embodiments of this application, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be described again.

[0060] like Figure 3 As shown, Figure 3This is a functional block diagram of a brain disease identification device provided in an embodiment of this application. The brain disease identification device 10 provided in this embodiment includes: an acquisition module 101, used to acquire user information of a user to be tested and a preset device identifier for capturing a two-dimensional brain image of the user to be tested; a collection module 102, used to collect the two-dimensional brain image of the user to be tested and convert the two-dimensional brain image into a three-dimensional brain image through a preset image conversion model; an image processing module 103, used to input the three-dimensional brain image into a pre-trained preset image processing model to obtain each brain segmentation region segmented by the model and at least one brain structural parameter of each brain segmentation region; and a deviation identification module 104, used to, based on the user information, The preset device identifier and at least one brain structural parameter of each brain segmentation region are fitted to calculate the mean and standard deviation of each brain structural parameter to calculate the percentile corresponding to each brain structural parameter. The percentile is used to describe the deviation of the brain structural parameter from the same reference standard of user information and preset device identifier. The prediction module 105 is used to input the percentile corresponding to each brain structural parameter into multiple prediction models to obtain the disease probability of the test user output by each prediction model. One prediction model corresponds to one brain disease. The evaluation module 106 is used to generate a reference evaluation result for the test user according to the disease probability corresponding to each prediction model.

[0061] Based on the same application concept, see [link / reference] Figure 4 The diagram shown is a structural schematic of an electronic device provided in an embodiment of this application. The electronic device 20 includes a processor 201, a memory 202, and a bus 203. The memory 202 stores machine-readable instructions executable by the processor 201. When the electronic device 20 is running, the processor 201 and the memory 202 communicate through the bus 203. The machine-readable instructions are executed by the processor 201 to perform the steps of the brain disease recognition method described in any of the above embodiments.

[0062] Specifically, when the machine-readable instructions are executed by the processor 201, the following processing can be performed: acquiring user information of the user to be tested and a preset device identifier for capturing a two-dimensional brain image of the user to be tested; acquiring the two-dimensional brain image of the user to be tested and converting the two-dimensional brain image into a three-dimensional brain image through a preset image conversion model; inputting the three-dimensional brain image into a pre-trained image processing model to obtain each brain segmentation region of the three-dimensional brain image and at least one brain structural parameter of each brain segmentation region output by the model; fitting the mean and standard deviation of each brain structural parameter according to the user information, the preset device identifier, and at least one brain structural parameter of each brain segmentation region to calculate the percentile corresponding to each brain structural parameter, the percentile being used to describe the deviation of the brain structural parameter from the same reference standard of the user information and the preset device identifier; inputting the percentile corresponding to each brain structural parameter into multiple prediction models to obtain the disease probability of the user to be tested output by each prediction model, one prediction model corresponding to one brain disease; generating a reference evaluation result for the user to be tested according to the disease probabilities corresponding to each prediction model.

[0063] Based on the same concept, embodiments of this application also provide a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the steps of the brain disease identification method provided in the above embodiments.

[0064] Specifically, the storage medium can be a general-purpose storage medium, such as a portable disk or hard disk. When the computer program on the storage medium is run, it can execute the above-mentioned brain disease identification method. This method acquires a two-dimensional brain image of the user to be tested, identifies the user information and the preset device identifier for capturing the two-dimensional brain image, and uses a preset image conversion model to convert the two-dimensional brain image into a three-dimensional brain image. A pre-trained image processing model is used to segment the three-dimensional brain image to obtain each brain segmentation region and at least one brain structural parameter for each brain segmentation region. Based on the user information, the preset device identifier, and at least one brain structural parameter for each brain segmentation region, the mean and standard deviation of each brain structural parameter are fitted to calculate the percentile corresponding to each brain structural parameter. The percentile corresponding to each brain structural parameter is then input into the prediction model corresponding to each brain disease to obtain the probability of the user being diagnosed with each brain disease. Finally, a reference evaluation result for the user is generated based on the probability of each brain disease. This method solves the technical problems of difficulty in acquiring three-dimensional images and inaccurate manual identification in the prior art, thereby improving the technical effect of improving the accuracy of brain disease identification.

[0065] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and devices described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. In the several embodiments provided in this application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division; in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Another point is that the displayed or discussed mutual coupling or direct coupling or communication connection may be through some communication interfaces; the indirect coupling or communication connection of devices or units may be electrical, mechanical, or other forms.

[0066] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0067] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0068] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0069] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for identifying brain diseases, characterized in that, The method includes: Obtain user information of the user to be tested and a preset device identifier for capturing two-dimensional brain images of the user to be tested; Two-dimensional brain images of the user to be tested are acquired, and the two-dimensional brain images are converted into three-dimensional brain images through a preset graphics conversion model; The three-dimensional brain image is input into a pre-trained image processing model to obtain the model output of each brain segmentation region of the three-dimensional brain image and at least one brain structural parameter of each brain segmentation region; Based on the user information, the preset device identifier, and at least one brain structural parameter for each brain segmentation region, the mean and standard deviation of each brain structural parameter are fitted to calculate the percentile corresponding to each brain structural parameter. The percentile is used to describe the deviation of the brain structural parameter from the same reference standard of the user information and the preset device identifier. The percentile corresponding to each brain structure parameter is input into multiple prediction models to obtain the disease probability of the test user output by each prediction model. One prediction model corresponds to one brain disease. Based on the disease probabilities corresponding to each prediction model, a reference evaluation result is generated for the user to be tested.

2. The method according to claim 1, characterized in that, The preset graphics conversion model is obtained through the following methods: Collect multiple pairs of two-dimensional and three-dimensional brain images, with one data pair corresponding to one user; The diffusion model is trained by using the three-dimensional brain image of each data pair as the training target and the two-dimensional brain image of each data pair as the model input, so as to obtain the preset graphics conversion model.

3. The method according to claim 1, characterized in that, The image processing model is trained in the following manner: Obtain an image processing training set for training the image processing model, the image processing training set including three-dimensional brain images of multiple users; Three-dimensional brain images from multiple users are input into brain magnetic resonance imaging analysis software. Each three-dimensional brain image is segmented according to a preset brain partitioning template to obtain multiple brain segmentation regions, and at least one brain structural parameter corresponding to each brain segmentation region is obtained. Each three-dimensional brain image in the image processing training set is used as the input to the image processing model, and at least one brain structural parameter of each brain segmentation region corresponding to each three-dimensional brain image is used as the training label of the image processing model to train the image processing model.

4. The method according to claim 1, characterized in that, The mean and standard deviation of each brain structural parameter were fitted using the following method: Obtain modeling sample sets corresponding to multiple reference users for constructing reference standards. The modeling sample sets include the user information of each reference user, the preset device identifier, and at least one brain structural parameter corresponding to each brain segmentation region. A generalized additive positional scale shape model is constructed based on the user information of each reference user, the preset device identifier, and at least one brain structural parameter corresponding to each brain segmentation region. The user information of the user to be tested and the preset device identifier are input into the generalized additive positional scale shape model to obtain the mean and standard deviation of each brain structure parameter output by the model.

5. The method according to claim 1 or 4, characterized in that, The percentile for each brain structural parameter was calculated using the following method: In the above formula, This refers to the percentile corresponding to the i-th brain structural parameter. This refers to the value of the i-th brain structural parameter. This refers to the average value of the i-th brain structure parameter under the same user information and the preset device identifier. This refers to the standard deviation of the i-th brain structural parameter under the same user information and the preset device identifier. This refers to the cumulative distribution function of the standard normal distribution.

6. The method according to claim 1, characterized in that, Each prediction model is trained in the following way: The target group and the non-target group, which are diagnosed with the brain disease corresponding to the prediction model, are obtained. The target group and the non-target group each include multiple users. The prediction model is trained by using the percentile corresponding to each brain structural parameter of each user in the target group and the percentile corresponding to each brain structural parameter of each user in the target group as training samples, so that the prediction model can predict the probability that each user belongs to the target group.

7. The method according to claim 1 or 6, characterized in that, The prediction model is a support vector machine or a multilayer perceptron.

8. A brain disease identification device, characterized in that, The device includes: The acquisition module is used to acquire user information of the user to be tested and a preset device identifier for capturing two-dimensional brain images of the user to be tested; The acquisition module is used to acquire two-dimensional brain images of the user under test and convert the two-dimensional brain images into three-dimensional brain images through a preset graphics conversion model. An image processing module is used to input the three-dimensional brain image into a pre-trained image processing model to obtain each brain segmentation region of the three-dimensional brain image and at least one brain structural parameter of each brain segmentation region as output by the model. The deviation recognition module is used to fit the mean and standard deviation of each brain structural parameter based on the user information, the preset device identifier and at least one brain structural parameter of each brain segmentation region, so as to calculate the percentile corresponding to each brain structural parameter. The percentile is used to describe the deviation of the brain structural parameter from the same reference standard of the user information and the preset device identifier. The prediction module is used to input the percentile corresponding to each brain structure parameter into multiple prediction models to obtain the disease probability of the test user output by each prediction model. One prediction model corresponds to one brain disease. The evaluation module is used to generate reference evaluation results for the user under test according to the disease probabilities corresponding to each prediction model.

9. An electronic device, characterized in that, include: The device includes a processor, a memory, and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor communicates with the memory via the bus, and the machine-readable instructions are executed by the processor to perform the steps of the brain disease identification method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, performs the steps of the brain disease identification method as described in any one of claims 1 to 7.

Images