A data-driven acute phase assessment system for anti-LGI1 encephalitis

Abstract

This invention provides a data-driven acute-phase assessment system for LGI1 encephalitis, comprising: a data acquisition unit for acquiring FDG-PET brain imaging data of multiple LGI1 encephalitis patients and healthy subjects; an LGI1 encephalitis characteristic brain metabolic distribution pattern construction unit for obtaining brain metabolic components characterizing LGI1 encephalitis based on the radioactivity concentration values ​​of each pixel in the FDG-PET brain imaging data of the multiple LGI1 encephalitis patients and healthy subjects, and re-projecting the brain metabolic components back to the corresponding positions of each pixel in the brain imaging data to construct an LGI1 encephalitis characteristic brain metabolic distribution pattern; and an encephalitis severity assessment unit for obtaining the characterization result of the LGI1 encephalitis characteristic brain metabolic distribution pattern based on the FDG-PET brain imaging data of the patient to be assessed, as the acute-phase assessment result of LGI1 encephalitis in the patient to be assessed. This invention solves the problem that existing LGI1 encephalitis assessment methods do not jointly consider the covariant relationship between metabolic reduction and increase, resulting in limited assessment effectiveness.

Description

Technical Field

[0001] This invention belongs to the field of clinical medical technology, and in particular relates to a data-driven acute phase assessment system for anti-LGI1 encephalitis. Background Technology

[0002] The brain is composed of hundreds of billions of neurons, and the interconnected networks formed by these neurons form the basis for the complex and diverse functions of the brain. Research shows that the onset and progression of LGI1 encephalitis are not only related to the damage to the function of individual brain regions, but also to the disruption of the covariant balance of functions between different brain regions. In-depth research into the covariant relationships of metabolism in different brain regions during the onset of LGI1 encephalitis will help to further understand the pathogenesis of LGI1 encephalitis in the acute phase and may also help to identify new targets for disease intervention and treatment.

[0003] Currently, analyses of brain metabolism in LGI1 encephalitis patients primarily employ intergroup statistical comparisons to identify which brain regions exhibit metabolic damage or abnormally elevated metabolism compared to healthy controls. However, these analyses do not consider the covariance between decreased and increased metabolism, or whether this covariance affects prognosis. Furthermore, while FDG-PET images can reflect each subject's glucose consumption capacity, conventional data analysis methods typically select disease-free brain regions as reference areas for data standardization to normalize the physiological baselines of all subjects. For example, the cerebellum is often chosen as the reference area in Alzheimer's disease studies. Inappropriate reference area selection can severely impact the reproducibility and objectivity of the data analysis results. In LGI1 encephalitis research, it remains unclear which brain regions are unaffected by the disease; therefore, the effectiveness of current LGI1 encephalitis severity assessments needs improvement. Summary of the Invention

[0004] Based on the above analysis, the present invention aims to provide a data-driven acute phase assessment system for anti-LGI1 encephalitis, which addresses the problem that existing LGI1 encephalitis assessment systems do not consider the covariant relationship between metabolic reduction and increase, thus limiting the assessment effectiveness.

[0005] The objective of this invention is mainly achieved through the following technical solutions:

[0006] This invention provides a system for assessing the severity and prognosis of acute disease in patients with LGI1 encephalitis based on specific related patterns. The system includes:

[0007] The data acquisition unit is used to acquire FDG-PET brain imaging data from multiple LGI1 encephalitis patients and healthy subjects.

[0008] The LGI1 encephalitis characteristic brain metabolism distribution pattern construction unit is used to obtain brain metabolic components that characterize the characteristics of LGI1 encephalitis based on the radioactivity concentration values ​​of each pixel in the FDG-PET brain imaging data of multiple LGI1 encephalitis patients and healthy subjects, and to project the brain metabolic components back into the corresponding positions of each pixel in the brain imaging data to construct the LGI1 encephalitis characteristic brain metabolism distribution pattern.

[0009] The encephalitis severity assessment unit is used to obtain the characterization results of the brain metabolic distribution pattern of the LGI1 encephalitis based on the FDG-PET brain imaging data of the patient to be assessed, and to serve as the assessment result of the acute phase of LGI1 encephalitis of the patient to be assessed.

[0010] Furthermore, brain metabolic components characterizing LGI1 encephalitis were obtained using the following method:

[0011] The FDG-PET brain imaging data of multiple LGI1 encephalitis patients and healthy subjects were arranged sequentially according to the position index of the pixels;

[0012] The data for each patient and subject, arranged in the correct order, are combined column-wise to obtain a data matrix P.

[0013] The residual matrix IRP is obtained by centering the data matrix P.

[0014] Singular matrix decomposition was performed on the residual matrix IRP to obtain brain metabolic components that characterize LGI1 encephalitis.

[0015] Furthermore, after obtaining the data matrix P, the process also includes performing a log transformation on the data matrix P;

[0016] The residual matrix IRP is obtained by centering the data matrix after log transformation.

[0017] Furthermore, the centering includes row centering and column centering; the residual matrix IRP is obtained by the following method:

[0018] Subtracting the individual average value from the value of each pixel in the data matrix after log transformation yields the row-centered data.

[0019] The LGI1 encephalitis patients and healthy subjects were divided into a patient group and a healthy group, and the mean values ​​within each group were calculated.

[0020] For the row-centered data matrix, the residual matrix IRP is obtained by subtracting the average value of the corresponding column in the column from the value of each pixel.

[0021] Furthermore, the centralization is represented as:

[0022] ;

[0023] In this context, the subscript I represents the I-th subject, and V represents the V-th pixel. This represents the individual mean of the i-th subject, i.e., the within-row mean. This represents the group average.

[0024] Furthermore, singular matrix decomposition is performed on the residual matrix IRP to obtain brain metabolic components characterizing LGI1 encephalitis, represented as follows:

[0025] ;

[0026] in, This is the data matrix corresponding to the LGI1 encephalitis patient group in the residual matrix IRP. These are the brain metabolic components that have been broken down. Indicates the first The first subject's characterization A score for each brain metabolic component.

[0027] Furthermore, the most orthogonal brain metabolic components obtained after singular matrix decomposition These are the brain metabolic components that characterize LGI1 encephalitis.

[0028] Furthermore, the characterization results of the brain metabolic distribution patterns characteristic of LGI1 encephalitis using FDG-PET brain imaging data of the patients to be evaluated were obtained through the following methods:

[0029] The FDG-PET brain imaging data of the patients to be evaluated were subjected to a log transformation. The individual mean value of each pixel in the log-transformed data matrix was then subtracted from the value of the individual pixel, and the column mean value within the patient group was also subtracted to obtain the decentralized brain imaging data of the patients to be evaluated. ,

[0030] Based on decentralized brain imaging data The characterization result is obtained through the following formula:

[0031] ;

[0032] in, The characterization results are as follows. It is the most orthogonal brain metabolic component.

[0033] Furthermore, a data preprocessing unit is included before the LGI1 encephalitis characteristic brain metabolic distribution pattern construction unit, used to preprocess the acquired brain imaging data using the following methods:

[0034] Convert all raw DICOM data into analyzable 3D image files (NII data);

[0035] The DARTEL algorithm was used to standardize the NII data corresponding to the FDG-PET brain images of all subjects to the MNI space;

[0036] Uses a half-height width of 8mm 3 Gaussian smoothing is performed on the spatially standardized data using a Gaussian kernel.

[0037] Furthermore, the healthy subjects were matched with the LGI1 encephalitis patients in terms of age and sex.

[0038] The beneficial effects of this technical solution are:

[0039] This invention establishes a data-driven acute-phase assessment system for LGI1 encephalitis. Without requiring pre-defined reference brain regions, it can systematically obtain the metabolic abnormality patterns across the entire brain of encephalitis patients. This pattern includes not only metabolic changes in individual brain regions but also the interconnected patterns of these changes. Specifically, it characterizes which brain regions in LGI1 encephalitis patients exhibit increased metabolism and which exhibit decreased metabolism compared to healthy subjects; moreover, these increases and decreases must occur simultaneously to characterize the disease pattern of LGI1 encephalitis. Furthermore, by using the similarity score of this metabolic pattern characterization in LGI1 encephalitis patients, the severity of the disease can be assessed, and its prognosis can be accurately predicted, providing strong objective evidence for the systematic assessment of the acute phase of LGI1 encephalitis.

[0040] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description, claims, and drawings. Attached Figure Description

[0041] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts.

[0042] Figure 1 This is a schematic diagram of a data-driven acute phase assessment system for anti-LGI1 encephalitis according to an embodiment of the present invention;

[0043] Figure 2 This is a schematic diagram of the workflow of the LGI1 encephalitis characteristic brain metabolic distribution pattern construction unit in an embodiment of the present invention. Detailed Implementation

[0044] The preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not intended to limit the scope of the present invention.

[0045] One embodiment of the present invention provides a data-driven acute-phase assessment system for anti-LGI1 encephalitis, such as... Figure 1 As shown, it includes:

[0046] The data acquisition unit is used to acquire FDG-PET brain imaging data from multiple LGI1 encephalitis patients and healthy subjects.

[0047] The LGI1 encephalitis characteristic brain metabolic distribution pattern construction unit is used to compare the radioactivity concentration values ​​of each pixel in the FDG-PET brain imaging data of multiple LGI1 encephalitis patients and healthy subjects to obtain brain metabolic components that characterize LGI1 encephalitis, and to re-introduce the brain metabolic components back into the brain imaging data to construct the LGI1 encephalitis characteristic brain metabolic distribution pattern.

[0048] The encephalitis severity assessment unit is used to calculate the characterization results of the FDG-PET brain imaging data of the patient to be assessed for the characteristic brain metabolic distribution pattern of LGI1 encephalitis, and to assess the severity of LGI1 encephalitis of the patient to be assessed based on the characterization results.

[0049] Specifically, this embodiment uses a commercial PET / CT scanner to obtain... PET images. Patients are required to fast for 4-6 hours before the scan to ensure adequate blood glucose levels. and conduct For drug injection, the dosage is determined based on the patient's weight, and is approximately... After the drug injection, the patient needs to rest in a quiet room, avoiding light and sound stimulation or other activities that may affect brain metabolic distribution. After 50 minutes, when the radioactive reagent has largely stabilized in the brain, brain PET images are acquired. The acquisition time is 10 minutes, and reconstruction is performed using Ordered Subset Expectation-Maximization (OSEM), with simultaneous decay correction and dead-time correction. Finally, the acquired data is stored in standard DICOM format.

[0050] The data preprocessing unit is included before the LGI1 encephalitis characteristic brain metabolic distribution pattern construction unit, which is used to preprocess the acquired FDG-PET brain imaging data using the following methods:

[0051] Convert all raw DICOM format data into analyzable 3D image files – NII data;

[0052] The DARTEL algorithm was used to standardize the NII data corresponding to the FDG-PET brain images of all subjects to the MNI space;

[0053] Using half-height and width as Gaussian smoothing is applied to the spatially standardized data using Gaussian kernels to obtain preprocessed brain imaging data.

[0054] Furthermore, such as Figure 2 As shown, the brain metabolic distribution pattern construction unit for LGI1 encephalitis characteristics obtained brain metabolic components characterizing LGI1 encephalitis features through the following method:

[0055] The FDG-PET brain imaging data of multiple LGI1 encephalitis patients and healthy subjects were arranged sequentially according to the pixel location index (Voxels(N));

[0056] The data for each patient and subject, arranged in the correct order, are column-wise concatenated to obtain a data matrix P; then, a log transformation is performed on the data matrix P.

[0057] Based on the data matrix P after log transformation, row centering and column centering are performed to obtain the residual matrix IRP;

[0058] Singular matrix decomposition was performed on the residual matrix IRP to obtain brain metabolic components that characterize LGI1 encephalitis.

[0059] It should be noted that brain metabolic components refer to the pattern of brain metabolic changes in LGI1 encephalitis patients compared to healthy subjects. Specifically, some brain regions show significantly reduced metabolism, while others show significantly increased metabolism, and still others are unaffected by LGI1 encephalitis. If these brain tissue structural regions with reduced, increased, or unchanged metabolisms are distributed regularly and occur simultaneously, then the pattern of brain metabolic changes in these brain regions is considered to be a brain metabolic component that can characterize the characteristics of LGI1 encephalitis patients.

[0060] As a specific example, FDG-PET brain imaging data of 30 patients with LGI1 encephalitis were first randomly selected, and imaging data of 30 healthy subjects matched for their age and sex were also selected.

[0061] The PET image data of each subject are arranged in order according to the position index of the pixel. The value of each point is the original radioactivity concentration value of the image. Thus, the 3D image data of each subject can be transferred into a row vector data with a size of 1×M, where M is the number of pixels in the image.

[0062] Data from 60 subjects were column-wise concatenated, with one subject per row, resulting in 60 rows. Specifically, the first 30 rows contained data from LGI1 encephalitis patients, and the last 30 rows contained data from healthy subjects. The concatenated data matrix had a size of N×M, where N was the number of subjects (N = 60). This data matrix was denoted as... ;

[0063] For matrix Perform a log transformation to obtain the matrix. The log transform can convert nonlinear multiplicative noise in an image into linear additive noise.

[0064] right First, perform "row centering," which involves subtracting the average value of its row from the value of each point; then perform "column centering," which involves subtracting the average value of its column from the value of each point. This yields the residual matrix (IRP), as shown in the following formula:

[0065] ;

[0066] Where the subscript I represents subject I, and V represents the Vth pixel, This represents the individual mean of the i-th subject, i.e., the within-row mean. This represents the group average of the Vth pixel, i.e., the within-group average of the LGI1 encephalitis patient group or the healthy subject group.

[0067] This embodiment transforms multiplicative noise in an image into additive noise by performing a log transformation on the data matrix, and then effectively eliminates the noise in the image through a decentralization operation.

[0068] Furthermore, singular matrix decomposition is performed on the residual matrix IRP, decomposing it into the sum of products of multiple brain metabolic components and their coefficients. Specifically,

[0069] Singular matrix decomposition of IRP yields:

[0070] (1)

[0071] in: for The left singular matrix represents the characteristics of all column vectors. for The right singular matrix represents the features of all pixels. for The diagonal matrix has elements on the diagonal called singular values, which are arranged in descending order.

[0072] Based on the properties of singular matrix decomposition, It can be represented by IRP as equation (2):

[0073] (2)

[0074] Again Solving for the eigenvalues ​​of the square matrix yields equation (3):

[0075] (3)

[0076] By performing an equivalent transformation on equation (3), we obtain:

[0077] (4)

[0078] Substituting (2) into (4) yields:

[0079] ;

[0080] but: That is, the k-th brain metabolic component characterizing LGI1 encephalitis, denoted as... , The eigenvalues ​​of the k-th component;

[0081] Based on the properties of singular matrices ;

[0082] Then we have: ; ;

[0083] According to the definition of independent components (i.e., brain metabolic components), Corresponding right singular matrix If the column vector is , then the first... The first subject was on the The scores for each component can be expressed as:

[0084] ;

[0085] By identifying the most significant difference between LGI1 encephalitis patients and healthy subjects (i.e., the most substantial inter-group difference), the data matrix corresponding to the LGI1 encephalitis patient group in the residual matrix IRP can be represented as follows: ;

[0086] Right now: ;

[0087] in, This represents the brain metabolic components that have been broken down. Indicates the first The first subject's characterization The score for each brain metabolic component is used to indicate the degree to which the subject represents that component pattern. The higher the score, the better the subject's representation of that component pattern.

[0088] Among the M brain metabolic components, the most orthogonal component indicates that the brain metabolism of the corresponding patient and normal person has the greatest difference. Therefore, in this embodiment, the most orthogonal GIS component (brain metabolic component), that is, the component that best represents the characteristic metabolic pattern of LGI1 encephalitis, is selected and projected back into the brain image to determine the characteristic brain metabolic distribution pattern of LGI1 encephalitis.

[0089] The process of backprojecting brain images involves projecting the most orthogonal row vectors obtained back into the three-dimensional brain image space according to the positional order of the voxels when the original subject splicing matrix was constructed. The image matrix size, spatial resolution, etc., are kept consistent with the image before the splicing matrix was constructed. The image intensity value of the brain image after backprojection is the value of the most orthogonal component.

[0090] Brain metabolic distribution pattern, or the distribution pattern of brain metabolic components, refers to the distribution pattern of brain metabolic changes in LGI1 encephalitis patients compared to healthy subjects in the component characterization of brain images.

[0091] The representation result and the distribution pattern are different representations of the same feature. The representation result refers to the most orthogonal component obtained after singular matrix decomposition, which is a row vector with a length consistent with the number of pixels in the image. The distribution pattern refers to the three-dimensional visualized distribution pattern of brain metabolic changes obtained by back-projecting the representation result into the brain image according to the initial pixel position order. After back-projecting the pixels into the brain image, the regions where metabolism has changed can be located in three-dimensional space to obtain the characteristic brain metabolic distribution pattern of LGI1 encephalitis patients.

[0092] The severity of encephalitis can be assessed using the characteristic brain metabolic distribution pattern of LGI1 encephalitis. Specifically, the characterization results of the characteristic brain metabolic distribution pattern of LGI1 encephalitis using FDG-PET brain imaging data of the patients to be evaluated are obtained through the following method:

[0093] The FDG-PET brain imaging data of the patients to be evaluated were subjected to a log transformation. The individual mean value of each pixel in the log-transformed data matrix was then subtracted from the value of the individual pixel, and the column mean value within the patient group was also subtracted to obtain the decentralized brain imaging data of the patients to be evaluated. ;

[0094] Based on decentralized brain imaging data Based on the orthogonality of GIS, the characterization result is obtained using the following formula:

[0095] ;

[0096] in, The characterization results are as follows. It is the most orthogonal brain metabolic component.

[0097] This characterization result can assess the severity of LGI1 encephalitis and accurately predict its prognosis. The higher the characterization result value, the more severe the acute phase of the encephalitis in the patient.

[0098] In summary, the data-driven acute-phase assessment system for anti-LGI1 encephalitis provided by the embodiments of the present invention can systematically obtain the metabolic abnormality patterns of encephalitis patients throughout the entire brain without the need for pre-setting reference brain regions. This pattern includes not only metabolic changes in individual brain regions but also the interconnected patterns of changes in these regions. Furthermore, by using the similarity score of the metabolic pattern representation in LGI1 encephalitis patients, the severity of the disease can be assessed, and its prognosis can be accurately evaluated, providing a strong objective basis for the systematic assessment of the acute phase of LGI1 encephalitis. Moreover, this system is entirely data-driven and can establish the correlation of metabolic covariance in various brain regions during the acute phase of LGI1 encephalitis. Using the score of this correlation, the severity of the disease during the acute phase of LGI1 encephalitis can be assessed, and this assessment can accurately predict the patient's prognosis, contributing to the advancement of research on the pathogenesis and new therapeutic targets of acute-phase LGI1 encephalitis.

[0099] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware, and the program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0100] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A data-driven acute-phase assessment system for anti-LGI1 encephalitis, characterized in that, include: The data acquisition unit is used to acquire FDG-PET brain imaging data from multiple LGI1 encephalitis patients and healthy subjects. The LGI1 encephalitis characteristic brain metabolic distribution pattern construction unit is used to obtain brain metabolic components characterizing LGI1 encephalitis based on the radioactivity concentration values ​​of each pixel in the FDG-PET brain imaging data of multiple LGI1 encephalitis patients and healthy subjects. This includes: arranging the FDG-PET brain imaging data of multiple LGI1 encephalitis patients and healthy subjects sequentially according to the pixel position index; concatenating the sequentially arranged data for each patient and subject to obtain a data matrix P; centering the data matrix P to obtain a residual matrix IRP; and performing singular matrix decomposition on the residual matrix IRP to obtain the brain metabolic components characterizing LGI1 encephalitis. The most orthogonal brain metabolic components obtained after singular matrix decomposition are... This refers to the brain metabolic components that characterize the features of LGI1 encephalitis; the brain metabolic components are then projected back into the corresponding positions of each pixel in the brain imaging data to construct the brain metabolic distribution pattern characteristic of LGI1 encephalitis. The encephalitis severity assessment unit is used to obtain the characterization result of the characteristic brain metabolic distribution pattern of LGI1 encephalitis based on the FDG-PET brain imaging data of the patient to be assessed, as the assessment result of the acute phase of LGI1 encephalitis in the patient to be assessed; wherein, the characterization result of the characteristic brain metabolic distribution pattern of LGI1 encephalitis by the FDG-PET brain imaging data of the patient to be assessed is obtained by performing a log transformation on the FDG-PET brain imaging data of the patient to be assessed, and subtracting the individual mean value of each pixel in the log-transformed data matrix, and subtracting the column mean value within the patient group, to obtain the decentralized brain imaging data of the patient to be assessed. Based on decentralized brain imaging data The characterization result is obtained through the following formula: ； in, The characterization results are as follows. It is the most orthogonal brain metabolic component.

2. The data-driven acute phase assessment system for anti-LGI1 encephalitis according to claim 1, characterized in that, After obtaining the data matrix P, the process also includes performing a log transformation on the data matrix P; The residual matrix IRP is obtained by centering the data matrix after log transformation.

3. The data-driven acute phase assessment system for anti-LGI1 encephalitis according to claim 2, characterized in that, The centering includes row centering and column centering; the residual matrix IRP is obtained by the following method: Subtracting the individual average value from the value of each pixel in the data matrix after log transformation yields the row-centered data. The LGI1 encephalitis patients and healthy subjects were divided into a patient group and a healthy group, and the mean values ​​within each group were calculated. For the row-centered data matrix, the residual matrix IRP is obtained by subtracting the average value of the corresponding column in the column from the value of each pixel.

4. The data-driven acute phase assessment system for anti-LGI1 encephalitis according to claim 3, characterized in that, The centralization is represented as follows: ； In this context, the subscript I represents the I-th subject, and V represents the V-th pixel. This represents the individual mean of the i-th subject, i.e., the within-row mean. This represents the group average.

5. The data-driven acute phase assessment system for anti-LGI1 encephalitis according to claim 1, characterized in that, Singular matrix decomposition was performed on the residual matrix IRP to obtain the brain metabolic components characterizing LGI1 encephalitis, as follows: ； in, This is the data matrix corresponding to the LGI1 encephalitis patient group in the residual matrix IRP. These are the brain metabolic components that have been broken down. Indicates the first The first subject's characterization A score for each brain metabolic component.

6. The data-driven acute-phase assessment system for anti-LGI1 encephalitis according to claim 1, characterized in that, The data preprocessing unit is included before the LGI1 encephalitis characteristic brain metabolic distribution pattern construction unit, which is used to preprocess the acquired brain imaging data using the following methods: Convert all raw DICOM data into analyzable 3D image files (NII data); The DARTEL algorithm was used to standardize the NII data corresponding to the FDG-PET brain images of all subjects to the MNI space; Using half-height and width as Gaussian smoothing is performed on the spatially standardized data using a Gaussian kernel.

7. The data-driven acute phase assessment system for anti-LGI1 encephalitis according to claim 1, characterized in that, The healthy subjects were matched for age and sex with the LGI1 encephalitis patients.

Images