Processing Method and Device for Constructing a DWI Image Medical Atlas Based on Image Registration

Abstract

An embodiment of the present invention relates to a processing method and device for constructing a DWI image medical atlas based on image registration. The method includes: using a standard brain template T1 image as a template image to construct a multi-modal image library, identifying the registration matrices of the template image and each modal image in the multi-modal image library, and creating a standard image library; receiving a subject dataset, if the T1 subset exists, using the T1 subset as the current subset, if the T1 subset does not exist but the T2 subset exists, using the T2 subset as the current subset, and if only the DWI subset exists, using the DWI subset as the current subset; screening the registration matrix between the subject's DWI image and the template image based on the current subset, the standard image library, and the template image; constructing four types of medical atlas templates corresponding to the subject's DWI image based on the candidate matrix sequence and four types of medical atlas templates of the template image. Through the present invention, the segmentation accuracy and processing efficiency of the DWI image medical atlas can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and particularly relates to a processing method and device for constructing a DWI image medical atlas based on image registration. Background Art

[0002] Common magnetic resonance imaging examination images of the brain include T1-weighted images (abbreviated as T1 images), T2-weighted images (abbreviated as T2 images), and magnetic resonance diffusion weighted imaging (DWI) images (abbreviated as DWI images). The four types of typical atlases corresponding to the brain in the international medical atlas standard are brain structure atlas, brain blood supply area atlas, brain watershed atlas, and cerebral cortex atlas; most of the standard brain templates in the field of brain medicine are modeled based on the image format of T1 images; the four types of medical atlas templates corresponding to the standard brain template (brain structure atlas template, brain blood supply area atlas template, brain watershed atlas template, and cerebral cortex atlas template) are essentially four types of semantic maps marked based on the T1 images of the standard brain template. The reason for choosing T1 images to construct (segment) medical atlases is that the segmentation accuracy based on T1 images is higher than that of T2 images, and the segmentation accuracy based on T2 images is higher than that of DWI images.

[0003] Predicting the type of cerebral infarction based on DWI images is a common prediction method. The steps of this prediction method are generally as follows: first, construct (or segment) four types of medical atlases based on DWI images, then extract feature information from each type of medical atlas, and then predict the type of cerebral infarction based on all the atlas features extracted. The accuracy of this prediction method mainly depends on the segmentation accuracy of the four types of medical atlases. However, through practice, we found that the accuracy of directly segmenting the four types of medical atlases based on DWI images is not high, which makes the accuracy of the above prediction method unable to be effectively improved. Summary of the Invention

[0004] The object of the present invention is to provide a processing method, device, electronic device and computer-readable storage medium for constructing a DWI image medical atlas based on image registration in view of the defects of the prior art. The present invention first creates a standard image library with registration information through a standard brain template T1 image (T1 template image) and a multi-modal image library; then, when receiving any subject dataset, a corresponding T1 / T2 / DWI subset is selected from the current subject dataset in the order of T1 being the most optimal, T2 being the second best, and DWI being the last as the current subset, and a candidate matrix sequence of the image registration matrix between the DWI image and the T1 template image is screened based on the current subset, the standard image library and the T1 template image. Finally, the four types of medical atlas templates are registered according to the candidate matrix sequence to obtain the four types of medical atlases corresponding to the DWI image. By means of the present invention, the segmentation accuracy of the four types of medical atlases corresponding to the DWI image can be improved, thus helping to improve the prediction accuracy of cerebral infarction type prediction.

[0005] To achieve the above object, a first aspect of an embodiment of the present invention provides a processing method for constructing a DWI image medical atlas based on image registration, the method comprising:

[0006] Using a preset standard brain template T1 image as the corresponding T1 template image; constructing a multi-modal image library through data acquisition; identifying the image registration matrix of each modal image of each sample record in the T1 template image and the multi-modal image library and creating a corresponding standard image library based on the identification result; and performing quality assessment on the sample records in the standard image library and deleting the sample records with unqualified assessment.

[0007] Receiving a scanned image dataset of any subject as the corresponding subject dataset; and confirming whether there is a T1 subset in the subject dataset; if the T1 subset exists, using the T1 subset as the corresponding current subset; if the T1 subset does not exist, confirming whether there is a T2 subset in the subject dataset, if the T2 subset exists, using the T2 subset as the corresponding current subset, if the T2 subset does not exist, using the DWI subset of the subject dataset as the corresponding current subset; the subject dataset includes the DWI subset and / or the T1 subset and / or the T2 subset; the DWI subset is a mandatory subset, and the T1 and T2 subsets are optional subsets.

[0008] Screening a candidate matrix sequence of the image registration matrix between the subject DWI image in the subject dataset and the T1 template image based on the current subset, the standard image library and the T1 template image.

[0009] Construct the four types of medical atlas corresponding to the DWI image of the subject based on the candidate matrix sequence and the four types of medical atlas templates of the T1 template image to obtain the corresponding atlas set of the subject.

[0010] Preferably, the multi-modal image library includes a plurality of first sample records; the first sample records include first sample T1 images, first sample T2 images, first sample DWI images, and a first set of scanning parameters; the first set of scanning parameters is composed of T1 image scanning parameters, T2 image scanning parameters, and DWI image scanning parameters, and each type of image scanning parameters includes at least corresponding scanning device parameters, scanning software parameters, and scanning configuration parameters; the first sample T1 images, the first sample T2 images, and the first sample DWI images in each of the first sample records are pixel-aligned with each other;

[0011] The standard image library includes a plurality of second sample records; the second sample records include second sample T1 images, second sample T2 images, second sample DWI images, a second set of scanning parameters, a first T1 registration matrix, a first T2 registration matrix, and a first DWI registration matrix; each of the second sample records corresponds to one of the first sample records;

[0012] The DWI subset includes the DWI image of the subject and the corresponding DWI image scanning parameters; the T1 subset includes the T1 image of the subject and the corresponding T1 image scanning parameters; the T2 subset includes the T2 image of the subject and the corresponding T2 image scanning parameters; the DWI image of the subject, the T1 image of the subject, and the T2 image of the subject are pixel-aligned with each other;

[0013] The four types of medical atlas templates include a brain structure atlas template, a brain blood supply area atlas template, a brain watershed atlas template, and a cerebral cortex atlas template; the two-dimensional graphic sizes of the brain structure atlas template, the brain blood supply area atlas template, the brain watershed atlas template, and the cerebral cortex atlas template are all consistent with the two-dimensional graphic size of the T1 template image;

[0014] The brain structure atlas template includes a plurality of first template pixel points; each of the first template pixel points corresponds to a first template semantic type; the first template semantic types include a background point type and multiple types of brain structure foreground point types; each type of brain structure foreground point type matches a type of brain anatomical structure;

[0015] The brain blood supply area atlas template includes a plurality of second template pixel points; each of the second template pixel points corresponds to a second template semantic type; the second template semantic types include the background point type and multiple types of blood supply area foreground point types; each type of blood supply area foreground point type matches the blood supply area of a type of brain blood vessel;

[0016] The brain watershed atlas template includes a plurality of third template pixel points; each of the third template pixel points corresponds to a third template semantic type; the third template semantic types include the background point type and multiple types of watershed foreground point types; each type of the watershed foreground point types matches a type of brain blood vessel watershed region;

[0017] The cerebral cortex atlas template includes a plurality of fourth template pixel points; each of the fourth template pixel points corresponds to a fourth template semantic type; the fourth template semantic types include the background point type and multiple types of cerebral cortex foreground point types; each type of the cerebral cortex foreground point types matches a type of cerebral cortex region;

[0018] The subject atlas set includes a subject brain structure atlas, a subject brain blood supply area atlas, a subject brain watershed atlas, and a subject cerebral cortex atlas; the two-dimensional graphic sizes of the subject brain structure atlas, the subject brain blood supply area atlas, the subject brain watershed atlas, and the subject cerebral cortex atlas are all consistent with the two-dimensional graphic size of the subject's DWI image;

[0019] The subject brain structure atlas includes a plurality of first atlas pixel points; each of the first atlas pixel points corresponds to a first atlas semantic type; the first atlas semantic types include the background point type and multiple types of the brain structure foreground point types;

[0020] The subject brain blood supply area atlas includes a plurality of second atlas pixel points; each of the second atlas pixel points corresponds to a second atlas semantic type; the second atlas semantic types include the background point type and multiple types of the blood supply area foreground point types;

[0021] The subject brain watershed atlas includes a plurality of third atlas pixel points; each of the third atlas pixel points corresponds to a third atlas semantic type; the third atlas semantic types include the background point type and multiple types of the watershed foreground point types;

[0022] The subject cerebral cortex atlas includes a plurality of fourth atlas pixel points; each of the fourth atlas pixel points corresponds to a fourth atlas semantic type; the fourth atlas semantic types include the background point type and multiple types of the cerebral cortex foreground point types.

[0023] Preferably, the construction of the multi-modal image library through data acquisition specifically includes:

[0024] Recruit a plurality of volunteers to form a corresponding first volunteer set; the first volunteer set includes a plurality of first volunteers; the first volunteer is a volunteer who has been medically examined and confirmed that no cerebral infarction lesions have occurred in their brain;

[0025] Traverse all the first volunteers; during the traversal, regard the currently traversed first volunteer as the corresponding current volunteer; perform a magnetic resonance imaging scan on the current volunteer and generate a set of T1, T2, and DWI images with the same scanning pose, synchronous scanning time, and aligned scanning imaging pixels from the results of this scan as the corresponding first sample T1 image, first sample T2 image, and first sample DWI image; compose a corresponding first scanning parameter set from the T1 image scanning parameters, T2 image scanning parameters, and DWI image scanning parameters of this scan; and compose a corresponding first sample record from the obtained first sample T1 image, first sample T2 image, first sample DWI image, and first scanning parameter set.

[0026] At the end of the traversal, compose the corresponding multimodal image library from all the obtained first sample records.

[0027] Preferably, the step of identifying the image registration matrices of the T1 template image and the modal images of each sample record in the multimodal image library and creating a corresponding standard image library based on the identification results specifically includes:

[0028] Regard each of the first sample records in the multimodal image library as the corresponding current sample record.

[0029] Regard the T1 template image as the corresponding current floating image, regard the first sample T1 image of the current sample record as the corresponding current fixed image, and use a preset medical image registration analysis tool to analyze the registration matrix from the current floating image to the current fixed image according to the flexible registration analysis rules and regard the analyzed registration matrix as the corresponding first T1 registration matrix; the medical image registration analysis tool includes at least ANTS software and NiftyReg software.

[0030] Perform two matrix replications on the first T1 registration matrix to obtain the corresponding first T2 registration matrix and first DWI registration matrix; the first T2 registration matrix and the first DWI registration matrix are the same as the first T1 registration matrix.

[0031] And use the first sample T1 image, the first sample T2 image, the first sample DWI image, and the first set of scanning parameters recorded by the current sample as the corresponding second sample T1 image, second sample T2 image, second sample DWI image, and second set of scanning parameters; and use the corresponding second sample T1 image, second sample T2 image, second sample DWI image, second set of scanning parameters, the first T1 registration matrix, the first T2 registration matrix, and the first DWI registration matrix recorded by the current sample to form a corresponding second sample record;

[0032] And use the second sample records corresponding to all the first sample records to form the corresponding standard image library.

[0033] Preferably, performing quality assessment on the sample records in the standard image library and deleting the sample records with unqualified assessment specifically includes:

[0034] Use each of the second sample records in the standard image library as the corresponding current sample record; randomly select the first specified number M of brain experts from a preset expert team to form the corresponding current assessment team; send the current sample record and the T1 template image to the current assessment team for image registration quality evaluation to obtain the corresponding M expert scores; calculate the average value of the obtained M expert scores to obtain the corresponding first average value; and confirm whether the first average value does not exceed a preset score threshold. If it is confirmed that it does not exceed, then delete the current sample record as the corresponding unqualified assessment record from the standard image library; the expert team consists of multiple brain experts from one or more medical institutions in one or more regions; the first specified number M is a positive integer.

[0035] Preferably, based on the current subset, the standard image library, and the T1 template image, screening candidate matrices for the image registration matrix between the subject DWI image and the T1 template image in the subject dataset to obtain the corresponding candidate matrix sequence, specifically including:

[0036] Identify the current subset; if the current subset is a T1 subset, extract the subject T1 image and the T1 image scanning parameters of the current subset as the corresponding current image and current scanning parameters; if the current subset is a T2 subset, extract the subject T2 image and the T2 image scanning parameters of the current subset as the corresponding current image and current scanning parameters; if the current subset is a DWI subset, extract the subject DWI image and the DWI image scanning parameters of the current subset as the corresponding current image and current scanning parameters;

[0037] And use the current image as the corresponding current image to be recognized; and perform brain imaging omics feature recognition processing on the current image to be recognized to obtain the corresponding image brain features, and use the image brain features obtained this time as the corresponding current brain features;

[0038] And traverse all the second sample records in the standard image library; and during the traversal, use the currently traversed second sample record as the corresponding current sample record; and use the T1, T2, or DWI image scanning parameters in the second scan parameter set of the current sample record corresponding to the current subset as the corresponding current comparison scanning parameters; and use the second sample T1 image, the second sample T2 image, or the second sample DWI image in the current sample record corresponding to the current subset as the corresponding current comparison image; and identify whether the current comparison scanning parameters match the current scanning parameters; if they do not match, directly go to the next second sample record to continue the traversal; if they match, use the current comparison image as the new current image to be recognized, and perform brain imaging omics feature recognition processing on the new current image to be recognized to obtain the new image brain features, and use the image brain features obtained this time as the corresponding current comparison brain features, and respectively perform characteristic vector conversion on the current brain features and the current comparison brain features to obtain the corresponding first and second feature vectors, and calculate the vector similarity of the first and second feature vectors based on the vector similarity algorithm to obtain the corresponding first similarity, and identify whether the first similarity exceeds the preset similarity threshold. If the first similarity exceeds the similarity threshold, mark the current sample record as a candidate sample record and go to the next second sample record to continue the traversal. If the first similarity does not exceed the similarity threshold, directly go to the next second sample record to continue the traversal;

[0039] And at the end of the traversal, count the total number of candidate sample records to obtain the corresponding first statistical total; and use the current image as the corresponding current fixed image;

[0040] And identify the first statistical total;

[0041] If the total number of the first statistics is greater than or equal to a preset first odd number N, record the N candidate sample records with the highest first similarity as the corresponding final selected sample records; and use the first T1 registration matrix, the first T2 registration matrix, or the first DWI registration matrix corresponding to the current subset in each of the final selected sample records as a corresponding one-step registration matrix; and use the second sample T1 image, the second sample T2 image, or the second sample DWI image corresponding to the current subset in each of the final selected sample records as the corresponding current floating image, and analyze the registration matrix from the current floating image to the current fixed image according to the flexible registration analysis rule and use the analyzed registration matrix as the corresponding two-step registration matrix; and use the matrix obtained by multiplying the one-step and two-step registration matrices corresponding to each of the final selected sample records as a corresponding candidate matrix; and form a corresponding candidate matrix sequence with the N candidate matrices obtained; the first odd number N is an odd number greater than 1;

[0042] If the total number of the first statistics is equal to N - 1, record the N - 1 candidate sample records with the highest first similarity as the corresponding final selected sample records; and use the first T1 registration matrix, the first T2 registration matrix, or the first DWI registration matrix corresponding to the current subset in each of the final selected sample records as a corresponding one-step registration matrix; and use the second sample T1 image, the second sample T2 image, or the second sample DWI image corresponding to the current subset in each of the final selected sample records as the corresponding current floating image, and analyze the registration matrix from the current floating image to the current fixed image according to the flexible registration analysis rule and use the analyzed registration matrix as the corresponding two-step registration matrix; and use the matrix obtained by multiplying the one-step and two-step registration matrices corresponding to each of the final selected sample records as a corresponding candidate matrix; and use the T1 template image as the corresponding current floating image, and analyze the registration matrix from the current floating image to the current fixed image according to the flexible registration analysis rule and use the analyzed registration matrix as a corresponding candidate matrix; and form a corresponding candidate matrix sequence with the N candidate matrices obtained;

[0043] If the total number of the first statistics is less than N - 1, use the T1 template image as the corresponding current floating image, and analyze the registration matrix from the current floating image to the current fixed image according to the flexible registration analysis rule and use the analyzed registration matrix as a corresponding candidate matrix to form a corresponding candidate matrix sequence.

[0044] Furthermore, the brain imaging feature recognition process specifically includes:

[0045] According to the foreground / background division rule where the pixel points in the ventricular region are foreground pixel points and the pixel points in the brain tissue region are background pixel points, perform binary image conversion on the current image to be recognized in this processing based on the Otsu algorithm to obtain the corresponding first binary image; and extract the first-order features and morphological features of the radiomics features of the ventricular region image on the first binary image to obtain the corresponding ventricular radiomics features; each pixel point of the first binary image corresponds to a first binary semantics, and the first binary semantics includes ventricular foreground points and brain tissue background points;

[0046] And according to the foreground / background division rule where the pixel points in the brain tissue region are foreground pixel points and the pixel points in the ventricular region are background pixel points, perform binary image conversion on the current image to be recognized in this processing based on the Otsu algorithm to obtain the corresponding second binary image; and extract the first-order features and morphological features of the radiomics features of the brain tissue region image on the second binary image to obtain the corresponding brain tissue radiomics features; each pixel point of the second binary image corresponds to a second binary semantics, and the second binary semantics includes brain tissue foreground points and ventricular background points;

[0047] And output the image brain features composed of the ventricular and brain tissue radiomics features obtained this time as the processing result of this time.

[0048] Preferably, the construction of the four types of medical atlas corresponding to the DWI image of the subject based on the four types of medical atlas templates of the candidate matrix sequence and the T1 template image specifically includes:

[0049] Identify the total number of candidate matrices in the candidate matrix sequence;

[0050] If the total number of candidate matrices is 1, then use the only candidate matrix in the candidate matrix sequence as the corresponding current registration matrix; and register the four types of templates of the four types of medical atlas templates based on the current registration matrix to obtain a corresponding set of four types of registered atlases; and use the first registered atlas, the second registered atlas, the third registered atlas, and the fourth registered atlas of the four types of registered atlases as the corresponding subject's brain blood supply area atlas, the subject's brain watershed atlas, and the subject's cerebral cortex atlas to form the corresponding subject atlas set; the four types of registered atlases include the first, second, third, and fourth registered atlases; the first registered atlas includes a plurality of first registered atlas pixel points, each of the first registered atlas pixel points corresponds to a first registered atlas semantic type, and the first registered atlas semantic type includes the background point type and multiple types of brain structure foreground point types; the second registered atlas includes a plurality of second registered atlas pixel points, each of the second registered atlas pixel points corresponds to a second registered atlas semantic type, and the second registered atlas semantic type includes the background point type and multiple types of blood supply area foreground point types; the third registered atlas includes a plurality of third registered atlas pixel points, each of the third registered atlas pixel points corresponds to a third registered atlas semantic type, and the third registered atlas semantic type includes the background point type and multiple types of watershed foreground point types; the fourth registered atlas includes a plurality of fourth registered atlas pixel points, each of the fourth registered atlas pixel points corresponds to a fourth registered atlas semantic type, and the fourth registered atlas semantic type includes the background point type and multiple types of cerebral cortex foreground point types;

[0051] If the total number of candidate matrices is equal to the preset first odd number N, then use each candidate matrix in the candidate matrix sequence as the corresponding current registration matrix one by one; and register the four types of templates of the four types of medical atlas templates based on the current registration matrix to obtain a corresponding set of the four types of registered atlases; and construct the subject atlas corresponding to the obtained N sets of the four types of registered atlases based on the voting mechanism.

[0052] Further, the registration of the four types of templates of the four types of medical atlas templates based on the current registration matrix specifically includes:

[0053] Take the brain structure atlas template, the brain blood supply area atlas template, the brain watershed atlas template, and the cerebral cortex atlas template in the four types of medical atlas templates one by one as the corresponding current floating images; register the current floating images with the current registration matrix to obtain the corresponding current fixed images; and use the current fixed image corresponding to the brain structure atlas template as the corresponding first registered atlas, the current fixed image corresponding to the brain blood supply area atlas template as the corresponding second registered atlas, the current fixed image corresponding to the brain watershed atlas template as the corresponding third registered atlas, and the current fixed image corresponding to the cerebral cortex atlas template as the corresponding fourth registered atlas; and form a group of corresponding four types of registered atlases from the first, second, third, and fourth registered atlases obtained this time, and output the four types of registered atlases obtained this time as the registration result of this time.

[0054] Further, constructing the subject atlas corresponding to the obtained N groups of the four types of registered atlases based on the voting mechanism specifically includes:

[0055] Initialize a subject brain structure atlas, a subject brain blood supply area atlas, a subject brain watershed atlas, and a subject cerebral cortex atlas; the first atlas semantic types of all the first atlas pixel points of the initialized subject brain structure atlas are all the background point types; the second atlas semantic types of all the second atlas pixel points of the initialized subject brain blood supply area atlas are all the background point types; the third atlas semantic types of all the third atlas pixel points of the initialized subject brain watershed atlas are all the background point types; the fourth atlas semantic types of all the fourth atlas pixel points of the initialized subject cerebral cortex atlas are all the background point types;

[0056] And form a corresponding first registered atlas set from the N first registered atlases in the N groups of the four types of registered atlases, a corresponding second registered atlas set from the N second registered atlases, a corresponding third registered atlas set from the N third registered atlases, and a corresponding fourth registered atlas set from the N fourth registered atlases;

[0057] And round up the product of the preset first voting ratio and the first odd number N to obtain the corresponding first vote threshold; the first voting ratio is a ratio value ranging from 0 to 1.

[0058] Traverse all the first atlas pixels of the brain structure atlas of the subject; during the traversal, regard the currently traversed first atlas pixel as the corresponding current pixel; regard the first registered atlas pixels on each of the first registered atlases in the first registered atlas set that match the current pixel as the corresponding first matching points; form a corresponding first type ballot box from the first registered atlas semantic types of the N first matching points; count the number of each first registered atlas semantic type in the first type ballot box and use the statistical result as the corresponding first type vote count; regard the first type vote count with the largest value as the corresponding first maximum vote count; when the first maximum vote count is greater than or equal to the first vote threshold, reset the first atlas semantic type of the current pixel to the first registered atlas semantic type corresponding to the first maximum vote count;

[0059] Traverse all the second atlas pixels of the brain blood supply area atlas of the subject; during the traversal, regard the currently traversed second atlas pixel as the corresponding current pixel; regard the second registered atlas pixels on each of the second registered atlases in the second registered atlas set that match the current pixel as the corresponding second matching points; form a corresponding second type ballot box from the second registered atlas semantic types of the N second matching points; count the number of each second registered atlas semantic type in the second type ballot box and use the statistical result as the corresponding second type vote count; regard the second type vote count with the largest value as the corresponding second maximum vote count; when the second maximum vote count is greater than or equal to the first vote threshold, reset the second atlas semantic type of the current pixel to the second registered atlas semantic type corresponding to the second maximum vote count;

[0060] Traverse all the third atlas pixels of the brain watershed atlas of the subject; during the traversal, regard the currently traversed third atlas pixel as the corresponding current pixel; regard the third registered atlas pixels on each of the third registered atlases in the third registered atlas set that match the current pixel as the corresponding third matching points; form a corresponding third type ballot box from the third registered atlas semantic types of the N third matching points; count the number of each third registered atlas semantic type in the third type ballot box and use the statistical result as the corresponding third type vote count; regard the third type vote count with the largest value as the corresponding third maximum vote count; when the third maximum vote count is greater than or equal to the first vote threshold, reset the third atlas semantic type of the current pixel to the third registered atlas semantic type corresponding to the third maximum vote count;

[0061] Traverse all the fourth atlas pixels of the brain cortex atlas of the subject; during the traversal, use the currently traversed fourth atlas pixel as the corresponding current pixel; use the fourth registration atlas pixels on each of the fourth registration atlases in the fourth registration atlas set that match the current pixel as the corresponding fourth matching points; form a corresponding fourth type ballot box from the fourth registration atlas semantic types of the N fourth matching points; count the number of each fourth registration atlas semantic type in the fourth type ballot box and use the statistical result as the corresponding fourth type vote count; use the fourth type vote count with the largest value as the corresponding fourth maximum vote count; when the fourth maximum vote count is greater than or equal to the first vote threshold, reset the fourth atlas semantic type of the current pixel to the fourth registration atlas semantic type corresponding to the fourth maximum vote count;

[0062] Form a corresponding subject atlas set from the brain blood supply area atlas, the brain watershed atlas, and the brain cortex atlas of the subject after the reset is completed.

[0063] A second aspect of the embodiments of the present invention provides an apparatus for implementing the processing method of constructing a DWI image medical atlas based on image registration described in the first aspect above. The apparatus includes: a preprocessing module, a data receiving module, a candidate matrix screening module, and a medical atlas construction module;

[0064] The preprocessing module is used to use a preset standard brain template T1 image as the corresponding T1 template image; construct a multi-modal image library through data acquisition; identify the image registration matrices of the T1 template image and each modal image of each sample record in the multi-modal image library and create a corresponding standard image library in combination with the identification results; perform quality assessment on the sample records in the standard image library and delete the sample records with unqualified evaluations;

[0065] The data receiving module is used to receive the scan image dataset of any subject as the corresponding subject dataset; confirm whether there is a T1 subset in the subject dataset; if the T1 subset exists, use the T1 subset as the corresponding current subset; if the T1 subset does not exist, confirm whether there is a T2 subset in the subject dataset. If the T2 subset exists, use the T2 subset as the corresponding current subset. If the T2 subset does not exist, use the DWI subset of the subject dataset as the corresponding current subset; the subject dataset includes the DWI subset and / or the T1 subset and / or the T2 subset; the DWI subset is a required subset, and the T1 and T2 subsets are optional subsets;

[0066] The candidate matrix screening module is used to screen candidate matrices for the image registration matrix between the subject's DWI image and the T1 template image in the subject dataset based on the current subset, the standard image library, and the T1 template image, so as to obtain a corresponding candidate matrix sequence;

[0067] The medical atlas construction module is used to construct four types of medical atlases corresponding to the subject's DWI image based on the candidate matrix sequence and four types of medical atlas templates of the T1 template image, so as to obtain a corresponding subject atlas set.

[0068] A third aspect of the embodiments of the present invention provides an electronic device, including: a memory, a processor, and a transceiver;

[0069] The processor is used to be coupled with the memory, read and execute the instructions in the memory to implement the method steps described in the first aspect above;

[0070] The transceiver is coupled with the processor, and the processor controls the transceiver to perform message sending and receiving.

[0071] A fourth aspect of the embodiments of the present invention provides a computer-readable storage medium storing computer instructions, and when the computer instructions are executed by a computer, the computer is caused to execute the instructions of the method described in the first aspect above.

[0072] The embodiments of the present invention provide a processing method, device, electronic device, and computer-readable storage medium for constructing a DWI image medical atlas based on image registration. As can be seen from the above content, the embodiments of the present invention pre-create a standard image library with registration information through a standard brain template T1 image (T1 template image) and a multi-modal image library; then, when receiving any subject dataset, a corresponding T1 / T2 / DWI subset is selected from the current subject dataset in the order of T1 being the best, T2 being the second best, and DWI being the last as the current subset, and candidate matrix screening is performed on the image registration matrix between the DWI image and the T1 template image based on the current subset, the standard image library, and the T1 template image. Finally, the four types of medical atlases corresponding to the DWI image are obtained by registering the four types of medical atlas templates according to the candidate matrix sequence. Through the embodiments of the present invention, the segmentation accuracy of the four types of medical atlases corresponding to the DWI image can be improved, which helps to improve the prediction accuracy of cerebral infarction type prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] Figure 1 It is a schematic diagram of a processing method for constructing a DWI image medical atlas based on image registration provided by Embodiment 1 of the present invention;

[0074] Figure 2It is a module structure diagram of a processing device for constructing a DWI image medical atlas based on image registration provided in the second embodiment of the present invention;

[0075] Figure 3 It is a schematic structural diagram of an electronic device provided in the third embodiment of the present invention. Specific embodiments

[0076] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0077] The first embodiment of the present invention provides a processing method for constructing a DWI image medical atlas based on image registration. As Figure 1 shown in the schematic diagram of the processing method for constructing a DWI image medical atlas based on image registration provided in the first embodiment of the present invention, the method mainly includes the following steps:

[0078] Step 1: Use the preset standard brain template T1 image as the corresponding T1 template image; construct a multi-modal image library through data collection; identify the image registration matrices of the T1 template image and each modal image of each sample record in the multi-modal image library and create a corresponding standard image library based on the identification results; perform quality assessment on the sample records in the standard image library and delete the sample records with unqualified assessment;

[0079] Specifically including: Step 11: Use the preset standard brain template T1 image as the corresponding T1 template image;

[0080] Step 12: Construct a multi-modal image library through data collection;

[0081] Among them, the multi-modal image library includes multiple first sample records; the first sample record includes a first sample T1 image, a first sample T2 image, a first sample DWI image and a first set of scanning parameters; the first set of scanning parameters is composed of T1 image scanning parameters, T2 image scanning parameters and DWI image scanning parameters, and each type of image scanning parameter includes at least corresponding scanning device parameters, scanning software parameters and scanning configuration parameters; the first sample T1 image, the first sample T2 image and the first sample DWI image in each first sample record are pixel-aligned with each other;

[0082] Specifically including: Step 121: Recruit multiple volunteers to form a corresponding first volunteer set;

[0083] Among them, the first volunteer set includes multiple first volunteers; the first volunteer is a volunteer who has been confirmed through a medical examination that there are no cerebral infarction lesions in their brain;

[0084] Step 122, traverse all the first volunteers; and during the traversal, regard the currently traversed first volunteer as the corresponding current volunteer; perform a magnetic resonance imaging scan on the current volunteer and generate a set of T1, T2, and DWI images with the same scan pose, synchronized scan time, and aligned scan imaging pixels from the results of this scan as the corresponding first sample T1 image, first sample T2 image, and first sample DWI image; and compose a corresponding first scan parameter set from the T1 image scan parameters, T2 image scan parameters, and DWI image scan parameters corresponding to this scan; and compose a corresponding first sample record from the obtained first sample T1 image, first sample T2 image, first sample DWI image, and first scan parameter set;

[0085] Step 123, and at the end of the traversal, compose the corresponding multi-modal image library from all the obtained first sample records;

[0086] Step 13, identify the image registration matrices of each modal image of the T1 template image and each sample record in the multi-modal image library and create the corresponding standard image library in combination with the identification results;

[0087] Among them, the standard image library includes multiple second sample records; the second sample record includes a second sample T1 image, a second sample T2 image, a second sample DWI image, a second scan parameter set, a first T1 registration matrix, a first T2 registration matrix, and a first DWI registration matrix; each second sample record corresponds to a first sample record;

[0088] Specifically include: Step 131, regard each first sample record in the multi-modal image library as the corresponding current sample record;

[0089] Step 132, regard the T1 template image as the corresponding current floating image, regard the first sample T1 image of the current sample record as the corresponding current fixed image, and use a preset medical image registration analysis tool to analyze the registration matrix from the current floating image to the current fixed image according to the flexible registration analysis rule and regard the analyzed registration matrix as the corresponding first T1 registration matrix;

[0090] Among them, the medical image registration analysis tool at least includes ANTS software and NiftyReg software;

[0091] Step 133, perform two matrix copies on the first T1 registration matrix to obtain the corresponding first T2 registration matrix and first DWI registration matrix;

[0092] Here, the first T2 registration matrix and the first DWI registration matrix are consistent with the first T1 registration matrix;

[0093] Step 134, and use the first sample T1 image, the first sample T2 image, the first sample DWI image, and the first set of scanning parameters recorded by the current sample as the corresponding second sample T1 image, second sample T2 image, second sample DWI image, and second set of scanning parameters; and form a corresponding second sample record from the corresponding second sample T1 image, second sample T2 image, second sample DWI image, second set of scanning parameters, the first T1 registration matrix, the first T2 registration matrix, and the first DWI registration matrix recorded by the current sample;

[0094] Step 135, and form a corresponding standard image library from the second sample records corresponding to all the first sample records;

[0095] Step 14, and perform quality assessment on the sample records in the standard image library and delete the sample records with unqualified assessment;

[0096] Specifically, it includes: taking each second sample record in the standard image library as the corresponding current sample record; randomly selecting the first specified number M of brain experts from a preset expert team to form the corresponding current assessment team; sending the current sample record and the T1 template image to the current assessment team for image registration quality evaluation to obtain the corresponding M expert scores; calculating the average of the obtained M expert scores to obtain the corresponding first average score; and confirming whether the first average score does not exceed the preset score threshold. If it is confirmed that it does not exceed, then delete the current sample record as the corresponding unqualified assessment record from the standard image library;

[0097] Among them, the expert team is composed of multiple brain experts from one or more medical institutions in one or more regions; the first specified number M is a positive integer; the score threshold is a preset score value.

[0098] Step 2, receive the scanning image dataset of any subject as the corresponding subject dataset; and confirm whether there is a T1 subset in the subject dataset; if there is a T1 subset, then use the T1 subset as the corresponding current subset; if there is no T1 subset, then confirm whether there is a T2 subset in the subject dataset. If there is a T2 subset, then use the T2 subset as the corresponding current subset. If there is no T2 subset, then use the DWI subset of the subject dataset as the corresponding current subset.

[0099] Here, the subject dataset includes a DWI subset and / or a T1 subset and / or a T2 subset; the DWI subset is a mandatory subset, and the T1 and T2 subsets are optional subsets; the DWI subset includes the subject's DWI images and the corresponding DWI image scanning parameters; the T1 subset includes the subject's T1 images and the corresponding T1 image scanning parameters; the T2 subset includes the subject's T2 images and the corresponding T2 image scanning parameters; the subject's DWI images, the subject's T1 images, and the subject's T2 images are pixel-aligned with each other.

[0100] Step 3: Based on the current subset, the standard image library, and the T1 template image, screen the candidate matrix for the image registration matrix between the subject's DWI image and the T1 template image in the subject dataset to obtain the corresponding candidate matrix sequence.

[0101] Specifically, it includes: Step 31: Identify the current subset; if the current subset is a T1 subset, extract the subject's T1 image and the T1 image scanning parameters of the current subset as the corresponding current image and current scanning parameters; if the current subset is a T2 subset, extract the subject's T2 image and the T2 image scanning parameters of the current subset as the corresponding current image and current scanning parameters; if the current subset is a DWI subset, extract the subject's DWI image and the DWI image scanning parameters of the current subset as the corresponding current image and current scanning parameters.

[0102] Step 32: And use the current image as the corresponding current image to be recognized; perform brain imageomics feature recognition processing on the current image to be recognized to obtain the corresponding image brain features, and use the image brain features obtained this time as the corresponding current brain features.

[0103] Here, the processing steps for performing brain imageomics feature recognition processing on the current image to be recognized in the current Step 32 specifically include:

[0104] Step A1: According to the foreground / background division rule with the pixel points in the ventricle area as foreground pixel points and the pixel points in the brain tissue area as background pixel points, perform binary image conversion on the current image to be recognized in this processing based on the Otsu algorithm to obtain the corresponding first binary image; and extract the first-order features and morphological features of the imageomics features in the ventricle area on the first binary image to obtain the corresponding ventricle imageomics features.

[0105] Among them, each pixel point of the first binary image corresponds to a first binary semantics, and the first binary semantics includes ventricle foreground points and brain tissue background points.

[0106] Step A2, and according to the foreground / background division rule with the pixel points of the brain tissue region as foreground pixel points and the pixel points of the ventricle region as background pixel points, perform binary image conversion on the current image to be recognized in this processing based on the Otsu algorithm to obtain the corresponding second binary image; and extract the first-order features and morphological features of the brain tissue region imaging omics features on the second binary image to obtain the corresponding brain tissue imaging omics features;

[0107] Among them, each pixel point of the second binary image corresponds to a second binary semantics, and the second binary semantics includes brain tissue foreground points and ventricle background points;

[0108] Step A3, and output the imaging brain features composed of the ventricle and brain tissue imaging omics features obtained this time as the processing result of this time;

[0109] Step 33, and traverse all the second sample records in the standard image library; and during the traversal, regard the currently traversed second sample record as the corresponding current sample record; and regard the T1, T2 or DWI image scanning parameters in the second scanning parameter set of the current sample record corresponding to the current subset as the corresponding current comparison scanning parameters; and regard the second sample T1 image, second sample T2 image or second sample DWI image in the current sample record corresponding to the current subset as the corresponding current comparison image; and identify whether the current comparison scanning parameters match the current scanning parameters; if they do not match, directly go to the next second sample record to continue the traversal; if they match, regard the current comparison image as the new current image to be recognized, and perform brain imaging omics feature recognition processing on the new current image to be recognized to obtain new imaging brain features, and regard the imaging brain features obtained this time as the corresponding current comparison brain features, and perform feature vector conversion on the current brain features and the current comparison brain features respectively to obtain the corresponding first and second feature vectors, and calculate the vector similarity of the first and second feature vectors based on the vector similarity algorithm to obtain the corresponding first similarity, and identify whether the first similarity exceeds the preset similarity threshold. If the first similarity exceeds the similarity threshold, mark the current sample record as a candidate sample record and go to the next second sample record to continue the traversal. If the first similarity does not exceed the similarity threshold, directly go to the next second sample record to continue the traversal;

[0110] Here, the similarity threshold is a preset threshold parameter;

[0111] The processing steps for performing brain imaging omics feature recognition processing on the new current image to be recognized in the current step 33 are the same as those for brain imaging omics feature recognition processing in step 32;

[0112] Step 34, and at the end of the traversal, count the total number of records in the candidate sample records to obtain the corresponding first total count; and use the current image as the corresponding current fixed image;

[0113] Step 35, and identify the first total count;

[0114] Step 36, if the first total count is greater than or equal to the preset first odd number N, then record the first N candidate sample records with the top first similarity as the corresponding final selected sample records; and use the first T1 registration matrix, the first T2 registration matrix, or the first DWI registration matrix corresponding to the current subset in each final selected sample record as a corresponding one-step registration matrix; and use the second sample T1 image, the second sample T2 image, or the second sample DWI image corresponding to the current subset in each final selected sample record as the corresponding current floating image, and analyze the registration matrix from the current floating image to the current fixed image according to the flexible registration analysis rule and use the analyzed registration matrix as the corresponding two-step registration matrix; and use the matrix obtained by multiplying the one-step and two-step registration matrices corresponding to each final selected sample record as a corresponding candidate matrix; and form a corresponding candidate matrix sequence from the obtained N candidate matrices;

[0115] Wherein, the first odd number N is an odd number greater than 1;

[0116] Step 37, if the first total count is equal to N - 1, then record the first N - 1 candidate sample records with the top first similarity as the corresponding final selected sample records; and use the first T1 registration matrix, the first T2 registration matrix, or the first DWI registration matrix corresponding to the current subset in each final selected sample record as a corresponding one-step registration matrix; and use the second sample T1 image, the second sample T2 image, or the second sample DWI image corresponding to the current subset in each final selected sample record as the corresponding current floating image, and analyze the registration matrix from the current floating image to the current fixed image according to the flexible registration analysis rule and use the analyzed registration matrix as the corresponding two-step registration matrix; and use the matrix obtained by multiplying the one-step and two-step registration matrices corresponding to each final selected sample record as a corresponding candidate matrix; and use the T1 template image as the corresponding current floating image, and analyze the registration matrix from the current floating image to the current fixed image according to the flexible registration analysis rule and use the analyzed registration matrix as a corresponding candidate matrix; and form a corresponding candidate matrix sequence from the obtained N candidate matrices;

[0117] Step 38, if the first total count is less than N - 1, then use the T1 template image as the corresponding current floating image, and analyze the registration matrix from the current floating image to the current fixed image according to the flexible registration analysis rule and use the analyzed registration matrix as a corresponding candidate matrix to form a corresponding candidate matrix sequence.

[0118] Step 4: Based on the candidate matrix sequence and the four types of medical atlas templates of the T1 template image, construct the four types of medical atlases corresponding to the DWI image of the subject to obtain the corresponding subject atlas set;

[0119] Specifically, it includes: Step 41: Identify the total number of candidate matrices in the candidate matrix sequence;

[0120] Step 42: If the total number of candidate matrices is 1, use the only candidate matrix in the candidate matrix sequence as the corresponding current registration matrix; and register the four types of templates of the four types of medical atlas templates based on the current registration matrix to obtain a set of corresponding four types of registered atlases; and use the first registered atlas, the second registered atlas, the third registered atlas, and the fourth registered atlas of the four types of registered atlases as the corresponding subject brain blood supply area atlas, the subject brain watershed atlas, and the subject cerebral cortex atlas to form the corresponding subject atlas set;

[0121] Specifically, it includes: Step 421: If the total number of candidate matrices is 1, use the only candidate matrix in the candidate matrix sequence as the corresponding current registration matrix;

[0122] Step 422: And register the four types of templates of the four types of medical atlas templates based on the current registration matrix to obtain a set of corresponding four types of registered atlases;

[0123] Among them, the four types of medical atlas templates include a brain structure atlas template, a brain blood supply area atlas template, a brain watershed atlas template, and a cerebral cortex atlas template; the two-dimensional graphic sizes of the brain structure atlas template, the brain blood supply area atlas template, the brain watershed atlas template, and the cerebral cortex atlas template are all consistent with the two-dimensional graphic size of the T1 template image;

[0124] The brain structure atlas template includes multiple first template pixel points; each first template pixel point corresponds to a first template semantic type; the first template semantic type includes a background point type and multiple types of brain structure foreground point types; each type of brain structure foreground point type matches a type of brain anatomical structure;

[0125] The brain blood supply area atlas template includes multiple second template pixel points; each second template pixel point corresponds to a second template semantic type; the second template semantic type includes a background point type and multiple types of blood supply area foreground point types; each type of blood supply area foreground point type matches the blood supply area of a type of brain blood vessel;

[0126] The brain watershed atlas template includes multiple third template pixel points; each third template pixel point corresponds to a third template semantic type; the third template semantic type includes a background point type and multiple types of watershed foreground point types; each type of watershed foreground point type matches a type of brain blood vessel watershed area;

[0127] The cerebral cortex atlas template includes multiple fourth template pixel points; each fourth template pixel point corresponds to a fourth template semantic type; the fourth template semantic types include background point types and multiple types of cerebral cortex foreground point types; each type of cerebral cortex foreground point type matches a type of cerebral cortex region;

[0128] The four types of registration atlases include the first, second, third, and fourth registration atlases;

[0129] The first registration atlas includes multiple first registration atlas pixel points, each first registration atlas pixel point corresponds to a first registration atlas semantic type, and the first registration atlas semantic types include background point types and multiple types of brain structure foreground point types;

[0130] The second registration atlas includes multiple second registration atlas pixel points, each second registration atlas pixel point corresponds to a second registration atlas semantic type, and the second registration atlas semantic types include background point types and multiple types of blood supply area foreground point types;

[0131] The third registration atlas includes multiple third registration atlas pixel points, each third registration atlas pixel point corresponds to a third registration atlas semantic type, and the third registration atlas semantic types include background point types and multiple types of watershed foreground point types;

[0132] The fourth registration atlas includes multiple fourth registration atlas pixel points, each fourth registration atlas pixel point corresponds to a fourth registration atlas semantic type, and the fourth registration atlas semantic types include background point types and multiple types of cerebral cortex foreground point types;

[0133] The current step 422 specifically includes:

[0134] Specifically includes: taking the brain structure atlas template, the brain blood supply area atlas template, the brain watershed atlas template, and the cerebral cortex atlas template in the four types of medical atlas templates as the corresponding current floating maps one by one; and registering the current floating maps with the current registration matrix to obtain the corresponding current fixed maps; and taking the current fixed map corresponding to the brain structure atlas template as the corresponding first registration atlas, the current fixed map corresponding to the brain blood supply area atlas template as the corresponding second registration atlas, the current fixed map corresponding to the brain watershed atlas template as the corresponding third registration atlas, and the current fixed map corresponding to the cerebral cortex atlas template as the corresponding fourth registration atlas; and forming a group of corresponding four types of registration atlases from the first, second, third, and fourth registration atlases obtained this time, and outputting the four types of registration atlases obtained this time as the registration result of this time;

[0135] Step 423, and taking the first registration atlas, the second registration atlas, the third registration atlas, and the fourth registration atlas of the four types of registration atlases as the corresponding subject's brain blood supply area atlas, the subject's brain watershed atlas, and the subject's cerebral cortex atlas to form the corresponding subject atlas set;

[0136] Among them, the subject atlas set includes the subject's brain structure atlas, the subject's brain blood supply area atlas, the subject's brain watershed atlas and the subject's cerebral cortex atlas; the two-dimensional graphic sizes of the subject's brain structure atlas, the subject's brain blood supply area atlas, the subject's brain watershed atlas and the subject's cerebral cortex atlas are consistent with the two-dimensional graphic sizes of the subject's DWI images;

[0137] The subject's brain structure atlas includes a plurality of first atlas pixel points; each first atlas pixel point corresponds to a first atlas semantic type; the first atlas semantic type includes a background point type and multiple brain structure foreground point types;

[0138] The subject's brain blood supply area map includes a plurality of second map pixel points; each second map pixel point corresponds to a second map semantic type; the second map semantic type includes a background point type and multiple types of blood supply area foreground point types;

[0139] The subject's brain watershed map includes a plurality of third map pixel points; each third map pixel point corresponds to a third map semantic type; the third map semantic type includes a background point type and multiple watershed foreground point types;

[0140] The subject's cerebral cortex atlas includes a plurality of fourth atlas pixel points; each fourth atlas pixel point corresponds to a fourth atlas semantic type; the fourth atlas semantic type includes a background point type and multiple types of cerebral cortex foreground point types;

[0141] Step 43, if the total number of candidate matrices is equal to the first odd number N, then each candidate matrix in the candidate matrix sequence is used as the corresponding current registration matrix one by one; and based on the current registration matrix, the four types of templates of the four types of medical atlas templates are registered to obtain a set of corresponding four types of registered atlases; and based on the voting mechanism, the subject atlas is constructed according to the obtained N groups of four types of registered atlases to obtain the corresponding subject atlas set;

[0142] Specifically, it includes: step 431, if the total number of candidate matrices is equal to the first odd number N, then taking each candidate matrix in the candidate matrix sequence as the corresponding current registration matrix one by one;

[0143] Step 432, registering the four types of templates of the four types of medical atlas templates based on the current registration matrix to obtain a set of corresponding four types of registered atlases;

[0144] Here, the current step 432 is processed in the same manner as the aforementioned step 422;

[0145] Step 433, based on the voting mechanism, subject atlases are constructed according to the obtained N groups of four-category registration atlases to obtain a corresponding subject atlas set;

[0146] Specifically, it includes: Step 4331, initializing a brain structure atlas of a subject, a brain blood supply area atlas of a subject, a brain watershed atlas of a subject, and a cerebral cortex atlas of a subject;

[0147] Here, the first atlas semantic type of all the first atlas pixels in the initialized brain structure atlas of the subject is the background point type; the second atlas semantic type of all the second atlas pixels in the initialized brain blood supply area atlas of the subject is the background point type; the third atlas semantic type of all the third atlas pixels in the initialized brain watershed atlas of the subject is the background point type; the fourth atlas semantic type of all the fourth atlas pixels in the initialized cerebral cortex atlas of the subject is the background point type;

[0148] Step 4332, and forming a corresponding first registration atlas set from N first registration atlases in N groups of four types of registration atlases, a corresponding second registration atlas set from N second registration atlases, a corresponding third registration atlas set from N third registration atlases, and a corresponding fourth registration atlas set from N fourth registration atlases;

[0149] Step 4333, and rounding up the product of the preset first voting ratio and the first odd number N to obtain the corresponding first vote threshold;

[0150] Among them, the first voting ratio is a ratio value ranging from 0 to 1;

[0151] Step 4334, and traversing all the first atlas pixels in the brain structure atlas of the subject; and during the traversal, taking the currently traversed first atlas pixel as the corresponding current pixel; and taking the first registration atlas pixels on each first registration atlas in the first registration atlas set that match the current pixel as the corresponding first matching points; and forming a corresponding first type ballot box from the first registration atlas semantic types of N first matching points; and counting the number of each first registration atlas semantic type in the first type ballot box and taking the statistical result as the corresponding first type vote count; and taking the first type vote count with the largest value as the corresponding first maximum vote count; and when the first maximum vote count is greater than or equal to the first vote threshold, resetting the first atlas semantic type of the current pixel to the first registration atlas semantic type corresponding to the first maximum vote count;

[0152] Step 4335, traverse all the second atlas pixels of the subject's brain blood supply area atlas; during the traversal, regard the currently traversed second atlas pixel as the corresponding current pixel; regard the second registered atlas pixels on each second registered atlas in the second registration atlas set that match the current pixel as the corresponding second matching points; form a corresponding second type ballot box from the second registered atlas semantic types of the N second matching points; count the number of each second registered atlas semantic type in the second type ballot box and use the statistical result as the corresponding second type vote count; regard the second type vote count with the largest value as the corresponding second largest vote count; when the second largest vote count is greater than or equal to the first vote threshold, reset the second atlas semantic type of the current pixel to the second registered atlas semantic type corresponding to the second largest vote count;

[0153] Step 4336, traverse all the third atlas pixels of the subject's brain watershed atlas; during the traversal, regard the currently traversed third atlas pixel as the corresponding current pixel; regard the third registered atlas pixels on each third registered atlas in the third registration atlas set that match the current pixel as the corresponding third matching points; form a corresponding third type ballot box from the third registered atlas semantic types of the N third matching points; count the number of each third registered atlas semantic type in the third type ballot box and use the statistical result as the corresponding third type vote count; regard the third type vote count with the largest value as the corresponding third largest vote count; when the third largest vote count is greater than or equal to the first vote threshold, reset the third atlas semantic type of the current pixel to the third registered atlas semantic type corresponding to the third largest vote count;

[0154] Step 4337, traverse all the fourth atlas pixels of the subject's cerebral cortex atlas; during the traversal, regard the currently traversed fourth atlas pixel as the corresponding current pixel; regard the fourth registered atlas pixels on each fourth registered atlas in the fourth registration atlas set that match the current pixel as the corresponding fourth matching points; form a corresponding fourth type ballot box from the fourth registered atlas semantic types of the N fourth matching points; count the number of each fourth registered atlas semantic type in the fourth type ballot box and use the statistical result as the corresponding fourth type vote count; regard the fourth type vote count with the largest value as the corresponding fourth largest vote count; when the fourth largest vote count is greater than or equal to the first vote threshold, reset the fourth atlas semantic type of the current pixel to the fourth registered atlas semantic type corresponding to the fourth largest vote count;

[0155] Step 4338, form a corresponding subject atlas set from the subject's brain blood supply area atlas, the subject's brain watershed atlas, and the subject's cerebral cortex atlas after the reset is completed.

[0156] Figure 2The module structure diagram of a processing device for constructing a DWI image medical atlas based on image registration provided in the second embodiment of the present invention. This device is a terminal device or a server for implementing the foregoing method embodiments, or may also be a device that enables the foregoing terminal device or server to implement the foregoing method embodiments. For example, this device may be a device or a chip system of the foregoing terminal device or server. As Figure 2 shown, the device includes: a preprocessing module 201, a data receiving module 202, a candidate matrix screening module 203, and a medical atlas construction module 204.

[0157] The preprocessing module 201 is used to use a preset standard brain template T1 image as the corresponding T1 template image; construct a multimodal image library through data acquisition; identify the image registration matrices of the T1 template image and each modality image of each sample record in the multimodal image library and create a corresponding standard image library based on the identification results; and perform quality assessment on the sample records in the standard image library and delete the sample records with unqualified assessment.

[0158] The data receiving module 202 is used to receive the scan image dataset of any subject as the corresponding subject dataset; confirm whether there is a T1 subset in the subject dataset; if there is a T1 subset, use the T1 subset as the corresponding current subset; if there is no T1 subset, confirm whether there is a T2 subset in the subject dataset. If there is a T2 subset, use the T2 subset as the corresponding current subset. If there is no T2 subset, use the DWI subset of the subject dataset as the corresponding current subset; the subject dataset includes a DWI subset and / or a T1 subset and / or a T2 subset; the DWI subset is a required subset, and the T1 and T2 subsets are optional subsets.

[0159] The candidate matrix screening module 203 is used to screen candidate matrices for the image registration matrix between the subject DWI image and the T1 template image in the subject dataset based on the current subset, the standard image library, and the T1 template image to obtain a corresponding candidate matrix sequence.

[0160] The medical atlas construction module 204 is used to construct four types of medical atlases corresponding to the subject DWI image based on the candidate matrix sequence and four types of medical atlas templates of the T1 template image to obtain a corresponding subject atlas set.

[0161] A processing device for constructing a DWI image medical atlas based on image registration provided in the embodiment of the present invention can execute the method steps in the foregoing method embodiment, and its implementation principle and technical effect are similar, which will not be elaborated here.

[0162] It should be noted that it should be understood that the division of each module of the above device is only a division of logical functions. In actual implementation, it can be fully or partially integrated into a physical entity, or physically separated. And these modules can all be implemented in the form of software called by processing elements; they can also all be implemented in the form of hardware; it is also possible that some modules are implemented in the form of software called by processing elements, and some modules are implemented in the form of hardware. For example, the data receiving module can be a separately established processing element, or can be integrated in a certain chip of the above device. In addition, it can also be stored in the memory of the above device in the form of program code, and the function of the above determined module is called and executed by a certain processing element of the above device. The implementation of other modules is similar. In addition, all or part of these modules can be integrated together or can be independently implemented. The processing element described here can be an integrated circuit with signal processing capabilities. In the implementation process, each step of the above method or each of the above modules can be completed by the integrated logic circuit in the processor element or the instruction in the form of software.

[0163] For example, the above modules can be one or more integrated circuits configured to implement the above method, such as: one or more Application Specific Integrated Circuits (ASICs), or one or more Digital Signal Processors (DSPs), or one or more Field Programmable Gate Arrays (FPGAs), etc. Again, when a certain module above is implemented in the form of a processing element scheduling program code, the processing element can be a general-purpose processor, such as a Central Processing Unit (CPU) or other processors that can call program code. Again, these modules can be integrated together and implemented in the form of a System-on-a-chip (SOC).

[0164] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions described in the foregoing method embodiments are generated in whole or in part. The above computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The above computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the above computer instructions may be transmitted from one website, computer, server, or data center to another website, computer, server, or data center in a wired manner (such as coaxial cable, optical fiber, Digital Subscriber Line (DSL)) or wirelessly (such as infrared, wireless, Bluetooth, microwave, etc.). The above computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more integrated available media. The above available medium may be a magnetic medium (such as a floppy disk, hard disk, magnetic tape), an optical medium (such as a DVD), or a semiconductor medium (such as a solid state disk (SSD)), etc.

[0165] Figure 3 FIG. 4 is a schematic structural diagram of an electronic device provided in Embodiment 3 of the present invention. The electronic device may be a terminal device or a server for implementing the method of the foregoing embodiments, or may be a terminal device or a server for implementing the method of the foregoing embodiments and connected to the foregoing terminal device or server. As Figure 3 shown, the electronic device may include: a processor 301 (such as a CPU), a memory 302, and a transceiver 303; the transceiver 303 is coupled to the processor 301, and the processor 301 controls the transceiver operations of the transceiver 303. Various instructions may be stored in the memory 302 for completing various processing functions and implementing the processing steps described in the foregoing method embodiments. Preferably, the electronic device according to the embodiments of the present invention further includes: a power supply 304, a system bus 305, and a communication port 306. The system bus 305 is used to implement communication connections between components. The above communication port 306 is used for the electronic device to connect and communicate with other peripherals.

[0166] In Figure 3The system bus 305 mentioned above can be a Peripheral Component Interconnect (PCI) bus, an Extended Industry Standard Architecture (EISA) bus, or the like. The system bus can be divided into an address bus, a data bus, a control bus, etc. For the sake of convenience in representation, only a thick line is used in the figure, but it does not mean that there is only one bus or one type of bus. The communication interface is used to implement communication between the database access device and other devices (such as clients, read-write libraries, and read-only libraries). The memory may include Random Access Memory (RAM), and may also include non-volatile memory, such as at least one disk memory.

[0167] The aforementioned processor can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), a Graphics Processing Unit (GPU), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0168] It should be noted that the embodiments of the present invention also provide a computer-readable storage medium, in which instructions are stored. When the instructions run on a computer, the computer is caused to execute the methods and processes provided in the above embodiments.

[0169] The embodiments of the present invention also provide a chip for running instructions, and the chip is used to execute the processing steps described in the foregoing method embodiments.

[0170] An embodiment of the present invention provides a processing method, device, electronic device, and computer-readable storage medium for constructing a DWI image medical atlas based on image registration. As can be seen from the above, in the embodiment of the present invention, a standard image library with registration information is created in advance through a standard brain template T1 image (T1 template image) and a multi-modal image library; then, when receiving any subject dataset, a corresponding T1 / T2 / DWI subset is selected from the current subject dataset in the order of T1 being the most optimal, T2 being the second, and DWI being the last as the current subset, and candidate matrix screening is performed on the image registration matrix between the DWI image and the T1 template image based on the current subset, the standard image library, and the T1 template image. Finally, the four types of medical atlas templates are registered according to the candidate matrix sequence to obtain the four types of medical atlases corresponding to the DWI image. Through the embodiment of the present invention, the segmentation accuracy of the four types of medical atlases corresponding to the DWI image can be improved, thereby helping to improve the prediction accuracy of cerebral infarction type prediction.

[0171] Those skilled in the art should also be able to further realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of the two. To clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described according to functions in the above description. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0172] The steps of the method or algorithm described in combination with the embodiments disclosed herein can be implemented by hardware, software modules executed by a processor, or a combination of the two. The software modules can be placed in a random access memory (RAM), memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0173] The specific embodiments described above have further elaborated on the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above is only the specific embodiment of the present invention and is not used to limit the protection scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A processing method for constructing a DWI image medical atlas based on image registration, characterized in that, The method includes: Taking the preset standard brain template T1 image as the corresponding T1 template image; constructing a multi-modal image library through data acquisition; identifying the image registration matrices of the T1 template image and each modal image of each sample record in the multi-modal image library and creating a corresponding standard image library based on the identification results; evaluating the quality of the sample records in the standard image library and deleting the sample records with unqualified evaluation; Receiving the scanned image dataset of any subject as the corresponding subject dataset; confirming whether there is a T1 subset in the subject dataset; if the T1 subset exists, taking the T1 subset as the corresponding current subset; if the T1 subset does not exist, confirming whether there is a T2 subset in the subject dataset, if the T2 subset exists, taking the T2 subset as the corresponding current subset, if the T2 subset does not exist, taking the DWI subset of the subject dataset as the corresponding current subset; the subject dataset includes the DWI subset and / or the T1 subset and / or the T2 subset; the DWI subset is a required subset, and the T1 and T2 subsets are optional subsets; Based on the current subset, the standard image library, and the T1 template image, screening candidate matrices for the image registration matrix between the subject DWI image and the T1 template image in the subject dataset to obtain a corresponding candidate matrix sequence; Based on the candidate matrix sequence and four types of medical atlas templates of the T1 template image, constructing four types of medical atlases corresponding to the subject DWI image to obtain a corresponding subject atlas set.

2. The processing method for constructing a DWI image medical atlas based on image registration according to claim 1, wherein The multi-modal image library includes multiple first sample records; the first sample record includes a first sample T1 image, a first sample T2 image, a first sample DWI image, and a first scan parameter set; the first scan parameter set consists of T1 image scan parameters, T2 image scan parameters, and DWI image scan parameters, and each type of image scan parameter includes at least corresponding scan device parameters, scan software parameters, and scan configuration parameters; the first sample T1 image, the first sample T2 image, and the first sample DWI image in each first sample record are pixel-aligned with each other; The standard image library includes multiple second sample records; the second sample record includes a second sample T1 image, a second sample T2 image, a second sample DWI image, a second scan parameter set, a first T1 registration matrix, a first T2 registration matrix, and a first DWI registration matrix; each second sample record corresponds to one first sample record; The DWI subset includes the subject's DWI image and the corresponding DWI image scanning parameters; the T1 subset includes the subject's T1 image and the corresponding T1 image scanning parameters; the T2 subset includes the subject's T2 image and the corresponding T2 image scanning parameters; the subject's DWI image, the subject's T1 image, and the subject's T2 image are pixel-aligned with each other; The four types of medical atlas templates include a brain structure atlas template, a brain blood supply area atlas template, a brain watershed atlas template, and a cerebral cortex atlas template; The two-dimensional graphic sizes of the brain structure atlas template, the brain blood supply area atlas template, the brain watershed atlas template, and the cerebral cortex atlas template are all consistent with the two-dimensional graphic size of the T1 template image; The brain structure atlas template includes a plurality of first template pixel points; each of the first template pixel points corresponds to a first template semantic type; The first template semantic type includes a background point type and multiple types of brain structure foreground point types; Each type of the brain structure foreground point type matches a type of brain anatomical structure; The brain blood supply area atlas template includes a plurality of second template pixel points; each of the second template pixel points corresponds to a second template semantic type; The second template semantic type includes the background point type and multiple types of blood supply area foreground point types; each type of the blood supply area foreground point type matches the blood supply area of a type of brain blood vessel; The brain watershed atlas template includes a plurality of third template pixel points; each of the third template pixel points corresponds to a third template semantic type; the third template semantic type includes the background point type and multiple types of watershed foreground point types; Each type of the watershed foreground point type matches a type of brain blood vessel watershed area; The cerebral cortex atlas template includes a plurality of fourth template pixel points; each of the fourth template pixel points corresponds to a fourth template semantic type; the fourth template semantic type includes the background point type and multiple types of cerebral cortex foreground point types; Each type of the cerebral cortex foreground point type matches a type of cerebral cortex area; The subject atlas set includes a subject brain structure atlas, a subject brain blood supply area atlas, a subject brain watershed atlas, and a subject cerebral cortex atlas; the two-dimensional graphic sizes of the subject brain structure atlas, the subject brain blood supply area atlas, the subject brain watershed atlas, and the subject cerebral cortex atlas are all consistent with the two-dimensional graphic size of the subject DWI image; The subject brain structure atlas includes a plurality of first atlas pixel points; each of the first atlas pixel points corresponds to a first atlas semantic type; the first atlas semantic type includes the background point type and multiple types of the brain structure foreground point types; The subject brain blood supply area atlas includes a plurality of second atlas pixel points; each of the second atlas pixel points corresponds to a second atlas semantic type; the second atlas semantic type includes the background point type and multiple types of the blood supply area foreground point types; The subject's brain watershed atlas includes a plurality of third atlas pixel points; each of the third atlas pixel points corresponds to a third atlas semantic type; the third atlas semantic type includes the background point type and multiple types of the watershed foreground point types; The subject's cerebral cortex atlas includes a plurality of fourth atlas pixel points; each of the fourth atlas pixel points corresponds to a fourth atlas semantic type; The fourth atlas semantic type includes the background point type and multiple types of the cerebral cortex foreground point types.

3. The processing method for constructing a DWI image medical atlas based on image registration according to claim 2, wherein, The construction of the multimodal image library through data collection specifically includes: Recruiting a plurality of volunteers to form a corresponding first volunteer set; the first volunteer set includes a plurality of first volunteers; the first volunteer is a volunteer who has been medically examined and confirmed that no cerebral infarction lesions have occurred in their brain; And traversing all the first volunteers; and during the traversal, taking the currently traversed first volunteer as the corresponding current volunteer; and performing a magnetic resonance imaging scan on the current volunteer and generating a set of T1, T2, and DWI images with the same scan pose, synchronous scan time, and aligned scan imaging pixels as the corresponding first sample T1 image, the first sample T2 image, and the first sample DWI image; and forming a corresponding first scan parameter set from the T1 image scan parameters, T2 image scan parameters, and DWI image scan parameters corresponding to this scan; and forming a corresponding first sample record from the obtained first sample T1 image, the first sample T2 image, the first sample DWI image, and the first scan parameter set; And at the end of the traversal, forming the corresponding multimodal image library from all the obtained first sample records.

4. The processing method for constructing a DWI imaging medical atlas based on image registration according to claim 2, wherein The identification of the image registration matrices of the T1 template image and the modal images of each sample record in the multimodal image library and the creation of the corresponding standard image library in combination with the identification results specifically include: Taking each of the first sample records in the multimodal image library as the corresponding current sample record; And taking the T1 template image as the corresponding current floating image, taking the first sample T1 image of the current sample record as the corresponding current fixed image, and using a preset medical image registration analysis tool to analyze the registration matrix from the current floating image to the current fixed image according to the flexible registration analysis rule and taking the analyzed registration matrix as the corresponding first T1 registration matrix; the medical image registration analysis tool at least includes ANTS software and NiftyReg software; And performing two matrix replications on the first T1 registration matrix to obtain the corresponding first T2 registration matrix and first DWI registration matrix; the first T2 registration matrix and the first DWI registration matrix are consistent with the first T1 registration matrix; The first sample T1 image, the first sample T2 image, the first sample DWI image, and the first set of scanning parameters recorded by the current sample are used as the corresponding second sample T1 image, second sample T2 image, second sample DWI image, and second set of scanning parameters; and the corresponding second sample T1 image, second sample T2 image, second sample DWI image, second set of scanning parameters, first T1 registration matrix, first T2 registration matrix, and first DWI registration matrix recorded by the current sample form a corresponding second sample record; And the corresponding second sample records corresponding to all the first sample records form the corresponding standard image library.

5. The processing method for constructing a DWI image medical atlas based on image registration according to claim 2, wherein, Performing quality assessment on the sample records in the standard image library and deleting the sample records with unqualified assessment, specifically including: Regarding each of the second sample records in the standard image library as the corresponding current sample record; randomly selecting the first specified number M of brain experts from a preset expert team to form the corresponding current assessment team; sending the current sample record and the T1 template image to the current assessment team for image registration quality evaluation to obtain the corresponding M expert scores; calculating the average score of the obtained M expert scores to obtain the corresponding first average score; and confirming whether the first average score does not exceed a preset score threshold. If it is confirmed that it does not exceed, the current sample record is deleted from the standard image library as the corresponding unqualified assessment record; the expert team consists of multiple brain experts from one or more medical institutions in one or more regions; the first specified number M is a positive integer.

6. The processing method for constructing a DWI image medical atlas based on image registration according to claim 2, characterized in that, Based on the current subset, the standard image library, and the T1 template image, screening candidate matrices for the image registration matrix between the DWI image of the subject in the subject dataset and the T1 template image to obtain the corresponding candidate matrix sequence, specifically including: Identifying the current subset; if the current subset is a T1 subset, extracting the subject T1 image and T1 image scanning parameters of the current subset as the corresponding current image and current scanning parameters; if the current subset is a T2 subset, extracting the subject T2 image and T2 image scanning parameters of the current subset as the corresponding current image and current scanning parameters; if the current subset is a DWI subset, extracting the subject DWI image and DWI image scanning parameters of the current subset as the corresponding current image and current scanning parameters; And regarding the current image as the corresponding current image to be recognized; performing brain imaging omics feature recognition processing on the current image to be recognized to obtain the corresponding image brain features, and regarding the image brain features obtained this time as the corresponding current brain features; Traverse all the second sample records in the standard image library; during the traversal, use the currently traversed second sample record as the corresponding current sample record; use the T1, T2, or DWI image scanning parameters in the second scanning parameter set of the current sample record corresponding to the current subset as the corresponding current comparison scanning parameters; use the second sample T1 image, the second sample T2 image, or the second sample DWI image in the current sample record corresponding to the current subset as the corresponding current comparison image; identify whether the current comparison scanning parameters match the current scanning parameters; if they do not match, directly proceed to the next second sample record for continued traversal; if they match, use the current comparison image as the new current image to be recognized, perform brain imaging omics feature recognition processing on the new current image to be recognized to obtain the new image brain features, use the image brain features obtained this time as the corresponding current comparison brain features, perform characteristic vector conversion on the current brain features and the current comparison brain features respectively to obtain the corresponding first and second feature vectors, calculate the vector similarity of the first and second feature vectors based on the vector similarity algorithm to obtain the corresponding first similarity, and identify whether the first similarity exceeds the preset similarity threshold. If the first similarity exceeds the similarity threshold, mark the current sample record as a candidate sample record and proceed to the next second sample record for continued traversal. If the first similarity does not exceed the similarity threshold, directly proceed to the next second sample record for continued traversal; At the end of the traversal, count the total number of candidate sample records to obtain the corresponding first total count; use the current image as the corresponding current fixed image; Identify the first total count; If the first total count is greater than or equal to the preset first odd number N, mark the N candidate sample records with the top first similarities as the corresponding final selected sample records; use the first T1 registration matrix, the first T2 registration matrix, or the first DWI registration matrix in each final selected sample record corresponding to the current subset as a corresponding one-step registration matrix; use the second sample T1 image, the second sample T2 image, or the second sample DWI image in each final selected sample record corresponding to the current subset as the corresponding current floating image, analyze the registration matrix from the current floating image to the current fixed image according to the flexible registration analysis rule, and use the analyzed registration matrix as the corresponding two-step registration matrix; use the matrix obtained by multiplying the one-step and two-step registration matrices corresponding to each final selected sample record as a corresponding candidate matrix; form the corresponding candidate matrix sequence from the N obtained candidate matrices; the first odd number N is an odd number greater than 1; If the first total count is equal to N - 1, record the top N - 1 candidate sample records in terms of the first similarity as the corresponding final selected sample records; and use the first T1 registration matrix, the first T2 registration matrix, or the first DWI registration matrix corresponding to the current subset in each of the final selected sample records as a corresponding one-step registration matrix; and use the second sample T1 image, the second sample T2 image, or the second sample DWI image corresponding to the current subset in each of the final selected sample records as the corresponding current floating image, and analyze the registration matrix from the current floating image to the current fixed image according to the flexible registration analysis rule and use the analyzed registration matrix as the corresponding two-step registration matrix; and use the matrix obtained by multiplying the one-step and two-step registration matrices corresponding to each of the final selected sample records as a corresponding candidate matrix; and use the T1 template image as the corresponding current floating image, and analyze the registration matrix from the current floating image to the current fixed image according to the flexible registration analysis rule and use the analyzed registration matrix as a corresponding candidate matrix; and form a corresponding candidate matrix sequence from the N obtained candidate matrices; If the first total count is less than N - 1, use the T1 template image as the corresponding current floating image, and analyze the registration matrix from the current floating image to the current fixed image according to the flexible registration analysis rule and use the analyzed registration matrix as a corresponding candidate matrix to form a corresponding candidate matrix sequence.

7. The processing method for constructing a DWI medical image atlas based on image registration according to claim 6, wherein The brain imaging genomics feature recognition process specifically includes: According to the foreground / background division rule with the pixel points in the ventricle region as foreground pixel points and the pixel points in the brain tissue region as background pixel points, perform binary image conversion on the current image to be recognized in this process based on the Otsu algorithm to obtain a corresponding first binary image; and extract the first-order features and morphological features of the imaging genomics features of the ventricle region image on the first binary image to obtain corresponding ventricle imaging genomics features; each pixel point of the first binary image corresponds to a first binary semantics, and the first binary semantics includes ventricle foreground points and brain tissue background points; And according to the foreground / background division rule with the pixel points in the brain tissue region as foreground pixel points and the pixel points in the ventricle region as background pixel points, perform binary image conversion on the current image to be recognized in this process based on the Otsu algorithm to obtain a corresponding second binary image; and extract the first-order features and morphological features of the imaging genomics features of the brain tissue region image on the second binary image to obtain corresponding brain tissue imaging genomics features; each pixel point of the second binary image corresponds to a second binary semantics, and the second binary semantics includes brain tissue foreground points and ventricle background points; And output the image brain features composed of the ventricle and brain tissue imaging genomics features obtained this time as the processing result of this time.

8. The processing method for constructing a DWI image medical atlas based on image registration according to claim 2, wherein Constructing four types of medical atlas corresponding to the DWI image of the subject based on the four types of medical atlas templates of the candidate matrix sequence and the T1 template image, specifically including: Identifying the total number of candidate matrices in the candidate matrix sequence; If the total number of candidate matrices is 1, taking the only candidate matrix in the candidate matrix sequence as the corresponding current registration matrix; registering the four types of templates of the four types of medical atlas templates based on the current registration matrix to obtain a set of corresponding four types of registered atlases; and taking the first registered atlas, the second registered atlas, the third registered atlas, and the fourth registered atlas of the four types of registered atlases as the corresponding subject brain blood supply area atlas, the subject brain watershed atlas, and the subject cerebral cortex atlas to form the corresponding subject atlas set; the four types of registered atlases include the first, second, third, and fourth registered atlases; the first registered atlas includes a plurality of first registered atlas pixel points, each first registered atlas pixel point corresponding to a first registered atlas semantic type, the first registered atlas semantic type including the background point type and multiple types of brain structure foreground point types; the second registered atlas includes a plurality of second registered atlas pixel points, each second registered atlas pixel point corresponding to a second registered atlas semantic type, the second registered atlas semantic type including the background point type and multiple types of blood supply area foreground point types; the third registered atlas includes a plurality of third registered atlas pixel points, each third registered atlas pixel point corresponding to a third registered atlas semantic type, the third registered atlas semantic type including the background point type and multiple types of watershed foreground point types; the fourth registered atlas includes a plurality of fourth registered atlas pixel points, each fourth registered atlas pixel point corresponding to a fourth registered atlas semantic type, the fourth registered atlas semantic type including the background point type and multiple types of cerebral cortex foreground point types; If the total number of candidate matrices is equal to the preset first odd number N, taking each candidate matrix in the candidate matrix sequence as the corresponding current registration matrix one by one; registering the four types of templates of the four types of medical atlas templates based on the current registration matrix to obtain a set of corresponding four types of registered atlases; and constructing the subject atlas corresponding to the obtained N groups of four types of registered atlases based on the voting mechanism.

9. The processing method for constructing a DWI image medical atlas based on image registration according to claim 8, characterized in that, The registration of the four types of templates of the four types of medical atlas templates based on the current registration matrix specifically includes: Take the brain structure atlas template, the brain blood supply area atlas template, the brain watershed atlas template, and the cerebral cortex atlas template in the four types of medical atlas templates one by one as the corresponding current floating images; register the current floating images with the current registration matrix to obtain the corresponding current fixed images; take the current fixed image corresponding to the brain structure atlas template as the corresponding first registered atlas, the current fixed image corresponding to the brain blood supply area atlas template as the corresponding second registered atlas, the current fixed image corresponding to the brain watershed atlas template as the corresponding third registered atlas, and the current fixed image corresponding to the cerebral cortex atlas template as the corresponding fourth registered atlas; and form a corresponding set of the four types of registered atlases from the first, second, third, and fourth registered atlases obtained this time, and output the four types of registered atlases obtained this time as the registration result of this time.

10. The processing method for constructing a DWI image medical atlas based on image registration according to claim 8, wherein The subject atlas construction based on the voting mechanism according to the obtained N sets of the four types of registered atlases to obtain the corresponding subject atlas set specifically includes: Initialize a subject brain structure atlas, a subject brain blood supply area atlas, a subject brain watershed atlas, and a subject cerebral cortex atlas; the first atlas semantic types of all the first atlas pixel points of the initialized subject brain structure atlas are all the background point types; the second atlas semantic types of all the second atlas pixel points of the initialized subject brain blood supply area atlas are all the background point types; the third atlas semantic types of all the third atlas pixel points of the initialized subject brain watershed atlas are all the background point types; the fourth atlas semantic types of all the fourth atlas pixel points of the initialized subject cerebral cortex atlas are all the background point types; And form a corresponding first registered atlas set from N first registered atlases in N sets of the four types of registered atlases, a corresponding second registered atlas set from N second registered atlases, a corresponding third registered atlas set from N third registered atlases, and a corresponding fourth registered atlas set from N fourth registered atlases; And round up the product of the preset first voting ratio and the first odd number N to obtain the corresponding first vote threshold; the first voting ratio is a ratio value ranging from 0 to 1; Traverse all the first atlas pixels of the subject's brain structure atlas; during the traversal, take the currently traversed first atlas pixel as the corresponding current pixel; take the first registered atlas pixels on each of the first registered atlases in the first registered atlas set that match the current pixel as the corresponding first matching points; form a corresponding first type ballot box from the first registered atlas semantic types of the N first matching points; count the number of each first registered atlas semantic type in the first type ballot box and take the statistical result as the corresponding first type vote count; take the first type vote count with the largest value as the corresponding first maximum vote count; when the first maximum vote count is greater than or equal to the first vote threshold, reset the first atlas semantic type of the current pixel to the first registered atlas semantic type corresponding to the first maximum vote count; Traverse all the second atlas pixels of the subject's brain blood supply area atlas; during the traversal, take the currently traversed second atlas pixel as the corresponding current pixel; take the second registered atlas pixels on each of the second registered atlases in the second registered atlas set that match the current pixel as the corresponding second matching points; form a corresponding second type ballot box from the second registered atlas semantic types of the N second matching points; count the number of each second registered atlas semantic type in the second type ballot box and take the statistical result as the corresponding second type vote count; take the second type vote count with the largest value as the corresponding second maximum vote count; when the second maximum vote count is greater than or equal to the first vote threshold, reset the second atlas semantic type of the current pixel to the second registered atlas semantic type corresponding to the second maximum vote count; Traverse all the third atlas pixels of the subject's brain watershed atlas; during the traversal, take the currently traversed third atlas pixel as the corresponding current pixel; take the third registered atlas pixels on each of the third registered atlases in the third registered atlas set that match the current pixel as the corresponding third matching points; form a corresponding third type ballot box from the third registered atlas semantic types of the N third matching points; count the number of each third registered atlas semantic type in the third type ballot box and take the statistical result as the corresponding third type vote count; take the third type vote count with the largest value as the corresponding third maximum vote count; when the third maximum vote count is greater than or equal to the first vote threshold, reset the third atlas semantic type of the current pixel to the third registered atlas semantic type corresponding to the third maximum vote count; Traverse all the fourth atlas pixels of the brain cortex atlas of the subject; during the traversal, use the currently traversed fourth atlas pixel as the corresponding current pixel; use the fourth registration atlas pixels on each of the fourth registration atlases in the fourth registration atlas set that match the current pixel as the corresponding fourth matching points; form a corresponding fourth type ballot box from the fourth registration atlas semantic types of the N fourth matching points; count the number of each fourth registration atlas semantic type in the fourth type ballot box and use the statistical result as the corresponding fourth type vote count; use the fourth type vote count with the largest value as the corresponding fourth maximum vote count; when the fourth maximum vote count is greater than or equal to the first vote threshold, reset the fourth atlas semantic type of the current pixel to the fourth registration atlas semantic type corresponding to the fourth maximum vote count; Form the corresponding subject atlas set from the brain blood supply area atlas, the brain watershed atlas, and the brain cortex atlas of the subject after the reset is completed.

11. An apparatus for performing the processing method of constructing a DWI image medical atlas based on image registration according to any one of claims 1-10, characterized in that, The device includes: a preprocessing module, a data receiving module, a candidate matrix screening module, and a medical atlas construction module; The preprocessing module is used to use a preset standard brain template T1 image as the corresponding T1 template image; construct a multi-modal image library through data acquisition; identify the image registration matrices of the T1 template image and each modal image of each sample record in the multi-modal image library and create a corresponding standard image library based on the identification results; evaluate the quality of the sample records in the standard image library and delete the sample records with unqualified evaluations; The data receiving module is used to receive the scan image dataset of any subject as the corresponding subject dataset; confirm whether there is a T1 subset in the subject dataset; if the T1 subset exists, use the T1 subset as the corresponding current subset; if the T1 subset does not exist, confirm whether there is a T2 subset in the subject dataset, if the T2 subset exists, use the T2 subset as the corresponding current subset, if the T2 subset does not exist, use the DWI subset of the subject dataset as the corresponding current subset; the subject dataset includes the DWI subset and / or the T1 subset and / or the T2 subset; the DWI subset is a required subset, and the T1 and T2 subsets are optional subsets; The candidate matrix screening module is used to screen candidate matrices for the image registration matrix between the DWI image of the subject and the T1 template image in the subject dataset based on the current subset, the standard image library, and the T1 template image to obtain a corresponding candidate matrix sequence; The medical atlas construction module is used to construct four types of medical atlases corresponding to the DWI image of the subject based on the candidate matrix sequence and four types of medical atlas templates of the T1 template image to obtain a corresponding subject atlas set.

12. An electronic device, characterized in that, Including: A memory, a processor, and a transceiver; The processor is used to be coupled with the memory, read and execute the instructions in the memory to implement the method according to any one of claims 1-10; The transceiver is coupled with the processor, and the processor controls the transceiver to perform message sending and receiving.

13. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions, and when the computer instructions are executed by a computer, the computer is caused to execute the method according to any one of claims 1-10.

Images