Processing method and device of a model for assessing the function of a lymphoid system

Abstract

Embodiments of the present application relate to a kind of lymphoid system function evaluation model processing method and device, the method comprises: in ICBM-DTI-81 brain white matter atlas selects candidate region set;And based on ICBM-DTI-81 atlas, candidate region set and by the first data set of all research objects lymphoid system function rating result and diffusion tensor image;Lymphoid system function evaluation model is constructed, and model is trained based on the first data set;After training, according to the registration result of ICBM-DTI-81 atlas and the diffusion tensor image of the brain of the measured person, the region of interest of candidate region set on the image of the measured person is marked, and according to all regions of interest, left and right brain ALPS index is calculated and thus generates index feature tensor, and index feature tensor is input into lymphoid system function evaluation model to obtain prediction vector for prediction.By the present application, evaluation accuracy and evaluation flexibility can be improved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a processing method and apparatus for a lymphatic system functional assessment model. Background Technology

[0002] The central nervous system (CNS) is the only organ system lacking lymphatic vessels to assist in the removal of interstitial metabolic waste products. The brain's glymphatic system (GS), composed of aquaporin 4 (AQP4) located on the terminal appendages of astrocytes and the perivascular space (PVS), plays a role in promoting the exchange and flow of cerebrospinal fluid (CSF) and interstitial fluid (ISF). GS dysfunction (such as functional decline or impairment) can lead to a decreased clearance rate of brain metabolic products. Previous studies have observed evidence of decreased GS function in patients with neurodegenerative diseases such as Parkinson's disease (PD), epilepsy, cerebrovascular disease, and Alzheimer's disease (AD). Therefore, assessing GS functional status can provide potential biomarkers for early intervention in neurodegenerative diseases.

[0003] Diffusion Tensor Image Analysis Along The Perivascular Space (DTI-ALPS) is a novel technique developed based on diffusion tensor imaging (DTI). It allows for non-invasive observation of water diffusion along the perivascular space using diffusion tensor images. This technique also provides the calculation principle for the ALPS index. Recent studies have shown a correlation between the ALPS index and GS functional status, and GS functional status can be assessed by observing the ALPS index.

[0004] However, analysis of the current assessment mechanisms reveals several problems: 1) Most current assessment mechanisms still rely on traditional statistical methods to collect and analyze samples to derive a series of fixed threshold ranges, and then assess the GS functional status corresponding to the ALPS index based on these fixed threshold ranges. This fixed threshold range assessment is clearly not flexible enough; 2) When calculating the ALPS index, it is necessary to manually set the Region of Interest (ROI), which may reduce the accuracy of the index due to limitations in human experience; 3) The number of ROIs is limited to only one or a pair of symmetrical regions of the left and right hemispheres, which may reduce the accuracy of the index due to limitations in the number of samples. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by providing a method, apparatus, electronic device, and computer-readable storage medium for processing a lymphoid system functional assessment model. This invention selects multiple fiber tract regions as a candidate region set from the ICBM-DTI-81 white matter atlas; constructs a lymphoid system functional assessment model based on a deep learning framework; selects multiple research subjects with normal / abnormal lymphoid system function according to rules; and has an expert panel conduct a three-level assessment of the GS function (normal, diminished, impaired) for each research subject. Based on the ICBM-DTI-81 atlas, the candidate region set, and the brain DTI images + GS function assessment results of all research subjects, a model training dataset, i.e., the first dataset, is constructed. Then, based on the first dataset, the model is further processed... A lymphatic system function assessment model is trained. During the assessment process based on this model, firstly, all regions of interest (ROIs) in the current image are automatically labeled according to the automatic registration results of the ICBM-DTI-81 atlas and the subject's brain DTI images. Then, according to the DTI-ALPS index calculation principle, the left and right hemisphere ALPS indices are calculated based on all ROIs, and corresponding model input vectors are generated based on the calculation process data and results. Finally, the lymphatic system function assessment model performs end-to-end three-category GS function prediction (normal, reduced, and impaired) based on the model input vectors. This invention can improve the accuracy of ROI localization, increase the flexibility and number of customized ROIs, and further enhance the assessment flexibility and convenience through a deep learning model.

[0006] To achieve the above objectives, a first aspect of the present invention provides a method for processing a lymphatic system-like functional assessment model, the method comprising:

[0007] Multiple fiber bundle regions were selected as candidate regions in the ICBM-DTI-81 white matter atlas according to the preset candidate region customization rules.

[0008] Multiple research subjects with normal and abnormal brain lymphoid system function were selected according to preset research subject selection rules; a preset expert group assessed the brain lymphoid system of each research subject at three levels: normal, diminished, or impaired; brain images of each research subject were acquired using diffusion tensor imaging technology; a subject dataset was constructed based on the functional level assessment results and diffusion tensor images of all research subjects; and a first dataset was constructed based on the ICBM-DTI-81 atlas, the candidate region set, and the subject dataset.

[0009] A deep learning model is constructed to predict the functional level of the brain's lymphoid system as a lymphoid system function assessment model. This model predicts the functional level of the brain's lymphoid system based on the exponential feature tensor A input to the model and outputs a corresponding prediction vector P. The exponential feature tensor A consists of left and right brain feature vectors a. L a R Composition; the left brain feature vector a L Including features feature feature feature And the index ALPS L The right brain feature vector a R Including features feature feature feature And the index ALPS R The prediction vector P includes three prediction probabilities p. i 1 ≤ index i ≤ 3, the predicted probability p i＝1、2、3 These correspond to normal function, diminished function, and impaired function, respectively, with the sum of the three predicted probabilities being 1.

[0010] The lymphoid system functional evaluation model is trained based on the first dataset;

[0011] After model training, the system receives a brain diffusion tensor image of any subject input by the user as the current image; and marks the regions of interest (ROIs) of the candidate region set on the current image based on the registration results of the ICBM-DTI-81 atlas and the current image; and calculates the left and right brain ALPS indices according to the DTI-ALPS index calculation principle based on all ROIs, generating the corresponding index feature tensor A based on the calculation process data and calculation results; and inputs the index feature tensor A into the lymphoid system function assessment model for prediction to obtain the corresponding prediction vector P, which is then fed back to the current user.

[0012] Preferably, the candidate region customization rules include projection fiber region rules and association fiber region rules. The projection fiber region rules require that, in the left and right hemispheres, candidate regions with projection fiber type are preferentially selected from the posterior corona radiata, superior corona radiata, posterior limb of internal capsule, sagittal layer, and corticospinal tract, which are related to cerebrospinal fluid flow, to form candidate region pairs. The association fiber region rules require that, in the left and right hemispheres, candidate regions with association fiber type are preferentially selected from the anterior thalamic radiation, fornix, sagittal layer, inferior frontal tract, superior longitudinal fasciculus, and arcuate fasciculus, which are related to cerebrospinal fluid flow, to form candidate region pairs.

[0013] The candidate region set includes one or more candidate region pairs; each candidate region pair consists of a pair of left and right candidate regions that are symmetrical about the left and right hemispheres; each candidate region is spherical in shape, and the region attributes corresponding to each candidate region include the fiber type, central coordinates, and sphere radius, wherein the fiber type includes projection fibers and association fibers; the left and right candidate regions of each candidate region pair have the same fiber type, the same central coordinates, and the same sphere radius;

[0014] The selection rules for the research subjects require that the number of research subjects with normal and abnormal brain lymphoid system function be equal; and that the abnormalities in brain lymphoid system function include some or all of the following disease types: Parkinson's disease, epilepsy, cerebrovascular disease, and Alzheimer's disease.

[0015] The object dataset includes multiple object data records; the object data records include object images and object functional levels; the object images are brain diffusion tensor images; the object functional levels include normal function, functional decline, and functional impairment.

[0016] The first dataset includes multiple first data records; the first data records include a training tensor A. TR and label vector P TR The training tensor A TR From the left brain feature vector a L and the right brain feature vector a R Composition; the label vector P TR Includes three label probabilities The three label probabilities correspond to normal function, reduced function, and dysfunction, respectively. Only one of the three label probabilities is 1, and the other two are 0.

[0017] Preferably, the model input of the lymphoid system functional assessment model is used to receive the exponential feature tensor A, and the model output is used to output the prediction vector P;

[0018] The lymphatic system functional assessment model includes a feature triage module, a left brain feature extraction module, a right brain feature extraction module, a feature fusion module, a fusion feature extraction module, and a prediction probability output module.

[0019] The input terminal of the feature splitting module is connected to the input terminal of the model, and the first and second output terminals are respectively connected to the input terminals of the left brain and right brain feature extraction modules; the output terminals of the left brain and right brain feature extraction modules are respectively connected to the first and second input terminals of the feature fusion module; the output terminal of the feature fusion module is connected to the input terminal of the fused feature extraction module; the output terminal of the fused feature extraction module is connected to the input terminal of the prediction probability output module; and the output terminal of the prediction probability output module is connected to the model output terminal.

[0020] The feature splitting module is used to split the left brain feature vector a of the exponential feature tensor A. L and the right brain feature vector a R Send them to the left brain and right brain feature extraction modules respectively;

[0021] The left-brain feature extraction module is implemented based on an MLP model, which consists of two or more fully connected layers connected sequentially, with a ReLU activation function connected to the output of each fully connected layer. The left-brain feature extraction module is used to process the left-brain feature vector a. L Feature extraction is performed to obtain the corresponding feature vector H. L Send to the feature fusion module;

[0022] The right-brain feature extraction module is implemented based on another MLP model. The current MLP model has the same model structure as the MLP model corresponding to the left-brain feature extraction module, but the parameters are independent. The right-brain feature extraction module is used to extract the right-brain feature vector a. R Feature extraction is performed to obtain the corresponding feature vector H. R Send the feature vector H to the feature fusion module. L H R The shape remains consistent;

[0023] The feature fusion module performs vector concatenation on the feature vector H. L H R Perform concatenation and use the resulting concatenated vector as the corresponding fusion vector H. C Send to the fusion feature extraction module;

[0024] The fusion feature extraction module is implemented based on another MLP model. The current MLP model consists of two or more fully connected layers connected sequentially, and the output of each fully connected layer is connected to a ReLU activation function. The fusion feature extraction module is used to process the fusion vector H. C The feature vector Z obtained by feature extraction is sent to the prediction probability output module; the vector length of the feature vector Z is 3.

[0025] The prediction probability output module is used to perform three-class classification probability prediction based on the feature vector Z using the Softmax function to obtain the corresponding three prediction probabilities p. i The corresponding prediction vector P is then formed and output.

[0026] Preferably, the step of constructing the first dataset based on the ICBM-DTI-81 map, the candidate region set, and the object dataset specifically includes:

[0027] A traversal is performed on all object data records in the object dataset; during this traversal, the currently traversed object data record is taken as the corresponding current record; and the object image of the current record is taken as the corresponding current image; the regions of interest (ROIs) of the candidate region set on the current image are marked according to the registration results of the ICBM-DTI-81 atlas and the current image; and the left and right brain ALPS indices are calculated based on all ROIs according to the DTI-ALPS index calculation principle, and the corresponding training tensor A is generated based on the calculation process data and calculation results. TR The system identifies the current functional level of the object; if the current functional level of the object is normal, it sets the corresponding three label probabilities. The values ​​are 1, 0, and 0; if the current functional level of the object is functionally degraded, then the corresponding three label probabilities are set. The values ​​are 0, 1, and 0; if the current functional level of the object is functional impairment, then the corresponding three label probabilities are set. The values ​​are 0, 0, and 1; and the three label probabilities obtained in this study form a corresponding label vector P. TR ; and the training tensor A corresponding to the current record. TR and the label vector P TR A corresponding first data record is formed; and at the end of this round of traversal, all the first data records obtained are combined to form the corresponding first dataset.

[0028] Preferably, training the lymphoid system functional assessment model based on the first dataset specifically includes:

[0029] Step 51: Based on a preset first segmentation ratio, the first dataset is randomly divided into two sub-datasets, denoted as the first training set and the first evaluation set.

[0030] Wherein, both the first training set and the first evaluation set are composed of multiple first data records; the ratio of the total number of records in the first training set and the first evaluation set satisfies the first segmentation ratio;

[0031] Step 52: Perform a traversal of all the first data records in the first training set; and during this traversal, take the currently traversed first data record as the corresponding current training record; and set the training tensor A of the current training record as... TR The corresponding exponential feature tensor A is input into the lymphoid system functional evaluation model for prediction, and the prediction vector P output by this prediction is used as the corresponding current prediction vector; and the current prediction vector and the label vector P of the current training record are used together. TR Form a corresponding first prediction-label pair; and at the end of this round of traversal, input all the obtained first prediction-label pairs into the preset first model loss function to calculate the corresponding first loss value;

[0032] The loss function of the first model is based on the multi-class cross-entropy loss function;

[0033] Step 53: Identify whether the first loss value meets the preset first loss value range; if the first loss value meets the first loss value range, proceed to step 54; if the first loss value does not meet the first loss value range, perform a round of modulation on the model parameters of the lymphatic system function evaluation model based on the preset first model optimizer in the direction of minimizing the first model loss function, and return to step 52 to continue training when the first round of modulation ends.

[0034] The first model optimizer includes at least the Adam optimizer and the SGD optimizer;

[0035] Step 54: Perform a traversal of all the first data records in the first evaluation set; and during this traversal, take the currently traversed first data record as the corresponding current evaluation record; and set the training tensor A of the current evaluation record... TR The corresponding exponential feature tensor A is input into the lymphoid system functional assessment model for prediction, and the prediction vector P output by this prediction is used as the corresponding current prediction vector; and the current prediction vector and the label vector P recorded in the current assessment are used together. TRA corresponding second prediction-label pair is formed; and at the end of this round of traversal, the first accuracy, first precision, first recall and first F1 score are calculated based on all the obtained second prediction-label pairs.

[0036] Step 55: Identify whether the first accuracy, first precision, first recall, and first F1 score all satisfy their respective first accuracy range, first precision range, first recall range, and first F1 score range; if not, return to step 52 to continue training; if yes, confirm that the model training is complete.

[0037] Preferably, the step of marking the region of interest in the current image based on the registration result of the ICBM-DTI-81 atlas and the current image specifically includes:

[0038] The ICBM-DTI-81 atlas is used as the corresponding standard brain template; the three-dimensional voxel space of the standard brain template is used as the corresponding first space; and the three-dimensional voxel space of the current image is used as the corresponding second space.

[0039] First, the FLIRT tool of the FSL toolkit is used to map the current image from the second space to the first space using linear registration to obtain the corresponding mapped image. The linear mapping matrix from the second space to the first space is used as the corresponding current transformation matrix, and the inverse of the current transformation matrix is ​​used as the corresponding current inverse transformation matrix. Then, the FNIRT tool of the FSL toolkit is used to perform nonlinear registration on the standard brain template and the mapped image in the first space to obtain the corresponding nonlinear deformation field. The FSL toolkit is a toolkit for data analysis of diffusion tensor images, the FLIRT tool is the linear registration tool of the FSL toolkit, and the FNIRT tool is the nonlinear registration tool of the FSL toolkit.

[0040] Based on the nonlinear deformation field, voxel-level deformation transformation is performed on the left and right candidate regions of each candidate region pair in the candidate region set to obtain the corresponding left and right deformed regions; and based on the current inverse transformation matrix, a linear mapping process from the first space to the second space is performed on each left and right deformed region to obtain the corresponding left and right mapped regions; and each left and right mapped region is taken as a corresponding region of interest; and voxel-level labeling is performed on each region of interest in the current image.

[0041] Preferably, the feature The average x-axis diffusion rate of all left brain projection fiber regions, where each left brain projection fiber region is a region of interest corresponding to a left candidate region of the fiber type being projection fiber; the feature The average y-axis diffusion rate of all the left brain projection fiber regions; the z-axis direction of each left brain projection fiber region is the main projection fiber direction of the current fiber region, the x-axis direction is the perivascular space direction perpendicular to the main projection fiber direction of the current fiber region, and the y-axis direction is the perpendicular direction of the xz plane formed by the main projection fiber direction of the current fiber region and the perivascular space direction.

[0042] The features The average x-axis diffusion rate of all left hemisphere associated fiber regions, where each left hemisphere associated fiber region is a region of interest corresponding to a left candidate region of associated fiber type; the feature The average z-axis diffusion rate of all the left brain association fiber regions; the y-axis direction of each left brain association fiber region is the main direction of the association fibers of the current fiber region, the x-axis direction is the direction of the perivascular space perpendicular to the main direction of the current association fibers, and the z-axis direction is the direction perpendicular to the xy plane formed by the main direction of the current association fibers and the direction of the perivascular space.

[0043] The features The average x-axis diffusion rate of all right brain projection fiber regions, where each right brain projection fiber region is a region of interest corresponding to a right candidate region of the fiber type being projection fiber; the feature The average y-axis diffusion rate of all the right brain projection fiber regions; the z-axis direction of each right brain projection fiber region is the main projection fiber direction of the current fiber region, the x-axis direction is the perivascular space direction perpendicular to the main projection fiber direction of the current fiber region, and the y-axis direction is the perpendicular direction of the xz plane formed by the main projection fiber direction and the perivascular space direction.

[0044] The features The average x-axis diffusion rate of all right hemisphere associated fiber regions, where each right hemisphere associated fiber region is a region of interest corresponding to a right candidate region of associated fiber type; the feature The average z-axis diffusion rate of all the right brain association fiber regions; the y-axis direction of each right brain association fiber region is the main direction of the association fibers of the current fiber region, the x-axis direction is the direction of the perivascular space perpendicular to the main direction of the current association fibers, and the z-axis direction is the direction perpendicular to the xy plane formed by the main direction of the current association fibers and the direction of the perivascular space.

[0045] The index ALPS L ALPS R for:

[0046]

[0047] Preferably, the step of calculating the left and right brain ALPS indices according to the DTI-ALPS index calculation principle based on all regions of interest and generating the corresponding index feature tensor A based on the calculation process data and calculation results specifically includes:

[0048] Each candidate region of interest corresponding to a left brain projection fiber is denoted as a corresponding left brain projection fiber region. Based on the x, y, and z axes of the current left brain projection fiber region, the x and y axis diffusion rates of each voxel point within the current left brain projection fiber region are calculated to obtain the corresponding voxel x-axis diffusion rate and voxel y-axis diffusion rate. The sum of the x-axis diffusion rates of all voxels in the current left brain projection fiber region is taken as the corresponding fiber region x-axis diffusion rate, and the sum of the y-axis diffusion rates of all voxels is taken as the corresponding fiber region y-axis diffusion rate. The average value of the x-axis diffusion rates of all left brain projection fiber regions is taken as the corresponding feature. The average y-axis diffusion rate of all the left brain projection fiber areas is taken as the corresponding feature.

[0049] Each candidate region of interest corresponding to a right-side candidate region of projection fiber type is denoted as a corresponding right brain projection fiber region. Based on the x, y, and z axes of the current right brain projection fiber region, the x and y axis diffusion rates of each voxel point within the current right brain projection fiber region are calculated to obtain the corresponding voxel x-axis diffusion rate and voxel y-axis diffusion rate. The sum of the x-axis diffusion rates of all voxels in the current right brain projection fiber region is taken as the corresponding fiber region x-axis diffusion rate, and the sum of the y-axis diffusion rates of all voxels is taken as the corresponding fiber region y-axis diffusion rate. The average value of the x-axis diffusion rates of all fiber regions in the right brain projection fiber region is taken as the corresponding feature. The average y-axis diffusion rate of all right brain projection fiber areas is taken as the corresponding feature.

[0050] Each candidate region of interest corresponding to a left-side fiber of type synoptic fiber is denoted as a corresponding left-brain synoptic fiber region. Based on the x, y, and z axes of the current left-brain synoptic fiber region, the x and z axis diffusion rates of each voxel point within the current left-brain synoptic fiber region are calculated to obtain the corresponding voxel x-axis diffusion rate and voxel z-axis diffusion rate. The sum of the x-axis diffusion rates of all voxels in the current left-brain synoptic fiber region is taken as the corresponding fiber region x-axis diffusion rate, and the sum of the z-axis diffusion rates of all voxels is taken as the corresponding fiber region z-axis diffusion rate. The average value of the x-axis diffusion rates of all fiber regions in the left-brain synoptic fiber region is taken as the corresponding feature. The average z-axis diffusion rate of all the aforementioned left brain association fiber regions is taken as the corresponding feature.

[0051] Each candidate region of interest corresponding to a right-side fiber of type synoptic fiber is denoted as a corresponding right-brain synoptic fiber region. Based on the x, y, and z axes of the current right-brain synoptic fiber region, the x and z axis diffusion rates of each voxel point within the current right-brain synoptic fiber region are calculated to obtain the corresponding voxel x-axis diffusion rate and voxel z-axis diffusion rate. The sum of the x-axis diffusion rates of all voxels in the current right-brain synoptic fiber region is taken as the corresponding fiber region x-axis diffusion rate, and the sum of the z-axis diffusion rates of all voxels is taken as the corresponding fiber region z-axis diffusion rate. The average value of the x-axis diffusion rates of all fiber regions in the right-brain synoptic fiber region is taken as the corresponding feature. The average z-axis diffusion rate of all the right brain association fiber regions is taken as the corresponding feature.

[0052] And based on the DTI-ALPS index calculation principle, and according to the characteristics obtained this time... The features The features and the features Calculate the corresponding exponent ALPS L ; and based on the DTI-ALPS index calculation principle, according to the characteristics obtained this time. The features The features and the features Calculate the corresponding exponent ALPS R ; and the features obtained in this study The features The features The features and the index ALPS L Form a corresponding left brain feature vector a L ; and the features obtained in this study The features The features The features and the index ALPS R Form a corresponding right brain feature vector a R ;

[0053] And the left brain feature vector a obtained in this study L and the right brain feature vector a R This forms a corresponding exponential feature tensor A.

[0054] A second aspect of the present invention provides an apparatus for implementing the processing method of the lymphoid system functional evaluation model described in the first aspect above, the apparatus comprising: a candidate region customization module, a model dataset preparation module, a model construction module, a model training module, and a lymphoid system functional evaluation module;

[0055] The candidate region customization module is used to select multiple fiber bundle regions as candidate region sets in the ICBM-DTI-81 white matter map according to preset candidate region customization rules.

[0056] The model dataset preparation module is used to select multiple research subjects with normal and abnormal brain lymphoid system functions according to preset research subject selection rules; and a preset expert group conducts a three-level functional assessment of the brain lymphoid system of each research subject, classifying it as normal, diminished, or impaired; and acquires brain images of each research subject using diffusion tensor imaging technology; and constructs an object dataset based on the functional level assessment results and diffusion tensor images of all research subjects; and constructs a first dataset based on the ICBM-DTI-81 atlas, the candidate region set, and the object dataset.

[0057] The model building module is used to construct a deep learning model for predicting the functional level of the brain's lymphoid system as a lymphoid system function assessment model. The lymphoid system function assessment model is used to predict the functional level of the brain's lymphoid system based on the exponential feature tensor A input to the model and output the corresponding prediction vector P. The exponential feature tensor A consists of left and right brain feature vectors a. L a R Composition; the left brain feature vector a L Including features feature feature feature And the index ALPS L The right brain feature vector a R Including features feature feature feature And the index ALPS R The prediction vector P includes three prediction probabilities p. i 1 ≤ index i ≤ 3, the predicted probability p i＝1、2、3 These correspond to normal function, diminished function, and impaired function, respectively, with the sum of the three predicted probabilities being 1.

[0058] The model training module is used to train the lymphatic system functional evaluation model based on the first dataset;

[0059] The lymphoid system functional assessment module is used to receive, after model training is completed, a brain diffusion tensor image of any subject input by the user as the current image; and to mark the regions of interest (ROIs) of the candidate region set on the current image according to the registration results of the ICBM-DTI-81 atlas and the current image; and to calculate the left and right brain ALPS indices according to the DTI-ALPS index calculation principle based on all ROIs, and generate the corresponding index feature tensor A based on the calculation process data and calculation results; and to input the index feature tensor A into the lymphoid system functional assessment model for prediction to obtain the corresponding prediction vector P and feed it back to the current user.

[0060] A third aspect of the present invention provides an electronic device, including: a memory, a processor, and a transceiver;

[0061] The processor is used to couple with the memory, read and execute instructions in the memory to implement the steps of the method described in the first aspect above;

[0062] The transceiver is coupled to the processor, and the processor controls the transceiver to send and receive messages.

[0063] A fourth aspect of the present invention provides a computer-readable storage medium storing computer instructions that, when executed by a computer, cause the computer to perform the instructions described in the first aspect.

[0064] This invention provides a method, apparatus, electronic device, and computer-readable storage medium for processing a lymphoid system functional assessment model. As described above, this invention can select multiple fiber tract regions as a candidate region set from the ICBM-DTI-81 white matter atlas; construct a lymphoid system functional assessment model based on a deep learning model framework; select multiple research subjects with normal / abnormal lymphoid system function according to rules; have an expert group conduct a three-level assessment of the GS function of each research subject (normal, reduced, impaired); and construct a model training dataset, i.e., the first dataset, based on the ICBM-DTI-81 atlas, the candidate region set, and the brain DTI images + GS function assessment results of all research subjects. Then, based on the second dataset... A dataset is used to train a lymphoid system function assessment model. During the assessment process based on this model, firstly, all regions of interest (ROIs) in the current image are automatically labeled according to the automatic registration results of the ICBM-DTI-81 atlas and the subject's brain DTI images. Then, according to the DTI-ALPS index calculation principle, the left and right hemisphere ALPS indices are calculated based on all ROIs, and corresponding model input vectors are generated based on the calculation process data and results. Finally, the lymphoid system function assessment model performs end-to-end three-category GS function prediction (normal, declining, and impaired) based on the model input vectors. This invention improves assessment accuracy by enhancing the accuracy, flexibility, and number of ROI locations, and by using a deep learning model to improve assessment flexibility and convenience. Attached Figure Description

[0065] Figure 1 This is a schematic diagram of a processing method for a lymphatic system functional assessment model provided in Embodiment 1 of the present invention;

[0066] Figure 2 This is a schematic diagram of the modules of the lymphatic system functional evaluation model provided in Embodiment 1 of the present invention;

[0067] Figure 3 This is a module structure diagram of a processing device for a lymphatic system functional assessment model provided in Embodiment 2 of the present invention;

[0068] Figure 4 This is a schematic diagram of the structure of an electronic device provided in Embodiment 3 of the present invention. Detailed Implementation

[0069] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0070] Embodiment 1 of the present invention provides a method for processing a lymphatic system-like functional assessment model, such as... Figure 1 The schematic diagram shows a processing method for a lymphatic system-like functional assessment model provided in Embodiment 1 of the present invention. The method mainly includes the following steps:

[0071] Step 1: Select multiple fiber bundle regions as candidate regions in the ICBM-DTI-81 white matter atlas according to the preset candidate region customization rules.

[0072] Here, the ICBM-DTI-81 white matter atlas is a standardized white matter anatomical atlas jointly developed by Johns Hopkins University (JHU) and the International Consortium for Brain Mapping (ICBM). Its core lies in achieving high-precision visualization and quantitative analysis of white matter fiber tracts through DTI technology. In this embodiment of the invention, the ICBM-DTI-81 white matter atlas is used both as a ROI expansion resource for DTI-ALPS technology and as a standard brain template for DTI image registration.

[0073] The candidate region customization rule of this invention is an ROI expansion rule based on the ROI selection rule of DTI-ALPS technology. The candidate region customization rule includes a projection fiber region rule and a conjunction fiber region rule. The projection fiber region rule requires that, in the left and right hemispheres, candidate regions with projection fiber type be selected from the posterior corona radiata, superior corona radiata, posterior limb of internal capsule, sagittal layer, and corticospinal tract, which are related to cerebrospinal fluid flow, to form a candidate region pair. The conjunction fiber region rule requires that, in the left and right hemispheres, candidate regions with conjunction fiber type be selected from the anterior thalamic radiation, fornix, sagittal layer, inferior frontal tract, superior longitudinal fasciculus, and arcuate fasciculus, which are related to cerebrospinal fluid flow, to form a candidate region pair.

[0074] The candidate region set of this invention includes one or more candidate region pairs; each candidate region pair consists of a pair of left and right candidate regions that are symmetrical between the left and right hemispheres; each candidate region is spherical in shape, and the region attributes corresponding to each candidate region include fiber type, central coordinates, and sphere radius. The fiber type includes projection fibers and association fibers; the left and right candidate regions of each candidate region pair have the same fiber type, symmetrical central coordinates, and equal sphere radii.

[0075] Step 2: Select multiple research subjects with normal and abnormal brain lymphoid system function according to the preset research subject selection rules; and have a preset expert group conduct a three-level functional assessment of the brain lymphoid system of each research subject, classifying it as normal, diminished, or impaired; and collect brain images of each research subject using diffusion tensor imaging technology; and construct a subject dataset based on the functional level assessment results and diffusion tensor images of all research subjects; and construct the first dataset based on the ICBM-DTI-81 atlas, candidate region set, and subject dataset.

[0076] Specifically, this includes: Step 21, selecting multiple research subjects with normal and abnormal brain lymphoid system functions according to preset research subject selection rules;

[0077] Here, the selection rules for research subjects in this embodiment of the invention require that the number of research subjects with normal and abnormal brain lymphoid system function be equal; and that the abnormal types of brain lymphoid system function include some or all of the disease types among Parkinson's disease, epilepsy, cerebrovascular disease, and Alzheimer's disease.

[0078] Step 22, and a pre-selected expert group will assess the brain lymphoid system of each research subject in three levels of function: normal, diminished, or impaired.

[0079] Here, the preset expert group in this embodiment of the invention is a pre-selected expert group composed of senior experts and scholars in the fields of central nervous system diseases, magnetic resonance imaging, and DTI technology; the three functional levels can also be understood as three functional levels, namely normal function, functional decline, and functional impairment.

[0080] Step 23, and data acquisition of brain images of each research subject using diffusion tensor imaging technology;

[0081] Here, the actual process involves collecting brain DTI images of each research subject within the most recent specified time period. If none are available, further imaging examinations based on DTI technology are required to obtain the corresponding brain DTI images.

[0082] Step 24, and construct the object dataset based on the functional level assessment results and diffusion tensor images of all research subjects;

[0083] Here, the object dataset of this embodiment of the invention includes multiple object data records; each object data record corresponds to a research object; each object data record includes an object image and an object functional level; the object image is a brain diffusion tensor image; the object functional level includes normal function, functional decline, and functional impairment;

[0084] Step 25, and construct the first dataset based on the ICBM-DTI-81 map, candidate region set, and object dataset;

[0085] The first dataset includes multiple first data records; each first data record includes a training tensor A. TR and label vector P TR Training tensor A TR From the left brain feature vector a L and right brain feature vector a R Composition; Label vector P TR Includes three label probabilities The three label probabilities correspond to normal function, reduced function, and dysfunction, respectively. Only one of the three label probabilities is 1, and the other two are 0.

[0086] Specifically, this includes: performing a traversal of all object data records in the object dataset; during this traversal, using the currently traversed object data record as the corresponding current record; using the object image of the current record as the corresponding current image; marking the regions of interest (ROIs) of the candidate region set on the current image based on the registration results of the ICBM-DTI-81 atlas and the current image; and calculating the left and right brain ALPS indices based on all ROIs according to the DTI-ALPS index calculation principle, and generating the corresponding training tensor A based on the calculation process data and calculation results. TR It also identifies the functional level of the currently recorded object; if the current object's functional level is normal, it sets the corresponding three label probabilities. The values ​​are 1, 0, and 0; if the current object's functionality level is reduced, then the corresponding three label probabilities are set. The values ​​are 0, 1, and 0; if the current object's functional level is functional impairment, then the corresponding three label probabilities are set. The values ​​are 0, 0, and 1; and a corresponding label vector P is formed from the three label probabilities obtained in this study. TR ; and the training tensor A corresponding to the current record. TR and label vector P TR Each set of data records forms a corresponding first data record; and at the end of this round of traversal, all the first data records obtained form the corresponding first dataset.

[0087] Specifically, based on the registration results between the ICBM-DTI-81 atlas and the current image, the regions of interest in the candidate region set on the current image are marked, including:

[0088] Step A1: Use the ICBM-DTI-81 atlas as the corresponding standard brain template; use the three-dimensional voxel space of the standard brain template as the corresponding first space; and use the three-dimensional voxel space of the current image as the corresponding second space.

[0089] Step A2: First, use the FLIRT tool in the FSL toolkit to map the current image from the second space to the first space using linear registration to obtain the corresponding mapped image. Then, use the linear mapping matrix from the second space to the first space as the corresponding current transformation matrix and the inverse matrix of the current transformation matrix as the corresponding current inverse transformation matrix. Next, use the FNIRT tool in the FSL toolkit to perform nonlinear registration on the standard brain template and the mapped image in the first space to obtain the corresponding nonlinear deformation field.

[0090] Here, the FSL (FMRIB Software Library) toolkit is a toolkit for data analysis of diffusion tensor images, the FLIRT (FMRIB's Linear Image Registration Tool) toolkit is the linear registration toolkit of the FSL toolkit, and the FNIRT (FMRIB's Non-linear Image Registration Tool) toolkit is the non-linear registration toolkit of the FSL toolkit;

[0091] Step A3 involves performing voxel-level deformation transformation on the left and right candidate regions of each candidate region pair in the candidate region set based on the nonlinear deformation field to obtain the corresponding left and right deformed regions; and performing linear mapping processing from the first space to the second space on each left and right deformed region based on the current inverse transformation matrix to obtain the corresponding left and right mapped regions; and taking each left and right mapped region as a corresponding region of interest; and marking each region of interest at the voxel level in the current image.

[0092] In addition, based on the DTI-ALPS index calculation principle, the ALPS indices of the left and right hemispheres are calculated according to all regions of interest, and the corresponding training tensor A is generated based on the calculation process data and calculation results. TR Specifically, it includes:

[0093] Step B1: The region of interest corresponding to each candidate region on the left side of the projective fiber type is denoted as a corresponding left brain projective fiber region. Based on the x, y, and z axes of the current left brain projective fiber region, the x-axis and y-axis diffusion rates of each voxel point within the current left brain projective fiber region are calculated to obtain the corresponding voxel x-axis diffusion rate and voxel y-axis diffusion rate. The sum of the x-axis diffusion rates of all voxels in the current left brain projective fiber region is taken as the corresponding fiber region x-axis diffusion rate, and the sum of the y-axis diffusion rates of all voxels is taken as the corresponding fiber region y-axis diffusion rate. The average value of the x-axis diffusion rates of all left brain projective fiber regions is taken as the corresponding feature. The average y-axis diffusion rate of all left brain projection fiber areas was used as the corresponding feature.

[0094] Step B2: The region of interest corresponding to each candidate region on the right side of the projective fiber type is recorded as a corresponding right brain projective fiber region. Based on the x, y, and z axes of the current right brain projective fiber region, the x and y axis diffusion rates of each voxel point within the current right brain projective fiber region are calculated to obtain the corresponding voxel x-axis diffusion rate and voxel y-axis diffusion rate. The sum of the x-axis diffusion rates of all voxels in the current right brain projective fiber region is taken as the corresponding fiber region x-axis diffusion rate, and the sum of the y-axis diffusion rates of all voxels is taken as the corresponding fiber region y-axis diffusion rate. The average value of the x-axis diffusion rates of all right brain projective fiber regions is taken as the corresponding feature. The average y-axis diffusion rate of all right brain projection fiber areas was used as the corresponding feature.

[0095] Step B3: The region of interest corresponding to each candidate region on the left side with fiber type "associative fiber" is recorded as a corresponding left hemisphere associative fiber region. Based on the x, y, and z axes of the current left hemisphere associative fiber region, the x-axis and z-axis diffusion rates of each voxel point within the current left hemisphere associative fiber region are calculated. The sum of the x-axis diffusion rates of all voxels in the current left hemisphere as the corresponding fiber region x-axis diffusion rate, and the sum of the z-axis diffusion rates of all voxels as the corresponding fiber region z-axis diffusion rate. The average value of the x-axis diffusion rates of all left hemisphere associative fiber regions is used as the corresponding feature. The average z-axis diffusion rate of all left brain association fiber areas was used as the corresponding feature.

[0096] Step B4: The region of interest corresponding to each candidate region on the right side with fiber type "associative fiber" is recorded as a corresponding right brain associative fiber region. Based on the x, y, and z axes of the current right brain associative fiber region, the x and z axis diffusion rates of each voxel point within the current right brain associative fiber region are calculated to obtain the corresponding voxel x-axis diffusion rate and voxel z-axis diffusion rate. The sum of the x-axis diffusion rates of all voxels in the current right brain associative fiber region x-axis diffusion rate, and the sum of the z-axis diffusion rates of all voxels as z-axis diffusion rate, are used as the corresponding fiber region z-axis diffusion rate. The average value of the x-axis diffusion rates of all right brain associative fiber regions is used as the corresponding feature. The average z-axis diffusion rate of all right brain association fiber areas was used as the corresponding feature.

[0097] Step B5, and based on the DTI-ALPS index calculation principle, according to the features obtained this time... feature feature and characteristics Calculate the corresponding exponent ALPS L ; and based on the DTI-ALPS index calculation principle, according to the characteristics obtained this time feature feature and characteristics Calculate the corresponding exponent ALPS R ; and the features obtained this time feature feature feature And the index ALPS L Form a corresponding left brain feature vector a L ; and the features obtained this time feature feature feature And the index ALPS R Form a corresponding right brain feature vector a R ;

[0098] Step B6, and the left brain feature vector a obtained this time. L and right brain feature vector a R Form a corresponding training tensor A TR .

[0099] Step 3: Construct a deep learning model to predict the functional level of the brain's lymphoid system as a lymphoid system function assessment model.

[0100] Here, the lymphoid system functional assessment model of this invention is used to predict the functional level of the brain's lymphoid system based on the exponential feature tensor A input to the model and output the corresponding prediction vector P. The exponential feature tensor A consists of the left and right brain feature vectors a L a R Composition; Left brain feature vector a L Including features feature feature feature And the index ALPS L Right brain feature vector a R Including features feature feature feature And the index ALPS R The prediction vector P includes three prediction probabilities p. i 1 ≤ index i ≤ 3, prediction probability p i＝1、2、3 These correspond to normal function, diminished function, and impaired function, respectively, with the sum of the three predicted probabilities being 1.

[0101] feature The average x-axis diffusion rate of all left brain projection fiber regions is represented, with each left brain projection fiber region being a region of interest corresponding to a left candidate region with projection fiber type; features The average y-axis diffusion rate of all left brain projection fiber areas; the z-axis direction of each left brain projection fiber area is the principal direction of the projection fiber of the current fiber area, the x-axis direction is the direction of the perivascular space perpendicular to the principal direction of the current projection fiber, and the y-axis direction is the direction perpendicular to the xz plane formed by the principal direction of the current projection fiber and the direction of the perivascular space.

[0102] feature The average x-axis diffusion rate of all left hemisphere associated fiber regions is represented, with each left hemisphere associated fiber region being a region of interest corresponding to a left candidate region with associated fiber type; features The z-axis diffusion rate is the average value of all left brain association fiber regions; the y-axis direction of each left brain association fiber region is the main direction of the association fibers of the current fiber region, the x-axis direction is the direction of the perivascular space perpendicular to the main direction of the current association fibers, and the z-axis direction is the direction perpendicular to the xy plane formed by the main direction of the current association fibers and the direction of the perivascular space.

[0103] feature The average x-axis diffusion rate of all right brain projection fiber regions is represented, with each right brain projection fiber region being a region of interest corresponding to a right-side candidate region with projection fiber type; features The average y-axis diffusion rate of all right brain projection fiber areas; the z-axis direction of each right brain projection fiber area is the main direction of the projection fiber of the current fiber area, the x-axis direction is the direction of the perivascular space perpendicular to the main direction of the current projection fiber, and the y-axis direction is the direction perpendicular to the xz plane formed by the main direction of the current projection fiber and the direction of the perivascular space.

[0104] feature The average x-axis diffusion rate of all right hemisphere associated fiber regions is represented, with each right hemisphere associated fiber region being a region of interest corresponding to a right-side candidate region with associated fiber type; features The z-axis diffusion rate is the average value of all right brain association fiber regions; the y-axis direction of each right brain association fiber region is the main direction of the association fibers of the current fiber region, the x-axis direction is the direction of the perivascular space perpendicular to the main direction of the current association fibers, and the z-axis direction is the direction perpendicular to the xy plane formed by the main direction of the current association fibers and the direction of the perivascular space.

[0105] ALPS Index L ALPS R The calculation method is as follows:

[0106]

[0107] like Figure 2 As shown in the schematic diagram of the lymphoid system functional assessment model provided in Embodiment 1 of the present invention, the model input terminal of the lymphoid system functional assessment model of the present invention is used to receive the exponential feature tensor A, and the model output terminal is used to output the prediction vector P. The model components of the lymphoid system functional assessment model include: a feature diversion module, a left brain feature extraction module, a right brain feature extraction module, a feature fusion module, a fusion feature extraction module, and a prediction probability output module.

[0108] The connection relationships of the model components in the lymphoid system functional assessment model are as follows: the input end of the feature diversion module is connected to the model input end; the first and second output ends are connected to the input ends of the left brain and right brain feature extraction modules, respectively; the output ends of the left brain and right brain feature extraction modules are connected to the first and second input ends of the feature fusion module, respectively; the output end of the feature fusion module is connected to the input end of the fusion feature extraction module; the output end of the fusion feature extraction module is connected to the input end of the prediction probability output module; and the output end of the prediction probability output module is connected to the model output end.

[0109] The functional components of the lymphoid system-like functional assessment model are shown below.

[0110] 1) Feature-based traffic splitting module:

[0111] The feature splitting module is used to split the left brain feature vector a of the exponential feature tensor A. Land right brain feature vector a R Send them to the feature extraction modules for the left and right hemispheres respectively.

[0112] 2) Left brain feature extraction module:

[0113] The left-brain feature extraction module is implemented based on an MLP model. The current MLP model consists of two or more fully connected layers connected sequentially, with a ReLU activation function connected to the output of each fully connected layer. The left-brain feature extraction module is used to extract the left-brain feature vector a. L Feature extraction is performed to obtain the corresponding feature vector H. L Send to the feature fusion module.

[0114] 3) Right Brain Feature Extraction Module:

[0115] The right-brain feature extraction module is implemented based on another MLP model. The current MLP model has the same model structure as the MLP model corresponding to the left-brain feature extraction module, but the parameters are independent. The right-brain feature extraction module is used to extract right-brain feature vector a. R Feature extraction is performed to obtain the corresponding feature vector H. R Send to the feature fusion module. Here, the feature vector H in this embodiment of the invention... L H R The shape remains consistent.

[0116] 4) Feature fusion module:

[0117] The feature fusion module concatenates the feature vector H according to the vector concatenation method. L H R Perform concatenation and use the resulting concatenated vector as the corresponding fusion vector H. C Send to the fusion feature extraction module.

[0118] 5) Feature extraction fusion module:

[0119] The fusion feature extraction module is implemented based on another MLP model. This MLP model consists of two or more fully connected layers connected sequentially, with each fully connected layer's output connected to a ReLU activation function. The fusion feature extraction module is used to process the fusion vector H... C The feature vector Z obtained by feature extraction is sent to the prediction probability output module. Here, in this embodiment of the invention, the vector length of the feature vector Z is 3.

[0120] 6) Prediction probability output module:

[0121] The prediction probability output module is used to perform three-class classification probability prediction based on the feature vector Z using the Softmax function, and obtain the corresponding three prediction probabilities p. i The corresponding prediction vector P is formed and output.

[0122] Step 4: Train the lymphatic system functional assessment model based on the first dataset;

[0123] Specifically, it includes: Step 41, randomly dividing the first dataset into two subsets based on a preset first segmentation ratio, denoted as the first training set and the first evaluation set;

[0124] Wherein, the first segmentation ratio is a pre-set ratio parameter, such as 8:2; both the first training set and the first evaluation set consist of multiple first data records; the ratio of the total number of records in the first training set and the first evaluation set satisfies the first segmentation ratio;

[0125] Step 42: Perform a full iteration on all the first data records in the first training set; during this iteration, use the currently iterated first data record as the corresponding current training record; and set the training tensor A of the current training record... TR The corresponding exponential feature tensor A is input into the lymphatic system functional assessment model for prediction, and the prediction vector P output from this prediction is used as the corresponding current prediction vector; and the current prediction vector and the label vector P of the current training record are used for prediction. TR Form a corresponding first prediction-label pair; and at the end of this round of traversal, input all the obtained first prediction-label pairs into the preset first model loss function to calculate the corresponding first loss value;

[0126] The first model's loss function is based on the multi-class cross-entropy loss function.

[0127] Step 43: Identify whether the first loss value meets the preset first loss value range; if the first loss value meets the first loss value range, proceed to step 44; if the first loss value does not meet the first loss value range, modulate the model parameters of the lymphatic system function evaluation model in one round based on the preset first model optimizer in the direction of minimizing the first model loss function, and return to step 42 to continue training when the first round of modulation ends.

[0128] Here, the first loss value range is a pre-set numerical range; the first model optimizer includes at least the Adam optimizer and the SGD optimizer;

[0129] Step 44: Perform a full iteration on all first data records in the first evaluation set; during this iteration, use the currently iterated first data record as the corresponding current evaluation record; and set the training tensor A of the current evaluation record... TR The corresponding exponential feature tensor A is input into the lymphatic system functional assessment model for prediction, and the prediction vector P output from this prediction is used as the corresponding current prediction vector; and the current prediction vector and the label vector P recorded in the current assessment are used together.TR A corresponding second prediction-label pair is formed; and at the end of this round of traversal, the first accuracy, first precision, first recall and first F1 score are calculated based on all the obtained second prediction-label pairs.

[0130] Step 45: Identify whether the first accuracy, first precision, first recall, and first F1 score all meet their respective first accuracy range, first precision range, first recall range, and first F1 score range; if not, return to step 42 to continue training; if yes, confirm that the model training is complete.

[0131] Here, the first accuracy range, the first precision range, the first recall range, and the first F1 score range are four pre-set numerical ranges.

[0132] Step 5: After model training is completed, the system receives the brain diffusion tensor image of any subject input by the user as the current image; and marks the regions of interest (ROIs) of the candidate region set on the current image based on the registration results of the ICBM-DTI-81 atlas and the current image; and calculates the left and right brain ALPS indices according to the DTI-ALPS index calculation principle based on all ROIs, and generates the corresponding exponential feature tensor A based on the calculation process data and calculation results; and inputs the exponential feature tensor A into the lymphatic system function assessment model for prediction to obtain the corresponding prediction vector P, which is then fed back to the current user.

[0133] Specifically, this includes: Step 51, after the model training is completed, receiving the brain diffusion tensor image of any subject input by the user as the current image;

[0134] Step 52, and mark the regions of interest in the current image based on the registration results of the ICBM-DTI-81 atlas and the current image;

[0135] Specifically, this includes: step 521, using the ICBM-DTI-81 atlas as the corresponding standard brain template; using the three-dimensional voxel space of the standard brain template as the corresponding first space; and using the three-dimensional voxel space of the current image as the corresponding second space;

[0136] Step 522: First, use the FLIRT tool in the FSL toolkit to map the current image from the second space to the first space using linear registration to obtain the corresponding mapped image. Then, use the linear mapping matrix from the second space to the first space as the corresponding current transformation matrix and the inverse matrix of the current transformation matrix as the corresponding current inverse transformation matrix. Next, use the FNIRT tool in the FSL toolkit to perform nonlinear registration on the standard brain template and the mapped image in the first space to obtain the corresponding nonlinear deformation field.

[0137] Step 523 involves performing voxel-level deformation transformation on the left and right candidate regions of each candidate region pair in the candidate region set based on the nonlinear deformation field to obtain the corresponding left and right deformed regions; and performing linear mapping from the first space to the second space on each left and right deformed region based on the current inverse transformation matrix to obtain the corresponding left and right mapped regions; and taking each left and right mapped region as a corresponding region of interest; and marking each region of interest at the voxel level in the current image.

[0138] Step 53, and calculate the left and right brain ALPS indices according to the DTI-ALPS index calculation principle based on all regions of interest, and generate the corresponding index feature tensor A based on the calculation process data and calculation results;

[0139] Specifically, this includes: Step 531, where the region of interest corresponding to each candidate region of the left brain with a fiber type of projection fiber is recorded as a corresponding left brain projection fiber region; and the x and y axis diffusion rates of each voxel point in the current left brain projection fiber region are calculated based on the x, y, and z axes to obtain the corresponding voxel x-axis diffusion rate and voxel y-axis diffusion rate, and the sum of the x-axis diffusion rates of all voxels in the current left brain projection fiber region is taken as the corresponding fiber region x-axis diffusion rate, and the sum of the y-axis diffusion rates of all voxels is taken as the corresponding fiber region y-axis diffusion rate; and the average value of the fiber region x-axis diffusion rates of all left brain projection fiber regions is taken as the corresponding feature. The average y-axis diffusion rate of all left brain projection fiber areas was used as the corresponding feature.

[0140] Step 532: The region of interest corresponding to each candidate region on the right side of the projective fiber type is recorded as a corresponding right brain projective fiber region. Based on the x, y, and z axes of the current right brain projective fiber region, the x and y axis diffusion rates of each voxel point within the current right brain projective fiber region are calculated to obtain the corresponding voxel x-axis diffusion rate and voxel y-axis diffusion rate. The sum of the x-axis diffusion rates of all voxels in the current right brain projective fiber region is taken as the corresponding fiber region x-axis diffusion rate, and the sum of the y-axis diffusion rates of all voxels is taken as the corresponding fiber region y-axis diffusion rate. The average value of the x-axis diffusion rates of all right brain projective fiber regions is taken as the corresponding feature. The average y-axis diffusion rate of all right brain projection fiber areas was used as the corresponding feature.

[0141] Step 533: The region of interest corresponding to each candidate region on the left side with fiber type of associative fiber is recorded as a corresponding left brain associative fiber region; the x-axis and z-axis diffusion rates of each voxel point in the current left brain associative fiber region are calculated based on the xyz axis of the current left brain associative fiber region, and the sum of the x-axis diffusion rates of all voxels in the current left brain associative fiber region x-axis diffusion rate, and the sum of the z-axis diffusion rates of all voxels in the current left brain associative fiber region z-axis diffusion rate; the average value of the x-axis diffusion rates of all left brain associative fiber regions is used as the corresponding feature. The average z-axis diffusion rate of all left brain association fiber areas was used as the corresponding feature.

[0142] Step 534: The region of interest corresponding to each candidate region on the right side with fiber type "associative fiber" is recorded as a corresponding right brain associative fiber region. Based on the x, y, and z axes of the current right brain associative fiber region, the x and z axis diffusion rates of each voxel point within the current right brain associative fiber region are calculated to obtain the corresponding voxel x-axis diffusion rate and voxel z-axis diffusion rate. The sum of the x-axis diffusion rates of all voxels in the current right brain associative fiber region x-axis diffusion rate, and the sum of the z-axis diffusion rates of all voxels as z-axis diffusion rate, are used as the corresponding fiber region z-axis diffusion rate. The average value of the x-axis diffusion rates of all fiber regions in the right brain associative fiber region is used as the corresponding feature. The average z-axis diffusion rate of all right brain association fiber areas was used as the corresponding feature.

[0143] Step 535, and according to the DTI-ALPS index calculation principle, based on the features obtained this time. feature feature and characteristics Calculate the corresponding exponent ALPS L ; and based on the DTI-ALPS index calculation principle, according to the characteristics obtained this time feature feature and characteristics Calculate the corresponding exponent ALPS R ; and the features obtained this time feature feature feature And the index ALPS L Form a corresponding left brain feature vector a L ; and the features obtained this time feature feature feature And the index ALPSR Form a corresponding right brain feature vector a R ;

[0144] Step 536, and from the left brain feature vector a obtained this time L and right brain feature vector a R This forms a corresponding exponential feature tensor A;

[0145] Step 54: Input the exponential feature tensor A into the lymphatic system function assessment model to make predictions and obtain the corresponding prediction vector P, which is then fed back to the current user.

[0146] Figure 3 This is a block diagram of a processing device for a lymphatic system functional assessment model provided in Embodiment 2 of the present invention. This device can be a terminal device or server implementing the aforementioned method embodiments, or it can be a device that enables the aforementioned terminal device or server to implement the aforementioned method embodiments. For example, the device can be a device or chip system of the aforementioned terminal device or server. Figure 3 As shown, the device includes: a candidate region customization module 201, a model dataset preparation module 202, a model construction module 203, a model training module 204, and a lymphatic system functional evaluation module 205.

[0147] The candidate region customization module 201 is used to select multiple fiber bundle regions as candidate region sets in the ICBM-DTI-81 brain white matter atlas according to preset candidate region customization rules.

[0148] The model dataset preparation module 202 is used to select multiple research subjects with normal and abnormal brain lymphoid system functions according to preset research subject selection rules; and a preset expert group conducts a three-level functional assessment of the brain lymphoid system of each research subject, classifying it as normal, diminished, or impaired; and collects brain images of each research subject using diffusion tensor imaging technology; and constructs an object dataset based on the functional level assessment results and diffusion tensor images of all research subjects; and constructs the first dataset based on the ICBM-DTI-81 atlas, candidate region set, and object dataset.

[0149] Model building module 203 is used to build a deep learning model for predicting the functional level of the brain's lymphoid system as a lymphoid system functional assessment model; the lymphoid system functional assessment model is used to predict the functional level of the brain's lymphoid system based on the exponential feature tensor A input to the model and output the corresponding prediction vector P; the exponential feature tensor A consists of the left and right brain feature vectors a L a R Composition; Left brain feature vector a L Including features feature feature feature And the index ALPS L Right brain feature vector a R Including features feature feature feature And the index ALPS R The prediction vector P includes three prediction probabilities p. i 1 ≤ index i ≤ 3, prediction probability p i＝1、2、3 These correspond to normal function, diminished function, and impaired function, respectively, with the sum of the three predicted probabilities being 1.

[0150] The model training module 204 is used to train the lymphatic system functional assessment model based on the first dataset.

[0151] The lymphoid system function assessment module 205 is used to receive the brain diffusion tensor image of any subject input by the user as the current image after the model training is completed; and to mark the regions of interest in the candidate region set on the current image according to the registration results of the ICBM-DTI-81 atlas and the current image; and to calculate the left and right brain ALPS indices according to the DTI-ALPS index calculation principle based on all regions of interest, and generate the corresponding exponential feature tensor A based on the calculation process data and calculation results; and to input the exponential feature tensor A into the lymphoid system function assessment model for prediction to obtain the corresponding prediction vector P and feed it back to the current user.

[0152] The processing device for a lymphatic system functional assessment model provided in this embodiment of the invention can execute the method steps in the above method embodiment. Its implementation principle and technical effect are similar, and will not be repeated here.

[0153] It should be noted that the division of the various modules in the above device is merely a logical functional division. In actual implementation, they can be fully or partially integrated into a single physical entity, or they can be physically separated. Furthermore, these modules can be implemented entirely in software via processing element calls; they can be fully implemented in hardware; or some modules can be implemented by processing element calls to software, while others are implemented in hardware. For example, the candidate region customization module can be a separate processing element, or it can be integrated into a chip in the above device. Alternatively, it can be stored as program code in the memory of the above device, and called and executed by a processing element of the device. The implementation of other modules is similar. Moreover, these modules can be fully or partially integrated together, or they can be implemented independently. 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 through integrated logic circuits in the hardware of the processor element or through software instructions.

[0154] For example, these modules can be one or more integrated circuits configured to implement the above methods, such as one or more Application Specific Integrated Circuits (ASICs), one or more Digital Signal Processors (DSPs), or one or more Field Programmable Gate Arrays (FPGAs). As another example, when a module is implemented using processing element scheduler code, the processing element can be a general-purpose processor, such as a Central Processing Unit (CPU) or other processor capable of calling program code. Furthermore, these modules can be integrated together as a System-on-a-Chip (SOC).

[0155] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. This computer program product includes one or more computer instructions. When these computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the foregoing method embodiments are generated. The computer described above can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The aforementioned computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the aforementioned computer instructions can be transmitted from one website, computer, server, or data center to another via wired (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL)) or wireless (e.g., infrared, wireless, Bluetooth, microwave, etc.) means. The aforementioned computer-readable storage medium can be any available medium that a computer can access, or a data storage device such as a server or data center that integrates one or more available media. The aforementioned available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state disks (SSDs)).

[0156] Figure 4This is a schematic diagram of an electronic device provided in Embodiment 3 of the present invention. This electronic device can be a terminal device or server implementing the methods of the aforementioned embodiments, or it can be a terminal device or server connected to the aforementioned terminal device or server implementing the methods of the aforementioned embodiments. Figure 4 As shown, the electronic device may include: a processor 301 (e.g., CPU), a memory 302, and a transceiver 303; the transceiver 303 is coupled to the processor 301, and the processor 301 controls the transmission and reception operations of the transceiver 303. The memory 302 may store various instructions for performing various processing functions and implementing the processing steps described in the foregoing embodiments. Preferably, the electronic device involved in 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 realize communication connections between components. The communication port 306 is used for communication between the electronic device and other peripherals.

[0157] exist Figure 4 The system bus 305 mentioned can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This system bus can be divided into address bus, data bus, control bus, etc. For ease of representation, it is represented by only one thick line in the figure, but this does not indicate that there is only one bus or one type of bus. The communication interface is used to enable communication between the database access device and other devices (e.g., clients, read-write libraries, and read-only libraries). Memory may include Random Access Memory (RAM) and may also include non-volatile memory, such as at least one disk storage device.

[0158] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), graphics processing units (GPUs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0159] It should be noted that the embodiments of the present invention also provide a computer-readable storage medium storing instructions that, when run on a computer, cause the computer to perform the methods and processes provided in the above embodiments.

[0160] This invention provides a method, apparatus, electronic device, and computer-readable storage medium for processing a lymphoid system functional assessment model. As described above, this invention can select multiple fiber tract regions as a candidate region set from the ICBM-DTI-81 white matter atlas; construct a lymphoid system functional assessment model based on a deep learning model framework; select multiple research subjects with normal / abnormal lymphoid system function according to rules; have an expert group conduct a three-level assessment of the GS function of each research subject (normal, reduced, impaired); and construct a model training dataset, i.e., the first dataset, based on the ICBM-DTI-81 atlas, the candidate region set, and the brain DTI images + GS function assessment results of all research subjects. Then, based on the second dataset... A dataset is used to train a lymphoid system function assessment model. During the assessment process based on this model, firstly, all regions of interest (ROIs) in the current image are automatically labeled according to the automatic registration results of the ICBM-DTI-81 atlas and the subject's brain DTI images. Then, according to the DTI-ALPS index calculation principle, the left and right hemisphere ALPS indices are calculated based on all ROIs, and corresponding model input vectors are generated based on the calculation process data and results. Finally, the lymphoid system function assessment model performs end-to-end three-category GS function prediction (normal, declining, and impaired) based on the model input vectors. This invention improves assessment accuracy by enhancing the accuracy, flexibility, and number of ROI locations, and by using a deep learning model to improve assessment flexibility and convenience.

[0161] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented in hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main 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.

[0162] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for processing a lymphatic system-like functional assessment model, characterized in that, The method includes: Multiple fiber bundle regions were selected as candidate regions in the ICBM-DTI-81 white matter atlas according to the preset candidate region customization rules. Multiple research subjects with normal and abnormal brain lymphoid system function were selected according to preset research subject selection rules; a preset expert group assessed the brain lymphoid system of each research subject at three levels: normal, diminished, or impaired; brain images of each research subject were acquired using diffusion tensor imaging technology; a subject dataset was constructed based on the functional level assessment results and diffusion tensor images of all research subjects; and a first dataset was constructed based on the ICBM-DTI-81 atlas, the candidate region set, and the subject dataset. A deep learning model is constructed to predict the functional level of the brain's lymphoid system as a lymphoid system function assessment model. This model predicts the functional level of the brain's lymphoid system based on the exponential feature tensor A input to the model and outputs a corresponding prediction vector P. The exponential feature tensor A consists of left and right brain feature vectors a. L a R Composition; the left brain feature vector a L Including features feature feature feature And the index ALPS L The right brain feature vector a R Including features feature feature feature And the index ALPS R The prediction vector P includes three prediction probabilities p. i 1 ≤ index i ≤ 3, the predicted probability p i＝1、2、3 These correspond to normal function, diminished function, and impaired function, respectively, with the sum of the three predicted probabilities being 1. The lymphoid system functional evaluation model is trained based on the first dataset; After model training, the system receives a brain diffusion tensor image of any subject input by the user as the current image; and marks the regions of interest (ROIs) of the candidate region set on the current image based on the registration results of the ICBM-DTI-81 atlas and the current image; and calculates the left and right brain ALPS indices according to the DTI-ALPS index calculation principle based on all ROIs, generating the corresponding index feature tensor A based on the calculation process data and calculation results; and inputs the index feature tensor A into the lymphoid system function assessment model for prediction to obtain the corresponding prediction vector P, which is then fed back to the current user.

2. The processing method for the lymphatic system-like functional assessment model according to claim 1, characterized in that, The candidate region customization rules include projection fiber region rules and association fiber region rules. The projection fiber region rules require that, in the left and right hemispheres, candidate regions with projection fiber type be selected from the posterior corona radiata, superior corona radiata, posterior limb of internal capsule, sagittal layer, and corticospinal tract, which are related to cerebrospinal fluid flow, to form candidate region pairs. The association fiber region rules require that, in the left and right hemispheres, candidate regions with association fiber type be selected from the anterior thalamic radiation, fornix, sagittal layer, inferior frontal tract, superior longitudinal fasciculus, and arcuate fasciculus, which are related to cerebrospinal fluid flow, to form candidate region pairs. The candidate region set includes one or more candidate region pairs; each candidate region pair consists of a pair of left and right candidate regions that are symmetrical about the left and right hemispheres; each candidate region is spherical in shape, and the region attributes corresponding to each candidate region include the fiber type, central coordinates, and sphere radius, wherein the fiber type includes projection fibers and association fibers; the left and right candidate regions of each candidate region pair have the same fiber type, the same central coordinates, and the same sphere radius; The selection rules for the research subjects require that the number of research subjects with normal and abnormal brain lymphoid system function be equal; and that the abnormalities in brain lymphoid system function include some or all of the following disease types: Parkinson's disease, epilepsy, cerebrovascular disease, and Alzheimer's disease. The object dataset includes multiple object data records; the object data records include object images and object functional levels; the object images are brain diffusion tensor images; the object functional levels include normal function, functional decline, and functional impairment. The first dataset includes multiple first data records; the first data records include a training tensor A. TR and label vector P TR The training tensor A TR From the left brain feature vector a L and the right brain feature vector a R Composition; the label vector P TR Includes three label probabilities The three label probabilities correspond to normal function, reduced function, and dysfunction, respectively. Only one of the three label probabilities is 1, and the other two are 0.

3. The processing method for the lymphatic system-like functional assessment model according to claim 1, characterized in that, The model input of the lymphoid system functional assessment model is used to receive the exponential feature tensor A, and the model output is used to output the prediction vector P. The lymphatic system functional assessment model includes a feature triage module, a left brain feature extraction module, a right brain feature extraction module, a feature fusion module, a fusion feature extraction module, and a prediction probability output module. The input terminal of the feature splitting module is connected to the input terminal of the model, and the first and second output terminals are respectively connected to the input terminals of the left brain and right brain feature extraction modules; the output terminals of the left brain and right brain feature extraction modules are respectively connected to the first and second input terminals of the feature fusion module; the output terminal of the feature fusion module is connected to the input terminal of the fused feature extraction module; the output terminal of the fused feature extraction module is connected to the input terminal of the prediction probability output module; and the output terminal of the prediction probability output module is connected to the model output terminal. The feature splitting module is used to split the left brain feature vector a of the exponential feature tensor A. L and the right brain feature vector a R Send them to the left brain and right brain feature extraction modules respectively; The left-brain feature extraction module is implemented based on an MLP model, which consists of two or more fully connected layers connected sequentially, with a ReLU activation function connected to the output of each fully connected layer. The left-brain feature extraction module is used to process the left-brain feature vector a. L Feature extraction is performed to obtain the corresponding feature vector H. L Send to the feature fusion module; The right-brain feature extraction module is implemented based on another MLP model. The current MLP model has the same model structure as the MLP model corresponding to the left-brain feature extraction module, but the parameters are independent. The right-brain feature extraction module is used to extract the right-brain feature vector a. R Feature extraction is performed to obtain the corresponding feature vector H. R Send the feature vector H to the feature fusion module. L H R The shape remains consistent; The feature fusion module performs vector concatenation on the feature vector H. L H R Perform concatenation and use the resulting concatenated vector as the corresponding fusion vector H. C Send to the fusion feature extraction module; The fusion feature extraction module is implemented based on another MLP model. The current MLP model consists of two or more fully connected layers connected sequentially, and the output of each fully connected layer is connected to a ReLU activation function. The fusion feature extraction module is used to process the fusion vector H. C The feature vector Z obtained by feature extraction is sent to the prediction probability output module; the vector length of the feature vector Z is 3. The prediction probability output module is used to perform three-class classification probability prediction based on the feature vector Z using the Softmax function to obtain the corresponding three prediction probabilities p. i The corresponding prediction vector P is then formed and output.

4. The processing method for the lymphatic system-like functional assessment model according to claim 2, characterized in that, The construction of the first dataset based on the ICBM-DTI-81 map, the candidate region set, and the object dataset specifically includes: A traversal is performed on all object data records in the object dataset; during this traversal, the currently traversed object data record is taken as the corresponding current record; and the object image of the current record is taken as the corresponding current image; the regions of interest (ROIs) of the candidate region set on the current image are marked according to the registration results of the ICBM-DTI-81 atlas and the current image; and the left and right brain ALPS indices are calculated based on all ROIs according to the DTI-ALPS index calculation principle, and the corresponding training tensor A is generated based on the calculation process data and calculation results. TR The system identifies the current functional level of the object; if the current functional level of the object is normal, it sets the corresponding three label probabilities. The values ​​are 1, 0, and 0; if the current functional level of the object is functionally degraded, then the corresponding three label probabilities are set. The values ​​are 0, 1, and 0; if the current functional level of the object is functional impairment, then the corresponding three label probabilities are set. The values ​​are 0, 0, and 1; and the three label probabilities obtained in this study form a corresponding label vector P. TR ; and the training tensor A corresponding to the current record. TR and the label vector P TR A corresponding first data record is formed; and at the end of this round of traversal, all the first data records obtained are combined to form the corresponding first dataset.

5. The processing method for the lymphatic system-like functional assessment model according to claim 2, characterized in that, The training of the lymphoid system functional assessment model based on the first dataset specifically includes: Step 51: Based on a preset first segmentation ratio, the first dataset is randomly divided into two sub-datasets, denoted as the first training set and the first evaluation set. Wherein, both the first training set and the first evaluation set are composed of multiple first data records; the ratio of the total number of records in the first training set and the first evaluation set satisfies the first segmentation ratio; Step 52: Perform a traversal of all the first data records in the first training set; and during this traversal, take the currently traversed first data record as the corresponding current training record; and set the training tensor A of the current training record as... TR The corresponding exponential feature tensor A is input into the lymphoid system functional evaluation model for prediction, and the prediction vector P output by this prediction is used as the corresponding current prediction vector; and the current prediction vector and the label vector P of the current training record are used together. TR Form a corresponding first prediction-label pair; and at the end of this round of traversal, input all the obtained first prediction-label pairs into the preset first model loss function to calculate the corresponding first loss value; The loss function of the first model is based on the multi-class cross-entropy loss function; Step 53: Identify whether the first loss value meets the preset first loss value range; if the first loss value meets the first loss value range, proceed to step 54; if the first loss value does not meet the first loss value range, perform a round of modulation on the model parameters of the lymphatic system function evaluation model based on the preset first model optimizer in the direction of minimizing the first model loss function, and return to step 52 to continue training when the first round of modulation ends. The first model optimizer includes at least the Adam optimizer and the SGD optimizer; Step 54: Perform a traversal of all the first data records in the first evaluation set; and during this traversal, take the currently traversed first data record as the corresponding current evaluation record; and set the training tensor A of the current evaluation record... TR The corresponding exponential feature tensor A is input into the lymphoid system functional assessment model for prediction, and the prediction vector P output by this prediction is used as the corresponding current prediction vector; and the current prediction vector and the label vector P recorded in the current assessment are used together. TR A corresponding second prediction-label pair is formed; and at the end of this round of traversal, the first accuracy, first precision, first recall and first F1 score are calculated based on all the obtained second prediction-label pairs. Step 55: Identify whether the first accuracy, first precision, first recall, and first F1 score all satisfy their respective first accuracy range, first precision range, first recall range, and first F1 score range; if not, return to step 52 to continue training; if yes, confirm that the model training is complete.

6. The processing method for the lymphatic system-like functional assessment model according to claim 2, characterized in that, The step of marking the regions of interest in the current image based on the registration results of the ICBM-DTI-81 map and the current image specifically includes: The ICBM-DTI-81 atlas is used as the corresponding standard brain template; the three-dimensional voxel space of the standard brain template is used as the corresponding first space; and the three-dimensional voxel space of the current image is used as the corresponding second space. First, the FLIRT tool of the FSL toolkit is used to map the current image from the second space to the first space using linear registration to obtain the corresponding mapped image. The linear mapping matrix from the second space to the first space is used as the corresponding current transformation matrix, and the inverse of the current transformation matrix is ​​used as the corresponding current inverse transformation matrix. Then, the FNIRT tool of the FSL toolkit is used to perform nonlinear registration on the standard brain template and the mapped image in the first space to obtain the corresponding nonlinear deformation field. The FSL toolkit is a toolkit for data analysis of diffusion tensor images, the FLIRT tool is the linear registration tool of the FSL toolkit, and the FNIRT tool is the nonlinear registration tool of the FSL toolkit. Based on the nonlinear deformation field, voxel-level deformation transformation is performed on the left and right candidate regions of each candidate region pair in the candidate region set to obtain the corresponding left and right deformed regions; and based on the current inverse transformation matrix, a linear mapping process from the first space to the second space is performed on each left and right deformed region to obtain the corresponding left and right mapped regions; and each left and right mapped region is taken as a corresponding region of interest; and voxel-level labeling is performed on each region of interest in the current image.

7. The processing method for the lymphatic system-like functional assessment model according to claim 2, characterized in that, The features The average x-axis diffusion rate of all left brain projection fiber regions, where each left brain projection fiber region is a region of interest corresponding to a left candidate region of the fiber type being projection fiber; the feature The average y-axis diffusion rate of all the left brain projection fiber regions; the z-axis direction of each left brain projection fiber region is the main projection fiber direction of the current fiber region, the x-axis direction is the perivascular space direction perpendicular to the main projection fiber direction of the current fiber region, and the y-axis direction is the perpendicular direction of the xz plane formed by the main projection fiber direction of the current fiber region and the perivascular space direction. The features The average x-axis diffusion rate of all left hemisphere associated fiber regions, where each left hemisphere associated fiber region is a region of interest corresponding to a left candidate region of associated fiber type; the feature The average z-axis diffusion rate of all the left brain association fiber regions; the y-axis direction of each left brain association fiber region is the main direction of the association fibers of the current fiber region, the x-axis direction is the direction of the perivascular space perpendicular to the main direction of the current association fibers, and the z-axis direction is the direction perpendicular to the xy plane formed by the main direction of the current association fibers and the direction of the perivascular space. The features The average x-axis diffusion rate of all right brain projection fiber regions, where each right brain projection fiber region is a region of interest corresponding to a right candidate region of the fiber type being projection fiber; the feature The average y-axis diffusion rate of all the right brain projection fiber regions; the z-axis direction of each right brain projection fiber region is the main projection fiber direction of the current fiber region, the x-axis direction is the perivascular space direction perpendicular to the main projection fiber direction of the current fiber region, and the y-axis direction is the perpendicular direction of the xz plane formed by the main projection fiber direction and the perivascular space direction. The features The average x-axis diffusion rate of all right hemisphere associated fiber regions, where each right hemisphere associated fiber region is a region of interest corresponding to a right candidate region of associated fiber type; the feature The average z-axis diffusion rate of all the right brain association fiber regions; the y-axis direction of each right brain association fiber region is the main direction of the association fibers of the current fiber region, the x-axis direction is the direction of the perivascular space perpendicular to the main direction of the current association fibers, and the z-axis direction is the direction perpendicular to the xy plane formed by the main direction of the current association fibers and the direction of the perivascular space. The index ALPS L ALPS R for:

8. The processing method for the lymphatic system-like functional assessment model according to claim 7, characterized in that, The method of calculating the DTI-ALPS index based on the principle of calculating the ALPS index of the left and right hemispheres according to all regions of interest, and generating the corresponding index feature tensor A based on the calculation process data and calculation results, specifically includes: Each candidate region of interest corresponding to a left brain projection fiber is denoted as a corresponding left brain projection fiber region. Based on the x, y, and z axes of the current left brain projection fiber region, the x and y axis diffusion rates of each voxel point within the current left brain projection fiber region are calculated to obtain the corresponding voxel x-axis diffusion rate and voxel y-axis diffusion rate. The sum of the x-axis diffusion rates of all voxels in the current left brain projection fiber region is taken as the corresponding fiber region x-axis diffusion rate, and the sum of the y-axis diffusion rates of all voxels is taken as the corresponding fiber region y-axis diffusion rate. The average value of the x-axis diffusion rates of all left brain projection fiber regions is taken as the corresponding feature. The average y-axis diffusion rate of all the left brain projection fiber areas is taken as the corresponding feature. Each candidate region of interest corresponding to a right-side candidate region of projection fiber type is denoted as a corresponding right brain projection fiber region. Based on the x, y, and z axes of the current right brain projection fiber region, the x and y axis diffusion rates of each voxel point within the current right brain projection fiber region are calculated to obtain the corresponding voxel x-axis diffusion rate and voxel y-axis diffusion rate. The sum of the x-axis diffusion rates of all voxels in the current right brain projection fiber region is taken as the corresponding fiber region x-axis diffusion rate, and the sum of the y-axis diffusion rates of all voxels is taken as the corresponding fiber region y-axis diffusion rate. The average value of the x-axis diffusion rates of all fiber regions in the right brain projection fiber region is taken as the corresponding feature. The average y-axis diffusion rate of all right brain projection fiber areas is taken as the corresponding feature. Each candidate region of interest corresponding to a left-side fiber of type synoptic fiber is denoted as a corresponding left-brain synoptic fiber region. Based on the x, y, and z axes of the current left-brain synoptic fiber region, the x and z axis diffusion rates of each voxel point within the current left-brain synoptic fiber region are calculated to obtain the corresponding voxel x-axis diffusion rate and voxel z-axis diffusion rate. The sum of the x-axis diffusion rates of all voxels in the current left-brain synoptic fiber region is taken as the corresponding fiber region x-axis diffusion rate, and the sum of the z-axis diffusion rates of all voxels is taken as the corresponding fiber region z-axis diffusion rate. The average value of the x-axis diffusion rates of all fiber regions in the left-brain synoptic fiber region is taken as the corresponding feature. The average z-axis diffusion rate of all the aforementioned left brain association fiber regions is taken as the corresponding feature. Each candidate region of interest corresponding to a right-side fiber of type synoptic fiber is denoted as a corresponding right-brain synoptic fiber region. Based on the x, y, and z axes of the current right-brain synoptic fiber region, the x and z axis diffusion rates of each voxel point within the current right-brain synoptic fiber region are calculated to obtain the corresponding voxel x-axis diffusion rate and voxel z-axis diffusion rate. The sum of the x-axis diffusion rates of all voxels in the current right-brain synoptic fiber region is taken as the corresponding fiber region x-axis diffusion rate, and the sum of the z-axis diffusion rates of all voxels is taken as the corresponding fiber region z-axis diffusion rate. The average value of the x-axis diffusion rates of all fiber regions in the right-brain synoptic fiber region is taken as the corresponding feature. The average z-axis diffusion rate of all the right brain association fiber regions is taken as the corresponding feature. And based on the DTI-ALPS index calculation principle, and according to the characteristics obtained this time... The features The features and the features Calculate the corresponding exponent ALPS L ; and based on the DTI-ALPS index calculation principle, according to the characteristics obtained this time. The features The features and the features Calculate the corresponding exponent ALPS R ; and the features obtained in this study The features The features The features and the index ALPS L Form a corresponding left brain feature vector a L ; and the features obtained in this study The features The features The features and the index ALPS R Form a corresponding right brain feature vector a R ; And the left brain feature vector a obtained in this study L and the right brain feature vector a R This forms a corresponding exponential feature tensor A.

9. An apparatus for performing a processing method for a lymphoid system functional assessment model according to any one of claims 1-8, characterized in that, The device includes: a candidate region customization module, a model dataset preparation module, a model construction module, a model training module, and a lymphatic system-like functional evaluation module; The candidate region customization module is used to select multiple fiber bundle regions as candidate region sets in the ICBM-DTI-81 white matter map according to preset candidate region customization rules. The model dataset preparation module is used to select multiple research subjects with normal and abnormal brain lymphoid system functions according to preset research subject selection rules; and a preset expert group conducts a three-level functional assessment of the brain lymphoid system of each research subject, classifying it as normal, diminished, or impaired; and acquires brain images of each research subject using diffusion tensor imaging technology; and constructs an object dataset based on the functional level assessment results and diffusion tensor images of all research subjects; and constructs a first dataset based on the ICBM-DTI-81 atlas, the candidate region set, and the object dataset. The model building module is used to construct a deep learning model for predicting the functional level of the brain's lymphoid system as a lymphoid system function assessment model. The lymphoid system function assessment model is used to predict the functional level of the brain's lymphoid system based on the exponential feature tensor A input to the model and output the corresponding prediction vector P. The exponential feature tensor A consists of left and right brain feature vectors a. L a R Composition; the left brain feature vector a L Including features feature feature feature And the index ALPS L The right brain feature vector a R Including features feature feature feature And the index ALPS R The prediction vector P includes three prediction probabilities p. i 1 ≤ index i ≤ 3, the predicted probability p i＝1、2、3 These correspond to normal function, diminished function, and impaired function, respectively, with the sum of the three predicted probabilities being 1. The model training module is used to train the lymphatic system functional evaluation model based on the first dataset; The lymphoid system functional assessment module is used to receive, after model training is completed, a brain diffusion tensor image of any subject input by the user as the current image; and to mark the regions of interest (ROIs) of the candidate region set on the current image according to the registration results of the ICBM-DTI-81 atlas and the current image; and to calculate the left and right brain ALPS indices according to the DTI-ALPS index calculation principle based on all ROIs, and generate the corresponding index feature tensor A based on the calculation process data and calculation results; and to input the index feature tensor A into the lymphoid system functional assessment model for prediction to obtain the corresponding prediction vector P and feed it back to the current user.

10. An electronic device, characterized in that, include: Memory, processor, and transceiver; The processor is configured to be coupled to the memory, read and execute instructions in the memory to implement the method according to any one of claims 1-8; The transceiver is coupled to the processor, and the processor controls the transceiver to send and receive messages.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that, when executed by a computer, cause the computer to perform the method described in any one of claims 1-8.

Images