Non-invasive brain lymphatic network evaluation method and system based on magnetic resonance imaging

Abstract

The application discloses a non-invasive brain lymphoid network evaluation method and system based on magnetic resonance imaging, and relates to the field of medical image processing.The system comprises a data acquisition module which acquires 3D-T1WI images, 3D-T2WI images, DWI images and DTI images; a data preprocessing module which pre-processes the images; and a multi-dimensional function analysis module which is used for determining multi-dimensional parameters of perivascular spaces according to the pre-processed 3D-T1WI images and the pre-processed 3D-T2WI images; separating perfusion and diffusion components of the pre-processed DWI images to obtain intra-voxel incoherent motion imaging parameters, and calculating free water fractions; and calculating a perivascular space diffusion tensor imaging index based on the pre-processed DTI images; and an evaluation and analysis module which evaluates according to the output results of the multi-dimensional function analysis module, so that the functional state of the brain lymphoid system can be evaluated non-invasively, comprehensively and systematically.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of medical image processing and neuroscience, and particularly relates to a non-invasive brain lymphoid network evaluation method and system based on magnetic resonance imaging. BACKGROUND

[0002] As the core pathway for removing metabolic waste and maintaining fluid homeostasis in the brain, the dysfunction of the lymphoid system is considered to be closely related to cognitive impairment, cerebral small vessel disease and other neurodegenerative diseases. Its main pathway is that the cerebrospinal fluid penetrates into the deep brain parenchyma through the perivascular space of the cerebral artery, exchanges with the interstitial fluid in the brain tissue interstitium through the regulation of the astrocyte foot process, and the metabolic products after the exchange are discharged outside the brain through the perivascular space of the vein. In the field of lymphoid system function research, imaging evaluation methods play a crucial role. However, the existing imaging evaluation methods of brain lymphoid system function are often limited to independent evaluation of a single stage of the above-mentioned pathway, lacking systematic and quantitative evaluation of the inflow pathway of the perivascular space of the artery, the exchange of interstitial fluid, and the outflow pathway of the perivascular space of the vein, which restricts the clinical application and the promotion of disease classification diagnosis. Therefore, it is urgent to establish a non-invasive, quantitative and comprehensive method for evaluating the function of the brain lymphoid system based on multi-modal magnetic resonance imaging. SUMMARY

[0003] The purpose of the present application is to provide a non-invasive brain lymphoid network evaluation method and system based on magnetic resonance imaging, which can non-invasively, comprehensively and systematically evaluate the functional status of the brain lymphoid system from the dynamic flow pathway of the liquid inflow of the perivascular space of the brain artery, the liquid exchange of the interstitial fluid, and the liquid outflow of the perivascular space of the brain vein.

[0004] To achieve the above-mentioned purpose, the present application provides the following solutions:

[0005] In a first aspect, the present application provides a non-invasive brain lymphoid network evaluation system based on magnetic resonance imaging, comprising:

[0006] A data acquisition module for acquiring 3D-T1WI images, 3D-T2WI images, DWI images and DTI images of a user;

[0007] A data preprocessing module for preprocessing the 3D-T1WI images, 3D-T2WI images, DWI images and DTI images to obtain preprocessed 3D-T1WI images, preprocessed 3D-T2WI images, preprocessed DWI images and preprocessed DTI images;

[0008] The multi-dimensional function analysis module includes an arterial perivascular space liquid inflow analysis unit, an interstitial fluid exchange analysis unit and a venous perivascular space liquid outflow analysis unit;

[0009] The cerebral arterial perivascular space fluid inflow analysis unit is configured to determine a perivascular space (PVS) multi-dimensional parameter based on the preprocessed 3D-T1WI image and the preprocessed 3D-T2WI image.

[0010] The tissue space fluid exchange analysis unit is configured to separate perfusion and diffusion components of the preprocessed DWI image by using an intravoxel incoherent motion imaging (IVIM) model to obtain IVIM model parameters, and perform free water imaging model analysis based on the preprocessed DTI image to obtain a free water fraction (FWF) parameter; the IVIM model parameters include a blood perfusion fraction, a pure diffusion coefficient, and a pseudo-diffusion coefficient.

[0011] The venous perivascular space fluid outflow analysis unit is configured to calculate an analysis along the perivascular space (ALPS) index based on the preprocessed DTI image.

[0012] The evaluation analysis module is configured to evaluate the functional state of the brain lymphoid system based on the PVS multi-dimensional parameter, the IVIM model parameter, the FWF parameter, and the ALPS index to obtain a brain lymphoid system function evaluation result of the user.

[0013] Optionally, when the PVS multi-dimensional parameter is PVS volume information, the cerebral arterial perivascular space fluid inflow analysis unit is configured to:

[0014] The PVS binary segmentation mask includes PVS volume information of each voxel.

[0015] The PVS binary segmentation mask is subjected to connected domain analysis to determine the number of voxels of each connected domain, and the PVS volume information of each brain region is calculated based on the number of voxels of each connected domain and the PVS volume information of a single voxel.

[0016] When the PVS multi-dimensional parameter is a spatial network feature of the PVS, the cerebral arterial perivascular space fluid inflow analysis unit is configured to:

[0017] The PVS binary segmentation mask is obtained by using the trained image segmentation model with the pre-processed 3D-T1WI image and the pre-processed 3D-T2WI image as inputs; the PVS binary segmentation mask comprises PVS volume information of each voxel;

[0018] The PVS centroid coordinates are determined according to the PVS binary segmentation mask;

[0019] A spatial adjacency network with the PVS centroid coordinates as nodes is constructed, and an edge is established between two nodes in the spatial adjacency network when a spatial connection radius between the two nodes is within a set threshold range, forming an undirected graph.

[0020] The spatial network features of the PVS are calculated based on the undirected graph; the spatial network features of the PVS comprise node connection degree, clustering coefficient and average shortest path length.

[0021] Optionally, the trained image segmentation model is a U-Net model or an nnU-Net model.

[0022] Optionally, the data preprocessing module is configured to:

[0023] The DWI image and the DTI image are subjected to first preprocessing to obtain a first pre-processed DWI image and a first pre-processed DTI image; the first preprocessing comprises denoising, motion correction, eddy current correction and magnetic susceptibility artifact correction.

[0024] The 3D-T1WI image and the 3D-T2WI image are subjected to non-uniform field correction to obtain a corrected 3D-T1WI image and a corrected 3D-T2WI image.

[0025] The first pre-processed DWI image, the first pre-processed DTI image and the corrected 3D-T2WI image are registered to the space of the corrected 3D-T1WI image to obtain a registered DWI image, a registered DTI image and a registered 3D-T2WI image.

[0026] The corrected 3D-T1WI image, the registered 3D-T2WI image, the registered DWI image and the registered DTI image are subjected to normalization processing to obtain a pre-processed 3D-T1WI image, a pre-processed 3D-T2WI image, a pre-processed DWI image and a pre-processed DTI image.

[0027] Optionally, the interstitial fluid exchange analysis unit is further configured to:

[0028] For each tissue interstitial fluid exchange performance parameter of each brain region, a mean value calculation is performed on the tissue interstitial fluid exchange performance parameters of all voxels in the brain region to obtain the tissue interstitial fluid exchange parameter value of the brain region; the tissue interstitial fluid exchange performance parameters are blood perfusion fraction, pure diffusion coefficient, pseudo-diffusion coefficient and FWF parameter;

[0029] A mean value calculation is performed on the tissue interstitial fluid exchange parameter values of all brain regions to obtain a whole brain tissue interstitial fluid exchange parameter value.

[0030] Optionally, the calculation formula of the ALPS index is as follows:

[0031] ;

[0032] Wherein, represents the ALPS index, represents the average value of the axial diffusion rate of the projection fiber region in the brain region, represents the average value of the axial diffusion rate of the association fiber region in the brain region, represents the average value of the axial diffusion rate of the projection fiber region, represents the average value of the axial diffusion rate of the association fiber region.

[0033] Optionally, in the aspect of evaluating the functional state of the brain lymphoid system according to the PVS multi-dimensional parameter, the IVIM model parameter, the FWF parameter and the ALPS index to obtain the brain lymphoid system function evaluation result of the user, the evaluation analysis module is used for:

[0034] Inputting the PVS multi-dimensional parameter, the IVIM model parameter, the FWF parameter and the ALPS index into the trained machine learning model to evaluate the functional state of the brain lymphoid system to obtain the brain lymphoid system function evaluation result of the user; the brain lymphoid system function evaluation result includes perivascular space inflow obstruction type, tissue interstitial exchange obstruction type, perivascular space outflow obstruction type, mixed type and normal.

[0035] Optionally, the machine learning model is a random forest model, a support vector machine or a deep learning network.

[0036] Optionally, the non-invasive brain lymphoid network evaluation system based on magnetic resonance imaging further comprises a report generation module; the report generation module is used for generating a comprehensive evaluation report; the comprehensive evaluation report includes the output result of the multi-dimensional function analysis module and the brain lymphoid system function evaluation result of the user.

[0037] ​​​​In a second aspect, the application provides a non-invasive brain lymphoid network evaluation method based on magnetic resonance imaging, comprising:

[0038] acquiring 3D-T1WI images, 3D-T2WI images, DWI images and DTI images of a user;

[0039] preprocessing the 3D-T1WI images, 3D-T2WI images, DWI images and DTI images to obtain preprocessed 3D-T1WI images, preprocessed 3D-T2WI images, preprocessed DWI images and preprocessed DTI images;

[0040] determining PVS multi-dimensional parameters according to the preprocessed 3D-T1WI images and the preprocessed 3D-T2WI images;

[0041] separating perfusion and diffusion components of the preprocessed DWI images by using an IVIM model to obtain IVIM model parameters; the IVIM model parameters include a vascular perfusion fraction, a pure diffusion coefficient and a pseudo-diffusion coefficient; and performing free water imaging model analysis according to the preprocessed DTI images to obtain FWF parameters;

[0042] calculating an ALPS index according to the preprocessed DTI images;

[0043] evaluating a functional state of a brain lymphoid system according to the PVS multi-dimensional parameters, the IVIM model parameters, the FWF parameters and the ALPS index to obtain a brain lymphoid system functional evaluation result of the user.

[0044] According to the specific embodiments provided in the application, the following technical effects are disclosed:

[0045] The application provides a non-invasive brain lymphoid network evaluation method and system based on magnetic resonance imaging, and the system comprises the following steps: a data acquisition module acquires 3D-T1WI images, 3D-T2WI images, DWI images and DTI images of a user; a data preprocessing module pre-processes the images; a multi-dimensional functional analysis module comprises the following steps: determining PVS multi-dimensional parameters according to the pre-processed 3D-T1WI images and the pre-processed 3D-T2WI images, so as to reflect the liquid inflow performance of the perivascular space; separating perfusion and diffusion components of the pre-processed DWI images to obtain IVIM model parameters, and analyzing FWF parameters of the pre-processed DTI images, so as to reflect the exchange performance of interstitial fluid in the tissue space by using the IVIM model parameters and the FWF parameters; calculating ALPS indexes in the pre-processed DTI images, so as to reflect the liquid outflow performance of the perivascular space; and an evaluation and analysis module evaluates according to the PVS multi-dimensional parameters, the IVIM model parameters, the FWF parameters and the ALPS indexes. The application realizes automatic extraction and quantification of the arterial PVS multi-dimensional parameters by joint multi-channel deep learning segmentation, evaluates the liquid exchange performance of the tissue space by combining the IVIM model and the free water imaging model, and quantitatively reflects the outflow performance of the perivascular space of the venous vessel by using the ALPS index. By comprehensively analyzing the three types of functional indexes, systematic, non-invasive, comprehensive and quantitative evaluation of the liquid inflow of the perivascular space of the arterial blood vessel, the liquid exchange of the tissue space and the liquid outflow of the perivascular space of the venous blood vessel is realized. BRIEF DESCRIPTION OF DRAWINGS

[0046] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed in the embodiments. Obviously, the drawings in the following description only constitute some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without any creative effort.

[0047] Figure 1 A functional module schematic diagram of a non-invasive brain lymphoid network evaluation system based on magnetic resonance imaging provided for Embodiment 1 of the present application;

[0048] Figure 2 An application scenario diagram of a non-invasive brain lymphoid network evaluation method based on magnetic resonance imaging provided for Embodiment 2 of the present application;

[0049] Figure 3 A flowchart of a non-invasive brain lymphoid network evaluation method based on magnetic resonance imaging provided for Embodiment 2 of the present application. DETAILED DESCRIPTION

[0050] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0051] Dynamic contrast-enhanced magnetic resonance imaging and static contrast-enhanced magnetic resonance imaging are common imaging methods for exploring the function of the brain lymphoid system, and can provide high-resolution images, but rely on the use of contrast agents, which not only increases the burden on patients, but also may cause safety problems. Arterial spin labeling imaging does not require the use of contrast agents, but the resolution is not ideal. In addition, susceptibility-weighted imaging and diffusion spectrum imaging can visualize the function of the brain lymphoid system under certain conditions, but they still have some limitations in practical application.

[0052] Therefore, it is urgent to develop a non-invasive brain lymphatic network evaluation method and system based on magnetic resonance imaging, which can realize comprehensive and systematic evaluation of the brain lymphatic system by integrating PVS segmentation analysis, IVIM and free water imaging model, and DTI-ALPS model, and provide scientific basis for early diagnosis and pathological mechanism research of nervous system diseases. The present application can overcome the limitations of the prior art, provide more comprehensive and accurate evaluation of the function of the brain lymphatic system, and help to promote early diagnosis and personalized treatment of nervous system diseases.

[0053] The above-mentioned purposes, features and advantages of the present application will be more apparent and easy to understand. The present application will be further described in detail below with reference to the drawings and specific embodiments.

[0054] Embodiment 1.

[0055] As shown in Figure 1 The present embodiment provides a non-invasive brain lymphatic network evaluation system based on magnetic resonance imaging, which includes a data acquisition module, a data preprocessing module, a multi-dimensional function analysis module, an evaluation analysis module and a report generation module.

[0056] (1) The data acquisition module is used to acquire the 3D-T1WI image, 3D-T2WI image, DWI image and DTI image of the user.

[0057] The data acquisition module of the present application is realized based on clinical magnetic resonance imaging technology. First, the original Dicom data of the user's head magnetic resonance scan is acquired, and the specific sequences include: 3D-T1WI image, 3D-T2WI image, DWI image and DTI image, and the acquisition requirements of each image are as follows: 3D-T1WI image and 3D-T2WI image: used for brain high-resolution structure imaging, multi-modal input of PVS segmentation, and the spatial resolution is recommended to be 1mm³. DWI image (multi-b value): collect b=0, 20, 50, 100, 200, 500, 1000 s / mm² and other multi-b value sequences, which is convenient for IVIM model analysis. DTI image: at least 30 scanning directions, b=0 and b=1000 s / mm², and the resolution is 2 mm³, which is convenient for ALPS index and free water imaging model fitting.

[0058] The data preprocessing module is used for preprocessing the 3D-T1WI image, the 3D-T2WI image, the DWI image and the DTI image, to obtain the preprocessed 3D-T1WI image, the preprocessed 3D-T2WI image, the preprocessed DWI image and the preprocessed DTI image.

[0059] Before preprocessing the 3D-T1WI image, the 3D-T2WI image, the DWI image and the DTI image, the 3D-T1WI image, the 3D-T2WI image, the DWI image and the DTI image are also subjected to image format conversion, specifically using dcm2niix and other tools to convert DICOM format in batches to NIFTI (.nii.gz), which is convenient for subsequent processing.

[0060] The data preprocessing module is used for: performing first preprocessing on the DWI image and the DTI image to obtain first preprocessed DWI image and first preprocessed DTI image; the first preprocessing includes denoising, motion correction, eddy current correction and magnetic susceptibility artifact correction; performing non-uniform field correction processing on the 3D-T1WI image and the 3D-T2WI image to obtain corrected 3D-T1WI image and corrected 3D-T2WI image; registering the first preprocessed DWI image, the first preprocessed DTI image and the corrected 3D-T2WI image to the space of the corrected 3D-T1WI image to obtain registered DWI image, registered DTI image and registered 3D-T2WI image; performing normalization processing on the corrected 3D-T1WI image, the registered 3D-T2WI image, the registered DWI image and the registered DTI image to obtain the preprocessed 3D-T1WI image, the preprocessed 3D-T2WI image, the preprocessed DWI image and the preprocessed DTI image.

[0061] 1) Head motion and distortion correction: denoising, motion correction, eddy current correction and susceptibility artifact correction are performed on DWI images and DTI images by the tool module of FSL to obtain first preprocessed DWI images and first preprocessed DTI images; if mild motion artifacts are found in the 3D-T1WI images and 3D-T2WI images, non-uniform field correction can be performed by using N4 Bias Field Correction and other tools to obtain corrected 3D-T1WI images and corrected 3D-T2WI images.

[0062] 2) Spatial registration: the corrected 3D-T2WI images are registered to the space of the corrected 3D-T1WI images by using FSL FLIRT / FNIRT, rigid registration (6 degrees of freedom) is used for preliminary alignment, and then FNIRT is used for nonlinear registration to obtain registered 3D-T2WI images.

[0063] The b0 images of the DTI and DWI data also need to be registered to the space of the corrected 3D-T1WI images to ensure the spatial consistency of the subsequent multi-modal fusion analysis. Therefore, the first preprocessed DWI images and the first preprocessed DTI images are registered to the space of the corrected 3D-T1WI images to obtain registered DWI images and registered DTI images.

[0064] 3) Intensity normalization: intensity normalization (z-score) is performed on the corrected 3D-T1WI images, the registered 3D-T2WI images, the registered DWI images and the registered DTI images to reduce batch effects and signal non-uniformity, all image signals are located on a unified scale, which provides standardized input for subsequent deep learning segmentation and helps stable training of deep learning models and multi-modal feature fusion.

[0065] (Three) Multi-dimensional functional analysis module includes cerebral arterial perivascular space liquid inflow analysis unit, tissue space liquid exchange analysis unit and venous perivascular space liquid outflow analysis unit.

[0066] (1) Cerebral arterial perivascular space liquid inflow analysis unit, for: determining PVS multi-dimensional parameters according to preprocessed 3D-T1WI images and preprocessed 3D-T2WI images.

[0067] Further, when the PVS multi-dimensional parameter is PVS volume information, the brain perivascular space fluid inflow analysis unit is configured to: take the preprocessed 3D-T1WI image and the preprocessed 3D-T2WI image as input, and use the trained image segmentation model to predict a PVS binary segmentation mask; the PVS binary segmentation mask includes PVS volume information of each voxel; perform connected component analysis on the PVS binary segmentation mask, determine the number of voxels in each connected component, and calculate PVS volume information of each brain region according to the number of voxels in each connected component and the PVS volume information of each voxel.

[0068] After automatic segmentation of PVS, in the obtained PVS binary segmentation mask, the voxel value of 1 represents PVS, and the voxel value of 0 represents non-PVS brain tissue.

[0069] The application uses a connected component analysis algorithm to independently analyze and quantify the PVS binary segmentation mask. All “1” voxels in the PVS binary segmentation mask are connected together, and these voxels are directly or indirectly connected together through a certain neighborhood (such as a six-neighborhood, an eighteen-neighborhood, or a twenty-six-neighborhood). Each connected component can be understood as an independent PVS.

[0070] Specifically, for the PVS binary segmentation mask, a connected component analysis algorithm is applied to automatically label each independent connected component, and each connected component is assigned a unique label number, so that all PVSs are separated one by one. The number of voxels contained in each connected component is counted, that is, the volume of each independent PVS (number of voxels x volume of a single voxel). In combination with a standard brain region partition template, each connected component can be assigned to a different brain region, and the number and total volume of PVSs in each brain region can be further counted. In a specific example, the connected component analysis algorithm can use a 3D version of flood fill and the measure.label function of the scikit-image library.

[0071] Suppose that the PVS binary segmentation mask obtained by automatic segmentation has 500 connected components. Among them, the first connected component has 100 voxels, the second connected component has 40 voxels, and so on. If the volume of each voxel is 1 mm3, then the volume of the first PVS region is 100 mm3, and the volume of the second PVS region is 40 mm3. Finally, the volume distribution of all PVSs and the PVS volume information of each brain region can be obtained.

[0072] When the PVS multidimensional parameter is the spatial network feature of the PVS, the cerebral arterial perivascular space fluid inflow analysis unit is configured to: take the preprocessed 3D-T1WI image and the preprocessed 3D-T2WI image as inputs, and use a trained image segmentation model to predict a PVS binary segmentation mask; the PVS binary segmentation mask includes PVS volume information of each voxel; determine the PVS centroid coordinates according to the PVS binary segmentation mask; construct a spatial adjacency network with the PVS centroid coordinates as nodes, and establish an edge between two nodes in the spatial adjacency network when the spatial connection radius between the two nodes is within a set threshold range, forming an undirected graph; and calculate the spatial network feature of the PVS based on the undirected graph; the spatial network feature of the PVS includes node connectivity, clustering coefficient, and average shortest path length.

[0073] The cerebral arterial perivascular space fluid inflow analysis unit described above is used for automatic PVS segmentation and multidimensional quantification in combination with modalities, and through multi-channel fusion of the preprocessed 3D-T1WI image and the preprocessed 3D-T2WI image, a deep learning framework is used to automatically extract the PVS volume, number, spatial distribution, and topological structure, and the system reflects the performance of the arterial perivascular space fluid inflow.

[0074] The preprocessed 3D-T1WI image and the preprocessed 3D-T2WI image are taken as dual-channel inputs and input to a trained image segmentation model (the image segmentation model can be an automated 3D U-Net framework based on PyTorch). Before this, the image segmentation model also needs to be trained using a sample set to obtain the trained image segmentation model, and the sample set includes a plurality of sample image groups and a sample PVS binary segmentation mask corresponding to each sample image group. The model input size is 64x64x64, the optimizer is Adam (initial learning rate 1e-4), and the loss function uses a combination of Dice loss and cross-entropy loss. The training data in the sample set is data subjected to data augmentation (rotation, mirroring, intensity disturbance, etc.) to improve the model generalization capability. During the training process, the artificially labeled PVS (sample PVS binary segmentation mask) is used as the gold standard, the training epoch is set to 300, five-fold cross-validation is used, the target segmentation Dice coefficient is greater than or equal to 0.85, that is, the Dice coefficient greater than or equal to 0.85 is used as the segmentation accuracy rate standard. After the training is completed, the trained image segmentation model automatically predicts the fusion image (the fusion image includes the preprocessed 3D-T1WI image and the preprocessed 3D-T2WI image) and outputs the PVS binary segmentation mask.

[0075] The trained image segmentation model is a U-Net model or an nnU-Net model, etc.

[0076] The perivascular space fluid inflow analysis unit is also used for post-processing segmentation and quantitative index extraction: the PVS binary segmentation mask obtained by segmentation is used to extract all independent PVS regions by using connected component analysis (based on the scipy and scikit-image libraries of Python), the number of voxels of each connected domain is counted, and the total volume of PVS is calculated by combining the single voxel volume (derived from the NIFTI header information). The total number is the number of connected domains. The statistical results can be regionally attributed according to the standard brain region template (MNI template partition), and the PVS is assigned to the frontal lobe, parietal lobe, basal ganglia, brainstem and other regions, thereby providing a basis for subsequent hierarchical analysis.

[0077] The perivascular space fluid inflow analysis unit can also obtain the spatial network characteristics of PVS through spatial clustering and topological structure analysis: the centroid coordinates of all PVSs are automatically extracted from the PVS binary segmentation mask (segmentation result of the trained image segmentation model). A spatial adjacency network with PVS centroids as nodes is constructed using NetworkX (an open-source graph network analysis package of Python), and edges are automatically added by setting a spatial connection radius (such as setting a threshold range of 3-5 mm) to form an undirected graph. Then, topological parameters such as node connectivity, clustering coefficient and average shortest path length are calculated to comprehensively evaluate the spatial network characteristics of PVS. The spatial clustering can be further evaluated and analyzed by using Ripley's K function to describe the uniformity and aggregation degree of PVS distribution in different brain regions. Visualization is performed by using matplotlib, seaborn and other Python libraries to generate PVS distribution and topological feature maps.

[0078] (2) The interstitial fluid exchange analysis unit is used to: separate perfusion and diffusion components from the preprocessed DWI image using the IVIM model to obtain IVIM model parameters, the IVIM model parameters including vascular perfusion fraction, pure diffusion coefficient and pseudo-diffusion coefficient; and analyze the preprocessed DTI image according to the free water imaging model to obtain FWF parameters.

[0079] The interstitial fluid exchange analysis unit is also used to: for each brain region and each interstitial fluid exchange performance parameter, the mean value of the interstitial fluid exchange performance parameters of all voxels in the brain region is calculated to obtain the interstitial fluid exchange parameter value of the brain region; the interstitial fluid exchange performance parameters are the vascular perfusion fraction, the pure diffusion coefficient, the pseudo-diffusion coefficient and the FWF parameter; and the mean value of the interstitial fluid exchange parameter values of all brain regions is calculated to obtain the whole brain interstitial fluid exchange parameter value.

[0080] The preprocessed DWI image (DWI multi-b value data) is used for IVIM model fitting. Matlab custom least squares method is used for fitting, and f, D, Parametric maps were generated to extract the mean, standard deviation, and histogram of the blood perfusion fraction, pure diffusion coefficient, and pseudo-diffusion coefficient in the whole brain and brain regions. The specific parameter calculation formula is as follows:

[0081] ;

[0082] wherein, is the DWI signal intensity when b is ; b is the diffusion sensitivity factor; S(0) is the DWI signal intensity when b=0 (reference signal when no diffusion gradient is added); f is the blood perfusion fraction, and D is the pure diffusion coefficient (true diffusion coefficient), is the pseudo-diffusion coefficient. represents the natural exponential function.

[0083] Free water generally refers to extracellular water, which can be directly reflected by the FWF parameter after data processing. Bound water has no direct reflection parameter, which can be indirectly inferred from the processed FA / MD. The free water distribution map is a pseudo-color display obtained by superimposing the FWF parameter value in the brain tissue. The analysis here is based on brain tissue voxels. The FWF parameter value of each voxel can be obtained in the obtained brain parenchyma, and then the spatial distribution can be seen. The average value can also be reported according to different brain regions.

[0084] The free water imaging parameter extraction process is: using Free-water imaging Toolbox, setting the fitting accuracy convergence threshold value 0.001, calculating the FWF parameter in each voxel, and then calculating the FWF parameter mean value and distribution characteristics of each brain region and the whole brain according to the FWF parameter in each voxel. The specific calculation formula is as follows:

[0085] S(b2)=S(0) [(1- ) exp(-b2 + exp(-b2 )];

[0086] wherein, S(b2) is the diffusion weighted imaging signal intensity when b is b2; b is the diffusion sensitivity factor; S(0) is the diffusion weighted imaging signal intensity when b=0; D_tissue is the apparent diffusion coefficient of water molecules in the tissue; D_water is the apparent diffusion coefficient of free water molecules; and FWF is used to reflect the interstitial fluid exchange capacity.

[0087] The mean of the blood perfusion fraction, the mean of the pure diffusion coefficient, the mean of the pseudo-diffusion coefficient, and the mean of the FWF parameter of the brain region are the interstitial fluid exchange parameter values of the brain region, and the mean of the blood perfusion fraction, the mean of the pure diffusion coefficient, the mean of the pseudo-diffusion coefficient, and the mean of the FWF parameter of the whole brain are the interstitial fluid exchange parameter values of the whole brain.

[0088] The interstitial fluid exchange analysis unit is also used for multi-parameter joint analysis, in particular: f, D, , FWF and other parameters are jointly modeled to analyze their differences and distribution characteristics in healthy controls and patients.

[0089] (3) A venous perivascular interstitial fluid outflow analysis unit is used to calculate the ALPS index based on the pre-processed DTI image.

[0090] Based on the diffusion tensor imaging technology, the ALPS index is calculated to evaluate the liquid diffusion function of the brain lymphoid system. By analyzing the diffusion anisotropy of water molecules in the perivascular space, the liquid clearance efficiency of the brain lymphoid system is quantified.

[0091] The ALPS index is calculated using a verified semi-automatic process. The FA map of all individuals is first registered to the FMRIB58_FA standard space through linear and nonlinear transformation. For the placement of the region of interest (ROI), a subject with the least deformation is selected. Using the color-coded FA map of the subject, 5mm-diameter spherical ROIs are placed in the projection area and commissure area at the body level of the bilateral lateral ventricles. Then the obtained ROIs are registered with the same FA template. The ROI positions of each participant are visually confirmed. If necessary, manual correction is performed by slightly moving the ROIs.

[0092] In the projection fiber region, the main fibers extend along the z-axis direction, while the x-axis and y-axis are perpendicular to the main fibers. In addition, in the commissure fiber region, the main fibers extend along the y-axis direction, and the x-axis and z-axis are perpendicular to the main fibers. Therefore, the ALPS index is obtained by the ratio of the average x-axis diffusivity of the projection fiber region to the average x-axis diffusivity of the commissure fiber region, divided by the ratio of the average y-axis diffusivity of the projection fiber region to the average z-axis diffusivity of the commissure fiber region.

[0093] The ALPS index calculation process is as follows: (1) Tensor fitting: The diffusion tensor is calculated based on the preprocessed DTI image using linear least squares or weighted least squares method to generate three eigenvalues ​​and three eigenvectors for each voxel; (2) Calculation of ALPS index: Regions of interest (ROI) are determined along three specific directions (X, Y, Z axes) of white matter fibers. A sphere with a diameter of 5 mm is selected as the ROI. The ADC values ​​on the fiber directions (such as x-axis, y-axis, z-axis) are extracted respectively, and the ALPS index is calculated using the following formula:

[0094] ;

[0095] in, Indicates the ALPS index. Indicates the regions of projection fibers in the brain. Average axial diffusivity For the brain region of the associated fiber region Average axial diffusivity For the projection fiber area Average axial diffusivity For the joint fiber region Average axial diffusivity.

[0096] (iv) Evaluation and analysis module, used to: evaluate the functional status of the brain lymphatic system based on PVS multidimensional parameters, IVIM model parameters, FWF parameters and ALPS index, and obtain the user's brain lymphatic system functional evaluation results.

[0097] In assessing the functional status of the brain-like lymphatic system based on PVS multidimensional parameters, IVIM model parameters, FWF parameters, and ALPS index to obtain the user's brain-like lymphatic system functional assessment results, the assessment and analysis module is used to: input the PVS multidimensional parameters, IVIM model parameters, FWF parameters, and ALPS index into a trained machine learning model to assess the functional status of the brain-like lymphatic system and obtain the user's brain-like lymphatic system functional assessment results; the brain-like lymphatic system functional assessment results include periarterial space inflow obstruction type, interstitial space exchange obstruction type, perivenous space outflow obstruction type, mixed type, and normal.

[0098] In the evaluation of the functional state of the brain lymphoid system, the PVS multi-dimensional parameters input into the trained machine learning model can be the PVS volume information of each brain region, or the spatial network characteristics of the PVS. Therefore, the input of the trained machine learning model is: the PVS volume information of each brain region, the interstitial fluid exchange parameter values of the brain region (including the mean value of the vascular perfusion fraction, the mean value of the pure diffusion coefficient, the mean value of the pseudo-diffusion coefficient, and the mean value of the FWF parameter), and the ALPS index. The input of the trained machine learning model can also be: the spatial network characteristics of the PVS, the interstitial fluid exchange parameter values of the brain region, and the ALPS index.

[0099] In this embodiment, the machine learning model is a random forest model, a support vector machine, or a deep learning network, which is not limited here.

[0100] In another example of the present application, the non-invasive brain lymphoid network evaluation system based on magnetic resonance imaging further comprises a normal value database construction module, which is used to: randomly select a plurality of healthy control group samples, and statistically analyze the feature vectors (including: PVS volume information of each brain region, interstitial fluid exchange parameter values of each brain region, ALPS index, and spatial network characteristics of PVS) of all the plurality of healthy control group samples. After analyzing the normality, the reference interval of each type of feature vector is represented by the mean value ± 2 standard deviations (Mean ± 2SD) or the 5th percentile to the 95th percentile (P5-P95) of each type of feature vector of all healthy control group samples. The PVS volume normal value reference interval, the IVIM and FWF parameter normal value reference interval, the ALPS index normal value reference interval, and the PVS spatial normal value reference interval are obtained. The normal value database is obtained, which is subsequently used as a reference standard for individualized evaluation of patients.

[0101] The PVS multi-dimensional parameters, IVIM model parameters, FWF parameters, and ALPS index are input into the trained machine learning model to evaluate the functional state of the brain lymphoid system, and the brain lymphoid system function evaluation result of the user is obtained, including the following contents:

[0102] (1) The PVS volume information of each brain region is compared with the PVS volume normal value reference interval, or the spatial network characteristics of the PVS are compared with the PVS spatial normal value reference interval. If the PVS volume information of any brain region is not within the PVS volume normal value reference interval or the spatial network characteristics of the PVS are not within the PVS spatial normal value reference interval, the first evaluation result is perivascular space inflow obstruction type, otherwise, the first evaluation result is normal.

[0103] (2) Comparing the interstitial fluid exchange parameter value of each brain region with the normal value reference interval of IVIM and FWF parameters, if the interstitial fluid exchange parameter value of any brain region is not in the normal value reference interval of IVIM and FWF parameters, the second evaluation result is interstitial space exchange disorder type, otherwise, the second evaluation result is normal.

[0104] (3) Comparing the ALPS index with the normal value reference interval of ALPS, if the ALPS index is not in the normal value reference interval of ALPS index, the third evaluation result is perivascular space outflow disorder type, otherwise, the third evaluation result is normal.

[0105] (4) According to the first evaluation result, the second evaluation result and the third evaluation result, the brain lymphoid system function evaluation result of the user is determined, specifically: when the first evaluation result, the second evaluation result and the third evaluation result are all normal, the brain lymphoid system function evaluation result of the user is normal; when any target evaluation result is disorder type and the other two evaluation results are normal, the brain lymphoid system function evaluation result of the user is the target evaluation result; when the first evaluation result, the second evaluation result and the third evaluation result at most have one normal, the brain lymphoid system function evaluation result of the user is mixed type.

[0106] The evaluation analysis module establishes a multi-modal fusion analysis framework, and the specific steps are as follows:

[0107] 1. Index integration: the above multi-dimensional quantitative characteristics (multi-dimensional quantitative characteristics include PVS multi-dimensional parameters, IVIM model parameters, FWF parameters and ALPS index) are jointly included in a unified analysis platform, and the scale of different indexes is unified through standardization and normalization operation.

[0108] 2. Individualized abnormality detection and evaluation: comparing the PVS multi-dimensional parameters, IVIM model parameters, FWF parameters and ALPS index of the user with the normal value database, determining the abnormality according to z-score or quantile, obtaining the indexes deviating from the normal range, and forming the individualized abnormality characteristic spectrum.

[0109] 3. Clinical typing diagnosis: using machine learning algorithms (such as random forest model, support vector machine or deep learning network) to perform clustering analysis and typing (such as perivascular space inflow disorder type, interstitial space exchange disorder type, perivascular space outflow disorder type, mixed type, normal, etc.) and severity evaluation on the multi-dimensional quantitative characteristics of the three major functional indexes.

[0110] 4. Result output and abnormality determination: generate a brain lymphoid function comprehensive score (i.e., the brain lymphoid system function evaluation result is the severity of perivascular space inflow disorder, interstitial space exchange disorder, perivascular space outflow disorder, and mixed type) or an abnormal probability atlas, which includes multi-dimensional quantitative characteristics of each brain region and the brain lymphoid system function evaluation result of each brain region, realizes non-invasive and quantitative brain lymphoid system function evaluation, and provides a visual report, which helps to early detect the risk of brain lymphoid function abnormality related diseases.

[0111] (five) a report generation module for automatically generating a comprehensive evaluation report; the comprehensive evaluation report includes the output results of the multi-dimensional function analysis module and the brain lymphoid system function evaluation results of the user. The report generation module provides a visual interface to display three-dimensional images and quantitative data of liquid diffusion, free water distribution, and PVS morphology.

[0112] In another example of the present application, the comprehensive evaluation report can further include: (1) quantitative results of multi-dimensional characteristic parameters of three major function indicators (ALPS index, IVIM model parameters, FWF parameters, PVS volume information, and spatial distribution characteristics of PVS); (2) quantitative pseudo-color maps and spatial distribution maps showing multi-dimensional quantitative characteristics; (3) comparison analysis maps of multi-dimensional quantitative characteristics and respective normal value reference intervals; (4) brain lymphoid system function evaluation results, and clinical interpretation and possible disease hints for abnormal indicators; (5) comprehensive scores and severity evaluations of brain lymphoid system function evaluation results; (6) targeted clinical suggestions and follow-up strategies.

[0113] The present application integrates DTI-ALPS analysis, free water imaging, IVIM model, and PVS analysis technology, realizes comprehensive, systematic, and non-invasive evaluation of the brain lymphoid system, and provides scientific basis for early auxiliary diagnosis and pathological mechanism research of nervous system diseases. Non-invasive: completely based on magnetic resonance imaging technology, without invasive operation, suitable for clinical and scientific research scenarios. Comprehensive: integrates DTI-ALPS, free water imaging, IVIM model, and PVS analysis to evaluate the functional status of the brain lymphoid system from multiple dimensions. Efficient: uses automated algorithms to quickly complete image processing and analysis, reducing manual intervention. Precise: based on deep learning and advanced image processing technology, provides high-precision quantitative analysis results.

[0114] The present application relates to the field of medical image processing and neuroscience, and discloses a non-invasive brain lymphoid network evaluation method and system based on multi-modal magnetic resonance imaging, which collects 3D-T1WI, 3D-T2WI, DWI (multi-b value), DTI, and other multi-modal magnetic resonance imaging data, and provides scientific basis for early screening and typing evaluation of neurodegenerative diseases, cerebral small vessel diseases, cognitive dysfunction, and other diseases.

[0115] The present application has the following advantages:

[0116] 1. The present application comprehensively analyzes the brain lymphoid network system from multiple dimensions and brain lymphoid functional pathways by integrating DTI-ALPS technology, free water imaging and IVIM model and PVS analysis technology, fully reflects the functional state of the brain lymphoid network system, improves the accuracy and comprehensiveness of the evaluation, realizes the comprehensive and systematic evaluation of the brain lymphoid system, overcomes the shortcomings of the prior art which only relies on invasive methods or indirect indicators, and provides a more comprehensive and accurate brain lymphoid system function evaluation method.

[0117] 2. The present application adopts multi-modal magnetic resonance imaging technology, combines DTI-ALPS, free water imaging and IVIM model and PVS analysis, can better characterize complex microstructure, improves the accuracy of brain lymphoid system function evaluation, and overcomes the limitation of insufficient resolution in related technology.

[0118] 3. The present application provides a systematic evaluation method by integrating multiple evaluation technologies, provides a scientific basis for early diagnosis and pathological mechanism research of nervous system diseases, and helps to promote early screening and personalized treatment of nervous system diseases.

[0119] Example 2.

[0120] The non-invasive brain lymphoid network evaluation method based on magnetic resonance imaging provided by the embodiments of the present application can be applied to, for example Figure 2The application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data required by the server 104 to process. The data storage system can be set up separately, or integrated on the server 104, or placed on the cloud or other servers. The terminal 102 can send the multi-modal image data to be processed to the server 104, and after the server 104 receives the multi-modal image data to be processed, for the multi-modal image data to be processed, the server 104 pre-processes the multi-modal image data to obtain pre-processed 3D-T1WI image, pre-processed 3D-T2WI image, pre-processed DWI image and pre-processed DTI image, determines the PVS multi-dimensional parameter according to the pre-processed 3D-T1WI image and the pre-processed 3D-T2WI image, separates the perfusion and diffusion components of the pre-processed DWI image using the IVIM model to obtain the IVIM model parameter, and analyzes the free water imaging model according to the pre-processed DTI image to obtain the FWF parameter, calculates the ALPS index according to the pre-processed DTI image, and evaluates the functional state of the brain lymphoid system according to the PVS multi-dimensional parameter, the IVIM model parameter, the FWF parameter and the ALPS index. The brain lymphoid system function evaluation result of the terminal 102. The server 104 can feed back the brain lymphoid system function evaluation result obtained for the multi-modal image data to the terminal 102. In addition, in some embodiments, the non-invasive brain lymphoid network evaluation method based on magnetic resonance imaging can also be realized by the server 104 or the terminal 102 alone, such as brain lymphoid network evaluation for multi-modal image data by the terminal 102 directly, or multi-modal image data can be obtained from the data storage system by the server 104, and brain lymphoid network evaluation for multi-modal image data.

[0121] Among them, the terminal 102 can be but not limited to various desktop computers and notebook computers. The server 104 can be realized by an independent server or a server cluster composed of multiple servers, and can also be a cloud server.

[0122] Based on the same inventive concept, the embodiments of the present application also provide a non-invasive brain lymphoid network evaluation method based on magnetic resonance imaging. The implementation scheme for solving the problem provided by the method is similar to the implementation scheme described in the above system, so the specific limitations in one or more non-invasive brain lymphoid network evaluation method embodiments based on magnetic resonance imaging provided below can refer to the limitations of the non-invasive brain lymphoid network evaluation system based on magnetic resonance imaging described above, which will not be repeated here.

[0123] In an exemplary embodiment, as Figure 3As shown, a non-invasive brain lymphoid network evaluation method based on magnetic resonance imaging is provided, which is executed by a computer device, specifically, can be executed by a terminal or a server alone, or can be executed by a terminal and a server together, in the embodiments of the present application, the method is applied to Figure 2 The server 104 in the method is taken as an example for illustration, and includes the following steps 201 to 206.

[0124] In step 201, 3D-T1WI images, 3D-T2WI images, DWI images and DTI images of a user are acquired.

[0125] In step 202, the 3D-T1WI images, 3D-T2WI images, DWI images and DTI images are preprocessed to obtain preprocessed 3D-T1WI images, preprocessed 3D-T2WI images, preprocessed DWI images and preprocessed DTI images.

[0126] In step 203, PVS multi-dimensional parameters are determined according to the preprocessed 3D-T1WI images and the preprocessed 3D-T2WI images.

[0127] In step 204, perfusion and diffusion component separation is performed on the preprocessed DWI images by using an IVIM model to obtain IVIM model parameters; the IVIM model parameters include a vascular perfusion fraction, a pure diffusion coefficient and a pseudo-diffusion coefficient; and free water imaging model analysis is performed on the preprocessed DTI images to obtain FWF parameters.

[0128] In step 205, an ALPS index is calculated according to the preprocessed DTI images.

[0129] In step 206, the functional state of the brain lymphoid system is evaluated according to the PVS multi-dimensional parameters, the IVIM model parameters, the FWF parameters and the ALPS index to obtain a brain lymphoid system function evaluation result of the user.

[0130] The application also provides an application scenario of the non-invasive brain lymphoid network evaluation method based on magnetic resonance imaging. Specifically, the non-invasive brain lymphoid network evaluation method based on magnetic resonance imaging can be applied in a brain lymphoid network evaluation scenario. The brain lymphoid network evaluation scenario includes an image acquisition link, a brain lymphoid network evaluation link, and a result display link. The multi-modal image data enters the brain lymphoid network evaluation link from the image acquisition link, and the corresponding brain lymphoid system function evaluation result is obtained through the man-machine cooperative mode, and enters the downstream result display link. The non-invasive brain lymphoid network evaluation system based on magnetic resonance imaging provided in the embodiment belongs to the brain lymphoid network evaluation link. Specifically, in the brain lymphoid network evaluation link process for the multi-modal image data, the multi-modal image data can be preprocessed to obtain a preprocessed 3D-T1WI image, a preprocessed 3D-T2WI image, a preprocessed DWI image, and a preprocessed DTI image. The PVS multi-dimensional parameters are determined according to the preprocessed 3D-T1WI image and the preprocessed 3D-T2WI image. The perfusion and diffusion components of the preprocessed DWI image are separated by using the IVIM model to obtain the IVIM model parameters. The FWF parameters are obtained by performing free water imaging model analysis on the preprocessed DTI image. The ALPS index is calculated according to the preprocessed DTI image. The function state of the brain lymphoid system is evaluated according to the PVS multi-dimensional parameters, the IVIM model parameters, the FWF parameters, and the ALPS index, and the brain lymphoid system function evaluation result of the user is obtained.

[0131] The technical features of the above embodiments can be combined in any manner. To make the description concise, not all possible combinations of the technical features in the above embodiments are described, but as long as the combinations of the technical features do not exist contradictions, they should be considered as the scope of the present application.

[0132] The principles and implementation modes of the present application are described by using specific examples. The above embodiments are only used to help understand the method and its core idea; meanwhile, for those skilled in the art, according to the idea of the present application, the specific implementation mode and application range can be changed. In conclusion, the content of the present application should not be understood as a limitation.

Claims

1. A non-invasive brain lymphoid network assessment system based on magnetic resonance imaging, characterized in that, include: The data acquisition module is used to acquire the user's 3D-T1WI images, 3D-T2WI images, DWI images, and DTI images; The data preprocessing module is used to preprocess the images from the data acquisition module to obtain preprocessed 3D-T1WI images, preprocessed 3D-T2WI images, preprocessed DWI images, and preprocessed DTI images. The multidimensional functional analysis module includes a cerebral artery perivascular space fluid inflow analysis unit, a tissue interstitial fluid exchange analysis unit, and a venous perivascular space fluid outflow analysis unit; The cerebral artery perivascular space fluid inflow analysis unit is used to: determine PVS multidimensional parameters based on preprocessed 3D-T1WI images and preprocessed 3D-T2WI images, wherein the PVS multidimensional parameters include PVS volume and PVS spatial network characteristics. The interstitial fluid exchange analysis unit is used to: separate perfusion and diffusion components in preprocessed DWI images using an IVIM model to obtain IVIM model parameters; the IVIM model parameters include vascular perfusion fraction, pure diffusion coefficient, and pseudo-diffusion coefficient; and perform free water imaging model analysis on preprocessed DTI images to obtain FWF parameters, wherein the FWF parameters are free water fraction parameters. The perivascular space fluid outflow analysis unit is used to: calculate the ALPS index based on the preprocessed DTI image; The normal value database construction module is used to: randomly select several healthy control group samples, statistically analyze the feature vectors of all healthy control group samples, and after analyzing the normality, obtain the normal value reference intervals of PVS volume, IVIM model parameters and FWF parameters, ALPS index and PVS spatial network characteristics, and obtain the normal value database. The evaluation and analysis module is used to input PVS multidimensional parameters, IVIM model parameters, FWF parameters, and ALPS index into a trained machine learning model to obtain the user's brain lymphatic system function evaluation results, including: Compare the PVS volume information of each brain region with the normal PVS volume reference range, or compare the PVS spatial network characteristics with the normal PVS spatial reference range. If the PVS volume information of any brain region is not within the normal PVS volume reference range or the PVS spatial network characteristics are not within the normal PVS spatial network characteristics reference range, then the first assessment result is periarterial space inflow obstruction type; otherwise, the first assessment result is normal. The interstitial fluid exchange parameter values ​​of each brain region were compared with the normal reference range of IVIM model parameters and FWF parameters. The mean values ​​of vascular perfusion fraction, pure diffusion coefficient, pseudo-diffusion coefficient, and FWF parameters of the brain region were used as the interstitial fluid exchange parameter values ​​of the brain region. If the interstitial fluid exchange parameter value of any brain region was outside the normal reference range of IVIM model parameters and FWF parameters, the second assessment result was interstitial fluid exchange disorder; otherwise, the second assessment result was normal. Compare the ALPS index with the normal reference range of ALPS. If the ALPS index is not within the normal reference range of ALPS, the third assessment result is perivenous space outflow obstruction type; otherwise, the third assessment result is normal. The user's brain lymphatic system function assessment results are determined based on the first, second, and third assessment results.

2. The non-invasive brain lymphoid network assessment system based on magnetic resonance imaging according to claim 1, characterized in that, When the PVS multidimensional parameters are PVS volume information, the fluid inflow analysis unit from the perivascular space of the cerebral arteries is used for: Using preprocessed 3D-T1WI and preprocessed 3D-T2WI images as input, a trained image segmentation model is used to predict a binary segmentation mask for the perivascular space. The binary segmentation mask for the perivascular space includes the perivascular space volume information of each voxel; Connected component analysis was performed on the binary segmentation mask of the perivascular space to determine the number of voxels in each connected component. Based on the number of voxels in each connected component and the perivascular space volume information of a single voxel, the PVS volume information of each brain region was calculated. When the PVS multidimensional parameters are PVS spatial network characteristics, the cerebral artery perivascular space fluid inflow analysis unit is used for: Using preprocessed 3D-T1WI and preprocessed 3D-T2WI images as input, a trained image segmentation model is used to predict a binary segmentation mask for the perivascular space; the binary segmentation mask for the perivascular space includes the perivascular space volume information of each voxel; The centroid coordinates of the perivascular space are determined based on the binary segmentation mask of the perivascular space. Construct a spatial adjacency network with the centroid coordinates of the perivascular space as nodes. When the spatial connection radius between any two nodes in the spatial adjacency network is within a set threshold range, an edge is established to form an undirected graph. Calculating PVS spatial network characteristics based on undirected graphs; PVS spatial network characteristics include node connectivity, clustering coefficient, and average shortest path length.

3. The non-invasive brain lymphoid network assessment system based on magnetic resonance imaging according to claim 2, characterized in that, The trained image segmentation model is a U-Net model or an nnU-Net model.

4. The non-invasive brain lymphoid network assessment system based on magnetic resonance imaging according to claim 1, characterized in that, The data preprocessing module is used for: The DWI image and DTI image are subjected to a first preprocessing to obtain a first preprocessed DWI image and a first preprocessed DTI image; the first preprocessing includes denoising, motion correction, eddy current correction and magnetic susceptibility artifact correction. Non-uniform field correction was performed on 3D-T1WI and 3D-T2WI images to obtain corrected 3D-T1WI and 3D-T2WI images; The first preprocessed DWI image, the first preprocessed DTI image, and the corrected 3D-T2WI image are registered to the space of the corrected 3D-T1WI image to obtain the registered DWI image, the registered DTI image, and the registered 3D-T2WI image. The corrected 3D-T1WI image, the registered 3D-T2WI image, the registered DWI image, and the registered DTI image were normalized to obtain the preprocessed 3D-T1WI image, the preprocessed 3D-T2WI image, the preprocessed DWI image, and the preprocessed DTI image.

5. The non-invasive brain lymphoid network assessment system based on magnetic resonance imaging according to claim 1, characterized in that, The formula for calculating the ALPS index is as follows: ； in, Indicates the ALPS index. Indicates the regions of projection fibers in the brain. Average axial diffusivity For the brain region of the associated fiber region Average axial diffusivity For the projection fiber area Average axial diffusivity For the joint fiber region Average axial diffusivity.

6. The non-invasive brain lymphoid network assessment system based on magnetic resonance imaging according to claim 1, characterized in that, The evaluation and analysis module is used for: The user's brain lymphatic system function assessment results were determined based on the first, second, and third assessment results, specifically: When the first, second, and third assessment results are all normal, the user's brain lymphatic system function assessment result is normal; when any target assessment result is disordered and the other two assessment results are normal, the user's brain lymphatic system function assessment result is the target assessment result; when at most one of the first, second, and third assessment results is normal, the user's brain lymphatic system function assessment result is mixed. The brain-like lymphatic system function assessment results include periarterial space inflow obstruction type, interstitial space exchange obstruction type, perivenous space outflow obstruction type, mixed type, and normal type.

7. The non-invasive brain lymphoid network assessment system based on magnetic resonance imaging according to claim 6, characterized in that, The machine learning model is a random forest model, support vector machine, or deep learning network.

8. The non-invasive brain lymphoid network assessment system based on magnetic resonance imaging according to claim 1, characterized in that, The non-invasive brain lymphoid network assessment system based on magnetic resonance imaging also includes a report generation module; the report generation module is used to generate a comprehensive assessment report; the comprehensive assessment report includes the output results of the multidimensional functional analysis module and the user's brain lymphoid system functional assessment results.

9. A non-invasive brain lymphoid network assessment method based on magnetic resonance imaging, characterized in that, include: Acquire the user's 3D-T1WI, 3D-T2WI, DWI, and DTI images; Preprocessing was performed on 3D-T1WI, 3D-T2WI, DWI and DTI images to obtain preprocessed 3D-T1WI, preprocessed 3D-T2WI, preprocessed DWI and preprocessed DTI images. The PVS multidimensional parameters are determined based on the preprocessed 3D-T1WI image and the preprocessed 3D-T2WI image. The PVS multidimensional parameters include PVS volume and PVS spatial network characteristics. The IVIM model was used to separate the perfusion and diffusion components of the preprocessed DWI images to obtain the IVIM model parameters. The IVIM model parameters include the vascular perfusion fraction, the pure diffusion coefficient, and the pseudo-diffusion coefficient. The free water imaging model was analyzed based on the preprocessed DTI images to obtain the FWF parameters, which are the free water fraction parameters. ALPS index is calculated based on the preprocessed DTI image; A number of healthy control group samples were randomly selected, and the feature vectors of all healthy control group samples were statistically analyzed. After normality analysis, the normal reference ranges of PVS volume, IVIM model parameters, FWF parameters, ALPS index and PVS spatial network characteristics were obtained. By inputting the PVS multidimensional parameters, IVIM model parameters, FWF parameters, and ALPS index into a trained machine learning model, the user's brain lymphatic system function assessment results are obtained, including: Compare the PVS volume information of each brain region with the normal PVS volume reference range, or compare the PVS spatial network characteristics with the normal PVS spatial reference range. If the PVS volume information of any brain region is not within the normal PVS volume reference range or the PVS spatial network characteristics are not within the normal PVS spatial network characteristics reference range, then the first assessment result is periarterial space inflow obstruction type; otherwise, the first assessment result is normal. The interstitial fluid exchange parameter values ​​of each brain region were compared with the normal reference range of IVIM model parameters and FWF parameters. The mean values ​​of vascular perfusion fraction, pure diffusion coefficient, pseudo-diffusion coefficient, and FWF parameters of the brain region were used as the interstitial fluid exchange parameter values ​​of the brain region. If the interstitial fluid exchange parameter value of any brain region was outside the normal reference range of IVIM model parameters and FWF parameters, the second assessment result was interstitial fluid exchange disorder; otherwise, the second assessment result was normal. Compare the ALPS index with the normal reference range of ALPS. If the ALPS index is not within the normal reference range of ALPS, the third assessment result is perivenous space outflow obstruction type; otherwise, the third assessment result is normal. The user's brain lymphatic system function assessment results are determined based on the first, second, and third assessment results.

Images