A disease intelligent diagnosis device based on brain structural connection identifier and application thereof

Abstract

This application provides a disease intelligent diagnostic device based on brain structural connectivity identifiers and its application. The device includes (1) a data acquisition module; (2) a data processing module: reconstructing the spin distribution function of the data and distributing the spin distribution function in a standard space; (3) a data projection module: projecting the spin distribution functions of patients and healthy individuals into the standard space to obtain the Z-value of the maximum direction within the voxel; and (4) a Connectome Identifier extraction module: reducing the dimensionality of the Z-value to form a 1-dimensional feature value. The device of this application can be used for the diagnosis of various brain diseases and the determination of lesion location and severity.

Description

Technical Field

[0001] This application belongs to the field of medical imaging data processing. Specifically, this application provides a disease intelligent diagnostic device based on brain structural connectivity identifiers and its application. Background Technology

[0002] The clinical application of diffusion tensor imaging (DTI) has been confirmed by numerous studies. For example, cadaveric anatomy experiments have demonstrated the effectiveness of DTI in tracking medium to large white matter fiber tracts. In subsequent studies, DTI has been used for preoperative assessment in neurosurgery, allowing for the depiction of the distribution and morphology of major white matter structures. Common white matter fiber tracts reconstructed by DTI include the pyramidal tract, visual pathway, and language pathway. By presenting these structures, DTI can further preserve patient function and significantly influence the development of surgical strategies for complex lesions. However, DTI has limitations. It cannot acquire and analyze intersecting fibers, accurately map the origin and termination points of white matter fibers, and often produces false positives and false negatives during tracking. This is inherent to the technical principles of DTI. Technically, DTI can only acquire the main restricted diffusion direction within a single voxel—the diffusion direction of water molecules after being encapsulated—and this is a relative value, not an absolute measurement. Specifically, tissue characteristics are measured using eigenvalues ​​(λ1, λ2, and λ3). Axial diffusion capability, represented by λ1, reflects the ability of water molecules to diffuse along the long axis of the fiber; radial eigenvalues ​​are divided into two perpendicular eigenvalues, λ2 and λ3. Fractional anisotropy (FA) is a commonly used diffusion scalar parameter in DTI technology, reflecting the degree of partial anisotropy. It represents the proportion of water molecules' ability to diffuse in a particular direction to the total diffusion capability, and is expressed by the formula:

[0003]

[0004] The FA value ranges between 0 and 1. Generally, fibers with better myelination have higher FA values. FA values ​​decrease when myelination is reduced or neurons are absent. In tissues with dense fibers, diffusion perpendicular to the fiber direction is inhibited, increasing the FA value; conversely, when fiber structures are damaged by degenerative diseases or other conditions, diffusion perpendicular to the fiber direction increases, decreasing the FA value. Therefore, assessing brain FA values ​​can evaluate structural changes in the tissue. Consequently, for complex fiber structures such as crossings and angulations within voxels, DTI does not collect this information from the beginning of data acquisition, resulting in a large amount of variation in neural tissue construction. A large sample size is required to reveal the functional state of the brain's networks when reflecting histological characteristics. Deterministic tracing based on this cannot reproduce the anatomical features of white matter fibers found in fiber bundle anatomy and histological techniques. Similarly, for white matter fibers embedded in edematous tissue, DTI, which only considers the average direction of voxels, cannot achieve sufficient tracing. Therefore, there is an urgent need for acquisition techniques that fully consider the unrestricted diffusion direction within voxels from the beginning of acquisition.

[0005] In recent years, emerging fiber tract mapping techniques such as high-angular-resolution diffusion imaging (HARDI) and DSI have emerged to address these issues. DIS is a diffusion-based white matter imaging technique. Compared to traditional diffusion tensor imaging (DTI), DSI offers more imaging directions and a higher b-value. Based on the probability density function (PDF) framework, it can describe the three-dimensional distribution of spinon micro-displacements within voxels, providing more accurate resolution of brain white matter fiber tracts. White matter research is a crucial component of neuroscience research. DTI technology has made in vivo white matter mapping possible and can be widely used in the diagnosis and treatment of brain diseases. DTI employs the generalized q-sampling imaging (GQI) algorithm, with the core model being the GQI model and the corresponding diffusion index parameter being quantitative anisotropy (QA). QA is calculated from the peak value of the spin distribution function (SDF). This algorithm analyzes all fiber orientations within a voxel, thus solving the problem of tracking crossed fibers. Specifically, this can be analyzed from two aspects. First, regarding the evaluation content, the tensor model of DTI evaluates diffusion. However, based on diffusion, due to the characteristics of Brownian motion, restricted and unrestricted diffusion cannot be completely separated. The analysis of crossed fibers, turning fibers, and fiber start and end points can only be derived using post-processing reconstruction models (such as probabilistic tracking). But for DTI with its poor signal-to-noise ratio, relying on models often fails to provide effective reconstruction. In contrast, the GQI algorithm used by DSI can distinguish between restricted and unrestricted motion from signal acquisition. Subsequent multi-directional identification only requires the simplest linear judgment. Therefore, in principle, DSI is more accurate than DTI. Secondly, regarding voxel signals, FA assesses the average diffusion signal of all fibers within the voxel, while QA assesses the diffusion signal of each individual fiber subpopulation within the voxel. That is, for each fiber subpopulation in a specific direction within the voxel, QA independently measures the diffusion direction and magnitude of that subpopulation. FA is reduced at fiber crossings or corners, while QA is almost unaffected.

[0006] For brain network disorders such as epilepsy, Alzheimer's disease, and Parkinson's disease, the application of artificial intelligence in localization diagnosis still requires further research. For example, other non-invasive methods for epilepsy localization, such as scalp EEG, cannot record EEG signals from deep brain regions and can only detect up to 6 cm deep. 2 Coordinated discharge; 18 F-FDG PET is subject to interference from low metabolism in distant sites and physiological low metabolism; invasive deep brain EEG has a limited number of sampling points, and requires prior inference in non-invasive assessment to ensure the accuracy of lesion detection. Therefore, there is currently no single modality that can non-invasively localize epilepsy. Summary of the Invention

[0007] This invention employs quantitative DSI data acquisition to construct a template of the structural connections in a healthy human brain, obtaining the intensity distribution of water molecule diffusion in each voxel (the smallest signal unit in 3D space, similar to a pixel in a camera photograph, specifically a numerical value corresponding to a coordinate in a 3D matrix). Then, DSI data from patients is acquired, and their 3D matrix is ​​projected onto the standard template to obtain the distribution matrix of the patient's water molecule diffusion data in a standard space, thus completing spatial standardization.

[0008] In the standard space, the data within the same voxel of the patient are decomposed into vectors according to the direction of the corresponding voxel in the standard template. The Z-transform is taken as the multiple of the standard deviation of the patient's vector value from the population's vector value in the most important direction. This is the eigenvalue in that direction. The distribution of the whole-brain eigenvalues ​​is the ConnectomeIdentifier.

[0009] In previous research, the inventors discovered that the Connectome Identifier for the same healthy subject at different time points exhibits individual specificity; that is, based on the Connectome Identifier of the brain connectome, subjects can be accurately identified without unblinding the data. Unpublished data from the inventors revealed that, when used in patients with temporal lobe epilepsy, the Connectome Identifier, combined with machine learning, can accurately identify whether the epilepsy is left-sided, right-sided, or bilateral. In Alzheimer's disease patients, the Connectome Identifier can identify the severity of Alzheimer's disease lesions at different stages.

[0010] On the one hand, this application provides a disease intelligent diagnostic device based on brain structural connectivity identifiers, the device comprising the following modules:

[0011] (1) Data acquisition module: used to acquire DSI data;

[0012] (2) Data processing module: The DSI data obtained in step (1) is reconstructed by Q-space differential isomorphism, so that the spin distribution function is reconstructed in the individual space, and the distribution of diffusion direction in each voxel in the standard space is projected.

[0013] (3) Data projection module: Project the spin distribution functions of one or more healthy individuals into the standard space and average them to obtain the mean and standard deviation of the length of the diffusion vector with the largest diffusion within the voxel; also project the spin distribution function of the patient individual into the standard space, take the direction with the largest diffusion within the voxel of a single subject, and subtract the mean length at the population level. mean Divide by the standard deviation length std The Z-value of the maximum direction within this voxel is obtained as follows: Z = (l - length) mean ) / length std ;

[0014] (4) Connectome Identifier extraction module: The Z values ​​arranged in a 3D matrix are encoded in the order of left-back-front to form a 1D feature value, which is the Connectome Identifier.

[0015] Furthermore, the data acquisition module includes an MRI instrument.

[0016] Furthermore, the DSI data refers to the patient's DSI data.

[0017] Furthermore, the DSI data includes the DSI data of the patient and the DSI data of one or more healthy individuals.

[0018] Furthermore, the DSI data is brain DSI data.

[0019] Furthermore, the data processing module performs the following functions: processing the DSI data obtained in step (1) using DSI-Studio software, inputting the DICOM format data in the original image, and packaging it into an SRC format file together with the b-value table; reading the SRC file to observe the brain parenchyma in the influence space, selecting a threshold that can cover the brain parenchyma without including the skull, and performing Q-space differential isomorphic reconstruction.

[0020] Furthermore, the functions performed by the data processing module include: using TOPUP to correct distortions caused by the magnetic field before the Q-space differential isomorphic reconstruction begins, and using Eddy current to correct head motion artifacts.

[0021] Furthermore, in the data processing module, the diffusion sampling length ratio during Q-space differential isomorphic reconstruction is 1.1, the mounting method is CDM, the spatial template is HCP1021, the directional distribution function interpolation parameter is 20 folds, the number of fibers processed within the voxel is 10, and the number of threads is 4.

[0022] Furthermore, the device also includes (5) a Connectome Identifier reading module for reading the Connectome Identifier.

[0023] On the other hand, the aforementioned device is used to diagnose brain diseases.

[0024] On the other hand, the aforementioned device is used to determine the location of lesions in brain diseases.

[0025] On the other hand, the aforementioned device is used to differentiate the severity of brain diseases.

[0026] Furthermore, the brain diseases mentioned are epilepsy, Parkinson's disease, and Alzheimer's disease.

[0027] On the other hand, this application provides a storage medium in which a program for performing the functions of the above-mentioned modules is recorded.

[0028] The DSI data of healthy individuals in this application can be obtained by MRI equipment in a diagnostic device, or can be obtained from a database or stored in a device or storage medium.

[0029] Beneficial effects:

[0030] Previous artificial intelligence identification methods were mainly based on diffusion scalars, considering the average value of water molecule diffusion in all directions within a voxel, without taking into account the direction of water molecule diffusion vectors and structural connectivity factors. This invention tracks the most important vector direction within the voxel and then analyzes the difference in the ratio between patients and healthy individuals in this direction. It considers both diffusion direction and diffusion intensity, which is more in line with the laws of disease development. Attached Figure Description

[0031] Figure 1 This demonstrates the basic principles of this application.

[0032] Figure 2 Example of a connection identifier for a patient.

[0033] Figure 3 An example of AI recognition performance for connecting identifiers. Detailed Implementation

[0034] Example 1: Data Acquisition

[0035] A SIGNA Premier 3.0 T MRI (GE, USA) with a 64-channel head coil was used. The patient was instructed to lie supine, quiet, and keep their head still. During the scan, the hands were placed at the sides of the body, and the entire head was scanned. The scanning sequences and parameters are as follows: Sagittal T1 MPRAGE: TE 2.69ms, TR 2477ms, flip angle 8°, 1.0mm voxel; DSI sequence: TE 84.2ms, TR 5589ms, FOV 224mm×224mm, voxel 2×2×2mm. 3 The maximum b value is 7000 s / mm 2 The total number of diffusion directions was 258; T2W FLAIR sequence: TE 105ms, TR 6300ms, 1.0mm voxel.

[0036] Example 2 Data Processing

[0037] Data from healthy individuals and subjects were obtained using the method described in Example 1.

[0038] The data acquired in Example 1 were processed using DSI-Studio software. DICOM format data from the original images were entered and packaged into an SRC format file along with a b-value table. The SRC file was read into the influence space to observe the extent of the brain parenchyma. A threshold that covers the brain parenchyma without including the skull was selected for reconstruction.

[0039] Before reconstruction began, TOPUP was used to correct distortions caused by the magnetic field, and Eddy current was used to correct head motion artifacts.

[0040] Q-space diffeomorphic reconstruction (QSDR) is performed to reconstruct the spin distribution function (SDF) in individual spaces, while further projecting the distribution of diffusion directions within each voxel in standard space (e.g., ...). Figure 1 As shown in the diagram, this means distributing the brain's morphology into a standard brain space. In this process, the diffusion sampling length ratio is 1.1, the mounting method is CDM, the spatial template is HCP1021, and the SDF is distributed in the standard space. The Orientation Distribution Function (ODF) interpolation parameter is set to 20-fold, the number of intravoxel fiber processing is set to 10, and the number of threads is set to 4. Half-Q space diffusion imaging is selected, and the output value is set to ODF, thus obtaining the SDF in the standard space.

[0041] Example 3: Data Projection and Connectome Identifier Extraction

[0042] By projecting the SDFs from the standard space of several healthy individuals together and averaging them, the mean and standard deviation of the length of the direction of maximum diffusion within the voxel are obtained.

[0043] The individual patient's SDF is also projected into the standard space, and the direction of maximum value within a single subject's voxel is taken, minus the population mean (length). mean ), divided by the standard deviation (length) std Remember the Z-value in the direction of maximum length within that voxel: Z = (l - length) mean ) / length std .

[0044] The Z values, arranged in a 3D matrix, are encoded in a left-back-front order to form 1-dimensional feature values ​​(e.g., Figure 2 As shown in the image, this is the Connectome Identifier.

[0045] Example 4: Application example of the connection identifier in this application

[0046] DSI data from 74 subjects were included. Connectome Identifiers were extracted from each subject. A support vector machine model was used, with 55 cases in the training set and 19 cases in the validation set. The ability of the Connectome Identifier index to identify left and right temporal lobe epilepsy and healthy volunteers was tested. The actual categories included 25 healthy individuals, 29 patients with left temporal lobe epilepsy, and 20 patients with right temporal lobe epilepsy. The test was performed according to the method described above. The results showed that healthy individuals and patients with right temporal lobe epilepsy were completely correctly identified in the validation set. Among patients with left temporal lobe epilepsy, 3 patients were incorrectly identified, and 7 patients were correctly identified (e.g., ...). Figure 3 (As shown). The concordance rate with the surgical resection results reached 84.2%, which is comparable to the level of manual interpretation, and the interpretation time was only 21 seconds.

Claims

1. A disease intelligent diagnostic device based on brain structural connectivity identifiers, the device comprising the following modules: (1) Data acquisition module: used to acquire DSI data; (2) Data processing module: The DSI data obtained in step (1) is reconstructed by Q-space differential isomorphism, so that the spin distribution function is reconstructed in the individual space, and the distribution of diffusion direction in each voxel in the standard space is projected. (3) Data projection module: Project the spin distribution functions of one or more healthy individuals into the standard space and average them to obtain the mean and standard deviation of the length of the diffusion vector with the largest diffusion within the voxel; also project the spin distribution function of the patient individual into the standard space, take the direction with the largest diffusion within the voxel of a single subject, and subtract the mean length at the population level. mean Divide by the standard deviation length std The Z-value of the maximum direction within this voxel is obtained as follows: Z = (l - length) mean ) / length std ; (4) Connectome Identifier extraction module: The Z values ​​arranged in a 3D matrix are encoded in the order of left-back-front to form a 1D feature value, which is the Connectome Identifier.

2. The apparatus according to claim 1, wherein the data acquisition module comprises an MRI instrument.

3. The apparatus according to claim 1 or 2, wherein the data processing module performs the following functions: The DSI data obtained in step (1) is processed using DSI-Studio software. The DICOM format data in the original image is entered and packaged into an SRC format file along with the b-value table. The SRC file is read into the range of brain parenchyma in the influence space. A threshold that can cover the brain parenchyma without including the skull is selected, and Q-space differential isomorphic reconstruction is performed.

4. The apparatus according to claim 1, wherein the data processing module performs the following functions: Before Q-space differential isomorphic reconstruction begins, TOPUP is used to correct distortions caused by the magnetic field, and Eddy current is used to correct head motion artifacts.

5. The apparatus according to claim 1, wherein during the operation of the data processing module, the diffusion sampling length ratio is 1.1, the mounting method is selected as CDM, the spatial template is HCP1021, the directional distribution function interpolation parameter is selected as 20 folds, the number of fibers processed within the voxel is selected as 10, and the number of threads is selected as 4.

6. The apparatus according to claim 1, further comprising (5) a Connectome Identifier reading module for reading the Connectome Identifier.

7. The device according to claim 1, wherein the DSI data is the patient's DSI data.

8. The apparatus according to claim 1, wherein the DSI data is the DSI data of a patient and the DSI data of one or more healthy individuals.

9. The apparatus according to claim 7, wherein the DSI data is brain DSI data.

10. The apparatus of claim 1, wherein the apparatus is used for diagnosing brain diseases.

11. The apparatus according to claim 1, wherein the apparatus is used to determine the location of lesions in brain diseases.

12. The apparatus of claim 1, wherein the apparatus is used to distinguish the severity of brain diseases.

13. The device according to claim 10, wherein the brain disease is epilepsy, Parkinson's disease, or Alzheimer's disease.

14. A storage medium having a program recorded for performing the following steps: (1) Obtain DSI data; (2) Data processing: DSI data is reconstructed by Q-space differential isomorphism, so that the spin distribution function of individual space is reconstructed at the same time, the distribution of diffusion direction in each voxel in standard space is projected. (3) Data projection: The spin distribution functions of one or more healthy individuals in the standard space are projected together and averaged to obtain the mean and standard deviation of the length of the diffusion vector with the largest diffusion within the voxel; the spin distribution function of the patient individual is also projected into the standard space, and the direction with the largest diffusion within the voxel of a single subject is taken, and the mean length at the population level is subtracted. mean Divide by the standard deviation length std The Z-value of the maximum direction within this voxel is obtained as follows: Z = (l - length) mean ) / length std ; (4) Connectome Identifier Extraction: The Z values ​​arranged in a 3D matrix are encoded in the order of left-back-front to form a 1-dimensional feature value, which is the Connectome Identifier.

15. The storage medium according to claim 14, wherein the data processing step includes processing the acquired data using DSI-Studio software, inputting DICOM format data from the original image, and packaging it together with the b-value table into an SRC format file; reading the SRC file to observe the brain parenchyma in the influence space, selecting a threshold that can cover the brain parenchyma without including the skull, and performing Q-space differential isomorphic reconstruction.

16. The storage medium according to claim 14 or 15, wherein the data processing includes: Before Q-space differential isomorphic reconstruction begins, TOPUP is used to correct distortions caused by the magnetic field, and Eddy current is used to correct head motion artifacts.

17. The storage medium according to claim 14, wherein the diffusion sampling length ratio is 1.1 during the data processing, the mounting method is CDM, the spatial template is HCP1021, the directional distribution function interpolation parameter is 20 folds, the number of fibers processed within the voxel is 10, and the number of threads is 4.

Images