A method and system for three-dimensional modeling and rendering of neural pathways

Abstract

The present application relates to the technical field of three-dimensional rendering, in particular to a nerve pathway three-dimensional modeling and rendering method and system, comprising the following steps: obtaining diffusion tensor data to solve the principal eigenvector field, generating fiber tracks based on divergence step length and vorticity correction, obtaining a geometric model by using curvature frequency to reconstruct abnormal sections, mapping vertex geometric curvature eigenvalues, attenuating transparency according to the curvature difference of adjacent depth layers, and synthesizing nerve pathway visualization images. The present application uses the principal eigenvector field to realize dynamic adjustment of spatial step length and correction of tracking direction by divergence and vorticity characteristics calculation, identifies high-frequency disturbance sections after statistical distribution of the change rate frequency of the discrete curvature derivative of the track point sequence, effectively corrects the model distortion caused by sudden changes in curvature, and performs saliency judgment according to the curvature value difference between depth layers in the rendering stage, controls the transparency in a regional manner, so that the image retains structural details while optimizing rendering efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of three-dimensional rendering, in particular to a neural pathway three-dimensional modeling and rendering method and system. BACKGROUND

[0002] The technical field of three-dimensional rendering involves the process of generating images of three-dimensional models through computer graphics technology, which includes three-dimensional geometric modeling, texture mapping, lighting calculation, visualization pipeline construction, rendering algorithm execution, image synthesis output, and other core matters. Three-dimensional rendering technology is widely used in game engines, film production, medical images, building visualization, and other scenarios. In this technical field, researchers construct scene geometry, set material properties and lighting conditions, and use rasterization, ray tracing, or physically-based rendering methods to complete image generation of the scene. In recent years, with the development of neural networks and deep learning technology, three-dimensional rendering has gradually introduced data-driven computing modes to improve rendering efficiency and detail performance.

[0003] Among them, the traditional neural pathway three-dimensional modeling and rendering method refers to a processing method that establishes a three-dimensional space model of the neural pathway based on the anatomical structure information of the neural tissue and performs image rendering. This method usually uses diffusion tensor imaging data in magnetic resonance imaging technology as the original input, uses trajectory tracking algorithms to infer and reconstruct the spatial direction of neural fiber bundles, and then constructs a geometric model based on the reconstruction results. In the modeling process, the dominant direction in the tensor field needs to be analyzed for continuity to achieve the stability of fiber tracking. After modeling, the traditional method usually performs rendering through graphics processing means based on volume rendering or curve rendering, including setting color mapping, transparency mapping, and lighting direction of the fiber bundle, thereby generating a visualization image for medical auxiliary analysis. This method depends on image preprocessing accuracy and tracking algorithm parameter settings, and has limitations in restoring models in areas with varying curvature. SUMMARY

[0004] The purpose of the present application is to solve the shortcomings in the prior art and to provide a neural pathway three-dimensional modeling and rendering method and system.

[0005] To achieve the above purpose, the present application adopts the following technical solution, a neural pathway three-dimensional modeling and rendering method, comprising the following steps:

[0006] S1: Obtain diffusion tensor imaging data of the brain neural tissue, calculate the diffusion tensor matrix, decompose the diffusion tensor matrix to obtain the principal eigenvector, and construct the principal eigenvector field;

[0007] S2: traverse the main eigenvector field to calculate divergence, compare the divergence with a stability threshold, adjust a spatial integral step parameter, generate a correction component by cross-multiplying the rotation vector with the tracking direction, generate a neural fiber trajectory point sequence based on the spatial integral step parameter and the correction component;

[0008] S3: intercept a local trajectory segment of the neural fiber trajectory point sequence, convert a curvature derivative change rate of the local trajectory segment into a frequency distribution feature, identify an abnormal fiber segment, fit a direction vector of the abnormal fiber segment, and generate a neural pathway geometric model;

[0009] S4: calculate a second derivative vector module length of a vertex of the neural pathway geometric model, generate a geometric curvature characteristic value, map the geometric curvature characteristic value to a vertex attribute buffer, and obtain a vertex attribute mapping processing result;

[0010] S5: based on the vertex attribute mapping processing result, render the neural pathway geometric model, calculate a difference value of geometric curvature characteristic values of adjacent depth layer fragments, if the difference value exceeds a preset difference threshold, attenuate a transparency value, and synthesize a neural pathway visualized image based on the transparency value.

[0011] The application improves that the main eigenvector field includes voxel space coordinates, three-dimensional vector components and local field divergence, the neural fiber trajectory point sequence includes three-dimensional vertex coordinates, path tangent vectors and neighborhood point indexes, the neural pathway geometric model includes fiber bundle topological connection relationships, grid vertex normals and curved texture coordinates, the vertex attribute mapping processing result includes vertex geometric curvature characteristic values, vertex depth values and vertex color attributes, and the neural pathway visualized image includes pixel color values, pixel depth values and pixel transparency values.

[0012] The application improves that the acquisition step of the main eigenvector field is specifically:

[0013] S111: acquire diffusion tensor imaging data of brain neural tissue, traverse signal intensity of each voxel in the diffusion tensor imaging data, extract diffusion weighted signal attenuation values of each voxel in multiple gradient directions, construct a diffusion signal distribution matrix including spatial position coordinates and direction signal intensity, call a least square method to perform linear fitting operation on the diffusion signal distribution matrix, and generate a diffusion tensor matrix of a three-dimensional voxel space;

[0014] S112: read a symmetric positive definite second-order tensor corresponding to each voxel in the diffusion tensor matrix, calculate a characteristic equation root of the symmetric positive definite second-order tensor, solve the characteristic equation to obtain three real characteristic values and arrange them in descending order according to the numerical value, take a characteristic vector direction corresponding to a maximum characteristic value as a local main diffusion direction of a neural fiber, and establish a voxel eigenvector set including a characteristic value sequence and a main diffusion direction vector;

[0015] S113: Extract the main diffusion direction vector of each voxel position based on the voxel feature vector set, construct a vector field data structure continuously distributed in a three-dimensional space, calculate the dot product between the main diffusion direction vectors of adjacent voxels, determine the consistency weight of the local trend, perform smoothing interpolation processing on the singular point area in the vector field according to the consistency weight, and generate a main feature vector field reflecting the local trend of the nerve fiber.

[0016] The application improves that the acquisition step of the nerve fiber trajectory point sequence is specifically:

[0017] S211: Traverse the voxel grid of the main feature vector field, calculate the vector field divergence scalar at each grid point using the central difference method, compare the vector field divergence scalar with the preset stability threshold, if the divergence scalar is higher than the stability threshold, then the basic step length is reduced in proportion according to the size of the divergence scalar, otherwise the basic step length remains unchanged, and an adaptive integration step length parameter is established to adapt to the local field twist degree;

[0018] S212: Call the curl operator to extract the local rotation feature of the main feature vector field at the current coordinates, construct a three-dimensional curl vector, obtain the tracking tangent vector of the current nerve fiber, perform vector cross multiplication operation on the three-dimensional curl vector and the tracking tangent vector, obtain the normal correction vector perpendicular to the current path, superimpose the normal correction vector to the tracking tangent vector and perform normalization processing, and generate a geometric correction growth direction;

[0019] S213: Take the current fiber endpoint as the starting coordinates, call the adaptive integration step length parameter as the spatial displacement module length, combine the geometric correction growth direction as the direction component of the displacement vector, perform fourth-order Runge-Kutta integration operation, calculate the three-dimensional space coordinates of the next trajectory point, append the calculated new coordinates to the current path array and repeat the iteration process, and generate a continuous nerve fiber trajectory point sequence.

[0020] The application improves that the acquisition step of the nerve pathway geometric model is specifically:

[0021] S311: Perform discrete differentiation operation along the path on the nerve fiber trajectory point sequence, calculate the tangent angle of adjacent trajectory points, extract the point-by-point curvature value, segmentally intercept the curvature value sequence using a fixed length sliding window, calculate the curvature change rate of the curvature value in the window, perform discrete Fourier transform to convert the spatial change rate sequence into frequency domain energy distribution data, and generate a local curvature frequency distribution feature.

[0022] S312: Traverse the spectral components in the local curvature frequency distribution features, calculate the energy integral ratio of the high-frequency band, compare the energy integral ratio with the preset disturbance discrimination threshold, identify the window index position where the integral ratio exceeds the disturbance discrimination threshold, extract the coordinates of continuous trajectory points in the corresponding index interval, and define the disturbance abnormal section caused by numerical oscillation resulting in geometric distortion.

[0023] S313: Extract the tangential direction vectors of the connection points at both ends of the perturbation abnormal segment, establish a parameterized curve equation with the connection point position and tangential vector as boundary constraints, call the least squares method to iteratively solve the optimal fitting coefficients of the equation, construct a smooth interpolation curve and replace the original abnormal trajectory point data, integrate the optimized fiber path data, and generate a neural pathway geometric model.

[0024] The present invention is improved in that the process of setting the disturbance discrimination threshold is specifically as follows: a set of neural fiber trajectory data that has been verified to be geometrically smooth is selected as a reference set; a discrete Fourier transform is performed on the data in the reference set to extract the distribution sequence of the energy proportion of the high-frequency band; the mathematical expectation value and standard deviation of the distribution sequence are calculated; and the mathematical expectation value is superimposed with the standard deviation by a preset multiple as the disturbance discrimination threshold.

[0025] The present invention is improved in that the step of obtaining the vertex attribute mapping processing result is specifically as follows:

[0026] S411: Perform a linear traversal of the vertex topology data of the neural pathway geometric model, identify the connectivity relationship of each vertex, obtain the coordinates of its predecessor and successor nodes on the path, construct a local spatial difference unit including the coordinates of three points, call the central difference numerical differentiation algorithm to perform vector addition and subtraction and step size square division operations on the local spatial difference unit, solve the second-order differential vector of each vertex position in three-dimensional space, and generate a vertex second-order derivative vector sequence;

[0027] S412: Read the three-dimensional vector components of each vertex in the second derivative vector sequence of the vertices, perform Euclidean norm operation of the square root of the sum of squares on the vector components to obtain the vector magnitude, define the vector magnitude as a curvature scalar characterizing the degree of local geometric curvature of the nerve fiber, call the linear normalization function to map the curvature scalar value to the standard numerical range of zero to one, and construct a geometric curvature feature value sequence that corresponds one-to-one with the vertex index order;

[0028] S413: Activate the vertex attribute buffer object of the graphics processing unit, parse the memory layout structure of the neural pathway geometric model, establish the data binding mapping logic between the vertex index list and the geometric curvature feature value sequence, write the feature value data into the custom attribute channel of the buffer, complete the memory interleaving and encapsulation of geometric topology data and physical attribute data, and obtain the vertex attribute mapping processing result.

[0029] The present invention is improved in that the specific steps for obtaining the neural pathway visualization image are as follows:

[0030] S511: Based on the vertex attribute mapping processing result, configure a frame buffer object with dual depth stripping, start multiple rendering loops for the neural pathway geometry model, and simultaneously sample the current depth texture and the next layer depth texture during the fragment shading stage, extract the curvature attribute data at the corresponding texture coordinates, and obtain inter-layer curvature feature pairs.

[0031] S512: Based on the interlayer curvature feature pair, obtain the curvature values ​​of the current depth layer and the next depth layer and the depth interval scalar between the two layers, and calculate the attenuated fragment transparency by combining the visual density parameter.

[0032] S513: Set the color blending mode from back to front, call the attenuated fragment transparency as the source factor to adjust the fragment color, accumulate the adjusted color components into the color buffer, and synthesize a neural pathway visualization image.

[0033] The present invention is improved in that the formula for obtaining the attenuated fragment transparency is specifically as follows:

[0034] ;

[0035] in, Represents the fragment transparency after attenuation. This represents the initial opacity channel value of the fragment. Represents the normalized damping coefficient of visual density. The normalized geometric curvature eigenvalues ​​represent the fragments at the current depth layer. The normalized geometric curvature eigenvalue represents the fragment of the next depth layer. This represents the normalized depth interval between the current fragment and the next depth layer fragment.

[0036] A three-dimensional modeling and rendering system for neural pathways, the system being used to implement the aforementioned three-dimensional modeling and rendering method for neural pathways, the system comprising:

[0037] The main feature analysis module acquires diffusion tensor imaging data of brain neural tissue, calculates the diffusion tensor matrix, decomposes the diffusion tensor matrix to obtain the main feature vector, and constructs the main feature vector field.

[0038] The trajectory correction processing module traverses the principal feature vector field to calculate the divergence, compares the divergence with the stability threshold, adjusts the spatial integration step size parameter, performs a cross product operation between the curl vector and the tracking direction to generate correction components, and generates a sequence of nerve fiber trajectory points based on the spatial integration step size parameter and the correction components.

[0039] The neural model construction module extracts local trajectory segments from the sequence of neural fiber trajectory points, converts the curvature derivative change rate of the local trajectory segments into frequency distribution features, identifies abnormal fiber segments, fits the direction vector of the abnormal fiber segments, and generates a geometric model of the neural pathway.

[0040] The vertex mapping implementation module calculates the magnitude of the second derivative vector of the vertex of the neural pathway geometric model, generates geometric curvature feature values, maps the geometric curvature feature values ​​to the vertex attribute buffer, and obtains the vertex attribute mapping processing result.

[0041] The rendering transparency adjustment module renders the neural pathway geometric model based on the vertex attribute mapping processing results, calculates the difference in geometric curvature feature values ​​of adjacent depth layer fragments, and attenuates the transparency value if the difference exceeds a preset difference threshold, and synthesizes a neural pathway visualization image based on the transparency value.

[0042] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0043] In this invention, the spatial step size is dynamically adjusted and the tracking direction is corrected by calculating the divergence and curl features of the principal eigenvector field, thereby enhancing the stability and continuity of the path in complex regions. After statistical analysis of the frequency distribution of the discrete curvature derivative rate of change of the trajectory point sequence, high-frequency disturbance segments are identified, and the geometric shape of abnormal segments is reconstructed by fitting the direction vector, effectively correcting the model distortion caused by curvature abrupt changes. The vertex data is used to obtain the geometric curvature value through center difference calculation and mapped to the attribute buffer to form local detail feature expression. During the rendering stage, a significance judgment is performed based on the difference in curvature values ​​between depth layers, and the attenuation function is called to control the transparency regionally, thereby improving the visual expressiveness of areas with drastic curvature changes and reducing the rendering intensity of structurally repetitive areas. This allows the image to retain structural details while optimizing rendering efficiency and enhancing the medical auxiliary interpretation effect of the 3D model. Attached Figure Description

[0044] Figure 1 This is a flowchart of the method of the present invention;

[0045] Figure 2 This is a flowchart of the process for obtaining the principal feature vector field in this invention;

[0046] Figure 3 This is a flowchart illustrating the process of obtaining a sequence of nerve fiber trajectory points according to the present invention;

[0047] Figure 4 This is a flowchart illustrating the process of obtaining a neural pathway geometric model for this invention.

[0048] Figure 5 This is a flowchart illustrating the process of obtaining vertex attribute mapping results according to the present invention;

[0049] Figure 6This is a flowchart illustrating the process of obtaining visual images of neural pathways for this invention. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0051] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0052] Please see Figure 1 This invention provides a technical solution, a method for three-dimensional modeling and rendering of neural pathways, comprising the following steps:

[0053] S1: Acquire diffusion tensor imaging data of brain neural tissue, perform tensor decomposition on diffusion tensor imaging data to extract the diffusion tensor matrix of voxels, perform eigenvalue decomposition operation on diffusion tensor matrix to obtain principal eigenvalues ​​and corresponding principal eigenvectors, and construct a principal eigenvector field reflecting the local orientation of nerve fibers based on the principal eigenvectors.

[0054] S2: Traverse the voxel coordinates in the principal eigenvector field, apply the divergence operator to calculate the vector field divergence value at the current coordinate, compare the vector field divergence value with the stability threshold, dynamically adjust the spatial integration step size parameter based on the comparison result, calculate the curl vector of the principal eigenvector field and perform a cross product operation with the current tracking direction to generate a correction component, and generate a neural fiber trajectory point sequence based on the adjusted spatial integration step size parameter and the growth direction after superimposing the correction component.

[0055] S3: Perform discrete curvature calculation along the path on the sequence of nerve fiber trajectory points, extract local trajectory segments based on a fixed-length sliding window, statistically analyze the first-order rate of change sequence of curvature derivatives within the local trajectory segments and convert it into frequency distribution features, identify abnormal fiber segments with high-frequency perturbations based on the frequency distribution features, and use the least squares method to curve fit and reconstruct the direction vectors at both ends of the abnormal fiber segments to generate a geometric model of the neural pathway.

[0056] S4: Traverse the vertex data of the neural pathway geometric model, calculate the second derivative vector of each vertex using the central difference method, calculate the magnitude of the second derivative vector to generate geometric curvature feature values, map the geometric curvature feature values ​​to the vertex attribute buffer of the neural pathway geometric model, and obtain the vertex attribute mapping processing result.

[0057] S5: Based on the vertex attribute mapping processing results, perform multiple depth stripping rendering operations on the neural pathway geometric model. In the rasterization stage, obtain the geometric curvature feature values ​​of the current fragment and the fragment of the next depth layer, calculate the difference between the two geometric curvature feature values, and if the difference exceeds the preset difference threshold, call the inverse proportional function to decay the transparency channel value of the current fragment according to the difference. Based on the processed transparency channel value, synthesize the neural pathway visualization image.

[0058] The principal eigenvector field includes voxel space coordinates, three-dimensional vector components, and local field divergence; the neural fiber trajectory point sequence includes three-dimensional vertex coordinates, path tangential vectors, and neighborhood point indices; the neural pathway geometric model includes fiber bundle topological connectivity, mesh vertex normals, and surface texture coordinates; the vertex attribute mapping processing results include vertex geometric curvature eigenvalues, vertex depth values, and vertex color attributes; and the neural pathway visualization image includes pixel color values, pixel depth values, and pixel transparency values.

[0059] Please see Figure 2 The specific steps for obtaining the principal eigenvector field are as follows:

[0060] S111: Acquire diffusion tensor imaging data of brain neural tissue, traverse the signal intensity of each voxel in the diffusion tensor imaging data, extract the diffusion weighted signal attenuation value of each voxel in multiple gradient directions, construct a diffusion signal distribution matrix including spatial position coordinates and directional signal intensity, call the least squares method to perform linear fitting operation on the diffusion signal distribution matrix, and generate a diffusion tensor matrix in three-dimensional voxel space.

[0061] Select three-dimensional spatial coordinates as The voxel is used as the processing unit, and the physical size of the voxel is set to be... The voxel was read from the DTI scanning device at the reference. The signal intensity (without diffusion weighting) and the signal strength in six non-collinear gradient directions are shown in Table 1, and a diffusion signal distribution matrix is ​​constructed. At the same time, establish the corresponding gradient direction matrix. Using the log-linearized form of the Stejskal-Tanner equation ,in For the first Diffusion-weighted signals in each direction, For undiffusion-weighted signals, Let the diffusion sensitivity coefficient be set to , For the first A unit vector in the gradient direction, For the diffusion tensor to be solved, an overdetermined system of equations is constructed. ,in Includes tensors The 6 independent components Call the Singular Value Decomposition (SVD) algorithm on the coefficient matrix Perform pseudo-inverse operation, i.e. Substituting the actual observed signals from Table 1 into the calculation, the diffusion tensor matrix at the voxel location is obtained. The diffusion tensor matrix in the three-dimensional voxel space is generated.

[0062] Table 1. Voxel Signal Sampling Table for Diffusion Tensor Imaging

[0063] ;

[0064] Table 1 lists the sampling data of a single voxel in the baseline and six gradient directions, which are used for linear fitting calculations.

[0065] S112: Read the symmetric positive definite second-order tensor corresponding to each voxel in the diffusion tensor matrix, calculate the characteristic equation root of the symmetric positive definite second-order tensor, solve the characteristic equation to obtain three real eigenvalues ​​and arrange them in descending order of numerical value, take the eigenvector direction corresponding to the largest eigenvalue as the local main diffusion direction of the nerve fiber, and establish a voxel eigenvector set including the eigenvalue sequence and the main diffusion direction vector.

[0066] Determine if the matrix satisfies the symmetric positive definite property; if so, construct its characteristic equation. ,in It is the identity matrix. For the eigenvalues, expanding the determinant yields information about... The cubic equation of By using Cardano's formula to solve the three real roots of the cubic equation, we obtain the set of eigenvalues. The specific calculated eigenvalues ​​are as follows: , , Sort the values ​​in descending order. Select the largest eigenvalue The corresponding eigenvector equation The eigenvectors are obtained by solving the problem using Gaussian elimination. , will vector After normalization, it is defined as the local principal diffusion direction of the nerve fiber of this voxel, and the sorted eigenvalue sequence is also defined. With the main diffusion direction vector Combine them to establish a voxel feature vector set.

[0067] S113: Extract the main diffusion direction vector of each voxel position based on the voxel feature vector set, construct a vector field data structure continuously distributed in three-dimensional space, calculate the dot product between the main diffusion direction vectors of adjacent voxels, determine the consistency weight of local orientation, and perform smooth interpolation processing on the singular region in the vector field according to the consistency weight to generate a main feature vector field that reflects the local orientation of nerve fibers.

[0068] exist Extract each coordinate point from the 3D voxel mesh The main diffusion direction vector at that location Construct a three-dimensional vector field data structure and traverse any adjacent voxel pairs in the vector field. and Calculate the dot product between the two. Set consistency weights for local trends ,like This indicates that the direction of the principal feature vector has been flipped by 180 degrees, at which point... Adjusted to To maintain directional continuity and record correction operations, for singular regions in the vector field (i.e., regions where the absolute value of the dot product is less than 0.5), interpolation is performed using a weighted average of the surrounding 26 neighborhood voxels. The weights are set to the reciprocal of the distance between the neighborhood voxels and the central voxel. For example, for coordinates... At the singularity, calculate the sum of its 26 neighborhood weighted vectors and perform normalization to generate a smooth principal feature vector field that reflects the local orientation of the nerve fiber.

[0069] Please see Figure 3 The specific steps for obtaining the neural fiber trajectory point sequence are as follows:

[0070] S211: Traverse the voxel grid of the principal eigenvector field, calculate the vector field divergence scalar at each grid point using the central difference method, and compare the vector field divergence scalar with the preset stability threshold. If the divergence scalar is higher than the stability threshold, the basic step size is reduced proportionally according to the magnitude of the divergence scalar; otherwise, the basic step size remains unchanged. Establish an adaptive integration step size parameter that adapts to the degree of local field distortion.

[0071] The process of setting the stability threshold is as follows: by traversing the divergence magnitude of all effective voxels in the vector field, the arithmetic mean and standard deviation of the divergence values ​​of the entire field are calculated, and the sum of the arithmetic mean and the standard deviation of the preset multiple is used as the stability threshold.

[0072] For grid points The divergence scalar at this point is calculated using the central difference method. ,in All voxels have a side length of 1 mm. For example, if the divergence modulus is calculated to be 0.35 at a certain point, the stability threshold setting process is performed, selecting all effective voxels in the vector field (totaling...). (each), calculate the arithmetic mean of its divergence values. Calculated Calculate the standard deviation Calculated If the preset multiplier is set to 2.0, then the stability threshold is... The current point dispersion of 0.35 is compared with the threshold of 0.31, because... The system determines that the field distortion in this region is significant, and adjusts the preset base step size accordingly. To scale down proportionally, the calculation formula is: If the divergence is less than the threshold, it remains unchanged at 0.5 mm, thus establishing an adaptive integral step size parameter that adapts to the degree of local field distortion.

[0073] S212: Call the curl operator to extract the local rotation features of the principal feature vector field at the current coordinates, construct a three-dimensional curl vector, obtain the tracking tangential vector of the current nerve fiber, perform a vector cross product operation on the three-dimensional curl vector and the tracking tangential vector to obtain a normal correction vector perpendicular to the current path, superimpose the normal correction vector on the tracking tangential vector and perform normalization processing to generate the geometric correction growth direction;

[0074] Calling the curl operator Extract the principal feature vector field at the current coordinates Calculate the components of the curl vector based on the local rotational characteristics at the location: , , Substituting the local field data, the three-dimensional curl vector is calculated. Obtain the tracking tangential vector of the current nerve fiber. (Normalized), perform vector cross product operation. The normal correction vector is calculated. The corrective force is superimposed on the tracking tangential vector with a preset corrective strength coefficient of 0.1, i.e. ,right Perform Euclidean norm normalization to generate geometric correction growth directions.

[0075] S213: Using the current fiber endpoint as the starting coordinate, the adaptive integral step size parameter is called as the spatial displacement modulus, and the geometric correction growth direction is combined as the direction component of the displacement vector. The fourth-order Runge-Kutta integral operation is performed to calculate the three-dimensional spatial coordinates of the next trajectory point. The calculated new coordinates are added to the current path array and the iteration process is repeated to generate a continuous sequence of nerve fiber trajectory points.

[0076] With the current fiber endpoint Using the starting coordinates, call the adaptive integration step size parameter. As the spatial displacement modulus, combined with the geometric correction growth direction As the directional component of the displacement vector, the fourth-order Runge-Kutta integral is performed, first calculating the four intermediate slopes: , , , The weighted average yields the increment vector. Calculate the three-dimensional spatial coordinates of the next trajectory point. The new coordinates are appended to the current path array. If the path length does not reach the termination condition, the above iterative process is repeated to generate a continuous sequence of neural fiber trajectory points.

[0077] Please see Figure 4 The specific steps for obtaining the geometric model of the neural pathway are as follows:

[0078] S311: Perform discrete differential operation along the path on the sequence of nerve fiber trajectory points, calculate the angle between the tangents of adjacent trajectory points, extract the curvature values ​​point by point, segment the curvature value sequence using a fixed-length sliding window, calculate the rate of change of curvature values ​​within the window, perform discrete Fourier transform to convert the rate of change sequence in the spatial domain into energy distribution data in the frequency domain, and generate local curvature frequency distribution features.

[0079] Nerve fiber trajectory point sequence Perform discrete differentiation along the path to calculate the angle between the tangents of adjacent trajectory points. ,in Extracting point-by-point curvature values The curvature numerical sequence is segmented using a fixed-length sliding window with 32 sampling points. For the first... Curvature sequence within a window Calculate its rate of change of curvature The Fast Fourier Transform (FFT) algorithm is used to convert the spatial rate of change sequence into frequency domain energy distribution data, resulting in a spectral sequence. Extract its amplitude As an energy distribution, it generates local curvature frequency distribution characteristics.

[0080] S312: Traverse the spectral components in the local curvature frequency distribution features, calculate the energy integral ratio of the high-frequency band, compare the energy integral ratio with the preset disturbance discrimination threshold, identify the window index position where the integral ratio exceeds the disturbance discrimination threshold, extract the coordinates of continuous trajectory points in the corresponding index interval, and define the disturbance abnormal section caused by numerical oscillation resulting in geometric distortion.

[0081] The specific process of setting the disturbance discrimination threshold is as follows: select a set of neural fiber trajectory data that has been verified to be geometrically smooth as a reference set, perform discrete Fourier transform on the data in the reference set, extract the distribution sequence of the energy proportion of the high frequency band, calculate the mathematical expectation and standard deviation of the distribution sequence, and use the mathematical expectation plus the standard deviation of the preset multiple as the disturbance discrimination threshold.

[0082] Set the high-frequency band as the normalized frequency Calculate the percentage of energy integral within the specified interval. Calculated by setting a certain window The process of setting the perturbation discrimination threshold involves selecting a set of neural fiber trajectory data (sample size) that has been manually verified by experts to be geometrically smooth. Using this as a reference set, the same FFT transformation is performed on the reference set data, and the proportion of high-frequency energy is extracted to calculate the mathematical expectation value of the distribution sequence. Standard deviation Set the preset multiplier to 3.0 (based on 3). (Principles) Calculate the disturbance discrimination threshold The current window's integral percentage of 0.45 is compared with the threshold of 0.27. The system identifies the trajectory index range corresponding to the window (e.g., index 120 to 152) as an anomaly, and defines the abnormal disturbance segment caused by numerical oscillation leading to geometric distortion.

[0083] S313: Extract the tangential direction vectors of the connection points at both ends of the perturbation abnormal segment, establish a parameterized curve equation with the connection point position and tangential vector as boundary constraints, call the least squares method to iteratively solve the optimal fitting coefficient of the equation, construct a smooth interpolation curve and replace the original abnormal trajectory point data, integrate the optimized fiber path data, and generate a neural pathway geometric model.

[0084] Extract the connection points at both ends of the disturbance anomaly segment (index 120 to 152). (Index 119) and Coordinates and tangential direction vector of (index 153) and Establish the equation of the cubic Hermite interpolation curve. ,in To normalize the parameters, the least squares method is used to approximate the points within the original outlier segment to optimize the parameters. The distribution of the data is analyzed, the optimal fitting coefficients are calculated, and 33 smooth interpolation coordinate points are generated by resampling at 0.5mm intervals. These points replace the original abnormal trajectory point data, and the optimized fiber path data is integrated to generate a neural pathway geometric model.

[0085] Please seeFigure 5 The specific steps for obtaining the vertex attribute mapping processing results are as follows:

[0086] S411: Perform linear traversal of the vertex topology data of the neural pathway geometric model, identify the connectivity relationship of each vertex, obtain the coordinates of its predecessor and successor nodes on the path, construct a local spatial difference unit including the coordinates of three points, call the central difference numerical differentiation algorithm to perform vector addition and subtraction and step size square division operations on the local spatial difference unit, solve the second-order differential vector of each vertex position in three-dimensional space, and generate a vertex second-order derivative vector sequence;

[0087] A linear traversal is performed on the topological data of 1500 vertices contained in the neural pathway geometric model, for the index... vertex Identify its predecessor node and successor nodes coordinates, for example , , Construct local spatial difference units, step size Taking the average Euclidean distance of two adjacent points as 0.55 mm, the second-order differential vector is calculated using the central difference numerical differentiation algorithm. Substituting the numerical values, we get Components , Components , Components The vector of that point in the sequence of second derivative vectors of the vertex is obtained as follows: .

[0088] S412: Read the three-dimensional vector components of each vertex in the second derivative vector sequence of the vertices, perform Euclidean norm operation of the square root of the sum of squares on the vector components, obtain the vector magnitude, define the vector magnitude as a curvature scalar characterizing the degree of local geometric curvature of the nerve fiber, call the linear normalization function to map the curvature scalar value to the standard numerical range of zero to one, and construct a geometric curvature feature value sequence that corresponds one-to-one with the vertex index order.

[0089] Read the second derivative vector sequence of the vertex for that vertex vector components Perform Euclidean norm operations to obtain the vector magnitude. The model length is defined as a curvature scalar characterizing the degree of local geometric curvature of the nerve fiber, and the maximum value of the curvature scalar is obtained by traversing the entire model. and minimum value Call the linear normalization function Mapping the current curvature scalar 0.467 to obtain the standard value. Construct a sequence of geometric curvature feature values ​​that correspond one-to-one with the vertex index order.

[0090] S413: Activate the vertex attribute buffer object of the graphics processing unit, parse the memory layout structure of the neural pathway geometric model, establish the data binding mapping logic between the vertex index list and the geometric curvature feature value sequence, write the feature value data into the custom attribute channel of the buffer, complete the memory interleaving and encapsulation of geometric topology data and physical attribute data, and obtain the vertex attribute mapping processing result.

[0091] Activate the vertex attribute buffer object of the graphics processing unit, parse the memory layout structure of the neural pathway geometry model, specify the attribute index channel ID as 3 (custom attribute), call the API function glVertexAttribPointer, set the data type to GL_FLOAT, the step size to sizeof(float), write the calculated normalized feature value data (such as 0.1868) into the custom attribute channel of the buffer according to the IEEE754 single-precision floating-point standard, complete the memory interleaving and encapsulation of geometric topology data and physical attribute data, and obtain the vertex attribute mapping processing result.

[0092] Please see Figure 6 The specific steps for obtaining visualization images of neural pathways are as follows:

[0093] S511: Configure a frame buffer object with dual depth stripping based on vertex attribute mapping processing results, start multiple rendering loops for the neural pathway geometry model, and simultaneously sample the current depth texture and the next layer depth texture during the fragment shading stage to extract curvature attribute data at the corresponding texture coordinates and obtain inter-layer curvature feature pairs.

[0094] Based on a framebuffer object with dual depth stripping, and two depth attachments, Depth_Tex_1 and Depth_Tex_2, four rendering loops are initiated for the neural pathway geometry model. During the fragment shader stage, the current depth texture is synchronously sampled using the texelFetch instruction (the first...). (layer) and the next layer of depth texture (the first layer) (Layer) in screen coordinates The data at the specified location is used, and the curvature attribute data at the corresponding texture coordinates is extracted using MRT (Multiple Render Target) technology to obtain the inter-layer curvature feature pairs. .

[0095] S512: Based on interlayer curvature feature pairs, obtain the curvature values ​​of the current depth layer and the next depth layer, as well as the depth interval scalar between the two layers. Combined with visual density parameters, the formula is used:

[0096] ;

[0097] The attenuated fragment opacity is calculated.

[0098] in, Represents the fragment transparency after attenuation. Representing the initial transparency channel value of the fragment, the raw alpha channel data is read from the texture properties of the neural pathway geometry model. Represents the visual density normalized damping coefficient, a dimensionless constant preset according to the scene rendering configuration to control the rate of transparency decay. The normalized geometric curvature feature value representing the fragment of the current depth layer is obtained by sampling the geometric curvature data of the vertex attribute buffer of the current rasterized depth layer. The normalized geometric curvature feature value representing the fragment of the next depth layer is obtained by sampling the geometric curvature data at the corresponding position of the texture of the next rasterized depth layer. This represents the normalized depth interval between the current fragment and the next depth layer fragment. It is the value obtained after calculating the difference between the current fragment depth value and the next layer fragment depth value and normalizing it.

[0099] Obtain the curvature values ​​of the current depth layer and the next depth layer, as well as the depth interval scalar between the two layers. Combine this with the visual density parameter and use the formula:

[0100] ;

[0101] in, Set it to 0.8 (meaning an initial opacity of 80%). Set to 0.3, this coefficient is a dimensionless constant preset according to the scene rendering configuration to control the rate of transparency decay. Its value range is usually between 0.1 and 0.5, which is used to prevent the denominator from being too small and causing numerical overflow and to control the decay benchmark. The value is 0.1868 (from S412). The value is 0.5 (the next layer sample value), and the difference between the two is -0.3132; The depth difference is set to 0.2 for the current layer and 0.25 for the next layer, resulting in a difference of 0.05, which remains 0.05 after normalization; the denominator term in the formula... The combined distance between the feature space (curvature difference) and the physical space (depth distance) was calculated using the Pythagorean theorem, and its value is... This composite distance is used to quantify the geometric and spatial separation between the two layers of fragments; the attenuated fragment transparency is calculated. The results show that when the curvature difference between two layers of fragments is large or the distance between them is large, the contribution of the later layer fragments to the occlusion of the earlier layer is significantly reduced, and the transparency decreases from 0.8 to 0.389, thereby enhancing the perspective effect of the internal structure.

[0102] The formula integrates geometric differences and depth distance information, constructing a fragment spatial separation index using the Pythagorean theorem. The design logic for attenuating transparency is based on the visual occlusion mechanism in volume rendering. In the field of graphics rendering, transparency control often relies on the geometric differences and distances between fragments. For example, in medical image volume rendering, a greater curvature indicates more drastic fiber changes, and its perspective structure should be emphasized. This application employs a visual weight attenuation strategy, significantly reducing the transparency of fragments with large geometric / depth differences to enhance layered visibility.

[0103] S513: Set the color blending mode from back to front, call the attenuated fragment transparency as the source factor to adjust the fragment color, accumulate the adjusted color components into the color buffer, and synthesize the neural pathway visualization image.

[0104] Set the color blending mode, i.e., glBlendFunc(GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA), and call the attenuated fragment opacity. As the source factor for the RGB color components of the current fragment (For example Adjustments are made and the mixed colors are calculated. ,in The adjusted color components are accumulated into the color buffer to represent the background color that has been accumulated in the color buffer. After completing the loop of all peeling layers, a neural pathway visualization image is synthesized.

[0105] The color components are adjusted using Aout; the adjusted results are then written to the frame buffer. This part clarifies the formula's position and calling path in the graphics rendering process, demonstrating its operability and implementability.

[0106] A 3D modeling and rendering system for neural pathways, used to implement the aforementioned 3D modeling and rendering method for neural pathways, the system comprising:

[0107] The main feature analysis module acquires diffusion tensor imaging data of brain neural tissue, calculates the diffusion tensor matrix, decomposes the diffusion tensor matrix to obtain the main feature vector, and constructs the main feature vector field.

[0108] The trajectory correction processing module traverses the principal feature vector field to calculate the divergence, compares the divergence with the stability threshold, adjusts the spatial integration step size parameter, performs a cross product operation between the curl vector and the tracking direction to generate correction components, and generates a sequence of nerve fiber trajectory points based on the spatial integration step size parameter and the correction components.

[0109] The neural model construction module extracts local trajectory segments from the sequence of neural fiber trajectory points, transforms the curvature derivative change rate of the local trajectory segments into frequency distribution features, identifies abnormal fiber segments, fits the direction vector of the abnormal fiber segments, and generates a geometric model of the neural pathway.

[0110] The vertex mapping implementation module calculates the magnitude of the second derivative vector of the vertex in the neural pathway geometric model, generates geometric curvature feature values, maps the geometric curvature feature values ​​to the vertex attribute buffer, and obtains the vertex attribute mapping processing result.

[0111] The rendering transparency adjustment module renders the neural pathway geometric model based on the vertex attribute mapping processing results, calculates the difference in geometric curvature feature values ​​of adjacent depth layer fragments, and attenuates the transparency value if the difference exceeds a preset difference threshold. Based on the transparency value, a visual image of the neural pathway is synthesized.

[0112] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for three-dimensional modeling and rendering of neural pathways, characterized in that, Includes the following steps: S1: Obtain diffusion tensor imaging data of brain neural tissue, solve for the diffusion tensor matrix, decompose the diffusion tensor matrix to obtain the principal eigenvectors, and construct the principal eigenvector field; S2: Traverse the principal feature vector field to calculate the divergence. If the divergence scalar is higher than the stability threshold, reduce the basic step size proportionally according to the magnitude of the divergence scalar. Otherwise, keep the basic step size unchanged. Extract the local rotation features of the principal feature vector field at the current coordinates, obtain the tracking tangential vector of the current nerve fiber, perform a cross product operation between the curl vector and the tracking direction to generate a correction component, and generate a sequence of nerve fiber trajectory points based on the spatial integral step size parameter and the correction component. S3: Extract local trajectory segments from the neural fiber trajectory point sequence, convert the curvature derivative change rate of the local trajectory segments into frequency distribution features, identify abnormal fiber segments, fit the direction vector of the abnormal fiber segments, and generate a neural pathway geometric model. S4: Calculate the magnitude of the second derivative vector of the vertex of the neural pathway geometric model, generate geometric curvature feature values, map the geometric curvature feature values ​​to the vertex attribute buffer, and obtain the vertex attribute mapping processing result; S5: Based on the vertex attribute mapping processing result, the neural pathway geometric model is rendered, the difference in geometric curvature feature values ​​of adjacent depth layer fragments is calculated, and if the difference exceeds a preset difference threshold, the transparency value is attenuated, and a neural pathway visualization image is synthesized based on the transparency value.

2. The method for three-dimensional modeling and rendering of neural pathways according to claim 1, characterized in that, The principal feature vector field includes voxel space coordinates, three-dimensional vector components, and local field divergence; the neural fiber trajectory point sequence includes three-dimensional vertex coordinates, path tangential vectors, and neighborhood point indices; the neural pathway geometric model includes fiber bundle topological connectivity, mesh vertex normals, and surface texture coordinates; the vertex attribute mapping processing result includes vertex geometric curvature feature values, vertex depth values, and vertex color attributes; and the neural pathway visualization image includes pixel color values, pixel depth values, and pixel transparency values.

3. The method for three-dimensional modeling and rendering of neural pathways according to claim 2, characterized in that, The specific steps for obtaining the principal feature vector field are as follows: S111: Acquire diffusion tensor imaging data of brain neural tissue, traverse the signal intensity of each voxel in the diffusion tensor imaging data, extract the diffusion weighted signal attenuation value of each voxel in multiple gradient directions, construct a diffusion signal distribution matrix including spatial position coordinates and directional signal intensity, call the least squares method to perform linear fitting operation on the diffusion signal distribution matrix, and generate a diffusion tensor matrix in three-dimensional voxel space. S112: Read the symmetric positive definite second-order tensor corresponding to each voxel in the diffusion tensor matrix, calculate the characteristic equation root of the symmetric positive definite second-order tensor, solve the characteristic equation to obtain three real eigenvalues ​​and arrange them in descending order of numerical value, take the eigenvector direction corresponding to the largest eigenvalue as the local main diffusion direction of the nerve fiber, and establish a voxel eigenvector set including the eigenvalue sequence and the main diffusion direction vector. S113: Based on the voxel feature vector set, extract the main diffusion direction vector of each voxel position, construct a vector field data structure continuously distributed in three-dimensional space, calculate the dot product between the main diffusion direction vectors of adjacent voxels, determine the consistency weight of the local orientation, and perform smooth interpolation processing on the singular region in the vector field according to the consistency weight to generate a main feature vector field reflecting the local orientation of nerve fibers.

4. The method for three-dimensional modeling and rendering of neural pathways according to claim 3, characterized in that, The specific steps for obtaining the neural fiber trajectory point sequence are as follows: S211: Traverse the voxel grid of the principal eigenvector field, calculate the vector field divergence scalar at each grid point using the central difference method, and compare the vector field divergence scalar with a preset stability threshold. If the divergence scalar is higher than the stability threshold, reduce the basic step size proportionally according to the magnitude of the divergence scalar; otherwise, keep the basic step size unchanged, and establish an adaptive integration step size parameter that adapts to the degree of local field distortion. S212: Call the curl operator to extract the local rotation features of the principal feature vector field at the current coordinates, construct a three-dimensional curl vector, obtain the tracking tangential vector of the current nerve fiber, perform a vector cross product operation on the three-dimensional curl vector and the tracking tangential vector to obtain a normal correction vector perpendicular to the current path, superimpose the normal correction vector on the tracking tangential vector and perform normalization processing to generate a geometric correction growth direction; S213: Using the current fiber endpoint as the starting coordinate, the adaptive integral step size parameter is called as the spatial displacement modulus, and the geometric correction growth direction is combined as the direction component of the displacement vector. The fourth-order Runge-Kutta integral operation is performed to calculate the three-dimensional spatial coordinates of the next trajectory point. The calculated new coordinates are added to the current path array and the iteration process is repeated to generate a continuous sequence of nerve fiber trajectory points.

5. The method for three-dimensional modeling and rendering of neural pathways according to claim 4, characterized in that, The specific steps for obtaining the geometric model of the neural pathway are as follows: S311: Perform discrete differential operation along the path on the nerve fiber trajectory point sequence, calculate the angle between the tangents of adjacent trajectory points, extract the curvature values ​​point by point, segment the curvature value sequence using a fixed-length sliding window, calculate the rate of change of curvature values ​​within the window, perform discrete Fourier transform to convert the spatial rate of change sequence into frequency domain energy distribution data, and generate local curvature frequency distribution features. S312: Traverse the spectral components in the local curvature frequency distribution features, calculate the energy integral ratio of the high-frequency band, compare the energy integral ratio with the preset disturbance discrimination threshold, identify the window index position where the integral ratio exceeds the disturbance discrimination threshold, extract the coordinates of continuous trajectory points in the corresponding index interval, and define the disturbance abnormal section caused by numerical oscillation resulting in geometric distortion. S313: Extract the tangential direction vectors of the connection points at both ends of the perturbation abnormal segment, establish a parameterized curve equation with the connection point position and tangential vector as boundary constraints, call the least squares method to iteratively solve the optimal fitting coefficients of the equation, construct a smooth interpolation curve and replace the original abnormal trajectory point data, integrate the optimized fiber path data, and generate a neural pathway geometric model.

6. The method for three-dimensional modeling and rendering of neural pathways according to claim 5, characterized in that, The process of setting the disturbance discrimination threshold is as follows: a set of neural fiber trajectory data that has been verified to be geometrically smooth is selected as a reference set; a discrete Fourier transform is performed on the data in the reference set to extract the distribution sequence of the energy proportion of the high-frequency band; the mathematical expectation value and standard deviation of the distribution sequence are calculated; and the mathematical expectation value is superimposed with the standard deviation by a preset multiple as the disturbance discrimination threshold.

7. The method for three-dimensional modeling and rendering of neural pathways according to claim 6, characterized in that, The specific steps for obtaining the vertex attribute mapping processing result are as follows: S411: Perform a linear traversal of the vertex topology data of the neural pathway geometric model, identify the connectivity relationship of each vertex, obtain the coordinates of its predecessor and successor nodes on the path, construct a local spatial difference unit including the coordinates of three points, call the central difference numerical differentiation algorithm to perform vector addition and subtraction and step size square division operations on the local spatial difference unit, solve the second-order differential vector of each vertex position in three-dimensional space, and generate a vertex second-order derivative vector sequence; S412: Read the three-dimensional vector components of each vertex in the second derivative vector sequence of the vertices, perform Euclidean norm operation of the square root of the sum of squares on the vector components to obtain the vector magnitude, define the vector magnitude as a curvature scalar characterizing the degree of local geometric curvature of the nerve fiber, call the linear normalization function to map the curvature scalar value to the standard numerical range of zero to one, and construct a geometric curvature feature value sequence that corresponds one-to-one with the vertex index order; S413: Activate the vertex attribute buffer object of the graphics processing unit, parse the memory layout structure of the neural pathway geometric model, establish the data binding mapping logic between the vertex index list and the geometric curvature feature value sequence, write the feature value data into the custom attribute channel of the buffer, complete the memory interleaving and encapsulation of geometric topology data and physical attribute data, and obtain the vertex attribute mapping processing result.

8. The method for three-dimensional modeling and rendering of neural pathways according to claim 7, characterized in that, The specific steps for obtaining the visualization images of the neural pathways are as follows: S511: Based on the vertex attribute mapping processing result, configure a frame buffer object with dual depth stripping, start multiple rendering loops for the neural pathway geometry model, and simultaneously sample the current depth texture and the next layer depth texture during the fragment shading stage, extract the curvature attribute data at the corresponding texture coordinates, and obtain inter-layer curvature feature pairs. S512: Based on the interlayer curvature feature pair, obtain the curvature values ​​of the current depth layer and the next depth layer and the depth interval scalar between the two layers, and calculate the attenuated fragment transparency by combining the visual density parameter. S513: Set the color blending mode from back to front, call the attenuated fragment transparency as the source factor to adjust the fragment color, accumulate the adjusted color components into the color buffer, and synthesize a neural pathway visualization image.

9. The method for three-dimensional modeling and rendering of neural pathways according to claim 8, characterized in that, The specific formula for obtaining the attenuated fragment opacity is as follows: ； in, Represents the fragment transparency after attenuation. This represents the initial opacity channel value of the fragment. Represents the normalized damping coefficient of visual density. The normalized geometric curvature eigenvalues ​​represent the fragments at the current depth layer. The normalized geometric curvature eigenvalue represents the fragment of the next depth layer. This represents the normalized depth interval between the current fragment and the next depth layer fragment.

10. A three-dimensional modeling and rendering system for neural pathways, characterized in that, The system is used to implement the neural pathway three-dimensional modeling and rendering method according to any one of claims 1-9, and the system includes: The main feature analysis module acquires diffusion tensor imaging data of brain neural tissue, calculates the diffusion tensor matrix, decomposes the diffusion tensor matrix to obtain the main feature vector, and constructs the main feature vector field. The trajectory correction processing module traverses the principal feature vector field to calculate the divergence, compares the divergence with the stability threshold, adjusts the spatial integration step size parameter, performs a cross product operation between the curl vector and the tracking direction to generate correction components, and generates a sequence of nerve fiber trajectory points based on the spatial integration step size parameter and the correction components. The neural model construction module extracts local trajectory segments from the sequence of neural fiber trajectory points, converts the curvature derivative change rate of the local trajectory segments into frequency distribution features, identifies abnormal fiber segments, fits the direction vector of the abnormal fiber segments, and generates a geometric model of the neural pathway. The vertex mapping implementation module calculates the magnitude of the second derivative vector of the vertex of the neural pathway geometric model, generates geometric curvature feature values, maps the geometric curvature feature values ​​to the vertex attribute buffer, and obtains the vertex attribute mapping processing result. The rendering transparency adjustment module renders the neural pathway geometric model based on the vertex attribute mapping processing results, calculates the difference in geometric curvature feature values ​​of adjacent depth layer fragments, and attenuates the transparency value if the difference exceeds a preset difference threshold, and synthesizes a neural pathway visualization image based on the transparency value.

Images