Virtual three-dimensional reconstruction system and method for digestive tract
Patent Information
- Application Number
- CN202510874070.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-27
- Publication Date
- 2025-10-17
Smart Images
Figure CN120807831A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of 3D modeling, in particular to a virtual three-dimensional reconstruction system and method for a digestive tract. BACKGROUND
[0002] D modeling is a technology that constructs the geometric shape and structural characteristics of an object in three-dimensional space through digital means, and is widely used in fields such as industrial design, digital twin, medical imaging, virtual reality, game development, and film production.
[0003] The prior art mainly uses a general three-dimensional modeling framework in the modeling process, which can express the spatial form, but it is difficult to realize hierarchical geometric restoration for medical structures, usually only the external contour is retained, and there is a lack of independent modeling capability for internal and external structures, resulting in insufficient precision. In terms of material expression, traditional 3D models mostly stay in the stage of visual texture assignment, and it is difficult to truly map the physical properties such as thickness and hardness existing in the entity structure to the model structure. This deficiency makes the model only have a display function and is difficult to participate in deformation simulation or mechanical analysis in a real scene. Therefore, improvement is needed. SUMMARY
[0004] The purpose of the present application is to solve the shortcomings in the prior art and to provide a virtual three-dimensional reconstruction system and method for a digestive tract.
[0005] In order to achieve the above-mentioned purpose, the present application adopts the following technical scheme: the virtual three-dimensional reconstruction system for the digestive tract comprises:
[0006] An initial geometric construction module, based on a sequence of tomographic images of medical images, identifies the contour boundary points of the digestive tract region in the sequence of tomographic images, performs three-dimensional space point connection and initial surface mesh generation, and smoothes the generated initial surface mesh to establish an initial lumen surface model;
[0007] A layered shell construction module, based on the geometric data of the initial lumen surface model and a preset first wall layer target thickness value of the digestive tract, performs offset calculation along the normal direction of each vertex of the initial lumen surface model, obtains first layer offset vertex coordinates, connects the coordinate points to generate a first wall layer outer surface mesh based on the first layer offset vertex coordinates, and establishes a multi-layer digestive tract shell model;
[0008] A material thickness calibration module, based on the hierarchical geometric data of the multi-layer digestive tract shell model, selects a target layer, calculates the distance between the inner surface mesh vertex and the corresponding outer surface mesh vertex in the target layer, obtains a target layer vertex thickness value set, and based on the target layer vertex thickness value set, assigns a preset material identifier and the calculated thickness value to the layer mesh unit to establish an attributed digestive tract entity model;
[0009] a dynamic morphological evolution module, based on the geometry and attribute data of the attributed digestive tract entity model and the intraluminal target pressure value, adjusting the three-dimensional coordinates of the vertices of the inner surface of the attributed digestive tract entity model to obtain a pressure-adjusted geometric morphology, based on the pressure-adjusted geometric morphology and in accordance with the peristaltic wave characteristics in the physiological parameter set, performing a sequence displacement value calculation on the model vertices in accordance with the peristaltic wave propagation law to establish a dynamic deformation geometry sequence.
[0010] Preferably, the step of obtaining the initial luminal surface model is:
[0011] reading a sequence of tomographic images of the medical image, performing image segmentation on each frame of image, identifying the digestive tract region in the image and extracting all boundary contour points to form a contour boundary point set of the digestive tract region;
[0012] based on the contour boundary point set of the digestive tract region, sequentially pairing the three-dimensional coordinates of the corresponding boundary points in adjacent image frames, and connecting the curves based on the point pairs to form an initial surface mesh by collecting all the connection results;
[0013] according to the initial surface mesh, performing geometric curvature analysis on each group of triangular facets, marking the sharp connection region and calling the position relationship of the boundary points of the continuous adjacent facets for smooth interpolation to form an initial luminal surface model.
[0014] Preferably, the step of obtaining the first layer offset vertex coordinates is:
[0015] based on the initial luminal surface model, extracting all triangular facets constituting the initial luminal surface model, parsing each triangular facet into three vertex coordinate combinations according to the vertex number, calculating the normal vector of the triangular facet by three-dimensional vector cross multiplication, and accumulating the normal vector value of the three vertices corresponding to the triangular facet respectively, after completing the accumulation of all triangular facets, dividing the accumulated normal vector of each vertex by the corresponding accumulation times to generate a unit normal vector set of each vertex of the initial luminal surface model;
[0016] according to the unit normal vector set of each vertex of the initial luminal surface model, reading the three-dimensional coordinate value corresponding to each vertex, and calling a preset target thickness value of the first wall layer of the digestive tract, multiplying each unit normal vector by the target thickness value of the first wall layer of the digestive tract to obtain the offset displacement vector of each vertex, and then adding the offset displacement vector and the original three-dimensional coordinate value by coordinate axis dimension by component to generate the new three-dimensional space coordinates of each vertex after offset;
[0017] based on the new three-dimensional space coordinates of each vertex after offset, rearranging and organizing the points in accordance with the index order of the original initial luminal surface model vertex number, eliminating the decimal residual error caused by floating point error in the coordinate expression and performing uniform precision reduction to generate the first layer offset vertex coordinates.
[0018] Preferably, the step of obtaining the multi-layer digestive tract shell model is:
[0019] Based on the first layer offset vertex coordinates, the original connection relationship of the corresponding matching triangular patches is matched and the vertex index mapping is established in the order of the triangular patch topology structure in the initial lumen surface model, and each group of three offset vertex coordinates is called to perform spatial connection, thereby constructing a set of triangular patches with consistent structure and generating a first wall layer outer surface mesh;
[0020] According to the first wall layer outer surface mesh, all vertex coordinates of the mesh are extracted as a new round of offset reference coordinate set, and each subsequent wall layer target thickness value is read in turn, and offset operation in the unit normal direction of each vertex is performed for each target thickness value. A new vertex set of each level is constructed by accumulating offset coordinates, and a vertex coordinate set of all wall layers is obtained;
[0021] Based on the vertex coordinate set of all wall layers, the triangular patch structure is reused to generate the corresponding surface mesh layer by layer according to the connection order of the triangular patches of the previous layer corresponding to each layer offset coordinate set, and the surface meshes of each layer are topologically closed and organized and integrated in order of level, thereby generating a multi-layer digestive tract shell model.
[0022] Preferably, the step of obtaining the target layer vertex thickness value set is:
[0023] Based on the hierarchical geometric data of the multi-layer digestive tract shell model, the structure attribute identifier of each layer is traversed and the contained vertex group number, vertex index range, connection order and hierarchical number are read, matching is performed according to the input target layer number, the model target layer corresponding to the number is locked, and the three-dimensional coordinates of all inner surface mesh vertices and outer surface mesh vertices in the model target layer are extracted as a structured coordinate pair set, thereby generating a model target layer vertex coordinate pair set;
[0024] According to the model target layer vertex coordinate pair set, the three-dimensional coordinates of each group of inner surface mesh vertices and corresponding outer surface mesh vertices are traversed in a loop, the coordinate difference values in the X-axis, Y-axis and Z-axis directions are calculated for each group of coordinate pairs, and the sum of the squares of the difference values in the three directions is calculated to construct a sequence of real geometric distance values between vertices. Then, all distance values are organized into an original thickness value sequence according to the vertex number index correspondence relationship, thereby generating a model target layer three-dimensional space thickness distance sequence;
[0025] Based on the model target layer three-dimensional space thickness distance sequence, each distance value is rounded according to the rounding rule, and after the processing is completed, the original vertex index order is reordered, thereby generating a target layer vertex thickness value set.
[0026] Preferably, the step of obtaining the attribute digestive tract entity model is:
[0027] Based on the target layer vertex thickness value set, read the mesh cell structure information corresponding to the model target layer, traverse each cell index in the mesh cell, sequentially obtain the associated three or four vertex numbers, and match and read all vertex numbers in the target layer vertex thickness value set to form a thickness value set corresponding to each mesh cell, and generate a target layer mesh cell thickness mapping set;
[0028] According to the target layer mesh cell thickness mapping set, read the preset material identifier corresponding to the model target layer, combine the material identifier of each mesh cell with the average thickness value calculated in the mesh cell into an attribute tag group through unit-by-unit binding, and embed all tag groups into the target layer structure in the order of mesh cell number to generate a target layer attribute mesh cell set;
[0029] Based on the target layer attribute mesh cell set, sequentially repeat the attribute tag group binding process for all remaining layers in the multi-layer digestive tract shell model, integrate all processed attribute mesh cell sets layer by layer, and establish an attribute digestive tract entity model.
[0030] Preferably, the step of obtaining the pressure-adjusted geometry is:
[0031] Based on the geometric and attribute data of the attribute digestive tract entity model, read the inner surface vertex coordinates, corresponding vertex normal direction, material identifier, target layer thickness value and layer structure number, and unify all vertex spatial positions, attribute values and structure association information into a structured calculation basic unit set to generate an inner surface vertex comprehensive information set;
[0032] According to the inner surface vertex comprehensive information set, read the target pressure value in the lumen, map the material identifier corresponding to each vertex to the material elastic response coefficient, and correct the elastic response coefficient in combination with the target layer thickness value, judge the displacement response increment under the action of unit pressure of each vertex, and directionally offset the three-dimensional coordinates of each vertex along the normal direction to generate a vertex coordinate set corrected by the lumen pressure response;
[0033] Based on the vertex coordinate set corrected by the lumen pressure response, reassemble the spatial position relationship of all vertex coordinates according to the hierarchical order and the original topological structure, and keep the original mesh patch connection unchanged to form an updated model consistent with the structure of the attribute digestive tract entity model, and generate a pressure-adjusted geometry.
[0034] Preferably, the step of obtaining the dynamic deformation geometry sequence is:
[0035] According to the pressure-adjusted geometry, the peristaltic wave propagation speed, the propagation path length of the vertex along the central axis direction and the initial phase angle in the physiological parameter set are called to calculate the normal displacement value of each vertex at a given time;
[0036] Based on the normal displacement value, the spatial coordinate of each vertex corresponding to the normal displacement value is updated along the unit normal direction, and the update is performed multiple times at time step intervals, and the three-dimensional coordinate state of all vertices is summarized at each time frame to generate a dynamic deformation geometry sequence.
[0037] The present application provides a virtual three-dimensional reconstruction method of the digestive tract, comprising the following steps:
[0038] Based on the tomographic image sequence of medical images, the outline boundary points of the digestive tract region in the tomographic image sequence are identified, three-dimensional space point connection and initial surface mesh generation are performed, and the generated initial surface mesh is smoothed to establish an initial lumen surface model;
[0039] Based on the geometric data of the initial lumen surface model and the preset first wall layer target thickness value of the digestive tract, offset calculation is performed along the normal direction of each vertex of the initial lumen surface model, to obtain the first layer offset vertex coordinates, and based on the first layer offset vertex coordinates, the first wall layer outer surface mesh is connected by the coordinate points to establish a multi-layer digestive tract shell model;
[0040] Based on the hierarchical geometric data of the multi-layer digestive tract shell model, the target layer of the model is selected, the distance between the inner surface mesh vertex and the corresponding outer surface mesh vertex in the target layer is calculated, the target layer vertex thickness value set is obtained, and based on the target layer vertex thickness value set, the preset material identifier and the calculated thickness value are jointly assigned to the layer grid unit to establish an attributed digestive tract entity model;
[0041] Based on the geometric and attribute data of the attributed digestive tract entity model and the target pressure value in the lumen, the three-dimensional coordinates of the vertices on the inner surface of the attributed digestive tract entity model are adjusted to obtain a pressure-adjusted geometry, based on the pressure-adjusted geometry, and according to the peristaltic wave characteristics in the physiological parameter set, the sequence displacement value calculation conforming to the peristaltic wave propagation rule is performed on the model vertices to establish a dynamic deformation geometry sequence.
[0042] Compared with the prior art, the present application has the advantages and positive effects that:
[0043] In the present application, the boundary points of the digestive tract region are extracted based on the sequence of tomographic images, and are constructed as a spatial point cloud, and then an initial surface model is formed in a geometric connection manner, and on this basis, a multi-layer wall layer coordinate offset operation is performed according to the normal direction of the vertex, which can realize the geometric expression of the complex cavity structure in multiple layers and continuity. By introducing a thickness calculation operation in each layer, the distance between the inner and outer vertices is mapped to a specific geometric index, which is then bound to the material identifier, so that each layer of grid cells has a clear physical property expression capability. In the construction process, a normal direction coordinate adjustment mechanism based on three-dimensional structure is introduced, combined with the internal pressure value of the lumen, to simulate the stress response of the spatial structure, so that the three-dimensional structure has deformation adaptability. At the same time, in the time evolution, the physiological characteristics of the peristaltic wave are read to bind the time-varying displacement result to the vertex coordinates, realizing continuous and physiological rhythm-compliant dynamic response. Through the step-by-step processing mode from contour recognition to multi-layer expansion, from static structure to physical response, from single-frame modeling to sequence animation, not only the spatial accuracy and hierarchical integrity of the anatomical structure restoration are improved, but also the flexibility and dynamic adaptation ability of the model to physiological state simulation are enhanced, supporting the functional expansion requirements in multiple scenes and multiple scales. BRIEF DESCRIPTION OF DRAWINGS
[0044] Figure 1 The system flowchart of the present application. DETAILED DESCRIPTION
[0045] In order to make the purpose, technical scheme and advantages of the present application clearer, the present application will be further described in detail below in combination with the drawings and examples. It should be understood that the specific examples described herein are only used to explain the present application and do not limit the present application.
[0046] Please refer to Figure 1 The present application provides a technical scheme: a virtual three-dimensional reconstruction system of the digestive tract comprises:
[0047] An initial geometry construction module, based on the sequence of tomographic images of medical images, identifies the contour boundary points of the digestive tract region in the sequence of tomographic images, performs three-dimensional spatial point connection and initial surface grid generation, and smoothes the generated initial surface grid to establish an initial lumen surface model;
[0048] A layered shell construction module, based on the geometric data of the initial lumen surface model and a preset first wall layer target thickness value of the digestive tract, performs offset calculation along the normal direction of each vertex of the initial lumen surface model to obtain first layer offset vertex coordinates, and based on the first layer offset vertex coordinates, connects the coordinate points to generate a first wall layer outer surface grid to establish a multi-layer digestive tract shell model;
[0049] The material thickness calibration module selects a target layer based on hierarchical geometry data of the multi-layer digestive tract shell model, calculates distances between grid vertices on the inner surface of the target layer and corresponding grid vertices on the outer surface, obtains a set of vertex thickness values of the target layer, assigns a preset material identifier and a calculated thickness value to the layer grid cells based on the set of vertex thickness values, and establishes an attributed digestive tract entity model.
[0050] The dynamic morphological evolution module adjusts three-dimensional coordinates of vertices on the inner surface of the attributed digestive tract entity model based on geometry and attribute data of the attributed digestive tract entity model and a target pressure value in the lumen, obtains a pressure-adjusted geometry, performs a sequence displacement value calculation on the model vertices based on the pressure-adjusted geometry and according to peristaltic wave characteristics in the physiological parameter set, and establishes a dynamic deformation geometry sequence.
[0051] The initial lumen surface model is obtained by:
[0052] Reading a sequence of tomographic images of a medical image, performing image segmentation on each frame of image, identifying a digestive tract region in the image and extracting all boundary contour points to form a contour boundary point set of the digestive tract region;
[0053] Based on the contour boundary point set of the digestive tract region, corresponding boundary points in adjacent image frames are sequentially paired in three-dimensional coordinates, and surface connection is performed based on the point pairs to form an initial surface grid by collecting all connection results.
[0054] According to the initial surface grid, geometric curvature analysis is performed on each group of triangular facets, sharp connection regions are marked, and continuous adjacent facet boundary point position relationships are called for smooth interpolation to form an initial lumen surface model.
[0055] Specifically, a sequence of tomographic images of a medical image is read, and specifically, first, a pre-processing operation is performed on each frame of tomographic images in the sequence, including applying a Gaussian filter to attenuate image noise, wherein the standard deviation σ of the Gaussian filter is empirically set according to the signal-to-noise ratio of the image, for example, σ can be set to 1.0, and then histogram equalization is performed to enhance image contrast, facilitating subsequent tissue region identification. Next, a U-Net model based on deep learning is used for image segmentation. The U-Net model structure includes an encoder path and a decoder path. The encoder path is composed of five down-sampling modules, each module containing two consecutive 3x3 convolution layers, each convolution layer followed by a ReLU activation function, and a 2x2 max-pooling layer to reduce the spatial dimension of the feature map. The decoder path also contains five up-sampling modules, each module starting with a 2x2 deconvolution layer (or up-sampling layer), the output of which is spliced with the feature map of the corresponding level of the encoder path through a jump connection. The spliced feature map is then passed through two consecutive 3x3 convolution layers and a ReLU activation function. The last layer is a 1x1 convolution layer, which outputs a pixel-level probability map through a Sigmoid activation function, indicating the probability of each pixel belonging to the digestive tract region. The training data set of the U-Net model contains, for example, 2000 medical image tomographic images whose digestive tract regions (such as the stomach, small intestine, colon, etc.) have been manually outlined by experienced radiologists and their corresponding binary mask images. The input model is a single tomographic image of 512x512 pixels in grayscale, with pixel values normalized to the [0, 1] interval. The output is a probability map of the same size. During the training process, the loss function used is the weighted sum of the Dice loss and the binary cross-entropy loss, for example, the weights are each 0.5, the optimizer is Adam, and the initial learning rate is set to 1x10 -4And learning rate decay strategy is adopted, for example, when the validation set loss does not decrease for 3 consecutive cycles, the learning rate is multiplied by 0.1, the batch size is set to 8, the training iteration period is set to 150, or until the validation set Dice coefficient no longer significantly improves, data augmentation techniques are also applied during training, including random rotation (-15 to +15 degrees), horizontal flip, and slight elastic deformation, after training, the U-Net model is applied to each input frame of the tomographic image to generate a probability map of the digestive tract region, then the probability map is thresholded to generate a binary mask image, the probability threshold is set as follows: in a validation set containing 50 images, by testing different thresholds between 0.1 and 0.9 (step 0.05), select the threshold that can make the average Dice similarity coefficient between the segmentation result and the expert manual delineation result the highest, for example, if the threshold is 0.55, the average Dice coefficient is 0.92, and the coefficients corresponding to other thresholds are lower, then 0.55 is used as the final threshold, any pixel point in the probability map with a value greater than 0.55 is identified as a digestive tract region, and less than or equal to 0.55 is background region, finally, the generated binary mask image is applied to the contour tracking algorithm proposed by Suzuki et al., to accurately extract the boundary point pixel coordinates of each closed contour of the digestive tract region, forming a contour boundary point set of the digestive tract region.
[0056] Based on the contour boundary point set of the digestive tract region, first, the two-dimensional contour boundary points (x, y coordinates) extracted in each frame of the tomographic image are assigned to three-dimensional space coordinates, and the specific method is to calculate the z-axis coordinate of each point according to the layer sequence of the tomographic image in the sequence and the known layer spacing (for example, the layer thickness of CT scanning, such as 1 millimeter), for example, the z coordinate of the nth frame image is (n-1) x layer spacing, thereby converting the two-dimensional point set into a three-dimensional point set, next, the digestive tract contour boundary point sets in adjacent image frames (for example, the i th frame and the i+1 th frame) are paired, and the pairing process first calculates the Euclidean distance of each point on the i th frame contour to all points on the i+1 th frame contour, and selects the point with the smallest distance as the initial corresponding point, in order to improve the accuracy of pairing, a search radius limit is introduced, and the setting standard of the search radius is based on the statistical analysis of the average distance of at least 20 mature model contour points, and the average point distance is taken as 1.5 times the search radius, for example, if the average point distance is 2.0 millimeters, the search radius is set to 1.5 x 2.0 millimeters = 3.0 millimeters, that is, only within 3.0 millimeters radius range after the current point is projected to the adjacent layer, the nearest point is searched, if there is a significant difference in the number of points of adjacent contours, for example, one contour has 100 points and the other has 150 points, the sampling points are increased on the contour with fewer points through linear interpolation to make the number of points close to that of the contour with more points, or the sampling points are uniformly reduced on the contour with more points to ensure the uniformity of subsequent connection, after the point pairing is completed, the surface connection is carried out based on these three-dimensional coordinate points, specifically, the triangular subdivision technology is used to construct triangular patches between the corresponding points of adjacent contours, a commonly used method is the point-by-point advancing method, which starts from a pair of starting corresponding points on the two adjacent contours, and sequentially connects two points of the current contour segment and one corresponding point on the adjacent contour to form a triangle, and then connects two points of the adjacent contour segment and the next corresponding point on the current contour to form another triangle, and so on, until the triangular patches are constructed around the entire contour, ensuring that the normal vector directions of all generated triangular patches are basically consistent, for example, all point outward, finally, all triangular patches generated by connecting adjacent contour points are combined to form an initial surface mesh.
[0057] According to the initial surface mesh, first, the geometric curvature analysis is carried out on each group of triangular patches constituting the mesh, and specifically to each vertex, the average curvature H is estimated by analyzing the geometric shape of the triangular patches around the vertex, for example, the discrete average curvature approximation value at each vertex defined by its adjacent triangular patches can be calculated, or the two angles a j and β j opposite to the edge connecting the vertex, and the vector composed of the vertex and its adjacent point v j are calculated, and the average curvature H is estimated by the formula , wherein A mixedis the mixed area related to vertex v, N(v) is a set of one-ring neighborhood vertices of v, is the normal vector of vertex v, the curvature value of each vertex is obtained by this method, then, the sharp connection area is marked according to the calculated curvature value, the judgment standard of the sharp connection area is that the average curvature value of the vertex exceeds a preset curvature threshold value, the setting method of the curvature threshold value is: a group (for example, 10) of three-dimensional models of the digestive tract with good morphology confirmed by experts is selected as a reference, the average curvature value of all vertices of these reference models is calculated, a curvature value distribution is formed, and the 90th percentile of the distribution is taken as the sharp curvature threshold value, for example, if a group of calculated curvature value sequences is [0.1, 0.2, 0.3, 0.35, 0.4, 0.45, 0.5, 0.55, 0.6, 0.7], the 90th percentile is 0.65mm -1 , then the curvature threshold value T curvature is set to 0.65mm -1 , when the calculated curvature of a certain vertex is greater than 0.65mm -1 , the vertex and its adjacent area are marked as a sharp connection area, for these marked sharp connection areas, the position relationship of the boundary points of the continuous adjacent faces is called for smooth interpolation processing, the smooth interpolation adopts an iterative Laplace smoothing method, and the position of each vertex V k of the marked area is updated, and the update formula is , wherein V′ k is the updated position of the vertex V k , N(k) is a set of one-ring neighborhood vertices of the vertex V k , |N(k)| is the number of neighborhood vertices, and λ is a smoothing weight factor, the value of λ is set according to experience, for example, 0.4, so as to effectively smooth the sharp features while avoiding excessive smoothing to cause significant volume shrinkage or loss of details, the number of smoothing iterations is set as a fixed value, for example, 8 times, or the iteration is stopped when the maximum change amount of the position of all vertices in the iteration process is less than a minimum value (for example, 0.005mm), and the initial lumen surface model is formed after the smooth interpolation is performed on all the marked sharp connection areas.
[0058] The first layer offset vertex coordinate acquisition step is:
[0059] Based on the initial lumen surface model, all triangular facets constituting the initial lumen surface model are extracted, each triangular facet is parsed into three vertex coordinate combinations according to the vertex number, the normal vector of the triangular facet is calculated by three-dimensional vector cross multiplication, and the normal vector value of the three vertices corresponding to the triangular facet is accumulated respectively, after the accumulation of all triangular facets is completed, the accumulated normal vector of each vertex is divided by the corresponding accumulation number, and a unit normal vector set of each vertex of the initial lumen surface model is generated;
[0060] According to the unit normal vector set of each vertex of the initial lumen surface model, the corresponding three-dimensional coordinate value of each vertex is read, and a preset target thickness value of the first wall layer of the digestive tract is called, each unit normal vector is multiplied by the target thickness value of the first wall layer of the digestive tract to obtain an offset displacement vector of each vertex, and then the offset displacement vector and the original three-dimensional coordinate value are added by coordinate axis dimension by component to generate a new three-dimensional space coordinate of each vertex after offset;
[0061] Based on the new three-dimensional space coordinate of each vertex after offset, the points are rearranged in accordance with the index order of the original initial lumen surface model vertex number, the decimal residual error caused by the floating point error in the coordinate expression is removed, and the unified precision is reduced, to generate the first layer offset vertex coordinate.
[0062] Specifically, based on the initial lumen surface model, first, the model data structure is traversed to extract all triangular facets constituting the initial lumen surface model, each triangular facet is defined by three vertex indexes, through which the three-dimensional coordinates of the corresponding vertices can be found, for each triangular facet, for example, composed of vertices V1, V2, and V3, the corresponding three-dimensional coordinates are P1(x1, y1, z1), P2(x2, y2, z2), and P3(x3, y3, z3), then the vector from vertex P1 to P2 is calculated and the vector from vertex P1 to P3 is calculated The normal vector of the triangular facet is calculated by three-dimensional vector cross multiplication operation of the two vectors The calculation result is (A y B z -A z B y ,A z B x -A x B z ,A x B y -A y B x ), and the consistency of the direction of the normal vector should be ensured during calculation, for example, unified to point outward of the lumen, which can be determined by checking the dot product sign of the normal vector and the vector from the center of the triangular facet to the approximate center of the model, if the direction is not consistent, the normal vector is reversed, after calculating the normal vector of the triangular facet, the The accumulated normal vector memory of each of the three vertices V1, V2, and V3 corresponding to the triangle is added, and a counter is maintained for each vertex to record how many triangle normal vectors contribute to the calculation, i.e. each time a triangle normal vector is added to a vertex, the counter of the vertex is incremented by one. After all the triangle normal vectors in the initial lumen surface model are calculated and accumulated, each vertex is traversed, and the accumulated normal vector value is divided by the corresponding cumulative number (i.e. the number of triangles contributing to the vertex), to obtain the average normal vector of each vertex. Finally, the average normal vector of each vertex is normalized, i.e. each component of the average normal vector is divided by the length of the vector (Euclidean length), to generate a set of unit normal vectors for each vertex of the initial lumen surface model.
[0063] According to the set of unit normal vectors of each vertex of the initial lumen surface model and the vertex coordinate data of the initial lumen surface model, first, for each vertex in the model, read its original three-dimensional coordinate value P orig =(x orig ,y orig ,z orig ) and the corresponding unit normal vector and call a preset first wall layer target thickness value of the digestive tract, which is denoted as T1. The setting of the first wall layer target thickness value of the digestive tract is based on anatomical literature and histological data research on a specific digestive tract segment (e.g. stomach wall, small intestine wall). These researches usually provide the average thickness range of the corresponding wall layer (e.g. combined mucosa and submucosa) of a healthy adult, for example, if the literature indicates that the average thickness of the first wall layer (from the lumen to the outside) of a certain segment of the small intestine is between 1.5 mm and 2.5 mm, a representative value is selected as T1. The specific selection can consider the average situation of the batch of medical image data or use the reference value of a standard human model, for example, set T1 = 2.0 mm, which is a system parameter preconfigured. Then, multiply the unit normal vector of each vertex with the first wall layer target thickness value T1 of the digestive tract to obtain the three-dimensional offset displacement vector of the vertex Add the calculated offset displacement vector to the original three-dimensional coordinate value P orig of the vertex by coordinate axis dimension, and the calculation formula is Generate new three-dimensional space coordinates of each vertex after offset.
[0064] Based on the new three-dimensional coordinate set of each vertex after offset, it is first necessary to rearrange these newly calculated coordinates point by point according to the index order of the vertex numbers in the original initial lumen surface model. The specific operation is to create a new coordinate list whose length is the same as the number of vertices in the original model, and then traverse the vertex index of the original model from 0 to N-1 (N is the total number of vertices), and fill the newly calculated offset coordinate corresponding to the i-th original vertex into the i-th position of the new list, so as to ensure that the order of the new coordinate set is consistent with the topological structure of the original model. Next, process the coordinate expression in Small numerical errors that may be introduced by floating-point operations are called decimal residuals. For example, a coordinate value that should theoretically be 3.0 may be calculated as 3.0000000000000004 or 2.99999999999999996. In order to eliminate these decimal residuals and perform unified accuracy reduction, a unified coordinate accuracy standard is set. The standard is set based on the resolution of medical imaging equipment and the actual requirements of subsequent applications for model accuracy. Usually, a specific number of decimal places below the millimeter level is sufficient. For example, the coordinate accuracy is set to 4 decimal places (i.e. 10 -4 mm), for each new coordinate's X, Y, and Z components, the precision reduction operation is performed. The specific calculation method is that for any coordinate component value c, the reduced value c rounded By formula c rounded =
[0065] round(c×10 k ) / 10 k Calculated, where k is the number of decimal places you want to retain. In this example, k = 4. For example, if a calculated x-coordinate value is 15.2345789 mm, the reduced value is round(15.2345789×10000) /
[0066] 10000 = round (152345.789) / 10000 = 152346 / 10000 = 15.2346 mm. If the other z coordinate value is 9.8765123 mm, then the reduced value is round (9.8765123 × 10000) / 10000 = round (98765.123) /
[0067] 10000 = 98765 / 10000 = 9.8765 mm. After completing this unified precision reduction process for all coordinate components of all vertices, the final data set formed is the first-layer offset vertex coordinates.
[0068] The steps for obtaining the multi-layer digestive tract shell model are:
[0069] Based on the first layer offset vertex coordinates, according to the initial lumen surface model triangular facet topology structure sequence, the original connection relationship of the corresponding matching triangular facet is matched and the vertex index mapping is established, each group of three offset vertex coordinates is called to perform spatial connection, a triangular facet set with consistent structure is constructed, and a first wall layer outer surface grid is generated;
[0070] According to the first wall layer outer surface grid, all vertex coordinates of the grid are extracted as a new round of offset reference coordinate set, each subsequent wall layer target thickness value is read in turn, and offset operation in the unit normal direction of each vertex is performed for each target thickness value. A new vertex set of each level is constructed by accumulating offset coordinates, and a vertex coordinate set of all wall layers is obtained;
[0071] Based on the vertex coordinate set of all wall layers, according to the connection order of the triangular facets of the previous layer corresponding to each layer offset coordinate set, the triangular facet structure is reused to generate the corresponding surface grid layer by layer, and the surface grids of each layer are topologically closed and organized and integrated in order of level to generate a multi-layer digestive tract shell model.
[0072] Specifically, based on the first layer offset vertex coordinates, and referring to the triangular facet topology structure recorded in the initial lumen surface model, first, each triangular facet definition in the initial lumen surface model is traversed, each triangular facet is composed of three vertex indexes, for example, a triangular facet is defined by vertex indexes i, j, and k. These indexes point to specific vertices in the vertex list of the initial lumen surface model. Since the first layer offset vertex coordinates are arranged in the index order of the original initial lumen surface model vertex number, the vertex index i corresponds to the i-th vertex in the initial lumen surface model, and its new coordinate after the first layer offset is the i-th coordinate in the first layer offset vertex coordinate list. Similarly, vertex indexes j and k correspond to the j-th and k-th coordinates in the first layer offset vertex coordinate list, respectively. Then, for each triangular facet (e.g. defined by indexes i, j, and k) in the initial lumen surface model, the new three-dimensional space coordinates corresponding to the three indexes are extracted from the first layer offset vertex coordinate set, denoted as P′ i , P′ j , P′ k Then, the three new three-dimensional space coordinates P′ i , P′ j , P′ kA new triangular facet is constructed, which represents the surface morphology of the first wall layer of the digestive tract after the outward shift in spatial position, but the vertex connection relationship is completely consistent with the original triangular facet in the initial lumen surface model. All the triangular facets newly created in this way, that is, reusing the topological connection information of the initial lumen surface model and replacing it with the first layer offset vertex coordinates, are collected to form a complete triangular facet set. All the triangular facets in this set collectively constitute the outer boundary of the first wall layer, that is, the first wall layer outer surface mesh is generated.
[0073] According to the first wall layer outer surface mesh, first, the three-dimensional coordinates of all the vertices of the mesh are extracted to form a coordinate set, which serves as a new set of offset reference coordinates for calculating the subsequent second wall layer outer surface. At the same time, the unit normal vector of each vertex of the first wall layer outer surface mesh is calculated. The calculation method is as follows: for each triangular facet in the mesh, two edge vectors are calculated through its three vertex coordinates, and then the cross product of the two edge vectors is calculated to obtain the normal vector of the facet. Subsequently, the normal vector of the facet is normalized. For each vertex in the mesh, the unit normal vectors of all adjacent triangular facets are averaged (for example, weighted average, the weight can be the area of the triangular facet or the vertex angle of the vertex in the facet, or simple arithmetic average), to obtain the average normal vector of the vertex. Finally, the average normal vector is normalized to obtain the unit normal vector of the vertex. Next, the preset sequence of target thickness values of the subsequent wall layers is read in turn. These target thickness values, such as T2, T3, …, T M , are pre-stored as a parameter list. For each subsequent wall layer (from the second layer to the Mth layer), a vertex offset operation is performed once. Specifically, for the outer surface of the kth wall layer (where k≥2), the three-dimensional coordinates of its vertices are calculated based on the corresponding vertex coordinates of the outer surface of the (k-1)th wall layer and its unit normal vector , which are calculated by the formula , where T k is the target thickness value of the kth wall layer. is the unit normal vector at vertex v on the outer surface of the k-1th layer, by performing this offset calculation on all vertices in the current offset reference coordinate set, a new vertex coordinate set of the outer surface of the current level (kth layer) is constructed, after sequentially completing the vertex coordinate calculation of all subsequent wall layers, the vertex coordinate set of the outer surface of each layer (from the first layer to the Mth layer) is collected to obtain the vertex coordinate set of all wall layers.
[0074] Based on the vertex coordinate set of all wall layers, which contains the vertex coordinate data of each layer interface from the outer surface of the first wall layer to the outer surface of the outermost wall layer, and the number and arrangement order of the vertices of each layer interface are consistent with the initial lumen surface model, next, according to the connection order of the previous layer patch corresponding to each layer offset coordinate set, that is, the original triangular patch connection relationship defined in the initial lumen surface model, the triangular patch topology structure is reused layer by layer to generate the corresponding surface mesh for the outer interface of each wall layer. The specific operation is as follows: for the vertex coordinate set of the kth wall layer (k is from 1 to M, M is the total number of wall layers), traverse the triangular patch list in the initial lumen surface model, if a triangular patch in the original model is defined by vertex indices (v1, v2, v3), then take out the three vertex coordinates with indices v1, v2, v3 from the kth layer vertex coordinate set, and use these three coordinates to form a new triangular patch as a component of the kth layer outer surface. Repeat this operation for all original triangular patches to construct a complete surface mesh for the kth layer outer surface. In this way, a series of surface meshes representing the outer boundaries of the wall layers are obtained, denoted as S1, S2, …, SM. M At the same time, the initial lumen surface model itself can be regarded as the surface mesh of the 0th layer, denoted as S0, and then these surface meshes generated layer by layer are organized and integrated according to their physiological hierarchical order in the digestive tract wall. Specifically, the jth wall layer (for example, mucosa layer, muscle layer, etc.) is defined geometrically as the space region surrounded by its inner surface mesh S j-1 and its outer surface mesh S j In this way, the surface meshes of each layer are combined, and the adjacency relationship and hierarchical attribution between layers are clearly defined, for example, the first wall layer is defined by S0 and S1, the second wall layer is defined by S1 and S2, and so on, until the outermost wall layer is defined by S M-1 and S M After this hierarchical organization, a multi-layer digestive tract shell model containing the geometric information of all wall layers and their spatial relationships is finally generated.
[0075] The target layer vertex thickness value set is obtained by the following steps:
[0076] Based on the hierarchical geometry data of the multi-layer digestive tract shell model, the structural attribute identifier of each layer is traversed, and the number of vertex groups contained, the vertex index range, the connection order and the hierarchical number are read. According to the input target layer number, the matching is performed, the target layer of the model is locked, and the three-dimensional coordinates of all inner surface mesh vertices and outer surface mesh vertices in the target layer of the model are extracted as a structured coordinate pair set respectively, and a target layer vertex coordinate pair set of the model is generated.
[0077] According to the model target layer vertex coordinate pair set, the three-dimensional coordinates of each group of inner surface mesh vertices and corresponding outer surface mesh vertices are traversed, the coordinate difference values in the X, Y and Z axis directions are calculated for each group of coordinate pairs, the sum of the squares of the difference values in the three directions is calculated, and the real geometric distance value sequence between the vertices is constructed. Then, all distance values are organized into an original thickness value sequence according to the vertex number index correspondence relationship, and a three-dimensional space thickness distance sequence of the model target layer is generated.
[0078] Based on the three-dimensional space thickness distance sequence of the model target layer, each distance value is rounded according to the rounding rule, and after the processing is completed, it is re-ordered according to the original vertex index sequence, and a target layer vertex thickness value set is generated.
[0079] Specifically, based on the hierarchical geometry data of the multi-layer digestive tract shell model, the target layer to be processed needs to be determined first, which is achieved by receiving an "input target layer number". For example, if the multi-layer digestive tract shell model contains three wall layers, numbered as the first layer, the second layer, and the third layer, and the "input target layer number" is "the second layer", the program will lock the second layer as the "model target layer" for the current operation. Then, the structural attribute identifiers of each layer constituting the multi-layer digestive tract shell model are traversed. These identifiers record the number of vertex groups contained in each layer (usually consistent with the number of vertices of the initial lumen surface model), the index range or direct pointer of the vertex data in the "total wall layer vertex coordinate set", and the connection order of the triangular facets (this connection order is reused from the initial lumen surface model for all layers) and the "hierarchical number" of the layer. By comparing the "hierarchical number" of each stored layer with the "input target layer number", when they match, the "model target layer" is determined. Once the "model target layer" is locked, for example, the second layer is locked, the program will extract the vertex coordinates constituting the geometric boundary of the second layer. Specifically, the inner surface of the second layer is composed of the outer surface of the first layer, and its vertex coordinates can be obtained from the coordinate set of the first layer's outer surface in the "total wall layer vertex coordinate set". The outer surface of the second layer is its own vertex coordinates, also obtained from the coordinate set of the second layer's outer surface in the "total wall layer vertex coordinate set". Since the construction process of the multi-layer digestive tract shell model ensures a one-to-one correspondence between the vertices of the layers (i.e., the ith inner surface vertex corresponds to the ith outer surface vertex), for each vertex index of the "model target layer" (from 0 to N-1, N is the total number of vertices of each layer surface), its three-dimensional coordinates on the inner surface mesh P inner,i and on the outer surface mesh P outer,i are obtained respectively, and this pair of coordinates (P inner,i , P outer,i ) is stored as an element in a set. After extracting and pairing all vertices, the "model target layer vertex coordinate pair set" is formed.
[0080] According to the "model target layer vertex coordinate pair set", each element in the set is a pair of three-dimensional coordinates (P inner,i , P outer,i ), where P inner,i = (x in,i , y in,i , z in,i ) represents the coordinates of the ith vertex of the "model target layer" on its inner surface, and P outer,i = (x out,i , y out,i , z out,irepresent the corresponding coordinates of the same vertex on its outer surface, next, loop through each coordinate pair in this set, for the i-th coordinate pair (P inner,i ,P outer,i ), first calculate the coordinate difference in the X-axis, Y-axis and Z-axis direction respectively, i.e. Δx i =x out,i -x in,i , Δy i =y out,i -y in,i , and Δz i =z out,i -z in,i , then use these differences to calculate the real three-dimensional Euclidean distance between the corresponding vertices of the inner and outer surfaces, this distance d i is the local thickness of the "model target layer" at the i-th vertex position, and its calculation formula is: d i = where d i is the distance value calculated between the i-th vertex pair, repeat this distance calculation process for all coordinate pairs in the model target layer vertex coordinate pair set, thereby constructing a sequence containing the local thickness values of all vertex positions, this sequence is the real geometric distance value sequence between vertices, since the model target layer vertex coordinate pair set itself is organized in the original vertex index order (i.e. the vertex index order in the initial lumen surface model), therefore the calculated real geometric distance value sequence between vertices also naturally follows this order, then directly identify this sequence containing all the calculated distance values as the original thickness value sequence, to generate the three-dimensional space thickness distance sequence of the model target layer.
[0081] Based on the model target layer three-dimensional space thickness distance sequence, the sequence contains the original calculated thickness value of the "model target layer" at each vertex position, and these thickness values are arranged in the original vertex index order. Next, rounding processing is performed on each distance value (i.e. thickness value) in the sequence, and the rounding precision standard needs to be set in advance, which is mainly based on the accuracy requirement of the thickness data in the subsequent application scenario. For example, if the thickness value will be used for visualization or preliminary mechanical property assignment, retaining two or three decimal places is usually sufficient. For example, if the preset rounding precision is two decimal places (i.e. 0.01 mm), then for any distance value d in the model target layer three-dimensional space thickness distance sequence, the rounded value d' is obtained by multiplying d by 100, rounding the result, and then dividing by 100. For example, if the calculated distance value at a vertex is 2.4582 mm, rounding to two decimal places gives round(2.4582x100) / 100 = round(245.82) / 100 = 246 / 100 = 2.46 mm. If another value is 1.9965 mm, the rounding is round(1.9965x100) / 100 = round(199.65) / 100 = 200 / 100 = 2.00 mm. All distance values in the model target layer three-dimensional space thickness distance sequence are subjected to this rounding processing. Since the rounding operation is performed on each element, the relative order of the elements in the original sequence is not changed, so the rounded thickness value sequence still maintains the consistency with the original vertex index order, and no additional reordering operation is needed. The rounded thickness value sequence is the final generated target layer vertex thickness value set.
[0082] The obtaining step of the attribute digestive tract entity model is:
[0083] Based on the target layer vertex thickness value set, the grid cell structure information corresponding to the model target layer is read, each cell index in the grid cell is traversed, the associated three or four vertex numbers are obtained in turn, and all vertex numbers are matched and read in the target layer vertex thickness value set to form a thickness value set corresponding to each grid cell, and a target layer grid cell thickness mapping set is generated;
[0084] According to the target layer grid cell thickness mapping set, the preset material identifier corresponding to the model target layer is read, the material identifier of each grid cell is combined with the calculated average thickness value of the grid cell to form an attribute label group through unit-by-unit binding, and all label groups are embedded into the target layer structure according to the grid cell number order to generate a target layer attribute grid cell set;
[0085] Based on the target layer attribute grid unit set, the attribute label group binding process is repeatedly performed on all remaining layers in the multi-layer digestive tract shell model in turn, and all processed attribute grid unit sets are integrated layer by layer to establish an attribute digestive tract entity model.
[0086] Specifically, based on the target layer vertex thickness value set, which contains the accurate thickness of the current "model target layer" at each vertex position, first, the grid cell structure information corresponding to the "model target layer" needs to be read, which defines how the model surface is composed of a series of basic geometric units (usually triangular patches, but may also contain quadrilateral patches, i.e. grid cells). Specifically, it lists each grid cell and the vertex number (or index) that makes up the cell. Next, each cell index recorded in the grid cell structure information of the "model target layer" is traversed. For the currently traversed grid cell, for example, cell C k , first, the vertex numbers associated with the cell are obtained. If it is a triangular patch, three vertex numbers (e.g. v1, v2, v3) are obtained, and if it is a quadrilateral patch, four vertex numbers (e.g. v1, v2, v3, v4) are obtained. Then, using these obtained vertex numbers, a matching search is performed in the previously generated "target layer vertex thickness value set", and the thickness value corresponding to each vertex is read out, for example, for vertex v1, its thickness Th(v1) is read from the "target layer vertex thickness value set", for vertex v2, its thickness Th(v2) is read, and so on. Collect the thickness values of all vertices associated with a grid cell to form a small set containing these thickness values, for example, for triangular cell C k , its corresponding thickness value set is {Th(v1), Th(v2), Th(v3)}. Repeat this process for all grid cells in the "model target layer", and finally associate the identification of each grid cell (such as cell index) with its corresponding thickness value set to form a target layer grid cell thickness mapping set.
[0087] According to the target layer mesh cell thickness mapping set, which associates a set of vertex thickness values with each mesh cell of the "model target layer", the next step is to assign material information to these mesh cells. First, read the "preset material identifier" corresponding to the current "model target layer". The "preset material identifier" is a pre-defined code or text label that uniquely specifies the type of biological material that makes up the layer of the digestive tract wall. Its setting is based on the anatomical layered structure of the digestive tract. For example, if the "model target layer" is the first layer of the stomach wall (mucosa and submucosa), its "preset material identifier" may be set to "GASTRIC_MUCOSA_SUBMUCOSA_L1" according to an internal lookup table. This lookup table pre-stores standard material identifiers corresponding to different parts of the digestive tract (such as the esophagus, stomach, small intestine, large intestine) and their sub-layers (such as the mucosa layer, muscle layer, serosa layer). For example, the stomach mucosa layer identifier is "MAT_STOMACH_MUCOSA", and the muscle layer is "MAT_STOMACH_MUSCLE". These identifiers are entered by the system administrator based on anatomical and histological standard data. Then, for each mesh cell (e.g. triangle C k ) in the "target layer mesh cell thickness mapping set" and its corresponding set of vertex thickness values (e.g. {Th1, Th2, Th3}), calculate the average thickness value of the mesh cell. The calculation formula is the arithmetic mean of the thickness values of the vertices of the mesh cell. For a mesh cell with N vertices, the average thickness For a triangle (N = 3), it is Next, combine the previously read "preset material identifier" and the average thickness value of the mesh cell just calculated to form an attribute label group. For example, if the material identifier is "GASTRIC_MUCOSA_SUBMUCOSA_L1" and the average thickness is 2.15 mm, the attribute label group can be represented as {material: "GASTRIC_MUCOSA_SUBMUCOSA_L1", average thickness: 2.15 mm}. Finally, according to the numbering order of the mesh cells in the original model, embed the attribute label group formed by each mesh cell into the corresponding mesh cell data structure of the "model target layer", thereby generating the target layer attribute mesh cell set.
[0088] Based on the target layer attributed grid cell set, this set already contains the material identification and average thickness information of all grid cells in the currently processed "model target layer". Next, it is necessary to repeat the aforementioned attribute label group binding process for all remaining layers in the multi-layer digestive tract shell model that have not yet been attributed. Specifically, the system will traverse each other layer defined in the multi-layer digestive tract shell model (for example, if the first layer has just been processed, the second layer, the third layer, and so on will be processed next until all layers are attributed). For each selected "remaining layer", the system will first set it as a new "model target layer", and then strictly follow the first two steps to generate a corresponding "attributed grid cell set" for it, which includes: calculating the "target layer vertex thickness value set" based on the inner and outer surface vertex coordinates of the new "model target layer", and then generating the "target layer grid unit" based on this thickness value set and the grid cell structure of this layer. The "mesh thickness mapping set" is then generated. The unique "preset material identifier" of the new "model target layer" is read from the preset lookup table based on the physiological characteristics of the new "model target layer" (for example, if the muscle layer is currently being processed, the identifier may be "MAT_MUSCULARIS_EXTERNA"). The average thickness is calculated for each mesh cell in the layer. Finally, the material identifier and the average thickness are combined into an attribute label group and bound to each mesh cell to complete the attribute processing of the "remaining layers". After each "remaining layer" is processed and its "attributed mesh cell set" is obtained, it is integrated into the final model being constructed. This complete attribute process is repeated layer by layer and the results are integrated until all layers in the multi-layer digestive tract shell model carry their respective material and thickness attribute information. Finally, a complete attributed digestive tract solid model that includes the geometric and physical properties (material identification and thickness) of each layer is established.
[0089] The steps to obtain the pressure-adjusted geometry are:
[0090] Based on the geometric and attribute data of the attributed digestive tract solid model, the inner surface vertex coordinates, corresponding vertex normal directions, material identifiers, target layer thickness values, and hierarchical structure numbers are read. The spatial positions, attribute values, and structural association information of all vertices are unified into a set of structured calculation basic units to generate a comprehensive information set of inner surface vertices.
[0091] Based on the comprehensive information set of the inner surface vertices, the target pressure value in the lumen is read, the material identifier corresponding to each vertex is mapped to the material elastic response coefficient, and the elastic response coefficient is corrected in combination with the target layer thickness value. The displacement response increment under the unit pressure of each vertex is determined, and the three-dimensional coordinates of each vertex are directionally offset along the normal direction to generate a vertex coordinate set corrected for the lumen pressure response;
[0092] Based on the vertex coordinate set corrected by the luminal pressure response, the spatial position relationship of all vertex coordinates is reassembled in hierarchical order and original topology, and the original mesh patch connection is kept unchanged, forming an updated model consistent with the structure of the attributed digestive tract entity model, and generating the pressure-adjusted geometric morphology.
[0093] Specifically, based on the geometry and attribute data of the attributed digestive tract entity model, first, the information related to the inner surface directly contacting the lumen of the digestive tract needs to be extracted. The specific operation is to read the three-dimensional coordinates of all vertices of the surface representing the innermost lumen wall in the entity model (i.e., the initial lumen surface model), and at the same time, obtain the unit normal vectors corresponding to each of these inner surface vertices. These normal vectors have been calculated and ensured to have a direction (e.g., all pointing outward from the lumen or inward from the solid tissue) in the previous step. Next, for each vertex on the inner surface, the associated attribute information needs to be consulted from the attributed digestive tract entity model. Since each inner surface vertex belongs to one or more mesh elements that constitute the first wall layer of the digestive tract, the "material identifier" of the first wall layer at the position of the vertex (e.g., "MAT_MUCOSA_L1") and the local "target layer thickness value" of the first wall layer at the position of the vertex (which has been calculated for each vertex previously and stored in, for example, the "target layer vertex thickness value set") can be determined. In addition, the "hierarchical structure number" of the first wall layer (e.g., marked as layer 1) is recorded. By summarizing and organizing, for each vertex on the inner surface, the three-dimensional spatial position coordinates, the corresponding unit normal vector, the material identifier of the first wall layer to which it belongs, the thickness value of the first wall layer at the vertex, and the hierarchical number of the first wall layer are combined into a data unit. The collection of all these data units constitutes a structured calculation basis unit set for subsequent pressure response calculation, i.e., an inner surface vertex comprehensive information set is generated.
[0094] According to the inner surface vertex comprehensive information set, which provides position, normal, material, thickness, and layer number information for each inner surface vertex, first, a "target pressure value inside the lumen" is read from external input or system default, denoted as P lumen The setting of this pressure value needs to refer to specific physiological situations or research purposes. For example, when simulating the initial state of gastric distension, P lumen may be set to 1.2 kilopascals (kPa) according to literature. Then, for each vertex data in the inner surface vertex comprehensive information set, the recorded "material identifier" (e.g., "MAT_MUCOSA_L1") is mapped to a basic "material elastic response coefficient" by consulting a predefined "material attribute table". This coefficient is usually the Young's modulus E mat , for example, the "material attribute table" may contain entries such as: "MAT_MUCOSA_L1" corresponds to Young's modulus E mat=50kPa,
[0095] "MAT_MUSCLE_L2" corresponds to E mat = 100 kPa. These values are set based on biomechanical experimental data or related literature. Then, it is necessary to combine the "target layer thickness value" at the vertex (i.e., the thickness of the first wall layer Th v ) is used to correct the elastic response to calculate the normal displacement response increment δ of the vertex under unit pressure (1 kPa) v , the calculation method is, for example, that the displacement response increment is proportional to the material flexibility and inversely proportional to the thickness (or scaled by introducing a reference thickness), and its calculation formula can be expressed as: Among them, δ v is the displacement response increment of vertex v under unit pressure (unit: millimeter per kiloPascal, mm / kPa), P unit is the unit pressure value, here it is 1kPa, E mat,v is the Young's modulus of the material where the vertex v is located (unit: kPa), Th v is the actual thickness of the first wall layer at vertex v (unit: mm), and Th ref is a "reference thickness value" which is used to set the thickness correction term (Th ref / Th v ) dimensionless or as a parameter to adjust the response sensitivity, for example, Th can be set ref =1.0mm, if the material of a vertex E mat,v =50kPa, actual thickness Th v =2.0mm, then δ v =(1kPa /
[0096] 50kPa)·(1.0mm / 2.0mm)=0.02·0.5mm / kPa=0.01mm / kPa, and the unit pressure displacement response increment δ is obtained v After that, the actual normal displacement Δd v =δ v ·P lumen , for example, in the previous example, if P lumen =1.2kPa, then Δd v =
[0097] 0.01mm / kPa·1.2kPa=0.012mm. Finally, the calculated displacement Δd v Along the unit normal of the vertex Applied to its original 3D coordinate P v Up, that is, the new coordinate After completing this orientation offset calculation for all inner surface vertices, a set of vertex coordinates corrected for the lumen pressure response is generated.
[0098] Based on the vertex coordinate set corrected by the lumen pressure response, the set contains the new three-dimensional coordinates of all vertices on the inner surface of the digestive tract model after displacement under the action of the "target pressure value in the lumen". Next, it is necessary to use these updated inner surface vertex coordinates to construct the overall geometric shape reflecting the influence of pressure. Specifically, first replace the vertex coordinates corresponding to the inner surface (i.e., the inner boundary of the first layer) in the original "attributed digestive tract entity model" with these vertex coordinates corrected by the lumen pressure response. For the inner and outer surface vertex coordinates of all other levels (such as the second layer, the third layer, etc.) in the "attributed digestive tract entity model", keep their original coordinates unchanged before the pressure is applied in this step, that is, for example, the current pressure response mainly or only affects the shape of the innermost lumen, or the subsequent steps will deal with how the deformation propagates to the outer layer, and then, according to the original " The topological structure information of the "attributed digestive tract entity model", that is, the vertex connection relationship of each layer of mesh facets (for example, which vertices constitute a triangle facet), is reassembled for the spatial position relationship of the updated inner surface vertices and other unchanged vertices of each layer. Since only the vertex coordinates are updated, and the connection definition of all mesh facets (that is, which facet is composed of which vertex indices) remains completely consistent with the original topological structure, the new coordinate set can be used to smoothly construct the inner surface with the changed geometry, and together with other layers with unchanged coordinates, form a new model instance. This new model instance is completely consistent with the original "attributed digestive tract entity model" in structure (that is, the number of layers, topological connections of each layer), except that the geometry of its inner cavity surface is adjusted according to the pressure response, and finally the pressure-adjusted geometry is formed and generated.
[0099] The steps to obtain the dynamic deformation geometry sequence are:
[0100] According to the pressure-adjusted geometry, the peristaltic wave propagation velocity, the propagation path length of the vertex along the central axis, and the initial phase angle in the physiological parameter set are called to calculate the normal displacement value of each vertex at a given time. The calculation formula is:
[0101] D c (t) = A c ·sin(ψ c t-κ c +β c );
[0102] Among them, D c (t) is the normal displacement of the c-th vertex caused by the creeping wave at time t, A c is the maximum displacement amplitude corresponding to the digestive tract segment where the cth vertex is located, ψ c is the creeping angular frequency of the segment where the cth vertex is located, κ cis the propagation distance of the cth vertex from the center axis of the peristalsis starting point, β c is the initial phase offset value of the cth vertex corresponding to the peristalsis.
[0103] Based on the normal displacement value, the spatial coordinate of each vertex corresponding to the normal displacement value is updated along the unit normal direction, and multiple updates are performed at time step intervals. The three-dimensional coordinate state of all vertices is summarized at each time frame to generate a dynamic deformation geometry sequence.
[0104] Specifically, according to the pressure-adjusted geometric shape, the parameters required for simulating the peristalsis of the digestive tract in the "physiological parameter set" are called, including the peristaltic wave propagation-related characteristics specific to each segment or vertex of the digestive tract. Based on these parameters, the normal displacement value of each vertex of the "pressure-adjusted geometric shape" surface due to the peristaltic wave at a given time point t is calculated.
[0105] Formula: D c (t) = A c · sin(ψ c · t - κ c + β c ), the formula has the advantage that it uses simple harmonic vibration to describe the peristaltic wave of the digestive tract. It can simulate the main characteristics of the peristaltic wave, including the amplitude, frequency (indirectly through the angular frequency), propagation characteristics (through the spatial phase term κ c , which embodies the wavelength and position dependence), and the initial phase of the wave. By assigning different parameter values (A c , ψ c , κ c , β c ) to the vertices of different positions of the model (represented by the index c), the differences in peristaltic characteristics of different regions of the digestive tract (such as the esophagus, stomach, small intestine, and large intestine) and different positions within the same region can be simulated. For example, it can simulate the contraction wave propagating from the oral side to the anal side, as well as the contraction amplitude and frequency changes of different segments, while the phase part ψ c · t - κ c + β c controls the evolution of the wave in time and space, ensuring the realism and controllability of the simulation.
[0106] The parameter A c is obtained by the following steps: the parameter A cRepresents the maximum normal displacement amplitude corresponding to the digestive tract segment where the c-th vertex is located, in millimeters (mm). This value reflects the maximum degree of deformation of the tube wall caused by the peristaltic wave. Its acquisition mainly depends on dynamic medical imaging. The specific acquisition process includes: collecting high-quality dynamic image sequences of the relevant digestive tract segment (for example, if vertex c is located in a certain segment of the small intestine) during the peristalsis process. Perform image processing on each frame in the image sequence to identify and segment the contours of the digestive tract lumen or wall. By tracking the changes in specific points on the tube wall or the diameter of the lumen during the peristaltic cycle, record its displacement or radius data over time. Extract the displacement value with the maximum deviation from the baseline state from these time series data, which is the displacement amplitude of the position. Perform statistical analysis on the amplitude data measured at multiple positions in a specific digestive tract segment, and combine it with the specific position of vertex c in the segment to obtain A. c For example, by analyzing CINE-MRI images of the small intestine, it is observed that the maximum inward displacement of a certain segment of the intestinal wall during peristalsis is 2.8 mm, and the maximum outward displacement during relaxation is 3.2 mm (relative to the resting state). The larger absolute value or the average value of the two can be selected as the amplitude reference. In this example, the typical maximum displacement amplitude A of the segment where vertex c is located is set to c =3.0mm.
[0107] Parameter ψ c The steps to obtain the parameter ψ are as follows: c represents the peristaltic angular frequency of the digestive tract segment where the cth vertex is located, in radians per second (rad / s). It is related to the peristaltic wave frequency f c (Unit: Hertz Hz, i.e., number of waves per second) through the relationship ψ c =2πf c Convert to get the creep frequency f c It is usually obtained by the following method: Use gastrointestinal pressure measurement technology (such as high-resolution esophageal or anorectal manometry) to record the pressure wave activity of a specific segment. Analyze the pressure waveform, count the number of times the peristaltic wave (manifested as a characteristic pressure peak) passes within a certain period of time (such as several minutes), and calculate the average number of waves per minute. Convert the number of waves per minute to the number of waves per second (Hz). For example, by monitoring the pressure of a specific small intestine segment for 5 minutes, a total of 12 complete peristaltic pressure wave cycles are recorded, and the average frequency is 12 waves / (5×60 seconds)=12 / 300Hz=0.04Hz. The corresponding angular frequency ψ c =2π×0.04rad / s=0.08πrad / s≈0.2513rad / s. In this example, the creep angular frequency ψ is set for the segment where the vertex c is located. c =0.2513rad / s.
[0108] Parameter κ c The steps to obtain the parameter κ are as follows: cIn the formula, it represents the phase position of the cth vertex in space, and the unit is radian (rad). It is based on the “central axis propagation distance” (denoted as z) from the vertex c to the starting point of the creep wave propagation. c , in millimeters (mm) and the wavelength of the peristaltic wave (recorded as λ, in millimeters (mm)). The acquisition steps are as follows: determine the starting point of the peristaltic wave propagation (for example, the starting end of the duodenum) as the calculation z c The reference origin is extracted from the central axis of the "pressure-adjusted geometry". Calculate the point where the vertex c is projected onto this central axis in three-dimensional space, and then calculate the path length between the projection point along the central axis and the defined propagation starting point, which is z. c Determine the wavelength λ of the peristaltic wave. The wavelength can be obtained by analyzing the distance between two consecutive contraction rings or expansion zones in the dynamic image, or by the formula λ = v / f c Calculate, where v is the propagation velocity of the peristaltic wave (mm / s), f c is the frequency (Hz) obtained above. The peristaltic wave propagation velocity v can be measured from image analysis by tracking the rate of change of the wavefront propagation distance over time. For example, the peristaltic wave propagation velocity v of the small intestine segment where vertex c is located is set to be 15 mm / s, and the frequency f is c = 0.04Hz, then the wavelength λ = 15mm / s / 0.04Hz = 375mm. If the propagation path length z from the vertex c to the starting point of the creep along the central axis is c =150mm, then κ c =(2π / 375mm)·150mm=(2π·150) / 375rad=(300π) / 375rad=0.8πrad≈2.513rad. In this example, set κ c =
[0109] 2.513 radians.
[0110] Parameter β c The steps to obtain the parameter β are as follows: c Represents the initial phase offset of the peristaltic waveform corresponding to the cth vertex, in radians (rad). This parameter is used to adjust the starting state of vertex c in the peristaltic waveform at time t = 0, or to simulate multiple wave sources with different phases or complex waveform patterns. For example, in order to make the wave at the starting point of the propagation path (z c =0, so κ c = 0) begins to shrink inward (negative displacement), and β cSet to π / 2 (if sin(phase) is positive, it means outward, then sin(phase+π / 2) starts from the peak, sin(phase+π) starts from 0 to negative, sin(phase+3π / 2) starts from the negative peak). In this example, set all vertices to a uniform initial phase offset value β c = 0 radians, which means at t = 0, the displacement is only determined by the spatial phase term κ c .
[0111] The parameter t is obtained by the following steps: the parameter t represents the current time point for displacement calculation, and the unit is second (s). In the dynamic simulation process, the time t starts from 0 and increases according to the preset time step Δt. For example, if the total simulation time is 10 seconds and the time step Δt = 0.1 seconds, then t will take 0, 0.1, 0.2, …, 10.0 seconds in turn. For a single calculation example, a specific time point is selected. In this example, the time t = 5.0 s is set.
[0112] Calculation process: select a vertex c and use the above obtained parameter values for calculation. The given parameter values are: A c = 3.0 mm, ψ c = 0.2513 rad / s, κ c = 2.513 radians (this is a spatial phase term, calculated from z c and λ) β c = 0 radians, t = 5.0 s;
[0113] The calculation steps are as follows: calculate the time-dependent phase part:
[0114] ψ c ·t ψ c ·t = 0.2513 rad / s 5.0 s = 1.2565 radians; calculate the total phase angle (radians): θ = ψ c ·t - κ c + β c θ = 1.2565 radians 2.513 radians + 0 radians θ = -1.2565 radians; calculate the sine value: sin(θ) sin(-1.2565 radians). -1.2565 radians is approximately equal to -1.2565 x (180 / π) ° ≈ -72.0°.
[0115] sin(-1.2565) ≈ -0.9510; calculate the final normal displacement value: D c (t) = A c · sin(θ) D c(5.0s) = 3.0 mm · (-0.9510)D c (5.0s) ~ -2.853 mm;
[0116] This result shows that at the time t = 5.0 s, the normal displacement value of the selected vertex c caused by the peristaltic wave is about -2.853 mm. This negative value indicates that the vertex has been displaced in the opposite direction of its normal vector. If the normal direction is defined as pointing from the inside of the lumen to the outside, then a negative displacement means that the tube wall has contracted inward by 2.853 mm. This calculated D c (t) value is the direct basis for updating the three-dimensional coordinate of this vertex in the subsequent step. If D c (t) is positive, it means that the tube wall has expanded outward; if it is 0, it means that the vertex is at the equilibrium position at this time (relative to the amplitude of the peristaltic wave). This numerical result directly constitutes an element in the set of "normal displacement values".
[0117] Based on the normal displacement value (i.e. the D c (t) value calculated for each inner surface vertex c at a specific time t in the previous step, and the initial three-dimensional coordinates and unit normal vectors of each vertex provided by the "pressure-adjusted geometry", the position of each vertex will be updated in space coordinates next to simulate the dynamic deformation caused by the peristaltic wave. The specific operation is that for each vertex c on the inner surface (i.e. the surface in contact with the lumen) of the "pressure-adjusted geometry", first obtain its three-dimensional coordinates P c = (x c , y c , z c ) and its corresponding unit normal vector pointing to the outside of the lumen (or to the inside of the tissue, consistency must be maintained) at this geometry. At the same time, obtain the calculated normal displacement value D c (t) of this vertex at the current time t, then the updated three-dimensional coordinates P' c (t) of the vertex c at time t due to the action of the peristaltic wave is calculated by vector addition, and its formula is This means that the original coordinate point is moved along its normal direction by a distance of D c (t). If D c (t) is positive, it is moved in the direction (e.g. tube wall expansion), if it is negative, it is moved in the In the reverse direction (e.g. the tube wall contracts), this coordinate updating process is applied to all vertices on the inner surface, resulting in the three-dimensional morphology of the entire inner surface at time t. In order to generate a dynamic deformation process, the above process of calculating the normal displacement of all inner surface vertices at a specific time t and updating their coordinates is repeated iteratively at a preset "time step interval" Δt. The selection of the "time step interval" Δt should take into account the smoothness of the simulation and the computational efficiency. For example, if the typical period of peristaltic wave is 5 seconds, Δt can be set to 0.05 seconds to generate 100 frames of animation within one period, ensuring the smoothness of the dynamic effect. When setting the specific value, it is necessary to ensure that Δt is less than the characteristic time scale of the peristaltic wave (e.g. 1 / 20 to 1 / 50 of the period). For example, if the frequency of the peristaltic wave is 0.25 Hz (period 4 seconds), Δt can be set to 0.04 seconds (i.e. 25 frames per second). At each discrete time point (or "time frame") of the simulation, such as t0=0, t1=Δt, t2=2Δt, …, tk=kΔt, the three-dimensional coordinate state of all relevant vertices (usually the inner surface vertices, or possibly including other layer vertices inside the wall according to the complexity of the model) at that moment is calculated and summarized. The set of all vertex coordinates at each time frame is stored as an independent geometric snapshot. These geometric snapshots arranged in time sequence collectively constitute the dynamic deformation geometry sequence of the peristaltic process of the digestive tract. k = kΔt, the three-dimensional coordinate state of all relevant vertices (usually the inner surface vertices, or possibly including other layer vertices inside the wall according to the complexity of the model) at that moment is calculated and summarized. The set of all vertex coordinates at each time frame is stored as an independent geometric snapshot. These geometric snapshots arranged in time sequence collectively constitute the dynamic deformation geometry sequence of the peristaltic process of the digestive tract.
Claims
1. A virtual 3D reconstruction system for the digestive tract, characterized by: The system comprises: The initial geometry construction module, based on the tomographic image sequence of medical images, identifies the contour boundary points of the digestive tract region in the tomographic image sequence, connects the three-dimensional space points and generates the initial surface mesh, and then smoothes the generated initial surface mesh to establish the initial lumen surface model; A layered shell construction module, based on the geometric data of the initial lumen surface model and a preset target thickness value of the first wall layer of the digestive tract, performs an offset calculation along the normal direction of each vertex of the initial lumen surface model to obtain the coordinates of the offset vertices of the first layer. Based on the coordinates of the first layer of offset vertices, the coordinate points are connected to generate a mesh of the outer surface of the first wall layer, thereby establishing a multi-layer digestive tract shell model; a material thickness calibration module, which selects a target layer of the model based on the hierarchical geometric data of the multi-layer digestive tract shell model, calculates the distance between the inner surface mesh vertices of the target layer and the corresponding outer surface mesh vertices, obtains a target layer vertex thickness value set, and assigns a preset material identifier and the calculated thickness value to the mesh unit of the layer based on the target layer vertex thickness value set, thereby establishing an attributed digestive tract solid model; The dynamic morphological evolution module adjusts the three-dimensional coordinates of the vertices of the inner surface of the attributed digestive tract entity model based on the geometric and attribute data of the attributed digestive tract entity model and the target pressure value in the lumen to obtain the pressure-adjusted geometric shape. Based on the pressure-adjusted geometric shape and according to the peristaltic wave characteristics in the physiological parameter set, the model vertices are subjected to serialized displacement value calculations that conform to the peristaltic wave propagation law to establish a dynamic deformation geometric sequence.
2. The digestive tract virtual 3D reconstruction system according to claim 1, characterized in that: The steps for obtaining the initial lumen surface model are: Read the tomographic image sequence of medical images, perform image segmentation on each frame, identify the digestive tract area in the image and extract all boundary contour points to form a contour boundary point set of the digestive tract area; Based on the outline boundary point set of the digestive tract region, the corresponding boundary points in adjacent image frames are paired in three-dimensional coordinates, and surface connection is performed based on the point pairs, and all connection results are combined to form an initial surface mesh; According to the initial surface mesh, a geometric curvature analysis is performed on each group of triangular facets, sharp connection areas are marked, and the positional relationship of continuous adjacent facet boundary points is called for smooth interpolation to form an initial lumen surface model.
3. The digestive tract virtual 3D reconstruction system according to claim 1, characterized in that: The steps for obtaining the first layer offset vertex coordinates are: Based on the initial lumen surface model, all triangular facets constituting the initial lumen surface model are extracted, each triangular facet is parsed into a combination of three vertex coordinates according to the vertex number, a normal vector of the triangular facet is calculated by a three-dimensional vector cross multiplication method, and the normal vector values of the three vertices corresponding to the triangular facet are accumulated respectively. After the accumulation of all triangular facets is completed, the accumulated normal vector of each vertex is divided by the corresponding accumulation number to generate a unit normal vector set of each vertex of the initial lumen surface model; According to the unit normal vector set of each vertex of the initial lumen surface model, the three-dimensional coordinate value corresponding to each vertex is read, and the preset target thickness value of the first wall layer of the digestive tract is called. Each unit normal vector is multiplied by the target thickness value of the first wall layer of the digestive tract to obtain an offset displacement vector for each vertex, and then the offset displacement vector is added to the original three-dimensional coordinate value component by component according to the coordinate axis dimension to generate the new three-dimensional coordinate of each vertex after the offset; Based on the new three-dimensional coordinates of each vertex after offset, the vertex numbers of the original initial tubular surface model are rearranged point by point according to the index order, the decimal residuals caused by floating-point errors in the coordinate expression are eliminated, and a unified precision specification is performed to generate the first layer of offset vertex coordinates.
4. The digestive tract virtual 3D reconstruction system according to claim 1, characterized in that: The steps for obtaining the multi-layer digestive tract shell model are: Based on the offset vertex coordinates of the first layer, the original connection relationship of the triangular facets is matched according to the topological structure order of the triangular facets in the initial lumen surface model, and a vertex index mapping is established. Each set of three offset vertex coordinates is called for spatial connection to construct a set of triangular facets with a consistent structure, thereby generating the first wall layer outer surface mesh; Based on the outer surface mesh of the first wall layer, all vertex coordinates of the mesh are extracted as a new round of offset reference coordinate set, the target thickness values of subsequent wall layers are sequentially read, and an offset operation is performed on each target thickness value in the direction of the vertex unit normal. A new vertex set of each level is constructed by accumulating the offset coordinates to obtain the vertex coordinate set of all wall layers; Based on the vertex coordinate sets of all the wall layers, the triangular patch structure is reused layer by layer to generate the corresponding surface mesh according to the connection order of the previous layer of patches corresponding to each layer's offset coordinate set. The surface meshes of each layer are topologically closed and organized in hierarchical order to generate a multi-layer digestive tract shell model.
5. The digestive tract virtual 3D reconstruction system according to claim 1, characterized in that: The steps for obtaining the target layer vertex thickness value set are: Based on the hierarchical geometric data of the multi-layer digestive tract shell model, the structural attribute identifier of each layer is traversed and the number of vertex groups, vertex index range, connection order and layer number contained therein are read. Matching is performed according to the input target layer number, the model target layer corresponding to the number is locked, and the three-dimensional coordinates of all inner surface mesh vertices and outer surface mesh vertices in the model target layer are respectively extracted as structured coordinate pair sets to generate a model target layer vertex coordinate pair set; According to the set of vertex coordinate pairs of the model target layer, the three-dimensional coordinates of each group of inner surface mesh vertices and the corresponding outer surface mesh vertices are looped and traversed, and the coordinate differences in the X-axis, Y-axis, and Z-axis directions are calculated for each group of coordinate pairs. The squares of the differences in the three directions are summed and then squared to construct a sequence of true geometric distance values between vertices. All distance values are then organized into a sequence of original thickness values according to the corresponding relationship of vertex number indexes to generate a three-dimensional space thickness distance sequence of the model target layer; Based on the three-dimensional space thickness distance sequence of the target layer of the model, each distance value is rounded and regularized, and after the processing is completed, it is re-sorted according to the original vertex index order to generate a target layer vertex thickness value set.
6. The digestive tract virtual 3D reconstruction system according to claim 1, characterized in that: The steps for obtaining the attributed digestive tract entity model are: Based on the target layer vertex thickness value set, read the grid unit structure information corresponding to the target layer of the model, traverse each unit index in the grid unit, sequentially obtain the associated three or four vertex numbers, and match and read all vertex numbers in the target layer vertex thickness value set to form a thickness value set corresponding to each grid unit, and generate a target layer grid unit thickness mapping set; According to the target layer grid cell thickness mapping set, a preset material identifier corresponding to the model target layer is read, and the material identifier of each grid cell and the average thickness value calculated in the grid cell are combined into an attribute label group through cell-by-cell binding, and all the label groups are embedded into the target layer structure in the order of grid cell numbers to generate a target layer attributed grid cell set; Based on the target layer attributed grid unit set, the attribute label group binding process is repeatedly executed for all remaining layers in the multi-layer digestive tract shell model, and all processed attributed grid unit sets are integrated layer by layer to establish an attributed digestive tract entity model.
7. The digestive tract virtual 3D reconstruction system according to claim 1, characterized in that: The steps for obtaining the geometric shape after pressure adjustment are: Based on the geometry and attribute data of the attributed digestive tract solid model, the inner surface vertex coordinates, the corresponding vertex normal direction, the material identifier, the target layer thickness value and the hierarchical structure number are read, and the spatial position, attribute value and structural association information of all vertices are unified into a set of structured calculation basic units to generate a comprehensive information set of the inner surface vertices; Based on the comprehensive information set of the inner surface vertices, the target pressure value in the lumen is read, the material identifier corresponding to each vertex is mapped to the material elastic response coefficient, and the elastic response coefficient is corrected in combination with the target layer thickness value. The displacement response increment under the unit pressure of each vertex is determined, and the three-dimensional coordinates of each vertex are directionally offset along the normal direction to generate a vertex coordinate set corrected for the lumen pressure response; Based on the vertex coordinate set corrected by the lumen pressure response, the spatial position relationship of all vertex coordinates is reassembled in hierarchical order and the original topological structure, while keeping the original mesh surface connections unchanged, forming an updated model consistent with the structure of the attributed digestive tract entity model, and generating the pressure-adjusted geometric shape.
8. The digestive tract virtual 3D reconstruction system according to claim 1, characterized in that: The steps for obtaining the dynamic deformation geometry sequence are: According to the pressure-adjusted geometric shape, the peristaltic wave propagation velocity, the propagation path length of the vertex along the central axis, and the initial phase angle in the physiological parameter set are called to calculate the normal displacement value of each vertex at a given time; Based on the normal displacement value, the normal displacement value corresponding to each vertex is updated in the spatial coordinate along the unit normal direction, and is updated multiple times at time step intervals. The three-dimensional coordinate states of all vertices are summarized in each time frame to generate a dynamic deformation geometry sequence.
9. The method for virtual 3D reconstruction of the digestive tract according to any one of claims 1 to 8, characterized in that: The following steps are involved: Based on the tomographic image sequence of medical images, the contour boundary points of the digestive tract area in the tomographic image sequence are identified, three-dimensional space point connection and initial surface mesh generation are performed, and the generated initial surface mesh is smoothed to establish the initial lumen surface model; Based on the geometric data of the initial lumen surface model and the preset target thickness value of the first wall layer of the digestive tract, an offset calculation is performed along the normal direction of each vertex of the initial lumen surface model to obtain the coordinates of the first layer offset vertices. Based on the first layer offset vertex coordinates, the coordinate points are connected to generate the outer surface mesh of the first wall layer, and a multi-layer digestive tract shell model is established; Based on the hierarchical geometric data of the multi-layer digestive tract shell model, a target layer of the model is selected, the distance between the inner surface mesh vertex of the target layer and the corresponding outer surface mesh vertex is calculated, and a target layer vertex thickness value set is obtained. Based on the target layer vertex thickness value set, a preset material identifier and the calculated thickness value are assigned to the mesh unit of the layer to establish an attributed digestive tract solid model. Based on the geometry and attribute data of the attributed digestive tract entity model, as well as the target pressure value in the lumen, the three-dimensional coordinates of the vertices of the inner surface of the attributed digestive tract entity model are adjusted to obtain a pressure-adjusted geometric shape. Based on the pressure-adjusted geometric shape and in accordance with the peristaltic wave characteristics in the physiological parameter set, a serialized displacement value calculation is performed on the model vertices in accordance with the peristaltic wave propagation law to establish a dynamic deformation geometric sequence.
Citation Information
Patent Citations
Computer simulation scaling biopsy method and apparatus
CN101322652A
Customized alimentary canal support and moulding method and application method thereof
CN105640679A
Method for reconstructing three-dimensional image of tissue cavity
CN115841537A
Ansys-based intestinal wall and wall surface motion modeling method and device
CN119903703A
Method, system and sheet for three-dimensional tomographic image construction, and three-dimensional tomographic image constructing apparatus
JP2013027639A
Cited By
Digestive tract length calculation method and device based on in-vivo micro intelligent terminal system image, storage medium and equipment
CN121661122A
Complex contour model construction method and device
CN121937644A