Heart left ventricle modeling image ventricle wall thickness calculation processing method

Abstract

The invention discloses a heart left ventricle modeling image ventricle wall thickness calculation processing method, which comprises the following steps of: reconstructing a personalized three-dimensional left ventricle model based on a patient heart magnetic resonance image, solving a myocardial transmural gradient by adopting a Laplace-Dirichlet parameterization method, and dividing a ventricle wall into a plurality of layers; through a layered mapping strategy, determining a mapping point between the myocardium of the inner layer and the myocardium of the outer layer by adopting a method of combining cone space search and Laplace-Dirichlet value matching, and completing mapping between the outermost layer and the ventricular outer membrane by adopting a nearest neighbor algorithm; and finally, through layer thickness real-time accumulation and data post-processing, obtaining accurate left ventricular wall thickness distribution data. The method overcomes the defect that a rigid mapping method easily generates measurement deviation when facing pathological forms such as myocardial fibrosis and ventricular wall tumors, improves the precision and reliability of ventricular wall thickness measurement, and provides effective technical support for accurate diagnosis, treatment planning and dynamic monitoring of cardiovascular diseases.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of cardiac medical image processing and computational modeling, in particular, to a method for left ventricular three-dimensional modeling and accurate calculation of its wall thickness based on cardiac magnetic resonance imaging. BACKGROUND

[0002] In the clinical diagnosis and treatment of cardiovascular diseases, accurate quantitative analysis of left ventricular structure is the core of early disease diagnosis, dynamic monitoring of disease progression, treatment strategy development and prognosis evaluation. Left ventricular wall thickness, as a key morphological parameter, has been included in the clinical guidelines of international authoritative organizations such as the American Heart Association and the European Society of Cardiology. The measurement accuracy directly affects the accuracy of disease risk stratification and prognosis evaluation.

[0003] Currently, cardiac magnetic resonance imaging is considered the gold standard for myocardial structure assessment due to its high soft tissue resolution and three-dimensional imaging capability. Transthoracic echocardiography is a commonly used clinical examination method due to its strong real-time performance, good portability and high cost-effectiveness.

[0004] However, most existing left ventricular wall thickness measurement techniques rely on two-dimensional image manual tracing or simplified calculations based on finite element models. For example, in the prior art, the endocardial and epicardial boundaries of the CMR image are segmented, a finite element mesh is constructed, and segmentation is divided. The maximum value of the segment is taken as the global maximum wall thickness. This method is essentially a "rigid mapping" based on the endocardial and epicardial boundaries. When faced with complex myocardial structural changes due to myocardial fibrosis, aneurysm and other pathological conditions, significant measurement deviations are likely to occur, thereby limiting its application in precision medicine and cardiac electrophysiology model construction.

[0005] Therefore, there is an urgent need in the art for a new method that can adapt to complex myocardial morphological changes and provide high-precision, personalized left ventricular wall thickness measurement. SUMMARY

[0006] The present application aims to overcome the shortcomings of the prior art and provide a method for accurately and individually calculating the left ventricular wall thickness of the heart. This method is particularly suitable for scenarios where the myocardium undergoes irregular deformation under pathological conditions, thereby improving the accuracy and reliability of the measurement.

[0007] The method is based on a Laplace-Dirichlet parameterization method to solve the transmural gradient of myocardial tissue, and the application takes the division of the ventricular wall into three layers as an example for specific description, and the actual number of layers can be flexibly adjusted according to the anatomical structure characteristics of the ventricular wall and the measurement accuracy requirements. The core adjustment principle is to avoid the layering of the interlayer transmural phenomenon caused by thin layers, and to ensure the anatomical continuity and data effectiveness of the mapping between adjacent layers. In the mapping process, a layered mapping strategy is adopted: the endocardium to the outermost myocardial layer (the third layer in the example) is matched with the Laplace-Dirichlet value to determine the mapping point and the new mapping plane through conic space search, and the outermost myocardial layer to the epicardium adopts the nearest neighbor algorithm to determine the mapping point because the boundary Laplace-Dirichlet value is 1. Through conic space search, new mapping plane generation and layer thickness accumulation, full-layer mapping is realized. Finally, after low-pass filtering to remove outliers and inverse distance weighted interpolation processing, accurate left ventricular wall thickness data is obtained, solving the measurement deviation caused by abnormal myocardial deformation caused by diseases, and providing a reliable reference for clinical treatment and diagnosis.

[0008] The application provides a left ventricular modeling image wall thickness calculation processing method, which has significant advantages. In terms of measurement accuracy and model construction, the Laplace-Dirichlet method is used to assign a smooth and continuous value of 0 to 1 to the grid points of the ventricular model, accurately reflecting the hierarchical characteristics of myocardial cells, avoiding the errors of traditional discrete measurement, and capturing the spatial changes of myocardium in detail; based on this, the myocardium is divided into three layers, and a layered mapping method is used to realize full-layer construction, with synchronous calculation and real-time accumulation of layer thickness each time to ensure measurement integrity and accuracy. At the same time, combined with the Frontal-Delaunay algorithm and the MeshAdapt algorithm, a three-dimensional left ventricular cavity model is reconstructed, which not only retains the detailed features of the heart, but also adaptively refines the local grid, so that the model highly fits the real heart structure and reduces the measurement error; in the face of the mapping problem of the outer layer of the ventricle, the nearest neighbor algorithm is used to ensure the smooth completion of full-layer mapping.

[0009] In the field of clinical application, the method is based on the CMR-LGE image of the individual patient to segment the ventricular boundary and reconstruct the model, generate a personalized left ventricular model, accurately measure the wall thickness, provide data support for doctors to develop treatment plans that fit the disease, and realize precision medicine. At the disease diagnosis level, accurate wall thickness measurement data has decisive significance for early identification, disease assessment and prognosis of hypertrophic cardiomyopathy, myocardial infarction and other heart diseases; in the surgical treatment stage, research has shown that the degree of myocardial fibrosis is significantly positively correlated with the wall thickness, which makes the thickness measurement an important basis for evaluating the degree of myocardial disease. Detailed left ventricular wall thickness information before surgery can assist doctors in precise surgical simulation and program planning, accurately target the ablation site through wall thickness, reduce the risk of surgery, and predict the impact of surgery on heart function. In the whole process of disease treatment management, regular measurement of left ventricular wall thickness can dynamically monitor myocardial changes in real time, provide quantitative basis for treatment effect evaluation and strategy adjustment, and ultimately realize the dual improvement of treatment quality and patient quality of life.

[0010] Technical scheme of the present application

[0011] To achieve the above object, the present application adopts the following technical scheme:

[0012] A left ventricular modeling image wall thickness calculation processing method, comprising the following steps:

[0013] S1, myocardial stratification standard: dividing the ventricular wall into at least three layers according to a predetermined numerical range;

[0014] S2, mapping standard: dividing the myocardial structure into endocardium, multi-layer myocardial layer and epicardium, and using a stratified mapping method, wherein the multi-layer mapping process from the endocardium to the outermost myocardial layer is determined by a conic space search and a Laplace-Dirichlet value matching mapping point, and the mapping process from the outermost myocardial layer to the epicardium uses a nearest neighbor algorithm to determine the mapping point;

[0015] S3, personalized three-dimensional heart model reconstruction: segmenting the left ventricular boundary based on the cardiac magnetic resonance image, and reconstructing a high-resolution three-dimensional left ventricular cavity model;

[0016] S4, myocardial stratification: solving the myocardial tissue transmural gradient by a Laplace-Dirichlet parameterization method, and numerically assigning the myocardial finite element grid points to realize left ventricular stratification;

[0017] S5, ventricular wall thickness calculation: through stratified data attribution unique processing, mapping plane construction and search, new mapping plane generation and layer thickness calculation, layer thickness accumulation and data post-processing, left ventricular wall thickness data is obtained.

[0018] Preferably, in step S1, the preset numerical range is 0-1, the ventricular wall is divided into three layers, including the first layer of myocardium (adjacent to the endocardium), i.e. the myocardial layer starting from the endocardium, the second layer of myocardium and the third layer of myocardium (adjacent to the epicardium), i.e. the myocardial layer ending at the epicardium, and the numerical range of each layer is dynamically adjusted according to the actual proportion requirement.

[0019] Preferably, in step S2, the layer mapping method specifically includes:

[0020] Based on the triangular facet of the endocardium, the centroid and normal vector are calculated, and a conical search space is constructed with the normal vector as the axis and the centroid as the vertex;

[0021] In the target point set, the candidate point set with an included angle less than a set threshold value with the normal vector is screened;

[0022] According to the size of the candidate point set, the point closest to the Laplace-Dirichlet value is selected to construct a new mapping plane or calculate the midpoint to generate a new perpendicular line;

[0023] For the mapping from the outermost myocardial layer to the epicardium, the nearest neighbor algorithm with Euclidean distance measurement is used to select the mapping point.

[0024] Preferably, in step S3, the personalized three-dimensional heart model reconstruction includes:

[0025] S31, ventricular boundary segmentation: manually or automatically segmenting the left ventricular, endocardial and epicardial boundaries on the CMR-LGE image;

[0026] S32, model reconstruction: using the Frontal-Delaunay algorithm combined with the adaptive mesh division algorithm to redivide the coarse surface mesh to generate a high-resolution three-dimensional model that retains the detailed features of the heart and is locally refined.

[0027] Preferably, in step S4, the myocardial layering includes:

[0028] S41, model surface extraction: extracting the apical and basal surfaces, as well as the endocardial and epicardial surfaces;

[0029] S42, Laplace-Dirichlet parameterization: obtaining the scalar value distribution of the transmural direction of the myocardial tissue by solving the Laplace equation and applying the Dirichlet boundary condition;

[0030] S43, reference value integration: assigning numerical values to the grid points according to the transmural gradient direction to form a layered model.

[0031] Preferably, in the step S5, the layer data attribution uniqueness includes: setting data priority rules, endocardium > inner myocardium > outer myocardium > epicardium, traversing all layer data, removing overlapping points, and ensuring that each spatial point is uniquely attributed in each layer.

[0032] Preferably, in the step S5, the mapping plane construction and search includes:

[0033] calculating the centroid and unit normal vector of the triangular facet;

[0034] constructing a conical search space with the centroid as the vertex and the normal vector as the axis;

[0035] searching for a point set located in the conical body in the target point set as a candidate point.

[0036] Preferably, in the step S5, the new mapping plane generation and layer thickness calculation includes:

[0037] if the number of candidate points is greater than or equal to 3, selecting three points with the closest Laplace-Dirichlet value to construct a new plane, and calculating the distance between the centroid of the new plane and the original centroid as the layer thickness;

[0038] if the number of candidate points is 1 or 2, calculating the midpoint and connecting it with the centroid, and taking the distance between the midpoint and the centroid as the layer thickness;

[0039] if the candidate point is empty, keeping the original plane and marking the current layer thickness as 0.

[0040] Preferably, in the step S5, the layer thickness accumulation includes: starting from the endocardium, mapping and accumulating the thickness between layers in real time, and finally generating a data set containing the epicardial point index and the corresponding accumulated layer thickness.

[0041] Preferably, in the step S5, the data post-processing includes: calculating the arithmetic mean of multiple layer thickness data in the epicardial point cloud, removing outliers through low-pass filtering, and using inverse distance weighted interpolation algorithm to interpolate the entire epicardial region to obtain a smooth and accurate left ventricular wall thickness distribution map.

[0042] Advantages of the present application

[0043] Compared with the prior art, the present application has the following advantages:

[0044] High precision and individualization: based on patient-specific CMR image reconstruction of three-dimensional model, and using Laplace-Dirichlet parameterization method to capture the spatial layering characteristics of myocardium, avoiding the error of traditional discrete measurement.

[0045] Strong anti-deformation ability: the innovative stratified mapping strategy, combined with conic space search and nearest neighbor algorithm, effectively overcomes the rigid mapping deviation caused by pathological morphological changes, and the measurement result is more reliable.

[0046] High clinical practical value: the accurate wall thickness data provided has direct guiding significance for the diagnosis, operation planning (such as ablation target positioning) and treatment effect dynamic monitoring of diseases such as hypertrophic cardiomyopathy and myocardial infarction, and strongly supports the practice of precision medicine. BRIEF DESCRIPTION OF DRAWINGS

[0047] Figure 1 is the overall flowchart of the left ventricular modeling image wall thickness calculation processing method of the application.

[0048] Figure 2 is a schematic diagram of myocardial stratification; (a) shows the surface extraction result of the left ventricular model, (b) shows the Laplace-Dirichlet boundary condition value distribution of the myocardial tissue, and (c) shows the solution result of the myocardial tissue transmural gradient.

[0049] Figure 3 is the final effect diagram of myocardial stratification.

[0050] Figure 4 is a comparison diagram of wall thickness calculation results; (a) is a perspective view of the heart chamber model, (b) is a stratified perspective view of the heart chamber model, and (c) is a three-dimensional visualization result of the calculated wall thickness data. DETAILED DESCRIPTION

[0051] The following takes the heart magnetic resonance image of the collector as an example, and further describes the application combined with specific implementation steps, such as Figure 1 As shown in the figure, the method for ventricular wall stratification evaluation includes the following steps:

[0052] A left ventricular modeling image wall thickness calculation processing method includes the following steps:

[0053] S1, myocardial stratification standard: the ventricular wall is divided into at least three layers according to a preset numerical range; the preset numerical range is 0-1, for example, the ventricular wall is divided into three layers, including the first layer of myocardium (adjacent to the inner atrial membrane), i.e. the stratification of myocardium starting from the inner atrial membrane, the second layer of myocardium and the third layer of myocardium (adjacent to the outer atrial membrane), i.e. the stratification of myocardium ending at the outer atrial membrane, and the numerical range of each layer is dynamically adjusted according to the actual proportion demand.

[0054] More specific steps are:

[0055] Considering that the finite element mesh points from the endocardium to the epicardium of the ventricular model are uniformly assigned values, to avoid the ventricular wall being penetrated between layers due to excessively thin layers, and to meet the needs of studying the distribution of ventricular wall cells, the ventricular wall is divided into three layers based on a numerical range of 0-1. The layering result is determined according to layering requirements, such as a 2:3:5 ratio for the first, second, and third layers. Therefore, the range 0-0.2 is set for the first layer of the myocardium, 0.2-0.5 for the second layer, and 0.5-1 for the third layer. The specific number of layers in the myocardium can be modified as needed.

[0056] S2. Mapping Criteria: The myocardial structure is divided into the endocardium, multiple myocardial layers, and the pericardium. A layered mapping method is adopted. The multi-layer mapping process from the endocardium to the outermost myocardial layer is determined by cone space search and Laplace-Dirichlet value matching to determine the mapping point. The mapping process from the outermost myocardial layer to the pericardium is determined by the nearest neighbor algorithm to determine the mapping point.

[0057] The hierarchical mapping method specifically includes:

[0058] Based on the triangular facet of the endocardium, its centroid and normal vector are calculated, and a conical search space with the normal vector as the axis and the centroid as the vertex is constructed.

[0059] Filter the set of candidate points in the target point set whose angle with the normal vector is less than a set threshold;

[0060] Based on the size of the candidate point set, select the point with the closest Laplace-Dirichlet value to construct a new mapping plane or calculate the midpoint to generate a new perpendicular line;

[0061] For the mapping from the outermost myocardial layer to the pericardium, the nearest neighbor algorithm with Euclidean distance metric is used to select the mapping point.

[0062] The more specific steps are as follows:

[0063] To avoid errors that may occur when the rigid mapping based on the boundaries of the inner and outer membranes is used when the myocardium undergoes irregular deformation, this application divides the myocardial structure into five layers: the endocardium, three myocardial layers, and the endocardium, and proposes a layered mapping method.

[0064] Taking the endocardial lining-first layer mapping as an example, first take a triangular patch of the endocardial lining. , as well as Three points, and calculate their centroid (center of mass). Construct the normal vector perpendicular to the triangular facet of the endocardium. ,by Normal vector is the axis and the centroid. Based on the reference, a half vertex angle is formed. cone search space, the first layer of myocardial point set screening with the normal vector angle less than candidate point set If , from select three points closest to the Laplace-Dirichlet value, to constitute the new plane of the first layer-second layer mapping If , calculate the centroid of all points in , connect the centroid and the original endocardial triangular patch centroid to generate a new perpendicular line for subsequent mapping; if is empty, merge the triangular patch into the mapping plane of the first layer, and search again when the first layer-second layer mapping. The mapping of the first layer-second layer and the second layer-third layer is repeated according to the above rules, and the operations of perpendicular line construction, cone space search and Laplace-Dirichlet value screening are repeated.

[0065] In the third layer-ventricular epicardium mapping, since the Laplace-Dirichlet value of the ventricular epicardium region is 1, the optimal mapping point cannot be determined, so the nearest neighbor algorithm based on Euclidean distance measurement is used, the distance from the centroid of each mapping plane in the third layer to each point on the ventricular epicardium surface is calculated, and the point with the smallest distance is selected as the mapping point to realize the full layer mapping construction.

[0066] S3, personalized three-dimensional heart model reconstruction: segmenting the left ventricular boundary based on cardiac magnetic resonance image, and reconstructing a high-resolution three-dimensional left ventricular cavity model;

[0067] The personalized three-dimensional heart model reconstruction comprises the following steps:

[0068] S31, ventricular boundary segmentation: manually or automatically segmenting the left ventricular, endocardial and epicardial boundaries on the CMR-LGE image;

[0069] S32, model reconstruction: using the Frontal-Delaunay algorithm combined with the adaptive mesh division algorithm to redivide the rough surface mesh to generate a high-resolution three-dimensional model that retains the detailed features of the heart and is locally refined.

[0070] The personalized three-dimensional heart model reconstruction comprises the following steps:

[0071] The personalized three-dimensional heart model reconstruction comprises the following steps:

[0072] ​S31, ventricular boundary segmentation: for each patient data, an experienced expert manually or automatically segments the left ventricle, endocardial, epicardial boundary on the CMR-LGE image;

[0073] S32, model reconstruction: reconstruct a high-resolution three-dimensional left ventricular cavity model; take the Frontal-Delaunay algorithm to automatically interpolate the segmented heart cavity surface to obtain a high-resolution ventricular cavity data. Then the ventricular cavity model is meshed by finite elements to obtain a high-resolution ventricular finite element model, wherein the average side length of the mesh is about 400 mm. The model consists of 4 million to 6 million nodes and 20 million to 40 million tetrahedral elements.

[0074] S4, myocardial stratification: solve the myocardial transmural gradient by Laplace-Dirichlet parameterization method, and numerically assign the left ventricular finite element grid points to realize the left ventricular stratification;

[0075] The myocardial stratification comprises:

[0076] S41, model surface extraction: extract the apical, basal surface and endocardial, epicardial surface;

[0077] S42, Laplace-Dirichlet parameterization: by solving the Laplace equation and applying the Dirichlet boundary condition, the scalar value distribution of the myocardial tissue internal transmural direction is obtained;

[0078] S43, reference value integration: numerically assign the grid points according to the transmural gradient direction to form a stratified model.

[0079] In order to numerically assign each finite element grid point in the ventricular model for stratification, the gradient of the Laplace equation solution of the Dirichlet boundary condition is used to define the apical-basal direction and the transmural direction. The specific steps of the myocardial stratification are:

[0080] S41, model surface extraction; the algorithm using the Laplace-Dirichlet rule first needs to determine the model surface of the ventricular model, and the model surface extraction is divided into apical and basal surface extraction, and endocardial and epicardial surface extraction.

[0081] By slicing on the long axis of the heart, the apical and basal surfaces are extracted, and the basal surface is defined as ∂Ω base , and the apical surface is defined as ∂Ω apex . As Figure 2 (a) As shown in the left side of the figure, the green part is the basal surface, and the black part is the apical surface.

[0082] Subsequently, the endocardium and epicardium of the left ventricle were extracted, and they were sorted by the size of the point set. The contours of the membranes were obtained by projecting the points on the membrane surface to a plane perpendicular to the long axis, and their circularities were calculated to distinguish the endocardium (∂Ω lv ) and epicardium (∂Ω epi ) of the left ventricle. As shown in Fig. Figure 2 (a), the blue surface is the epicardium, and the pink surface is the endocardium of the left ventricle.

[0083] S42, Laplace-Dirichlet parameterization: the Laplace-Dirichlet parameterization method of myocardial tissue; a function named Laplace(Ω mesh , A, B) is introduced, which is used to solve the Laplace equation on a grid Ω mesh , and can return a scalar value according to the Dirichlet boundary condition on the surface A (maximum) and B (minimum). In cardiac modeling, the application of this function allows the calculation of the reference value of different positions from 0 to 1 in the intramural direction of the myocardial tissue.

[0084] S43, reference value integration: after the intramural gradient of the myocardial tissue is solved, the result is obtained, which represents the intramural gradient direction from the epicardium to the endocardium of the left ventricle, and the Dirichlet boundary condition is , , as shown in Fig. Figure 2 (b); after the intramural gradient of the myocardial tissue is solved, the result of the intramural gradient of the myocardial tissue is obtained, as shown in Fig. Figure 3 (c).

[0085] Considering that this study only needs to measure the left ventricular wall thickness, the numerical value of the myocardial finite element grid point can be assigned according to the result, and the layered model of the left ventricle is obtained, as shown in Fig. Figure 4 .

[0086] S5, ventricular wall thickness calculation: left ventricular wall thickness data is obtained by unique processing of layered data, mapping plane construction and search, new mapping plane generation and layer thickness calculation, layer thickness accumulation and data post-processing.

[0087] The unique processing of layered data includes: setting the data priority rule, endocardium > inner myocardium > outer myocardium > epicardium, traversing all layered data, removing overlapping points, and ensuring that each spatial point is uniquely attributed to each layer.

[0088] The mapping plane construction and search include:

[0089] Calculate the centroid and unit normal vector of each triangular facet;

[0090] Construct a conical search space with the centroid as the vertex and the normal vector as the axis;

[0091] Search for points in the target point set that are located in the conical space as candidate points.

[0092] The new mapping plane generation and layer thickness calculation includes:

[0093] If the number of candidate points is ≥3, select the three points with the closest Laplace-Dirichlet value to construct a new plane, and calculate the distance between the new plane centroid and the original centroid as the layer thickness;

[0094] If the number of candidate points is 1 or 2, calculate the midpoint and the line connecting the midpoint and the centroid, and take the distance between the midpoint and the centroid as the layer thickness;

[0095] If the number of candidate points is 0, keep the original plane and mark the current layer thickness as 0.

[0096] The layer thickness accumulation includes: starting from the endocardium, mapping and accumulating the thickness between each layer in real time, and finally generating a dataset containing the epicardial point index and the corresponding accumulated layer thickness.

[0097] The data post-processing includes: calculating the arithmetic mean of multiple layer thickness data in the epicardial point cloud, removing outliers through low-pass filtering, and using inverse distance weighted interpolation algorithm to interpolate the entire epicardial region to obtain a smooth and accurate left ventricular wall thickness distribution map.

[0098] The specific steps for calculating the ventricular wall thickness are:

[0099] Layered data attribution uniqueness: In the process of layering the three-dimensional left ventricular cavity model based on tetrahedral mesh, due to the topological characteristics of the data structure, there is inevitably a phenomenon of spatial point cross-layer overlap, i.e. the same spatial coordinate point belongs to different layers. The existence of such overlapping points can significantly increase the probability of mapping errors during ventricular wall thickness calculation, seriously affecting the accuracy and reliability of subsequent layer thickness calculation.

[0100] To effectively solve this problem, before calculating the ventricular wall thickness, the layered data needs to be systematically pre-processed. By developing a strict priority removal strategy, the consistency and accuracy of the data are ensured. The priority setting is: endocardium First layer Second layer Third layer, epicardium Third layer.

[0101] The specific operation process is as follows: Traverse all layered data in the 3D model. For each spatial coordinate point, if it exists simultaneously in tetrahedral elements of different layers, process it according to the aforementioned priority rules. For example, when point... When a point is contained within both the first and second layer tetrahedral elements, it will be... By removing points from the second layer of mapped point cloud data, each spatial point is ensured to be uniquely assigned during the layer thickness calculation process, laying the foundation for subsequent accurate layer thickness calculation and mapping analysis.

[0102] Mapping Plane Conic Space Construction and Point Search: For points in space , and The determined mapping plane is first determined according to the formula , , Calculate its centroid Through vector cross product The unit normal vector of the plane is obtained after normalization. With center of mass For vertex, normal vector Using the axis as the semi-vertical angle, construct a semi-vertical angle as Given a conical space, based on the criterion for judging the positional relationship of points in three-dimensional space, search for all point sets located within the cone within the target point set.

[0103] New mapping plane generation and layer thickness calculation: For the number of candidate points in the target point set that meet the search criteria... The Laplace-Dirichlet value corresponding to the point is used for processing to form a new mapping plane. The generation of the new mapping plane involves three cases: if... Using the Euclidean distance metric, three points with the closest Laplace-Dirichlet values ​​were selected to construct a new mapping plane, and the centroid of the new plane was calculated. With the original center of mass The Euclidean distance between them is taken as the layer thickness. When When calculating the point set Find the coordinates of the midpoint, and then connect the midpoint with the centroid of the original triangular facet. Connect the lines to form a new perpendicular line, and simultaneously calculate the midpoint and the centroid. The distance is used as the layer thickness. If If it is an empty set, then the original mapping plane is preserved. , , When the next layer is mapped, the conical space search will be re-executed, and the current layer thickness will be marked as 0.

[0104] Special mapping strategy and layer thickness calculation of ventricular outer layer: In the process of mapping the ventricular outer layer, the traditional conic space mapping method fails due to the characteristics of the Laplace-Dirichlet value of 1. For the mapping process of the third layer-ventricular epicardium, the nearest neighbor algorithm is used. By calculating the Euclidean distance between the centroid of each mapping plane in the third layer and each point on the ventricular epicardium, the nearest point is selected as the mapping point, thereby completing the mapping of the ventricular outer layer and ensuring the integrity of the full-layer mapping. At the same time, the distance between the centroid of each mapping plane in the third layer and the nearest point on the ventricular epicardium constitutes the interlayer thickness of the third layer-ventricular epicardium.

[0105] Layer thickness accumulation and data generation: Starting from the endocardium, each triangle patch on the endocardium undergoes four mapping processes: endocardium-first layer, first layer-second layer, and third layer-ventricular epicardium. After each mapping is completed, the interlayer thickness of this mapping can be calculated according to the mapping rules of new mapping plane generation and layer thickness calculation, special mapping strategy and layer thickness calculation of ventricular outer layer, and the layer thickness is accumulated in real time. The layer thickness is accumulated four times in the four mapping processes, and finally a dataset containing the epicardial point index and its corresponding accumulated layer thickness is generated.

[0106] Data post-processing: For the case of multiple layer thickness data in the epicardial point cloud, the arithmetic mean method is used to calculate the average value. By using the low-pass filtering algorithm, abnormal values are removed according to the set threshold condition, and the interference of extreme data is eliminated. Subsequently, the Inverse Distance Weighted Interpolation (IDW) algorithm is used to perform interpolation calculation based on the spatial distribution of the epicardial points and the layer thickness data, and finally the accurate left ventricular wall thickness data is obtained and visualized in Figure 4 (c). By comparing the model perspective of Figure 4 (a) and the layered graph of Figure 4 (b), the accuracy and reliability of the detection results are clearly shown.

[0107] Through the above process, the accurate left ventricular wall thickness distribution of the patient is finally obtained, providing a reliable quantitative basis for clinicians to evaluate the degree of myocardial hypertrophy and develop treatment plans.

[0108] The above-described embodiments only express several embodiments of the present application, and the description is more specific and detailed, but it should not be understood as limiting the scope of the present patent. It should be noted that for ordinary skilled persons in the art, without departing from the concept of the present application, several modifications and improvements can be made, which are within the scope of the present application. Therefore, the protection scope of the present patent should be subject to the appended claims.

Claims

1. A method of calculating a left ventricular model image wall thickness processing method of a heart, characterized by, The method comprises the following steps: S1, myocardial layering standard: dividing the ventricular wall into at least three layers according to a preset numerical range; S2, mapping standard: dividing the myocardial structure into the endocardium, the multi-layer myocardial layer and the epicardium, and using a layered mapping method, wherein the multi-layer mapping process from the endocardium to the outermost myocardial layer is determined by a conic space search and a Laplace-Dirichlet value matching to determine the mapping point, and the mapping process from the outermost myocardial layer to the epicardium uses a nearest neighbor algorithm to determine the mapping point; S3, personalized three-dimensional heart model reconstruction: segmenting the left ventricular boundary based on the cardiac magnetic resonance image, and reconstructing a high-resolution three-dimensional left ventricular cavity model; S4, myocardial layering: solving the myocardial tissue transmural gradient by a Laplace-Dirichlet parameterization method, and numerically assigning the left ventricular finite element grid points to realize left ventricular layering; S5, ventricular wall thickness calculation: obtaining the left ventricular wall thickness data by layering data attribution unique processing, mapping plane construction and search, new mapping plane generation and layer thickness calculation, layer thickness accumulation and data post-processing.

2. The method of claim 1, wherein, In the step S1, the preset numerical range is 0-1, and the ventricular wall is divided into three layers, including the first layer of myocardium (adjacent to the atrial endocardium), i.e. the atrial endocardium side starting myocardial layering, the second layer of myocardium and the third layer of myocardium (adjacent to the atrial epicardium), i.e. the atrial epicardium side ending myocardial layering, and the numerical range of each layer is dynamically adjusted according to the actual proportion requirement.

3. The method of claim 1, wherein, In the step S2, the layered mapping method specifically comprises: Based on the endocardial triangular facet, the centroid and normal vector are calculated, and a conic search space with the normal vector as the axis and the centroid as the vertex is constructed; In the target point set, the candidate point set with an included angle smaller than a set threshold with the normal vector is screened; According to the size of the candidate point set, the point closest to the Laplace-Dirichlet value is selected to construct a new mapping plane or calculate a midpoint to generate a new vertical line; For the mapping from the outermost myocardial layer to the epicardium, the nearest neighbor algorithm with Euclidean distance measurement is used to select the mapping point.

4. The method of claim 1, wherein, In the step S3, the personalized three-dimensional heart model reconstruction comprises: S31, ventricular boundary segmentation: manually or automatically segmenting the left ventricular, endocardial and epicardial boundaries on the CMR-LGE image; S32, model reconstruction: using the Frontal-Delaunay algorithm combined with the adaptive mesh division algorithm to redivide the rough surface mesh to generate a high-resolution three-dimensional model retaining the characteristics and local refinement of the heart; 5. The method of claim 1, wherein, In the step S4, the myocardial layering comprises: S41, model surface extraction: extracting the apical, basal surfaces and endocardial, epicardial surfaces; S42, Laplace-Dirichlet parameterization: obtaining the scalar value distribution of the myocardial tissue transmural direction by solving the Laplace equation and applying the Dirichlet boundary condition; S43, reference value integration: numerically assigning the grid points according to the transmural gradient direction to form a layered model.

6. The method of claim 1, wherein, The step S5 includes: setting data priority rules, endocardium > inner myocardium > outer myocardium > epicardium; traversing all the layered data; removing overlapping points; and ensuring that each spatial point is uniquely attributed to each layer.

7. The method of claim 1, wherein, The step S5 includes: calculating the centroid and unit normal vector of each triangular facet; constructing a conical search space with the centroid as the vertex and the normal vector as the axis; and searching for a point set located in the conical body in the target point set as candidate points.

8. The method of claim 1, wherein, The step S5 includes: if the number of candidate points is greater than or equal to 3, selecting three points with the closest Laplace-Dirichlet values to construct a new plane, calculating the distance between the centroid of the new plane and the original centroid as the layer thickness; if the number of candidate points is 1 or 2, calculating the midpoint and connecting it with the centroid, and taking the distance between the midpoint and the centroid as the layer thickness; if the number of candidate points is 0, keeping the original plane and marking the current layer thickness as 0.

9. The method of claim 1, wherein, The step S5 includes:

10. The method of claim 1, wherein, starting from the endocardium, mapping and accumulating the thickness between layers in real time, and finally generating a data set containing the index of the epicardium points and the corresponding accumulated layer thickness. The step S5 includes: calculating the arithmetic mean of the multiple layer thickness data in the epicardium point cloud, removing abnormal values through low-pass filtering, and using the inverse distance weighted interpolation algorithm to interpolate the entire epicardium region to obtain a smooth and accurate left ventricular wall thickness distribution map.

Images