Three-dimensional myocardial scar visualization system

Abstract

The embodiment of the invention discloses a three-dimensional myocardial scar visualization system. According to one specific embodiment, the system comprises a data processing end configured to perform heart structure and scar segmentation on target heart magnetic resonance data to obtain a binary tag graph for the heart and the myocardium; constructing an individualized ventricular network and a scar three-dimensional surface model based on the binary label graph; generating an initial myocardial scar three-dimensional model; generating lead implantation track information; generating a myocardial scar three-dimensional model according to the lead implantation track information; and the model application end is configured to perform model rendering on a target application page according to the model parameters corresponding to the myocardial scar three-dimensional model sent by the data processing end so as to provide an interactive operation function and generate a clinical report. According to the embodiment, accurate and visual three-dimensional visual decision support can be provided for preoperative planning and intraoperative navigation of heart diseases, so that the accuracy of surgical planning and the clinical working efficiency are remarkably improved.

Description

Technical Field

[0001] The embodiments disclosed herein relate to the field of computer technology, and more specifically to a three-dimensional myocardial scar visualization system. Background Technology

[0002] Traditional methods rely on physician experience and imaging techniques such as ultrasound and X-rays, but they cannot directly quantify the distribution and electrophysiological characteristics of myocardial scars, potentially leading to electrodes being misplaced in the scar area and increasing the risk of surgical failure. Two-dimensional cardiac magnetic resonance imaging (MRI) with delayed enhancement, a representative technique in cardiac MRI, is widely used in myocardial infarction and fibrosis (i.e., qualitative and quantitative evaluation of myocardial scars). Although systems using two-dimensional cardiac MRI with delayed enhancement can clearly identify myocardial scars and perform qualitative and quantitative evaluations, the following problems still exist: The first technical problem is that traditional systems lack the ability to integrate three-dimensional spatial information, which makes it impossible for the system to build an accurate three-dimensional anatomical model of the heart, resulting in a waste of computing and storage resources.

[0003] The second technical problem is that traditional systems fail to achieve fusion modeling of electrophysiological characteristics. They only acquire scar anatomy information and do not fuse electrophysiological characteristics with anatomical images, thus failing to provide automated three-dimensional path planning for lead implantation and wasting the system's computational and storage resources.

[0004] The information disclosed in this background section is only intended to enhance the understanding of the background of the inventive concept, and therefore may contain information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] The summary portion of this disclosure is intended to provide a brief overview of the concepts, which will be described in detail in the detailed description portion. This summary portion is not intended to identify key or essential features of the claimed technical solutions, nor is it intended to limit the scope of the claimed technical solutions.

[0006] Some embodiments of this disclosure propose a three-dimensional myocardial scar visualization system to address one or more of the technical problems mentioned in the background section above.

[0007] In a first aspect, some embodiments of this disclosure provide a three-dimensional myocardial scar visualization system, comprising: a data processing terminal configured to: segment the target cardiac magnetic resonance data into cardiac structure and scar to obtain a binary label map for the heart and myocardium; construct an individualized ventricular network and a three-dimensional surface model of the scar based on the aforementioned binary label map; generate an initial three-dimensional model of the myocardial scar based on the aforementioned individualized ventricular network, the aforementioned three-dimensional surface model of the scar, and acquired cardiac electrophysiological data; generate lead implantation trajectory information based on the aforementioned initial three-dimensional model of the myocardial scar; and generate a three-dimensional model of the myocardial scar based on the aforementioned lead implantation trajectory information; and a model application terminal configured to render the model on a target application page according to the model parameters corresponding to the three-dimensional model of the myocardial scar sent by the data processing terminal, so as to provide interactive operation functions and generate clinical reports.

[0008] The various embodiments of this disclosure have the following beneficial effects: The three-dimensional myocardial scar visualization system of some embodiments of this disclosure provides accurate and intuitive three-dimensional visualization decision support for preoperative planning and intraoperative navigation in cardiac diseases, thereby significantly improving the accuracy of surgical planning and clinical efficiency. Specifically, the reasons for deviations in cardiac interventional surgery planning, high intraoperative risks, and low clinical efficiency are: traditional image assessment relies on manual interpretation, lacks the fusion of three-dimensional quantitative information of scars and electrophysiological data, and surgical path planning is highly subjective, making it difficult to accurately avoid scar areas and dangerous anatomical structures. Based on this, the three-dimensional myocardial scar visualization system of some embodiments of this disclosure, firstly, at the data processing end, is configured to: segment the target cardiac magnetic resonance data into cardiac structures and scars, obtaining binary label maps for the heart and myocardium. Through deep automatic segmentation of cardiac structures and scar areas, the target cardiac magnetic resonance data is converted into machine-readable binary label maps, providing an accurate anatomical basis for subsequent three-dimensional reconstruction, significantly improving segmentation efficiency and consistency. Then, based on the above binary label maps, an individualized ventricular network and a three-dimensional surface model of the scar are constructed. Based on the segmentation results, an individualized ventricular network and scar surface model are constructed, achieving a leap from two-dimensional slices to three-dimensional structures, providing an accurate geometric carrier for spatial positioning and quantitative analysis. Then, based on the aforementioned individualized ventricular network, the aforementioned three-dimensional scar surface model, and the acquired cardiac electrophysiological data, an initial three-dimensional model of myocardial scar is generated. The electrophysiological data is fused with the anatomical model to generate an initial three-dimensional model of myocardial scar containing both structural and functional information, realizing the spatial correlation between scar tissue and electrophysiological characteristics, providing a complete data foundation for precision treatment. Next, based on the aforementioned initial three-dimensional model of myocardial scar, lead implantation trajectory information is generated. The lead implantation trajectory generated based on the initial three-dimensional model of myocardial scar accurately avoids scars and dangerous anatomical areas, reducing the risk of electrode misplacement and improving the safety and accuracy of cardiac interventional surgery planning. Finally, based on the aforementioned lead implantation trajectory information, a three-dimensional model of myocardial scar is generated. The lead implantation trajectory information is integrated into the initial three-dimensional model of myocardial scar to generate an enhanced model that can be directly used for surgical navigation, ensuring spatial consistency between anatomical structures, lesion information, and surgical plans. Finally, the model application is configured to render the model on the target application page based on the model parameters corresponding to the 3D myocardial scar model sent by the data processing terminal, providing interactive operation functions and generating clinical reports. The 3D myocardial scar model generated by the data processing terminal is transformed into a clinically usable tool. Through high-quality real-time rendering and intuitive interactive operation, doctors can explore the 3D heart model from any angle, deeply understand the spatial relationship between lesions and normal structures, and link the 3D view with 2D quantitative charts (e.g., bullseye charts), greatly improving the intuitiveness and depth of diagnostic assessment. Furthermore, by generating clinical reports, the data interpretation process for doctors is simplified, manual processing time is reduced, and clinical efficiency is improved. Attached Figure Description

[0009] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and elements are not necessarily drawn to scale.

[0010] Figure 1 This is a schematic diagram of the structure of some embodiments of the three-dimensional myocardial scar visualization system according to the present disclosure.

[0011] Figures 2 to 3 It can be the visualization interface of the three-dimensional myocardial scar visualization system disclosed herein. Detailed Implementation

[0012] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0013] It should also be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings. Unless otherwise specified, the embodiments and features described in this disclosure can be combined with each other.

[0014] It should be noted that the concepts of "first" and "second" mentioned in this disclosure are used only to distinguish different devices, modules or units, and are not used to limit the order of functions performed by these devices, modules or units or their interdependencies.

[0015] It should be noted that the terms "a" and "a plurality of" used in this disclosure are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0016] The names of messages or information exchanged between multiple devices in the embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of such messages or information.

[0017] This disclosure will now be described in detail with reference to the accompanying drawings and embodiments.

[0018] refer to Figure 1 The diagram shows a structural schematic of some embodiments of a three-dimensional myocardial scar visualization system according to the present disclosure. Figure 1 This includes the data processing end and the model application end.

[0019] In some embodiments, the data processing terminal described above can be configured as follows: First, the target cardiac magnetic resonance imaging data is segmented to obtain a binary label map of the heart and myocardium.

[0020] In some embodiments, the aforementioned data processing terminal can segment the target cardiac magnetic resonance data to obtain a binary label map for the heart and myocardium. The data processing terminal can be a processing terminal (e.g., a server) that preprocesses, segments, and models the target cardiac magnetic resonance data to output a three-dimensional model of myocardial scars. The target cardiac magnetic resonance data can be raw data obtained using a 3.0T high-field magnetic resonance imaging (MRI) device for CMR-cine (Cardiac Magnetic Resonance cine imaging), resting perfusion, T1 mapping (T1 value quantitative imaging), and 2D LGE (2-Dimensional Late Gadolinium Enhancement) multimodal scanning. The cardiac structures and scars can be cardiac anatomical structures (e.g., left ventricle, heart chambers) and areas of myocardial necrosis and fibrosis. The binary label map can be an image that distinguishes cardiac structures and scars using 0 / 1 pixels.

[0021] In some optional implementations of certain embodiments, the data processing terminal described above can perform cardiac structure and scar segmentation on the target cardiac magnetic resonance data to obtain a binary label map for the heart and myocardium: The first step is to extract a two-dimensional gadolinium delayed enhancement image sequence from the target cardiac magnetic resonance imaging (MRI) data to obtain the two-dimensional gadolinium delayed enhancement image sequence. This two-dimensional gadolinium delayed enhancement image sequence can be a two-dimensional image sequence extracted from the target cardiac MRI data.

[0022] The second step involves performing at least one of the following on the aforementioned two-dimensional gadolinium delayed-enhanced image sequence: noise suppression, registration, voxel intensity normalization, and enhancement, to obtain the processed image sequence. The processed image sequence can be standardized image data obtained after preprocessing such as noise suppression, image registration, voxel normalization, and contrast enhancement.

[0023] As an example, firstly, noise suppression is performed on the aforementioned two-dimensional gadolinium delayed enhancement image sequence (e.g., noise suppression using an adaptive filtering algorithm) to obtain a denoised two-dimensional gadolinium delayed enhancement image sequence. Then, registration is performed on the denoised two-dimensional gadolinium delayed enhancement image sequence (e.g., registration using a non-rigid transformation model to correct inter-layer misalignment caused by heartbeat and respiration) to obtain a registered two-dimensional gadolinium delayed enhancement image sequence. Next, voxel intensity normalization is performed on the registered two-dimensional gadolinium delayed enhancement image sequence (e.g., normalizing the grayscale values ​​of the registered two-dimensional gadolinium delayed enhancement image sequence to the same range) to obtain a normalized two-dimensional gadolinium delayed enhancement image sequence. Finally, enhancement processing is performed on the normalized two-dimensional gadolinium delayed enhancement image sequence (e.g., intensity stretching to make the contrast between the ventricular wall and the heart chamber more distinct) to obtain the processed image sequence.

[0024] The third step involves inputting the processed image sequence into a pre-trained segmentation model to obtain preliminary segmentation results. This pre-trained segmentation model can be a pre-trained nnU-Net deep learning model. The input to this model can be the processed image sequence, and the output can be the preliminary segmentation results. The nnU-Net deep learning model can include 2D U-Net cascades, 3D U-Net cascades, and U-Net cascades. The 2D U-Net cascade directly uses two-dimensional slices as input and is suitable for tasks with anisotropic resolution (high intra-layer resolution and large inter-layer spacing). The 3D U-Net cascade can process complete three-dimensional volume data (suitable for isotropic data, better capturing the three-dimensional contextual information between layers). The U-Net cascade is used for segmenting large-volume images. The preliminary segmentation results can be unoptimized heart and scar region segmentation data directly output by the pre-trained segmentation model, which may include minor noise and holes.

[0025] The fourth step involves performing morphological post-processing on the preliminary segmentation results to generate the aforementioned binary label map. In practice, morphological post-processing techniques (e.g., erosion, dilation, and combinations thereof) can be used to eliminate minor noise and fill voids to generate the binary label map.

[0026] Second, based on the aforementioned binary label map, an individualized ventricular network and a three-dimensional surface model of the scar are constructed.

[0027] In some embodiments, the data processing terminal can construct an individualized ventricular network and a three-dimensional scar surface model based on the aforementioned binary label map. The individualized ventricular network can be a three-dimensional geometric model obtained by deforming a standard mesh based on the aforementioned binary label map, used to describe the anatomical morphology of the left ventricle. The aforementioned three-dimensional scar surface model can be a three-dimensional surface characterizing the spatial morphology of the scar.

[0028] In some optional implementations of certain embodiments, the data processing terminal can construct a personalized ventricular network based on the aforementioned binary label map through the following steps: The first step is to extract left ventricular information from the aforementioned binary label map. This left ventricular information can be features extracted from the binary label map that define the spatial structure of the heart. This information may include: the apex, the base plane, and the long axis direction (the line connecting the apex (lowest point) and the center of the base (highest point) is the long axis).

[0029] In practice, firstly, based on the binary label image, pixel regions representing the left ventricular myocardium and heart chambers are identified. Then, image processing algorithms (e.g., contour extraction, skeletonization) are used to determine the geometric center, boundaries, and main orientation of each pixel region. Finally, relevant information of the left ventricle is extracted (e.g., the apex (usually the point with the smallest Z-coordinate) and the center of the base of the heart, and the major axis direction vector connecting the two is determined).

[0030] The second step involves constructing a three-dimensional cardiac local coordinate system based on the aforementioned left ventricular information. This three-dimensional cardiac local coordinate system can be established with the apex of the left ventricle as the origin and the major axis as the Z-axis, used to describe the inherent anatomical orientation of the heart. In practice, firstly, the extracted apex is used as the origin of the coordinate system. Then, the major axis vector pointing from the apex to the base of the heart is defined as the Z-axis of the coordinate system. Finally, in the minor axis plane perpendicular to the Z-axis, two mutually orthogonal vectors are determined as the X-axis and Y-axis based on the inherent orientation of the image (e.g., anterior-posterior, left-right), thus obtaining the three-dimensional cardiac local coordinate system.

[0031] The third step involves aligning the aforementioned three-dimensional local cardiac coordinate system with the standard four-chamber template coordinate system to obtain the aligned standard myocardial mesh. The standard four-chamber template coordinate system can be a cardiac reference coordinate system based on population statistics and with standard anatomical orientation. The aligned standard myocardial mesh can be a standard average shape mesh registered to the individual coordinate space through a 4×4 affine transformation. In practice, firstly, key corresponding feature points (e.g., the apex of the heart, the center points of each chamber) between the three-dimensional local cardiac coordinate system and the standard four-chamber template coordinate system are identified. Then, a 4×4 affine transformation matrix is ​​used to transform the three-dimensional local cardiac coordinate system to the standard template space of the standard four-chamber template coordinate system to generate the aligned standard myocardial mesh.

[0032] The fourth step involves determining the geometric differences between the aligned standard myocardial grid and the individual myocardial morphology, thereby obtaining geometric difference feature information. The individual myocardial morphology can be the unique three-dimensional geometric structure of the heart muscle, including ventricular cavity size, ventricular wall thickness, and apical morphology. The geometric differences can be the spatial distance differences between the aligned standard myocardial grid and the individual myocardial morphology. The geometric difference feature information can be data features generated after quantifying the geometric differences.

[0033] Fifth, the aforementioned geometric difference feature information is input into a pre-trained deep deformation network to obtain predicted deformation parameters. The input to the pre-trained deep deformation network can be geometric difference feature information, and the output can be predicted deformation parameters. This pre-trained deep deformation network can be a regression neural network. The regression neural network can include feature extraction layers, feature fusion layers, and regression layers. The predicted deformation parameters can be a set of displacement vectors output by the deep deformation network to guide the movement of each vertex in the aligned standard myocardial mesh, and can include a global affine transformation parameter set and a local non-rigid deformation parameter set.

[0034] Step 6: Based on the predicted deformation parameters, perform deformation operations on the aligned standard myocardial mesh to obtain an individualized ventricular network. In practice, based on the predicted deformation parameters, apply corresponding displacements to each vertex of the aligned standard myocardial mesh to perform non-rigid deformation operations to generate an individualized ventricular network.

[0035] In some optional implementations of certain embodiments, the data processing terminal can construct a three-dimensional surface model of the scar based on the aforementioned binary label map through the following steps: The first step involves resampling the aforementioned binary label image to obtain binary scar voxel data. This binary scar voxel data can be a three-dimensional binary array of scar tissue with improved resolution after resampling. In practice, trilinear interpolation resampling can be performed on the scar voxel data in the binary label image to generate the binary scar voxel data. This binary scar voxel data can also be spatial distribution data of myocardial scars stored as a three-dimensional pixel array, with each voxel labeled as either scar or non-scar tissue.

[0036] The second step is to extract an initial three-dimensional scar mesh from the aforementioned binary scar voxel data. This initial three-dimensional scar mesh can be a rough-surfaced three-dimensional mesh containing geometric artifacts, obtained directly from the binary voxel data using an isosurface extraction algorithm (e.g., Flying Edges).

[0037] The third step involves filling the holes in the initial 3D scar mesh to obtain a filled 3D scar mesh. This filled 3D scar mesh can be a mesh processed using a hole-filling algorithm (e.g., VTK's vtkFillHolesFilter) to ensure the model's closure and topological correctness.

[0038] The fourth step involves simplifying the 3D mesh of the filled scar to obtain a simplified 3D scar mesh. This simplified 3D scar mesh can be a lightweight mesh optimized using a simplification algorithm. For example, the simplification algorithm could be a mesh simplification algorithm based on a quadratic error metric.

[0039] The fifth step involves smoothing and normal correction operations on the simplified 3D scar mesh to obtain a 3D scar surface model. In practice, firstly, Laplacian smoothing or Taubin smoothing is applied to the simplified 3D scar mesh to suppress high-frequency noise on the surface, resulting in an optimized smooth mesh. Then, based on the optimized smooth mesh, the normal vector of each vertex is determined to ensure that the normal direction is perpendicular to the smoothed surface to obtain the 3D scar surface model. Figure 2 It can represent a three-dimensional surface model of a scar, with the scar area displayed as a red gradient. Among them, Figure 2 The "Transmural Extent" parameter indicates the extent of scar tissue penetration, with a scale from 0.00 (no damage) to 1.00 (full-thickness damage). "Normal Heart View" displays a standard anatomical image of the heart, providing a reference for anatomical location. The major axis vector pointing from the apex to the base of the heart is defined as the Z-axis of the coordinate system. In the minor axis plane perpendicular to the Z-axis, two mutually orthogonal vectors are determined as the X-axis and Y-axis based on the inherent orientation of the image (e.g., anterior-posterior, left-right).

[0040] Optionally, the above data processing terminal can be configured to perform the following steps: The first step is to determine the minimum distance from each vertex of the aforementioned 3D scar surface model to the endocardial surface, thus obtaining the scar transmural depth data. This transmural depth data can be a set of values ​​describing the distance from each vertex of the 3D scar model to the nearest endocardial surface, quantifying the transmural extent of the scar. In practice, firstly, the endocardial surface is extracted using an individualized ventricular network model. Then, for each vertex of the 3D scar surface model, the minimum Euclidean distance from each vertex to the endocardial surface is determined to obtain the scar transmural depth dataset.

[0041] The second step is to determine the color mapping data based on the scar thickness corresponding to the transmural scar depth data. The scar thickness can be the thickness information of the scar tissue within the heart wall obtained from the transmural scar depth data (e.g., 6 mm). The color mapping data can be a rule or result that maps numerical scar thickness data to visual colors. For example, a thickness of 0-3 mm might be mapped to blue, 3-6 mm to green, 6-9 mm to yellow, and greater than 9 mm to red. In practice, firstly, the transmural scar depth data (i.e., the thickness value) is normalized to a preset range (e.g., 0 to 1). Then, the normalized values ​​are applied to a predefined color mapping table (e.g., a gradient from blue to red). Finally, a corresponding RGB color value is generated for each vertex, forming the color mapping data.

[0042] The third step involves registering the individualized ventricular network with the pre-acquired standard AHA template based on the aforementioned color mapping data, thus obtaining the partition label mapping relationship. This partition label mapping relationship can be the correspondence between each point on the individualized ventricular network and a standard AHA partition. In practice, firstly, based on the color mapping data, scar thickness is visually displayed on the individualized ventricular network using color. Then, the individualized ventricular network is spatially aligned with the standard AHA template. Finally, a correspondence is established between each vertex on the surface of the individual heart model and a standard AHA partition, serving as the partition label mapping relationship.

[0043] The fourth step involves generating quantitative indicators for scar tissue based on the aforementioned partition label mapping relationship. These quantitative indicators, derived from the partition label mapping relationship, include the total scar volume, average thickness, and percentage of the scar tissue. In practice, firstly, based on the partition label mapping relationship, the vertices of the 3D scar surface model are grouped according to their respective AHA partitions. Then, for each partition, relevant data for all scar vertices within it are determined. For example, the total scar volume, average transmural depth (thickness), and percentage of the scar area relative to the myocardial area of ​​that partition are determined. Finally, the statistical results for each partition are summarized to generate the quantitative indicators for scar tissue in each partition.

[0044] The fifth step involves generating a bullseye map based on the aforementioned quantified scar indicators for each region. In practice, firstly, based on the quantified scar indicators for each region and the layout of the AHA bullseye map template, the 17 regions of the left ventricle are mapped to different sector regions in polar coordinates. Finally, according to the quantified indicator value of each region, the corresponding sector region is filled with the appropriate color intensity or color, thereby generating a bullseye map that visually displays the distribution of scars throughout the left ventricle. Figure 3It can represent the bullseye chart of AHA 17 partitions, with detailed statistical tables.

[0045] Optionally, the above model application can be configured to perform the following steps: The first step involves spatially associating and rendering the bullseye map with the aforementioned 3D model of myocardial scar to generate a visualization scene that provides interactive operation functions and generates clinical reports. The model application can be a terminal (e.g., a server) that receives the 3D model of myocardial scar and provides a human-computer interaction interface based on it. The visualization scene can be a unified visual environment integrating the 3D model, color mapping, and quantitative charts. For example, the visualization scene could be an interactive interface simultaneously displaying the 3D model of the heart, a color image of scar thickness, and a bullseye map. The interactive operation functions can be functions that allow dynamic interaction with the visualization scene via mouse, touch, etc. For example, the interactive operation functions could allow users to rotate and zoom the 3D model of myocardial scar, click on any area to view its quantitative data, or adjust the model's transparency. In practice, firstly, based on the 3D model of myocardial scar and the bullseye map data from the data processing terminal, two view windows are created in the graphics rendering engine. Then, using a spatial mapping algorithm, the association logic between the AHA partitions on the 3D model of myocardial scar and the corresponding sector areas on the bullseye map is established. Finally, the 3D model of myocardial scar and the bullseye map are rendered synchronously in a unified interface to generate a visualized scene.

[0046] Third, based on the aforementioned individualized ventricular network, the aforementioned three-dimensional surface model of the scar, and the acquired cardiac electrophysiological data, an initial three-dimensional model of myocardial scar is generated.

[0047] In some embodiments, the data processing terminal can generate an initial three-dimensional model of myocardial scar based on the individualized ventricular network, the three-dimensional surface model of the scar, and the acquired cardiac electrophysiological data. The cardiac electrophysiological data may be myocardial surface voltage data. The initial three-dimensional model of myocardial scar may be a primary three-dimensional cardiac model integrating anatomical structure, scar distribution, and electrophysiological characteristics.

[0048] In some optional implementations of certain embodiments, the data processing terminal can generate an initial three-dimensional model of myocardial scar based on the individualized ventricular network, the three-dimensional surface model of the scar, and the acquired cardiac electrophysiological data. The first step involves unifying the spatial coordinates of the individualized ventricular network and the 3D scar surface model to obtain aligned individualized ventricular networks and aligned 3D scar surface models. The aligned individualized ventricular network can be an individualized ventricular mesh model registered with the 3D scar surface model in a unified world coordinate system. The aligned 3D scar surface model can be a scar surface model registered with the individualized ventricular network in a unified world coordinate system. In practice, spatial transformation parameters stored in the DICOM image can be applied to the individualized ventricular network and the 3D scar surface model to transform their data from the image coordinate system to a unified world coordinate system, thus obtaining the aligned individualized ventricular network and aligned 3D scar surface model.

[0049] The second step involves superimposing the aligned individualized ventricular network and the aligned 3D scar surface model to obtain a superimposed model. This superimposed model can be a composite model formed by combining the aligned ventricular network and the scar surface model in three-dimensional space. In practice, a spatial superposition algorithm can be used to superimpose the aligned individualized ventricular network and the aligned 3D scar surface model to obtain the superimposed model.

[0050] The third step involves converting the aforementioned overlay model into a voxel representation with uniform resolution, yielding myocardial voxel data and scar voxel data. The myocardial voxel data can be represented as a discrete three-dimensional pixel (voxel) array, converted from the aligned individualized ventricular network. Similarly, the scar voxel data can be represented as a discrete three-dimensional pixel (voxel) array, converted from the aligned three-dimensional scar surface model. In practice, firstly, a uniform resolution three-dimensional voxel mesh (e.g., 0.5mm × 0.5mm × 0.5mm) can cover the entire overlay model. Then, each voxel is identified: if its center is within the myocardial mesh, it is labeled as a myocardial voxel; if it is within the scar model, it is labeled as a scar voxel. Finally, the myocardial voxel data and scar voxel data are generated separately.

[0051] The fourth step involves assigning a corresponding anatomical region label to each myocardial voxel in the aforementioned myocardial voxel data to generate voxel data with anatomical labels. These anatomical region labels can be identifiers assigned to each region of the myocardium by the American Heart Association (AHA) standards (e.g., 1 represents the anterior wall of the basal segment). The voxel data with anatomical labels can be voxel data where each myocardial voxel is assigned an AHA partition number. In practice, firstly, the standard AHA17 partition template is mapped to the space containing the myocardial voxel data. Then, for each myocardial voxel, the partition number of the partition in the AHA template whose center point falls is determined to generate voxel data with anatomical labels.

[0052] The fifth step involves identifying the overlapping areas between the voxel data with anatomical labels and the scar voxel data to obtain the anatomical distribution mapping information of the scar. This anatomical distribution mapping information can be a mapping of the spatial correspondence between the scar voxel data and the AHA standard partitions. In practice, both the scar voxel data and the myocardial voxel data with anatomical labels can be traversed. Then, for each scar voxel, the myocardial voxel that overlaps with its spatial location is found, and its AHA partition label is obtained. Finally, the AHA partitions to which each scar voxel belongs are statistically analyzed to form the anatomical distribution mapping information of the scar.

[0053] Step 6: Based on the aforementioned overlay model, the voxel data with anatomical labels, and the anatomical distribution mapping information, an initial integrated 3D model is generated. This initial integrated 3D model can be a 3D data model that integrates anatomical structures and scar distribution. In practice, the initial integrated 3D model can be obtained by integrating the overlay model, the voxel data with anatomical labels, and the anatomical distribution mapping information.

[0054] Step 7: Map the aforementioned cardiac electrophysiological data to the initial integrated 3D model to generate an initial 3D model of myocardial scar. In practice, firstly, acquire the cardiac electrophysiological data and its coordinates in 3D space. Then, align the spatial coordinates of the electrophysiological data with the surface of the initial integrated 3D model using point cloud registration or surface matching algorithms. Finally, assign the electrophysiological parameters as new attributes to the corresponding vertices of the model surface, thereby generating the final initial 3D model of myocardial scar.

[0055] Fourth, based on the above-mentioned initial three-dimensional model of myocardial scar, lead implantation trajectory information is generated.

[0056] In some embodiments, the data processing unit can generate lead implantation trajectory information based on the initial three-dimensional model of myocardial scar. This lead implantation trajectory information can be one or more data points describing the three-dimensional spatial path and angle of the lead from the implantation point to the endpoint. The lead implantation trajectory information may include: starting point coordinates, endpoint coordinates, path point sequence, and implantation angle.

[0057] In some optional implementations of certain embodiments, the data processing terminal can generate lead implantation trajectory information based on the initial three-dimensional model of myocardial scar: The first step involves determining the safe target region information for leads in the initial 3D model of myocardial scar based on preset safety threshold rules. These preset safety threshold rules can be multi-parameter threshold rules defining the safe region. These multi-parameter thresholds may include: tissue characteristic thresholds (e.g., regions with pixel signal intensity > normal myocardial mean + 5 standard deviations are considered dense scar cores and must be excluded; while regions with signal intensity ≤ (normal myocardial mean + 3 standard deviations) are considered healthy myocardium); anatomical location and morphological thresholds (e.g., the myocardial thickness of the target region must be greater than the safety threshold (e.g., greater than 3-5 mm)); functional thresholds (e.g., ventricular wall thickening rate must be >10%, meaning the myocardium in the target region must have contractile reserve); and region continuity or size thresholds (e.g., the identified target region must be a continuous pixel block, and the target region area must be greater than a minimum threshold (e.g., greater than 0.3-0.5 cm²)). The safe target region information can be a set of 3D myocardial region data that conforms to the safety threshold rules. For example, the safe target region information could be a set of implantable region coordinates for the mid-segment of the left ventricular lateral wall that conforms to the preset safety threshold rules.

[0058] As an example, the initial three-dimensional model of myocardial scar is screened based on a preset safety threshold rule to obtain information on the safe target area.

[0059] The second step involves extracting the geometric center coordinates and regional contour information of the aforementioned initial three-dimensional model of myocardial scar, based on the safety target region information, to obtain core positioning parameter information. The geometric center coordinates can be the coordinates of the center point determined based on the three-dimensional geometry of the safety target region. The regional contour information can be geometric data describing the boundary and shape of the safety target region in three-dimensional space. The core positioning parameter information can be geometric feature data extracted from the safety target region for lead path planning, and may include the geometric center coordinates and regional contour information. In practice, firstly, the geometric center of the safety target region is determined, obtaining the center point coordinates of the target region. Then, the vertices constituting the boundary of the target region and their connections are extracted to obtain the three-dimensional contour information of the target region, thereby generating the core positioning parameter information.

[0060] The third step involves using the geometric center coordinates of the core positioning parameters and the individualized ventricular network to determine the three-dimensional coordinate set of the lead trajectory starting point. This set can be the set of three-dimensional coordinates of the anatomical entry point or starting point of the device entering the heart. In practice, firstly, based on different clinical surgical approaches (e.g., via the coronary sinus or via the ventricular septum), a set of device entry points is predefined on the individualized ventricular network. Then, using the geometric center coordinates (i.e., the target point) from the core positioning parameters, combined with the physical characteristics of the device (e.g., its bendability radius), points with no or minimal obstruction on the line connecting to the target point are selected from the set of device entry points. Finally, one or more three-dimensional coordinate sets of the lead trajectory starting point are generated as the lead trajectory starting point three-dimensional coordinate set.

[0061] The fourth step involves generating initial linear trajectory information for the leads based on the aforementioned three-dimensional coordinate set of the lead trajectory starting point, the aforementioned geometric center coordinates, the aforementioned individualized ventricular network, and the aforementioned three-dimensional scar surface model. This initial linear trajectory information can be a straight path connecting the starting point and the target center. In practice, firstly, a starting point is selected from the three-dimensional coordinate set of the lead trajectory starting point and connected to the geometric center coordinates of the safe target area, generating a straight line vector in three-dimensional space. Then, this straight line is discretized into a series of dense path points. Finally, the straight path connecting the starting point and the target point is output as the initial linear trajectory information for the leads.

[0062] The fifth step involves generating obstacle avoidance adjustment trajectory information using the tissue type data from the initial linear trajectory information of the leads. This tissue type data can be classification information describing the characteristics of the myocardial tissue traversed by the initial linear trajectory (e.g., healthy myocardium, scar tissue, blood vessels) (e.g., a tissue sequence where the initial linear trajectory passes through a 5mm healthy myocardium and a 2mm scar edge region). The obstacle avoidance adjustment trajectory information can be a path obtained by optimizing the initial linear trajectory information to avoid damage to important structures (e.g., blood vessels, scar tissue). In practice, firstly, each path point traversed by the initial linear trajectory information is determined, and the tissue type (e.g., healthy myocardium, scar tissue, blood vessels) in the individualized ventricular network and scar 3D surface model is obtained. Then, according to preset obstacle avoidance rules (e.g., avoiding all scar areas and major blood vessels), the path segments crossing prohibited areas are optimized and adjusted, and detour paths are generated using curve fitting or path search algorithms. Finally, the obstacle avoidance adjustment trajectory information is output.

[0063] Step 6: Based on the obstacle avoidance adjustment trajectory information mentioned above, lead implantation trajectory information is generated. This lead implantation trajectory information includes optimal lead implantation trajectory information. The optimal lead implantation trajectory information can be the best implantation path determined by comprehensively considering path length, safety (obstacle avoidance), and stability. In practice, firstly, a comprehensive evaluation algorithm (e.g., balancing path length, distance from hazardous structures, and path smoothness) is used to determine the optimal lead implantation trajectory information from the obstacle avoidance adjustment trajectory information. Then, the obstacle avoidance adjustment trajectory information and the optimal lead implantation trajectory information are integrated as the lead implantation trajectory information.

[0064] Fifth, based on the above lead implantation trajectory information, a three-dimensional model of myocardial scar is generated.

[0065] In some embodiments, the data processing terminal can generate a three-dimensional model of myocardial scar based on the lead implantation trajectory information. This three-dimensional myocardial scar model can be a three-dimensional digital cardiac model integrating anatomical structure, scar distribution, electrophysiological characteristics, and lead implantation trajectory information.

[0066] As an example, a three-dimensional model of myocardial scar, including the lead implantation trajectory, can be generated by overlaying lead implantation trajectory information (including optimal lead implantation trajectory information) onto the corresponding spatial location of the initial three-dimensional model of myocardial scar with a dedicated visual identifier (e.g., a red highlighted line).

[0067] In some optional implementations of certain embodiments, the data processing terminal can generate a three-dimensional model of myocardial scar based on the aforementioned lead implantation trajectory information: In some embodiments, the model application is configured to render the model on a target application page based on the model parameters corresponding to the 3D myocardial scar model sent by the data processing terminal, thereby providing interactive operation functions and generating clinical reports. The model parameters can be a set of data attributes describing the 3D myocardial scar model, including geometric attributes, texture attributes, and functional annotations. For example, model parameters may include vertex coordinates of the 3D myocardial scar model, color mapping values ​​of the scar region, electrophysiological data, and spatial coordinate sequences of lead trajectories. The target application page can be a user interface for hosting and displaying the 3D visualization scene and interactive controls. The clinical report can be a document integrating key images, quantitative data tables, and surgical planning schemes. For example, the clinical report may include 3D model screenshots, bullseye diagrams, scar burden statistics tables, and lead implantation schemes corresponding to lead implantation trajectory information.

[0068] In addressing the technical challenges mentioned above, the application scenario of precise planning and navigation for interventional cardiac surgery often presents the following challenges: wasted system resources. Traditional 3D visualization systems are disconnected from clinical procedures, resulting in low efficiency in interpreting model parameters and failing to provide intuitive, interactive analysis and quantitative decision support for 3D cardiac pathology, thus wasting significant time and computational resources. Given the specific requirements of this application scenario—namely, the need for a system that intuitively presents the 3D correlation between scar tissue and trajectory, supports interactive analysis by physicians, and rapidly generates standardized clinical documentation—we have decided to adopt the following solution: In some optional implementations of certain embodiments, the model application can render the model on the target application page based on the model parameters corresponding to the 3D model of myocardial scar sent by the data processing terminal, in order to provide interactive operation functions and generate clinical reports. The first step is to generate a structured data dictionary corresponding to the aforementioned 3D model of myocardial scar. This structured data dictionary can be a standardized dataset of parameters from the 3D model of the myocardial scar. For example, it could be a set of key-value pairs including vertex coordinates, scar thickness, voltage values, and AHA partition labels. In practice, the raw data of the 3D model of myocardial scar (including vertex coordinates, patch indices, color attributes, and normal vectors) sent from the data processing terminal can be used. Then, the raw data is categorized and organized into key-value pairs according to a predefined schema to obtain the structured data dictionary.

[0069] The second step involves constructing a 3D interactive rendering engine based on the aforementioned structured data dictionary to obtain an interactive 3D view. This 3D interactive rendering engine can be a program developed using WebGL / OpenGL, used to convert the structured data dictionary into a real-time renderable 3D scene. The aforementioned interactive 3D view can be an initial 3D scene view that supports user interaction (rotation, zooming, etc. via mouse / touch). In practice, The third step is to generate the interactive control instruction set corresponding to the interactive 3D view. This instruction set can be a collection of instructions that defines the mapping relationship between user actions and view responses. In practice, first, the types of user interaction operations are defined (e.g., rotation, zoom, pan, profile viewing). Then, specific control parameters and response functions are set for each operation type. Finally, the operation types and parameters are organized into a structured interactive control instruction set.

[0070] The fourth step is to associate the aforementioned interactive control command set with the aforementioned interactive 3D view to obtain the associated interactive 3D view. This associated interactive 3D view can be a visual component with complete interactive functionality, formed by binding the interactive control command set with the interactive 3D view. In practice, the associated interactive 3D view can be generated by binding the interactive control command set with the event listener of the 3D interactive rendering engine.

[0071] The fifth step involves generating an updated 3D visualization scene based on the associated interactive 3D view, in response to the target user executing the interactive control commands in the aforementioned set of interactive control commands. The target user can be a medical professional (e.g., an interventional cardiologist) performing cardiac surgery planning and analysis. The updated 3D visualization scene can be a 3D visualization scene that updates in real-time in response to user interaction. In practice, when the target user executes a command (e.g., rotate the model to a specific angle), the 3D interactive rendering engine determines the new position and viewpoint of the camera in real-time based on the command parameters. Then, the 3D interactive rendering engine re-renders the 3D scene based on the new viewpoint parameters to generate the updated 3D visualization scene.

[0072] Step 6: Based on the updated 3D visualization scene described above, generate a structured report characterizing the features of myocardial scars. This structured report can be a standardized template that integrates medical documents and may include patient information, screenshots of the 3D model of the myocardial scar, an AHA-partitioned scar load table, and a bullseye diagram. First, quantify the scar features based on key views (e.g., screenshots) of the updated 3D visualization scene and relevant data from the current perspective. Then, automatically populate the key views and quantified data into the corresponding positions in the preset report template. Finally, generate a structured report (e.g., a PDF document) containing visual evidence and quantified indicators.

[0073] Step 7: Based on the lead implantation trajectory information corresponding to the target point selected by the target user, a surgical planning report is generated. This surgical planning report can be a detailed document of surgical steps and parameter recommendations generated based on the target point and trajectory finally confirmed by the user. In practice, firstly, the three-dimensional coordinates of the target point selected by the target user and its corresponding lead implantation trajectory information are used. Then, combined with a medical knowledge base, detailed surgical parameter recommendations, instrument selection, and obstacle avoidance reminders are generated based on the lead implantation trajectory information, and integrated into a surgical planning report.

[0074] Step 8: Integrate the structured report and surgical planning report into a clinical report. In practice, the structured report and surgical planning report can be combined to generate a clinical report.

[0075] The above-described steps, as an inventive point of this disclosure, solve the technical problem mentioned in the background: waste of system resources. Traditional 3D visualization systems are disconnected from clinical diagnosis and treatment processes, resulting in low efficiency in interpreting model parameters and failing to meet the needs of intuitive, interactive analysis and quantitative decision support for 3D cardiac pathology, thus wasting significant time and computational resources. The reasons for these technical problems are as follows: traditional systems do not achieve deep integration between the structured organization of model parameters and the rendering engine, the interactive logic is disconnected from the model, and there is a lack of a fully automated mechanism from visualization analysis to report generation. This invention, through the construction of a structured data dictionary, the development of a 3D interactive rendering engine, and the full-link integration of interaction and report generation, achieves the integration of interactive model visualization and automated clinical report generation, saving time and computational resources in interpreting models and compiling reports.

[0076] In addressing the aforementioned technical challenges in employing technical solutions, the specific application scenario—precise preoperative localization and surgical planning for radiofrequency ablation of ventricular arrhythmias—often presents the following technical issues: traditional systems require electrophysiological mapping and modeling of the entire cardiac region, neglecting the crucial boundary between scar tissue and normal myocardium. Redundant computation and ineffective modeling consume significant computing and storage resources, leading to wasted system resources. To meet the specific requirements of this application scenario—focusing on key areas for precise modeling, reducing redundant computation to conserve system resources, and ensuring accurate ablation target localization—we have decided to adopt the following solution: Optionally, the above data processing terminal can be configured to perform the following steps: The first step is to determine the boundary region between scar tissue and normal myocardium in the aforementioned three-dimensional scar surface model. This boundary region can be considered a transition zone between scar tissue and normal myocardium in the three-dimensional myocardial scar model. In practice, firstly, based on the three-dimensional scar surface model, the relative positions of the individualized ventricular network and the three-dimensional scar surface model are determined. Then, the distance from each vertex of the scar model surface to the nearest normal myocardial surface is determined. Finally, a transition zone with electrophysiological heterogeneity between the scar tissue core and normal myocardium is extracted as the boundary region.

[0077] The second step involves generating a conduction heterogeneity distribution map of the boundary region based on the aforementioned electrophysiological data and boundary region. This conduction heterogeneity distribution map can be a spatial distribution map quantifying the dispersion of electrical conduction within the boundary region. In practice, cardiac electrophysiological data (e.g., voltage maps) can be spatially registered with the boundary region. Then, within the boundary region, parameters such as local conduction velocity and conduction direction are determined point-by-point. Finally, these parameters are quantified and mapped to colors to generate a distribution map that visually displays the spatial differences in conduction performance within the boundary region, serving as the conduction heterogeneity distribution map.

[0078] The third step involves using the aforementioned conduction heterogeneity distribution map and preset slow conduction conditions to determine the potential slow conduction channel set. The preset slow conduction conditions can be thresholds for electrophysiological parameters defining slow conduction channels. For example, a conduction velocity <0.3 m / s could be a threshold. The potential slow conduction channel set can be a set of continuous three-dimensional paths that meet the slow conduction conditions. In practice, the preset slow conduction conditions (e.g., conduction velocity <0.3 m / s) are used as thresholds to binarize the conduction heterogeneity distribution map, filtering out pixels / vertices that meet the conditions. Then, connected component analysis or path search algorithms from image processing are used to identify continuous paths with a certain length and direction from the discrete slow conduction points. Finally, the potential slow conduction channel set is output.

[0079] The fourth step involves constructing a set of virtual re-entry loop models based on the aforementioned set of potential slow conduction channels. This set of virtual re-entry loop models can be a collection of simulated re-entry paths built upon slow conduction channels. In practice, computational electrophysiology principles can be used to simulate the possible conduction paths of the electrostimulating wave after it enters from one end of each potential slow conduction channel at the other end of the scar (e.g., bypassing the scar core and returning to the channel entrance). Finally, multiple complete circular excitation paths are constructed, forming the set of virtual re-entry loop models.

[0080] The fifth step is to determine the set of three-dimensional locations of the critical islets corresponding to the aforementioned set of virtual loop models. This set of three-dimensional locations of the critical islets can refer to the set of three-dimensional spatial coordinates of the narrowest, most vulnerable, and crucial conduction channel portion of the virtual loop for maintaining the loop's integrity. For example, in a figure-eight loop, the narrow region at the intersection of the two loops is the critical islet. In practice, firstly, for each virtual loop model, its geometry and conduction characteristics are determined. Then, the narrowest, slowest-conducting, and most crucial segment for maintaining the loop's integrity is identified within the virtual loop model to obtain the set of three-dimensional locations of the critical islets.

[0081] Step 6: Based on the aforementioned set of three-dimensional locations of the critical isthmus, generate the corresponding set of ablation target points. This set of ablation target points can be a set of three-dimensional coordinates for radiofrequency ablation, generated based on the set of three-dimensional locations of the critical isthmus. In practice, firstly, the three-dimensional coordinates of each critical isthmus location are used as the center. Then, based on the physical characteristics of the radiofrequency ablation catheter (e.g., the ablation lesion size is typically 3-5 mm), an ablation region with a certain spatial range (e.g., spherical) is generated. Finally, the geometric center coordinates and spatial range of each ablation region are determined, constituting the set of ablation target points.

[0082] Step 7: Integrate the aforementioned ablation target set and the aforementioned virtual reentry circuit model set into the aforementioned three-dimensional myocardial scar model to obtain the three-dimensional myocardial scar model. The aforementioned three-dimensional myocardial scar model can be a comprehensive three-dimensional cardiac model integrating anatomical structure, scar distribution, electrophysiological characteristics, and ablation target information. In practice, the virtual reentry circuit model set and the ablation target set can be used as a new data layer, and integrated into the three-dimensional myocardial scar model through spatial registration and data fusion algorithms to obtain the three-dimensional myocardial scar model.

[0083] The above-described operational steps, as an inventive point of this disclosure, solve the technical problem mentioned in the background art: "Traditional systems require electrophysiological mapping and modeling of the entire heart region, failing to focus on the critical area of ​​the boundary between scar tissue and normal myocardium. Redundant calculations and ineffective modeling consume a large amount of computing power and storage resources, leading to a waste of system resources." The reasons for the above technical problem are as follows: Traditional systems lack the ability to identify key areas causing arrhythmias and adopt a coarse-grained calculation mode of "uniform analysis of the entire heart." This indiscriminate global modeling not only performs a large amount of ineffective calculations on healthy myocardial regions but also causes the system to process and maintain far more data than necessary, resulting in a triple waste of computing resources, storage space, and processing time. The present invention establishes a progressive and precise calculation process from boundary identification, heterogeneity analysis, channel extraction to loop construction, which greatly reduces the system load and saves system computing time, memory usage, and storage space while ensuring or even improving the accuracy of ablation target localization.

[0084] In addressing the technical challenges mentioned above, the application scenario—a data storage management system for a 3D myocardial scar visualization system—involves the storage and dynamic scheduling of multiple types of model data. This often presents the following technical issues: traditional systems using fixed storage allocation cannot adapt to dynamic capacity changes and access frequency differences in multiple model data, leading to wasted storage resources or system lag. Considering the following requirements for this application scenario—dynamically matching storage needs across multiple models, optimizing storage performance based on access frequency, and avoiding resource redundancy and insufficiency—we have decided to adopt the following solution: Optionally, the above data processing terminal can be configured to perform the following steps: The first step involves obtaining the storage capacities corresponding to the aforementioned binary label map, individualized ventricular network, and 3D scar surface model in parallel to determine the estimated total storage requirement. This estimated total storage requirement can be the predicted sum of the storage capacities corresponding to the binary label map, individualized ventricular network, and 3D scar surface model. For example, the estimated total storage requirement is 336MB, including the binary label map (50MB), ventricular network (150MB), and scar model (80MB), plus a 20% buffer (56MB). In practice, multiple computation threads are started in parallel to simultaneously obtain the estimated disk file size or memory usage of the binary label map, individualized ventricular network, and 3D scar surface model. Then, the estimated disk file size or memory usage is summarized, and an additional buffer overhead (e.g., 20%) is added to obtain the estimated total storage requirement.

[0085] The second step involves generating a partitioned storage scheme based on the estimated total storage requirements and the data access frequencies corresponding to the binary label map, the individualized ventricular network, and the 3D scar surface model. The data access frequency can be the frequency with which the binary label map, the individualized ventricular network, and the 3D scar surface model are invoked in the workflow. The partitioned storage scheme can be information on allocating data to storage media with different performance levels based on access frequency. In practice, the data access frequencies corresponding to the binary label map, the individualized ventricular network, and the 3D scar surface model are determined (e.g., the individualized ventricular network is frequently accessed during deformation calculation, thus considered high-frequency; the binary label map is rarely accessed after segmentation, thus considered low-frequency). Then, a partitioned storage scheme is developed based on the corresponding data access frequencies. For example, high-speed storage space is reserved for high-frequency access data, while large-capacity but slower storage areas are allocated to low-frequency data.

[0086] The third step involves performing initial storage allocation on the individualized ventricular network, the 3D surface model of the scar, and the 3D model of the myocardial scar, according to the aforementioned partitioned storage scheme, to obtain initial storage allocation information. This initial storage allocation information can be a description of the location and size of the initial storage space allocated to the individualized ventricular network, the 3D surface model of the scar, and the 3D model of the myocardial scar, based on the partitioned storage scheme. In practice, guided by the partitioned storage scheme, free space blocks that meet the size and performance requirements are determined on the target storage device (e.g., memory or hard disk). Then, designated storage spaces are allocated to the individualized ventricular network, the 3D surface model of the scar, and the 3D model of the myocardial scar, respectively, and the starting address and size of each space are recorded as initial storage allocation information.

[0087] The fourth step is to generate a real-time occupancy report based on the initial storage allocation information described above. This real-time occupancy report can be a report on actual storage space usage. In practice, firstly, at fixed time intervals, the current actual occupancy size corresponding to the individualized ventricular network, the three-dimensional surface model of the scar, and the three-dimensional model of the myocardial scar is collected to generate a real-time occupancy report. The fifth step is to determine the capacity deviation information between the real-time occupancy report and the partitioned storage scheme information. This capacity deviation information can be a quantitative description of the difference between the real-time occupancy report and the partitioned storage scheme information. For example, the capacity deviation information could be: the real-time report shows that the 3D model of myocardial scar actually occupies 220MB, which is +20MB compared to the initially allocated 200MB.

[0088] Step 6: Based on the aforementioned capacity deviation information, generate adjusted storage plan information. This adjusted storage plan information can be updated resource allocation strategy information generated based on the capacity deviation information. In practice, firstly, based on the capacity deviation information, identify data objects with insufficient or excessive resources. Then, according to a preset adjustment strategy (e.g., expanding space for data with increasing demand, compressing or releasing idle data), formulate a new allocation plan as the adjusted storage plan information.

[0089] Step 7: Based on the adjusted storage scheme information described above, reallocate storage resources for the individualized ventricular network, the 3D surface model of the scar, and the 3D model of the myocardial scar. In practice, execute the instructions in the adjusted storage scheme information (e.g., allocate new, larger contiguous space for data objects that need expansion). Then, migrate the existing data to the new space and release the old space. Simultaneously, perform corresponding operations on data objects that need to be released or compressed. Finally, complete the reallocation of storage resources.

[0090] The above-described steps, as an inventive point of this disclosure, solve the aforementioned technical problem: "Traditional systems use fixed storage allocation, which cannot adapt to dynamic capacity changes and access frequency differences in multi-model data, leading to wasted storage resources or system lag." The reasons for this technical problem are as follows: Traditional systems lack the ability to dynamically perceive changes in multi-model data capacity and access patterns, and cannot establish a closed-loop management mechanism for storage demand prediction and real-time monitoring. This invention, by establishing a complete storage management chain from parallel capacity estimation and intelligent partitioning to dynamic reallocation, achieves adaptive optimization management of system storage resources, saving system crash recovery time caused by improper resource allocation, avoiding idle and wasted storage resources, and improving system resource utilization efficiency.

[0091] The various embodiments of this disclosure have the following beneficial effects: The three-dimensional myocardial scar visualization system of some embodiments of this disclosure provides accurate and intuitive three-dimensional visualization decision support for preoperative planning and intraoperative navigation in cardiac diseases, thereby significantly improving the accuracy of surgical planning and clinical efficiency. Specifically, the reasons for deviations in cardiac interventional surgery planning, high intraoperative risks, and low clinical efficiency are: traditional image assessment relies on manual interpretation, lacks the fusion of three-dimensional quantitative information of scars and electrophysiological data, and surgical path planning is highly subjective, making it difficult to accurately avoid scar areas and dangerous anatomical structures. Based on this, the three-dimensional myocardial scar visualization system of some embodiments of this disclosure, firstly, at the data processing end, is configured to: segment the target cardiac magnetic resonance data into cardiac structures and scars, obtaining binary label maps for the heart and myocardium. Through deep automatic segmentation of cardiac structures and scar areas, the target cardiac magnetic resonance data is converted into machine-readable binary label maps, providing an accurate anatomical basis for subsequent three-dimensional reconstruction, significantly improving segmentation efficiency and consistency. Then, based on the above binary label maps, an individualized ventricular network and a three-dimensional surface model of the scar are constructed. Based on the segmentation results, an individualized ventricular network and scar surface model are constructed, achieving a leap from two-dimensional slices to three-dimensional structures, providing an accurate geometric carrier for spatial positioning and quantitative analysis. Then, based on the aforementioned individualized ventricular network, the aforementioned three-dimensional scar surface model, and the acquired cardiac electrophysiological data, an initial three-dimensional model of myocardial scar is generated. The electrophysiological data is fused with the anatomical model to generate an initial three-dimensional model of myocardial scar containing both structural and functional information, realizing the spatial correlation between scar tissue and electrophysiological characteristics, providing a complete data foundation for precision treatment. Next, based on the aforementioned initial three-dimensional model of myocardial scar, lead implantation trajectory information is generated. The lead implantation trajectory generated based on the initial three-dimensional model of myocardial scar accurately avoids scars and dangerous anatomical areas, reducing the risk of electrode misplacement and improving the safety and accuracy of cardiac interventional surgery planning. Finally, based on the aforementioned lead implantation trajectory information, a three-dimensional model of myocardial scar is generated. The lead implantation trajectory information is integrated into the initial three-dimensional model of myocardial scar to generate an enhanced model that can be directly used for surgical navigation, ensuring spatial consistency between anatomical structures, lesion information, and surgical plans. Finally, the model application is configured to render the model on the target application page based on the model parameters corresponding to the 3D myocardial scar model sent by the data processing terminal, providing interactive operation functions and generating clinical reports. The 3D myocardial scar model generated by the data processing terminal is transformed into a clinically usable tool. Through high-quality real-time rendering and intuitive interactive operation, doctors can explore the 3D heart model from any angle, deeply understand the spatial relationship between lesions and normal structures, and link the 3D view with 2D quantitative charts (e.g., bullseye charts), greatly improving the intuitiveness and depth of diagnostic assessment. Furthermore, by generating clinical reports, the data interpretation process for doctors is simplified, manual processing time is reduced, and clinical efficiency is improved.

[0092] In some implementations, clients and servers can communicate using any currently known or future-developed network protocol such as HTTP (Hypertext Transfer Protocol) and can interconnect with digital data communication (e.g., communication networks) of any form or medium. Examples of communication networks include local area networks (“LANs”), wide area networks (“WANs”), the Internet (e.g., the Internet of Things), and peer-to-peer networks (e.g., ad hoc peer-to-peer networks), as well as any currently known or future-developed networks.

[0093] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0094] The functions described above in this document can be performed at least in part by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), system-on-a-chip (SoCs), complex programmable logic devices (CPLDs), and so on.

[0095] The above description is merely a selection of preferred embodiments of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in the embodiments of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described inventive concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions disclosed in the embodiments of this disclosure.

Claims

1. A three-dimensional myocardial scar visualization system, comprising: The data processing end is configured as follows: The target cardiac magnetic resonance data is segmented to obtain a binary label map of the heart and myocardium; Based on the aforementioned binary label map, an individualized ventricular network and a three-dimensional surface model of the scar are constructed. An initial three-dimensional model of myocardial scar is generated based on the individualized ventricular network, the three-dimensional surface model of the scar, and the acquired cardiac electrophysiological data. Based on the initial three-dimensional model of myocardial scar, lead implantation trajectory information is generated; Based on the lead implantation trajectory information, a three-dimensional model of myocardial scar is generated; The model application is configured to render the model on the target application page based on the model parameters corresponding to the 3D model of myocardial scar sent by the data processing terminal, so as to provide interactive operation functions and generate clinical reports.

2. The system according to claim 1, wherein, The data processing terminal is configured to: Two-dimensional gadolinium delayed enhancement image sequence is extracted from the target cardiac magnetic resonance data to obtain the two-dimensional gadolinium delayed enhancement image sequence; At least one of the following processes is performed on the two-dimensional gadolinium delayed enhancement image sequence: noise suppression, registration, voxel intensity normalization, and enhancement, to obtain the processed image sequence. The processed image sequence is input into a pre-trained segmentation model to obtain preliminary segmentation results; The preliminary segmentation results are subjected to morphological post-processing to generate a binary label map.

3. The system according to claim 1, wherein, The data processing terminal is configured to: Extract left ventricular related information from the binary label map; Based on the relevant information of the left ventricle, a three-dimensional local cardiac coordinate system is constructed; Align the three-dimensional local cardiac coordinate system with the standard four-chamber cardiac template coordinate system to obtain the aligned standard myocardial mesh; The geometric differences between the aligned standard myocardial grid and the individual myocardial morphology are determined to obtain geometric difference feature information; The geometric difference feature information is input into a pre-trained deep deformation network to obtain the predicted deformation parameters; Based on the predicted deformation parameters, a deformation operation is performed on the aligned standard myocardial mesh to obtain an individualized ventricular network.

4. The system according to claim 1, wherein, The data processing terminal is configured to: The binary label image is resampled to obtain binary scar tissue data. An initial three-dimensional scar mesh is extracted from the binary scar voxel data. The initial three-dimensional scar mesh is filled with holes to obtain a filled three-dimensional scar mesh. The three-dimensional mesh of the filled scar is simplified to obtain a simplified three-dimensional mesh of the scar. The simplified 3D scar mesh is smoothed and normalized to obtain a 3D surface model of the scar.

5. The system according to claim 1, wherein, The data processing terminal is configured to: The minimum distance from each vertex of the three-dimensional surface model of the scar to the endocardial surface is determined to obtain the transmural depth data of the scar. Based on the scar thickness corresponding to the scar transmural depth data, determine the color mapping data; Based on the color mapping data, the individualized ventricular network is registered with the pre-acquired standard AHA template to obtain the partition label mapping relationship; Based on the partition label mapping relationship, generate partition scar quantitative index information; Based on the quantification index information of the zonal scars, a bullseye diagram is generated.

6. The system according to claim 5, wherein, The model application is configured as follows: The bullseye image and the three-dimensional model of myocardial scar are spatially correlated and rendered to generate a visual scene, providing interactive operation functions and generating clinical reports.

7. The system according to claim 1, wherein, The data processing terminal is configured to: Spatial coordinate unification is performed on the individualized ventricular network and the three-dimensional surface model of the scar to obtain the aligned individualized ventricular network and the aligned three-dimensional surface model of the scar. The aligned individualized ventricular network and the aligned three-dimensional scar surface model are superimposed to obtain a superimposed model. The superimposed model is converted into a voxel representation with uniform resolution to obtain myocardial voxel data and scar voxel data; Assign a corresponding anatomical region label to each myocardial voxel in the myocardial voxel data to generate voxel data with anatomical labels; By identifying the overlapping areas between the voxel data containing anatomical tags and the scar voxel data, the anatomical distribution mapping information of the scar is obtained. Based on the superposition model, the voxel data with anatomical labels, and the anatomical distribution mapping information, an initial integrated 3D model is generated; The cardiac electrophysiological data are mapped to the initial integrated 3D model to generate an initial 3D model of myocardial scar.

8. The system according to claim 1, wherein, The data processing terminal is configured to: Based on preset safety threshold rules, the safety target region information for leads in the initial three-dimensional model of myocardial scar is determined; Based on the initial three-dimensional model of myocardial scar, the geometric center coordinates and region contour information of the safety target area are extracted to obtain core positioning parameter information; Using the geometric center coordinates of the core positioning parameter information and the individualized ventricular network, the three-dimensional coordinate set of the lead trajectory starting point is determined; Based on the three-dimensional coordinate set of the lead trajectory starting point, the coordinates of the geometric center, the individualized ventricular network, and the three-dimensional surface model of the scar, the initial straight line trajectory information of the lead is generated; Using the tissue type data of the initial straight-line trajectory information of the leads, obstacle avoidance adjustment trajectory information is generated; Based on the obstacle avoidance adjustment trajectory information, lead implantation trajectory information is generated, wherein the lead implantation trajectory information includes: optimal lead implantation trajectory information.

Images