Medical image reconstruction method and electronic equipment

Abstract

The embodiment of the invention discloses a medical image reconstruction method and electronic equipment. The method comprises the steps of copying medical image data stored in a first memory of a central processing unit into a second memory of a graphics processor in response to a medical image reconstruction instruction; for a kernel function used for realizing linear interpolation of the unified computing equipment architecture, executing the kernel function on the graphics processor so as to utilize a plurality of threads started by the graphics processor to carry out linear interpolation on the medical image data stored in the second memory in parallel to obtain interpolated data; returning the interpolated data stored in the second memory to the first memory; and generating a first three-dimensional grid on the basis of the interpolated data stored in the first memory, and reconstructing a medical image model corresponding to the medical image data according to the first three-dimensional grid. According to the technical scheme of the embodiment of the invention, the efficiency of medical image reconstruction is improved.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of image processing technology, and in particular to a medical image reconstruction method and electronic equipment. Background Art

[0002] Medical image reconstruction is the core support for modern medical diagnosis, treatment and scientific research.

[0003] However, the efficiency of medical image reconstruction is currently low and needs to be addressed urgently. Summary of the Invention

[0004] The embodiments of the present invention provide a medical image reconstruction method and electronic equipment, which improve the efficiency of medical image reconstruction.

[0005] According to one aspect of the present invention, a medical image reconstruction method is provided, which may include: in response to a medical image reconstruction instruction, copying medical image data stored in a first memory of a central processing unit to a second memory of a graphics processing unit; executing a kernel function for implementing linear interpolation of a unified computing device architecture on the graphics processing unit, so as to utilize multiple threads launched by the graphics processing unit to perform linear interpolation on the medical image data stored in the second memory in parallel to obtain interpolated data; transmitting the interpolated data stored in the second memory back to the first memory; generating a first three-dimensional grid based on the interpolated data stored in the first memory, and reconstructing a medical image model corresponding to the medical image data based on the first three-dimensional grid.

[0006] According to another aspect of the present invention, an electronic device is provided, which may include: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that when executed by the at least one processor, the medical image reconstruction method provided by any embodiment of the present invention is implemented.

[0007] The technical solution of the embodiment of the present invention is to copy the medical image data stored in the first memory of the central processing unit to the second memory of the graphics processing unit in response to the medical image reconstruction instruction, so that the graphics processing unit can perform linear interpolation on the medical image data; execute the kernel function for implementing linear interpolation of the unified computing device architecture on the graphics processing unit, so as to use multiple threads launched by the graphics processing unit to perform linear interpolation on the medical image data stored in the second memory in parallel to obtain interpolated data, and the multiple threads launched by the graphics processing unit can be used to efficiently and parallelly perform linear interpolation on the medical image data; return the interpolated data stored in the second memory to the first memory; generate a first three-dimensional grid based on the interpolated data stored in the first memory, and reconstruct a medical image model corresponding to the medical image data based on the first three-dimensional grid to achieve reconstruction of the medical image model. The above technical solution copies the medical image data to the second memory of the graphics processing unit, and then uses multiple threads launched by the graphics processing unit to efficiently and parallelly perform linear interpolation on the medical image data, which can improve the efficiency of interpolation of the medical image data, and the entire medical image reconstruction process does not need to rely on manual labor and can be performed automatically, thereby improving the efficiency of medical image reconstruction.

[0008] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present invention, nor is it intended to limit the scope of the present invention. Other features of the present invention will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0010] Figure 1 is a flowchart of a medical image reconstruction method provided according to an embodiment of the present invention;

[0011] Figure 2 is a flowchart of another medical image reconstruction method provided by an embodiment of the present invention;

[0012] Figure 3 is a flowchart of another medical image reconstruction method provided according to an embodiment of the present invention;

[0013] Figure 4 is a schematic diagram of a two-dimensional view of the aorta in experimental results in an optional example of another medical image reconstruction method provided by an embodiment of the present invention;

[0014] Figure 5 is a schematic diagram of a three-dimensional view of the aorta in experimental results in an optional example of another medical image reconstruction method provided by an embodiment of the present invention;

[0015] Figure 6 is a schematic diagram of a two-dimensional view of cerebral blood vessels in experimental results in an optional example of another medical image reconstruction method provided by an embodiment of the present invention;

[0016] Figure 7 is a schematic diagram of a three-dimensional view of cerebral blood vessels in an experimental result of an optional example in another medical image reconstruction method provided by an embodiment of the present invention;

[0017] Figure 8 is a structural block diagram of a medical image reconstruction device provided according to an embodiment of the present invention;

[0018] Figure 9 It is a structural diagram of an electronic device for implementing the medical image reconstruction method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0019] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0020] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. The situations of "target", "original", etc. are similar and will not be repeated here. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or that are inherent to these processes, methods, products or devices.

[0021] Before introducing the embodiments of the present invention, an exemplary description is given of the implementation process of the current solution for medical image reconstruction and the reasons why the problem of low efficiency of medical image reconstruction occurs, so as to better understand why the solution proposed in the embodiments of the present invention improves the efficiency of medical image reconstruction.

[0022] Currently, the reconstruction of medical imaging data into medical imaging models mainly relies on manual processing tools with graphical interfaces that require manual operation, such as 3DSlicer, Insight Toolkit-Segmenting and Navigation Platform (ITK-SNAP), and some supporting Digital Imaging and Communications in Medicine (DICOM) toolkits. However, the above-mentioned manual processing tools that require manual operation rely on manual operation of the Graphical User Interface (GUI). Most processes such as image loading, threshold adjustment, target area annotation and model export need to be completed manually by users. The degree of automation is low and it is difficult to achieve an automated pipeline. Therefore, the efficiency of medical image reconstruction is low, especially for batch data processing. In addition, the above-mentioned manual processing tools use the Central Processing Unit (CPU) for calculations such as interpolation and grid generation. However, when performing operations such as interpolation in medical image reconstruction, there is a lack of parallel computing optimization and the Graphics Processing Unit (GPU) is not fully utilized for acceleration. Therefore, the efficiency of operations such as interpolation will be limited by the CPU performance, resulting in a long time-consuming medical image reconstruction, which in turn leads to a low efficiency of medical image reconstruction, especially when processing large-volume and high-resolution medical image data (such as high-resolution magnetic resonance imaging (MRI)). This problem is particularly pronounced because the time consumed during manual imaging (MRI) or computed tomography (CT) scanning grows exponentially, making it difficult to meet real-time and batch processing requirements. In addition, operations such as interpolation in the above-mentioned manual processing tools need to be completed across multiple modules or even multiple software programs, making automation and process optimization difficult.

[0023] To address this issue, an embodiment of the present invention copies medical image data to the secondary memory of a graphics processor. Multiple threads initiated by the graphics processor then efficiently and concurrently perform linear interpolation on the medical image data, improving the efficiency of interpolation and, consequently, medical image reconstruction. This will be discussed in detail below.

[0024] Figure 1This is a flowchart of a medical image reconstruction method provided in an embodiment of the present invention. This embodiment is applicable to medical image reconstruction. The method can be performed by a medical image reconstruction device provided in an embodiment of the present invention. This device can be implemented using software and / or hardware and can be integrated into an electronic device, such as various user terminals or servers.

[0025] See also Figure 1 The method of the embodiment of the present invention specifically includes the following steps:

[0026] S110 , in response to a medical image reconstruction instruction, copy the medical image data stored in the first memory of the central processing unit to the second memory of the graphics processing unit.

[0027] The medical image reconstruction instruction may be understood as an instruction for instructing medical image reconstruction.

[0028] The first memory can be understood as the memory of the central processing unit.

[0029] Medical imaging data can be understood as imaging data in the medical field; for example, the medical imaging data can be in the Neuroimaging Informatics Technology Initiative (NIfTI) format (.nii / .nii.gz), and the medical imaging data can be in multiple versions of the NIfTI format to enable multiple versions of the NIfTI format input.

[0030] The second memory can be understood as the memory of the graphics processor.

[0031] In an embodiment of the present invention, the medical image data stored in the first memory may be copied to the second memory in response to a medical image reconstruction instruction.

[0032] In an embodiment of the present invention, before the medical image data stored in the first memory is copied to the second memory, the core class (vtkNIFTIImageReader) of the Visualization Toolkit (VTK) library for reading medical image data in a formatted format can be used to read the medical image data, and then the medical image data can be converted into a Numpy array to obtain array data for subsequent processing; according to a pre-set label value (Label ID), binarize the array data (for example, explicit binarization) to obtain mask data containing only the target tissue (the voxel value that meets the condition corresponding to the label value is set to 1, and the other voxel values ​​are set to 0, that is, generating 0-1 binary mask data). The target tissue can be understood as the tissue for which a medical imaging model needs to be established. The target tissue can be, for example, at least one of a blood vessel, an organ, a bone or a tissue, etc.; according to the mask data, the array data is processed (for example, only the data corresponding to the mask data is retained) to obtain a first processing result; the first processing result is converted into float32 type conversion data to ensure consistency of data type in subsequent interpolation and other operations, and the medical imaging data is updated according to the conversion data so that the medical imaging data only includes data corresponding to the target tissue, laying a consistency foundation for subsequent spatial processing and model reconstruction operations.

[0033] In an embodiment of the present invention, before the medical imaging data stored in the first memory is copied to the second memory, the spatial parameters of the medical imaging data (such as dimension, spacing and / or origin, etc.) can be automatically parsed, and the medical imaging data can be updated based on the obtained parsing results to make the medical imaging data adapt to subsequent linear interpolation and other operations.

[0034] S120. Execute a kernel function for implementing linear interpolation for a unified computing device architecture on a graphics processor, so as to utilize multiple threads started by the graphics processor to perform linear interpolation on the medical imaging data stored in the second memory in parallel to obtain interpolated data.

[0035] Among them, the Compute Unified Device Architecture (CUDA) can be understood as an architecture designed for GPU accelerated computing.

[0036] The kernel function can be understood as a kernel function of a unified computing device architecture for implementing linear interpolation.

[0037] The thread can be understood as a thread started by the graphics processor for implementing linear interpolation.

[0038] Linear interpolation can be understood as an interpolation method in which the interpolation function is a linear polynomial, and linear difference can also be understood as a method of upsampling medical imaging data; linear interpolation can be bilinear interpolation or trilinear interpolation, etc. Trilinear interpolation can be used to continuously estimate the input voxel value in space, avoiding step boundaries so that the interpolated data still maintains the original anatomical structure continuity. Therefore, the embodiment of the present invention can use trilinear interpolation.

[0039] The interpolated data can be understood as data obtained by performing linear interpolation on the medical imaging data.

[0040] In an embodiment of the present invention, a kernel function of a unified computing device architecture can be executed on a graphics processor to perform linear interpolation on the medical imaging data stored in the second memory in parallel using multiple threads started by the graphics processor to obtain interpolated data. The above process can improve the spatial resolution of the medical imaging data, make the boundaries of the subsequently reconstructed medical imaging model smoother, and make the three-dimensional reconstruction effect more delicate, while helping to maintain the consistency of the three-dimensional reconstruction volume.

[0041] In an embodiment of the present invention, linear interpolation is performed in parallel on medical image data stored in the second memory using multiple threads started by a graphics processor. This can be achieved by utilizing the parallel computing framework of PyTorch to decompose the difference calculation of the medical image data into GPU thread blocks. This process can reduce the complexity from O(N) (CPU) to O(N / P) (GPU), where P is the number of CUDA cores.

[0042] In the embodiment of the present invention, linear interpolation is performed on the medical image data. For example, the interpolation multiple s can be determined to obtain the medical image data. , through the formula , perform trilinear interpolation under interpolation multiples, where is the voxel output after interpolation (the interpolated data is obtained based on the number of output voxels and medical images), , , .

[0043] In an embodiment of the present invention, after obtaining the interpolated data, median filtering can be performed on the interpolated data. For example, an odd-numbered kernel (3 / 5 / 7) that can eliminate noise and retain edge features can be used to perform median filtering on the interpolated data to remove noise points or isolated small blocks in the interpolated data, smooth the target area corresponding to the target tissue, and retain boundary features to protect edge information, thereby improving the quality of medical image reconstruction. The size of the filter kernel involved in the median filtering process can be determined according to the capacity or can be pre-set, and the size can be, for example, 3x3x3 or 5x5x5. The process of the median filtering process can be nonlinear processing to be suitable for the interpolated data. The process of performing median filtering on the interpolated data may be to execute a kernel function for implementing median filtering on the graphics processor, so as to utilize multiple threads started by the graphics processor to perform median filtering on the interpolated data in parallel, and update the interpolated data according to the obtained second processing result; the process of performing median filtering on the interpolated data may also be to perform median filtering on the interpolated data stored in the second memory after returning it to the first memory, and then update the interpolated data according to the obtained third processing result.

[0044] S130: Transmit the interpolated data stored in the second memory back to the first memory.

[0045] In an embodiment of the present invention, the interpolated data stored in the second memory may be transmitted back to the first memory so that the central processing unit can continue to perform processing related to medical image reconstruction.

[0046] S140: Generate a first three-dimensional grid based on the interpolated data stored in the first memory, and reconstruct a medical image model corresponding to the medical image data based on the first three-dimensional grid.

[0047] The first three-dimensional grid can be understood as a three-dimensional grid generated based on the interpolated data.

[0048] The medical image model can be understood as a model corresponding to the medical image data reconstructed based on the first three-dimensional grid; the medical image model can be, for example, a Standard Tessellation Language (STL) three-dimensional model.

[0049] In an embodiment of the present invention, a first three-dimensional grid can be generated based on the interpolated data stored in the first memory. For example, the interpolated data can be expanded (opening operation) and / or eroded (closing operation), and based on the obtained operation results, a first three-dimensional grid is generated to strengthen the first three-dimensional grid structure, close small holes, and eliminate isolated points. It can be understood that the expansion operation can fill small holes in the interpolated data and connect small structures; the erosion operation can remove redundant isolated points or shrink boundaries in the interpolated data; the combination of expansion and erosion operations helps to improve the integrity and coherence of the medical image model obtained by subsequent reconstruction; the expansion and / or erosion operations can be anisotropic operations; the expansion can perform multiple adjustable iterations, and the erosion can also perform multiple adjustable iterations. The number of iterations corresponding to the expansion and erosion operations can be pre-set or determined according to the parameters of the central processing unit. The number of iterations can be 0-3 times. The iterative process of the expansion operation D can be, for example, , the iterative process of the corrosion operation E can be, for example, ; In the case of both dilation and erosion operations, the dilation operation can be performed first and then the erosion operation; the dilation and erosion operations can be performed based on the corresponding structural elements respectively, for example, the formula can be used , defining the three-dimensional shape based on the element radius r of the structural element , dilation and / or erosion operations are performed according to the three-dimensional shape to control the expansion and contraction amplitude of the surface. The structural element can iteratively control the amplitude of the morphological operation. The structural element can be, for example, a 3x3x3 structural element. The element radius can determine the neighborhood during the dilation or erosion operation. The element radius can be customized. For example, when the element radius is 1, it can be 6 neighborhoods. For example, when the element radius is 2, it can be 26 neighborhoods.

[0050] In an embodiment of the present invention, before generating the first three-dimensional grid based on the interpolated data stored in the first memory, the core filter class (vtkWindowedSinc) filter for three-dimensional data smoothing (denoising) in VTK can be combined to perform 50-1000 low-pass filtering iterations on the interpolated data to smooth and eliminate surface high-frequency noise. For example, the transfer function can be combined to The vtkWindowedSinc filter performs 50-1000 low-pass filtering iterations on the interpolated data, where =0.01 is the cutoff frequency, fstop=0.1. It should be noted that the relationship between the number of iterations N of the low-pass filter and the smoothing effect is: the roughness reduction rate (k is a material-related constant). The number of iterations N can be controlled by the smoothing iterations parameter to balance smoothing speed and quality.

[0051] In an embodiment of the present invention, before generating a first three-dimensional grid based on the interpolated data stored in the first memory, the interpolated data, which may be in Numpy format and has undergone multi-stage processing, can be converted into an image data object that can be directly used by the VTK framework, and the interpolated data is updated according to the conversion result to facilitate subsequent grid generation operations. For example, a container (vtkImageData) object for three-dimensional image data can be constructed, and the image size and spatial resolution (Spacing) of the image data object can be set. The spatial resolution can be determined based on the interpolation multiple. The origin coordinates (Origin) of such an image data object are consistent with the medical imaging data. The interpolated data in Numpy format is converted into a voxel (Scalar) array of VTK and set to the Scalars attribute of the vtkImageData object to obtain a conversion result based on the Scalar array and the vtkImageData object, and the interpolated data is updated according to the conversion result.

[0052] In the embodiment of the present invention, a medical image model can be reconstructed based on the first three-dimensional grid.

[0053] In an embodiment of the present invention, before reconstructing a medical image model corresponding to the medical image data based on the first three-dimensional mesh, mesh smoothing may be performed on the first three-dimensional mesh to remove burrs and irregularities on the surface of the first three-dimensional mesh, thereby improving the visualization and subsequent processing quality of the reconstructed medical image model. For example, the number of smoothing iterations and the filter bandwidth may be set (e.g., 0.01), non-manifold smoothing may be enabled (to ensure smoothing effects in complex topological situations), mesh smoothing may be performed on the first three-dimensional mesh under the conditions of the number of smoothing iterations, filter bandwidth, and non-manifold smoothing, and the first three-dimensional mesh may be updated based on the obtained third processing result. It should be noted that the number of iterations for mesh smoothing may be multiple, and the number of iterations may be pre-set or determined based on capacity, etc.

[0054] In an embodiment of the present invention, after the medical image model is reconstructed based on the first three-dimensional mesh, the medical image model can be exported as an STL standard format file for use by other three-dimensional design software or three-dimensional printing equipment. For example, the output file path can be set, and according to the output file path, the medical image model can be written into an .stl file and exported through the core class (vtkSTLWriter) module in VTK for writing three-dimensional mesh data into an STL format file.

[0055] It can be understood that the medical image reconstruction process of the technical solution of the embodiment of the present invention can perform linear interpolation of medical image data in parallel and efficiently, which can significantly reduce the calculation time, thereby improving the efficiency of interpolation of medical image data, and then improving the efficiency of medical image reconstruction, which is conducive to the real-time processing of large-scale data; and there is no need to collaborate across multiple modules or even multiple software to complete the various operations involved in medical image reconstruction, so it does not need to rely on manual labor and can be performed automatically, avoiding the tedious manual operation of manual processing tools, thereby improving the efficiency of medical image reconstruction.

[0056] The technical solution of the embodiment of the present invention is to copy the medical image data stored in the first memory of the central processing unit to the second memory of the graphics processing unit in response to the medical image reconstruction instruction, so that the graphics processing unit can perform linear interpolation on the medical image data; execute the kernel function for implementing linear interpolation of the unified computing device architecture on the graphics processing unit, so as to use multiple threads launched by the graphics processing unit to perform linear interpolation on the medical image data stored in the second memory in parallel to obtain interpolated data, and the multiple threads launched by the graphics processing unit can be used to efficiently and parallelly perform linear interpolation on the medical image data; return the interpolated data stored in the second memory to the first memory; generate a first three-dimensional grid based on the interpolated data stored in the first memory, and reconstruct a medical image model corresponding to the medical image data based on the first three-dimensional grid to achieve reconstruction of the medical image model. The above technical solution copies the medical image data to the second memory of the graphics processing unit, and then uses multiple threads launched by the graphics processing unit to efficiently and parallelly perform linear interpolation on the medical image data, which can improve the efficiency of interpolation of the medical image data, and the entire medical image reconstruction process does not need to rely on manual labor and can be performed automatically, thereby improving the efficiency of medical image reconstruction.

[0057] An optional technical solution reconstructs a medical image model corresponding to medical image data based on a first three-dimensional mesh, including: obtaining the curvature of each vertex in the first three-dimensional mesh, and obtaining the edge folding cost based on the curvature, so as to sort the edges in the first three-dimensional mesh according to the edge folding cost corresponding to each vertex to obtain a sorting result; based on each edge in the sorting result, simplifying the first three-dimensional mesh in turn, and when the simplification ratio of the first three-dimensional mesh reaches a preset ratio, reconstructing the medical image model corresponding to the medical image data based on the currently obtained simplified three-dimensional mesh.

[0058] The vertices may be understood as the vertices of the meshes in the first three-dimensional mesh.

[0059] The curvature may be understood as the curvature of the first three-dimensional mesh at the vertex.

[0060] It's important to note that edge collapse can merge two vertices on either side of an edge into a single vertex, while simultaneously deleting the associated edges and faces. Based on this, the edge collapse cost can be understood as the impact on the geometric accuracy or visual quality of the original 3D mesh when collapsing the edge corresponding to the vertex. Alternatively, it can be understood as the error in the new vertex after merging the vertex with other vertices.

[0061] In the embodiment of the present invention, the curvature of each vertex in the first three-dimensional grid can be obtained, and the edge folding cost can be obtained according to the curvature. For example, the edge folding cost function , the edge folding cost is obtained based on the curvature ,in, is the curvature, and are weight coefficients respectively.

[0062] The sorting result can be understood as the result obtained by sorting the edges in the first three-dimensional grid.

[0063] In an embodiment of the present invention, the edges in the first three-dimensional grid can be sorted according to the edge collapse cost corresponding to each vertex to obtain a sorting result. For example, the edges in the first three-dimensional grid can be sorted according to the edge collapse cost of the vertices corresponding to each edge in the first three-dimensional grid to obtain a sorting result.

[0064] An edge may be understood as an edge in the first three-dimensional mesh.

[0065] The simplification ratio can be understood as the simplification ratio of the first three-dimensional mesh.

[0066] The preset ratio can be understood as the preset ratio of the first three-dimensional mesh simplification requirement; the preset ratio can be, for example, 0.1~0.5; the preset ratio can be set directly (for example, through the mesh simplification ratio (target_reduction) parameter), or it can be determined by a preset number of retained patches (meshes), for example, by the formula , according to the number of retained patches , determine the preset ratio R, where is the total number of faces.

[0067] In an embodiment of the present invention, the first three-dimensional mesh can be simplified in sequence based on each edge in the sorting result. For example, the first three-dimensional mesh can be simplified in sequence based on each edge in the sorting result using an efficient three-dimensional mesh simplification filter (vtkDecimatePro) in VTK.

[0068] In an embodiment of the present invention, before simplifying the first three-dimensional mesh in sequence, PreserveTopologyOn() can be enabled to prevent the mesh topology of the first three-dimensional mesh from being destroyed during the simplification process, thereby preventing the shape of the reconstructed medical image model from being destroyed.

[0069] In the embodiment of the present invention, when the simplification ratio reaches a preset ratio, a medical image model can be reconstructed based on the currently obtained simplified three-dimensional mesh.

[0070] In an embodiment of the present invention, for each vertex in the first three-dimensional mesh, the curvature of the vertex is obtained, and the edge folding cost is obtained based on the curvature, so that the edges in the first three-dimensional mesh are sorted according to the edge folding cost corresponding to each vertex to obtain a sorting result; based on each edge in the sorting result, the first three-dimensional mesh is simplified in turn, and when the simplification ratio reaches a preset ratio, the medical image model is reconstructed based on the currently obtained simplified three-dimensional mesh. Compared with the related solutions that lack refined post-processing for complex medical image data, especially the simplification processing with a fixed reduction in the number of facets and lack of curvature optimization support, the above technical solution simplifies the first three-dimensional mesh, which can reduce the number of mesh facets, reduce the size of the first three-dimensional mesh and the processing difficulty, and simplify the first three-dimensional mesh according to the preset ratio, which can achieve refined simplified post-processing of complex medical image data, which is beneficial to balancing the accuracy and computational load of medical image reconstruction, and optimizing the storage and rendering efficiency of medical image models.

[0071] Another optional technical solution is to copy the medical imaging data stored in the first memory of the central processing unit to the second memory of the graphics processing unit, including: copying multiple medical imaging data stored in the first memory of the central processing unit to the second memory of the graphics processing unit respectively to realize parallel reconstruction of multiple medical imaging data.

[0072] In the embodiment of the present invention, multiple medical image data stored in the first memory can be copied to the second memory respectively to achieve parallel reconstruction of multiple medical image data. The above technical solution can achieve automatic batch reconstruction of medical image data.

[0073] Another optional technical solution is to copy the medical imaging data stored in the first memory of the central processing unit to the second memory of the graphics processing unit, including: checking the availability of the graphics processing unit; and if the graphics processing unit is available, copying the medical imaging data stored in the first memory of the central processing unit to the second memory of the graphics processing unit.

[0074] In an embodiment of the present invention, availability may be checked. For example, an anomaly detection mechanism may be set to check availability. The anomaly detection mechanism may also verify the dimension of the medical imaging data to prevent memory overflow.

[0075] In the embodiment of the present invention, the medical image data stored in the first memory may be copied to the second memory when the graphics processor is available.

[0076] In the embodiment of the present invention, the availability of CUDA may be explicitly checked, and if CUDA is available, the medical image data stored in the first memory may be copied to the second memory.

[0077] In an embodiment of the present invention, when it is detected that there is no graphics processor, the graphics processor is unavailable, or the CUDA availability is lower than a preset threshold, it can also automatically switch to CPU mode (reserved interface) so that the process of linear interpolation of medical image data can be performed directly by the CPU, thereby ensuring that degraded processing of medical image reconstruction is achieved when there is no graphics processor, the graphics processor is unavailable, or the CUDA availability is lower than a preset threshold.

[0078] In an embodiment of the present invention, availability can be checked; if the graphics processor is available, the medical image data stored in the first memory is copied to the second memory. The above technical solution, in which the medical image data is copied to the second memory if the graphics processor is available, can improve the success rate of linear interpolation of the medical image data.

[0079] Another optional technical solution is to store medical image data in a medical image file, and copy the medical image data stored in the first memory of the central processing unit to the second memory of the graphics processing unit, including: obtaining the extension of the medical image file stored in the first memory of the central processing unit; reading the medical image data in the medical image file based on a reader corresponding to the extension, and copying the medical image data to the second memory of the graphics processing unit.

[0080] The medical imaging file may be understood as a file storing medical imaging data.

[0081] The extension can be understood as an extension used to indicate the type and format of the medical image file.

[0082] The reader can be understood as a reader for reading medical imaging data.

[0083] In an embodiment of the present invention, an extension can be obtained; based on a reader corresponding to the extension, medical image data in the medical image file can be read and copied to a second memory. The above technical solution, which reads medical image data by extension and copies the medical image data to the second memory, can ensure that the medical image data is accurately copied to the second memory.

[0084] Figure 2 This is a flowchart of another medical image reconstruction method provided in an embodiment of the present invention. This embodiment is optimized based on the above-mentioned technical solutions. In this embodiment, optionally, the medical image reconstruction method further includes: obtaining the capacity of the second memory, and determining the interpolation multiple based on the capacity; using multiple threads started by the graphics processor to perform linear interpolation on the medical image data stored in the second memory in parallel to obtain interpolated data, including: using multiple threads started by the graphics processor to perform linear interpolation on the medical image data stored in the second memory under the interpolation multiple in parallel to obtain interpolated data. Among them, the explanations of the terms that are the same as or corresponding to the above-mentioned embodiments are not repeated here.

[0085] See also Figure 2 The method of this embodiment may specifically include the following steps:

[0086] S210 , in response to a medical image reconstruction instruction, copy the medical image data stored in the first memory of the central processing unit to the second memory of the graphics processing unit.

[0087] S220: Obtain the capacity of the second memory, and determine an interpolation multiple according to the capacity.

[0088] Here, the capacity can be understood as the capacity of the second memory.

[0089] The interpolation multiple can be understood as the density multiple of the interpolated data points relative to the original data points in the process of linear interpolation of medical image data; the interpolation multiple can be an integer or a non-integer, that is, linear interpolation supports non-integer scaling, thereby improving the flexibility of medical image reconstruction; the interpolation multiple can be determined according to the capacity or pre-set. Specifically, the interpolation multiple can be controlled by the user parameter scale_factor, which can be pre-set or determined according to the capacity; the interpolation multiple can be 1-4 times.

[0090] In the embodiment of the present invention, the capacity may be acquired, and the interpolation multiple may be determined according to the capacity.

[0091] S230. Execute a kernel function for implementing linear interpolation for a unified computing device architecture on a graphics processor, so as to utilize multiple threads started by the graphics processor to perform linear interpolation of the medical imaging data stored in the second memory in parallel under an interpolation multiple to obtain interpolated data.

[0092] In an embodiment of the present invention, multiple threads started by a graphics processor can be used to perform linear interpolation under an interpolation multiple on the medical image data stored in the second memory in parallel to obtain interpolated data.

[0093] S240: Transmit the interpolated data stored in the second memory back to the first memory.

[0094] S250: Generate a first three-dimensional grid based on the interpolated data stored in the first memory, and reconstruct a medical image model corresponding to the medical image data based on the first three-dimensional grid.

[0095] The technical solution of an embodiment of the present invention obtains the capacity of the second memory and determines an interpolation factor based on the capacity; and utilizes multiple threads initiated by a graphics processor to perform linear interpolation of the medical imaging data stored in the second memory in parallel under the interpolation factor to obtain interpolated data. Compared to related solutions that use fixed module settings, where linear interpolation is limited to the interface parameter range and lacks the ability to automatically adjust according to hardware configuration, the above technical solution, which determines the interpolation factor based on the capacity, can support dynamic adjustment of the interpolation factor to adapt to the capacity of the second memory, achieving flexible adaptation to different hardware environments and avoiding problems such as memory overflow or calculation anomalies caused by improper interpolation factor settings during the medical image reconstruction process. In particular, it can avoid the occurrence of such problems when the medical imaging data is extremely large.

[0096] Figure 3 This is a flowchart of another medical image reconstruction method provided in an embodiment of the present invention. This embodiment is optimized based on the above-mentioned technical solutions. In this embodiment, optionally, a first three-dimensional grid is generated based on the interpolated data stored in the first memory, including: obtaining the topological structure of the data to be corroded based on the interpolated data stored in the first memory, and performing a corrosion operation on the data to be corroded; checking the topological structure during the corrosion process, and if the topological structure changes, stopping the corrosion operation, and generating a first three-dimensional grid based on the currently obtained corroded data. The explanations of the terms that are the same as or corresponding to the above-mentioned embodiments are not repeated here.

[0097] See also Figure 3 The method of this embodiment may specifically include the following steps:

[0098] S310 , in response to a medical image reconstruction instruction, copy the medical image data stored in the first memory of the central processing unit to the second memory of the graphics processing unit.

[0099] S320. Execute a kernel function for implementing linear interpolation for a unified computing device architecture on a graphics processor, so as to utilize multiple threads started by the graphics processor to perform linear interpolation on the medical imaging data stored in the second memory in parallel to obtain interpolated data.

[0100] S330: Transmit the interpolated data stored in the second memory back to the first memory.

[0101] S340 , obtaining a topological structure of the data to be eroded based on the interpolated data stored in the first memory, and performing an erosion operation on the data to be eroded.

[0102] The data to be corroded can be understood as data to be subjected to corrosion operation.

[0103] The topological structure can be understood as the topological structure of the data to be eroded.

[0104] In the embodiment of the present invention, the data to be eroded may be obtained based on the interpolated data stored in the first memory, for example, the interpolated data may be used as the data to be eroded.

[0105] In the embodiment of the present invention, the topological structure of the data to be corroded may be obtained. For example, a structural analysis may be performed on the data to be corroded, and the topological structure may be determined according to the obtained analysis result.

[0106] In the embodiment of the present invention, an erosion operation may be performed on the data to be eroded.

[0107] S350 , checking the topology during the corrosion process, and stopping the corrosion operation if the topology changes, and generating a first three-dimensional grid based on the currently obtained corroded data.

[0108] In an embodiment of the present invention, if the erosion operation is continued when the topological structure changes, the subsequently generated first three-dimensional mesh will not correspond to the topological structure of the interpolated data. Therefore, when the topological structure changes, the erosion operation can be stopped and the first three-dimensional mesh can be generated based on the eroded data.

[0109] In an embodiment of the present invention, a first three-dimensional mesh is generated based on the corroded data. For example, the first three-dimensional mesh can be generated based on the corroded data by using a three-dimensional mesh extraction (Marching Cubes) algorithm of an isosurface. Specifically, a threshold for extracting the isosurface can be set (for example, 0.5) to identify the 1-0 boundary of the corroded data. The VTK marching cube filter (vtkMarchingCubes) algorithm is used to extract the isosurface of the threshold in the corroded data to automatically generate a continuous triangular facet mesh to obtain a first three-dimensional mesh that ensures surface continuity. This can achieve automatic extraction of a target surface (triangular mesh) from the corroded data, forming a preliminary modeling basis.

[0110] S360: Reconstruct a medical image model corresponding to the medical image data based on the first three-dimensional grid.

[0111] The technical solution of an embodiment of the present invention obtains the topological structure of the data to be eroded based on the interpolated data stored in the first memory, and performs an erosion operation on the data to be eroded. During the erosion process, the topological structure is checked, and if the topological structure changes, the erosion operation is stopped, and a first three-dimensional mesh is generated based on the currently obtained eroded data. Compared with related schemes such as threshold segmentation-based mesh generation algorithms (such as the VTK isosurface filter (vtkContourFilter)), which have defects in the mesh generation process, the above-mentioned technical solution provides support for generating a smooth first three-dimensional mesh through the topological structure. This can avoid problems such as reduced accuracy, surface discontinuities, and obvious jaggedness of the generated first three-dimensional mesh. It can also improve the surface quality of the medical image model reconstructed based on the first three-dimensional mesh and enhance the rationality of the topological structure of the first three-dimensional mesh, thereby helping to improve the surface roughness of the medical image model and avoid affecting subsequent diagnosis and surgical planning.

[0112] An optional technical solution for obtaining the topological structure of the data to be corroded includes: marking the connected areas in the data to be corroded and obtaining the number of connected areas, wherein the number is used to characterize the topological structure of the data to be corroded; checking the topological structure during the corrosion process and stopping the corrosion operation if the topological structure changes, including: checking the number during the corrosion process and stopping the corrosion operation if the number changes.

[0113] The connected area can be understood as an area with spatial connectivity within the data to be eroded.

[0114] The number can be understood as the number of connected regions, which can characterize the topological structure of the data to be eroded.

[0115] In this embodiment of the present invention, connected regions can be marked and their number can be determined. During the erosion process, the number of connected regions can be checked, and if a change in the number indicates a change in the topological structure, the erosion operation can be stopped. This technical solution, by stopping the erosion operation when the number of connected regions changes, ensures the generation of a smooth first three-dimensional mesh and also ensures the accuracy of the generated first three-dimensional mesh.

[0116] Another optional technical solution is that the corrosion operation is performed based on the structural element, and the structural element is created in the following manner: obtaining the corrosion radius corresponding to the corrosion operation; obtaining the structural element diameter based on the corrosion radius, and obtaining the center position based on the structural element diameter; generating a second three-dimensional grid based on the structural element diameter, and determining the distance between each grid point in the second three-dimensional grid and the center position; obtaining a spherical mask corresponding to the second three-dimensional grid based on the numerical relationship between each distance and the corrosion radius, and creating a structural element based on the spherical mask.

[0117] Among them, the structural element can be understood as the core tool for defining morphological operations (here, the erosion operation). It is essentially a small matrix that can achieve specific transformation effects by sliding and interacting with each other.

[0118] The corrosion radius can be understood as the maximum distance from the center to the edge of a structural element.

[0119] In the embodiment of the present invention, the corrosion radius can be obtained.

[0120] The diameter of a structural element can be understood as the maximum span of the structural element within the spatial range.

[0121] In the embodiment of the present invention, the diameter of the structural element is obtained according to the erosion radius. For example, the erosion radius can be used as the element radius, and the diameter of the structural element is determined according to the element radius.

[0122] The center position can be understood as the position of the center of the structural element, which can also represent the reference coordinates of the structural element in the data to be eroded.

[0123] In the embodiment of the present invention, the center position may be obtained according to the diameter of the structure element. For example, the diameter of the structure element may be obtained by floating-point calculation according to the erosion radius, and the center position may be obtained according to the diameter of the structure element.

[0124] The second three-dimensional grid can be understood as a three-dimensional grid generated according to the diameter of the structural element.

[0125] The distance can be understood as the distance between the grid point and the center position.

[0126] In the embodiment of the present invention, a second three-dimensional grid may be generated according to the diameter of the structural element, and the distance between each grid point in the second three-dimensional grid and the center position may be determined.

[0127] The numerical relationship can be understood as the numerical relationship between the distance and the corrosion radius.

[0128] The spherical mask can be understood as a mask corresponding to the valid area of ​​the spherical structure element.

[0129] In an embodiment of the present invention, a spherical mask can be obtained based on the numerical relationship between each distance and the corrosion radius. For example, grid points whose standard distances of the numerical relationship are less than or equal to the corrosion radius can be used as target points, and a spherical mask can be determined based on the obtained target points.

[0130] In an embodiment of the present invention, a structure element may be created based on a spherical mask.

[0131] In this embodiment of the present invention, the erosion radius can be obtained; the diameter of the structural element can be obtained based on the erosion radius, and the center position can be obtained based on the diameter of the structural element; a second three-dimensional grid can be generated based on the diameter of the structural element, and the distance between each grid point in the second three-dimensional grid and the center position can be determined; a spherical mask can be obtained based on the numerical relationship between each distance and the erosion radius, and a structural element can be created based on the spherical mask. The above technical solution can realize the adaptive creation of structural elements.

[0132] In order to better understand the technical solutions of the above-mentioned embodiments of the present invention, an optional example is provided herein. For example, the embodiments of the present invention adopt the binarization preprocessing of medical image data, anomaly detection mechanism, CUDA-based trilinear interpolation upsampling processing, median filtering processing, dilation and erosion processing, conversion of image data objects that can be directly used by the VTK framework, first three-dimensional mesh extraction based on isosurfaces, mesh smoothing processing, mesh simplification processing, and reconstruction and export of medical image models involved in the above-mentioned embodiments to achieve medical image reconstruction.

[0133] The above scheme can be tested to provide a theoretical basis for parameter tuning in the medical image reconstruction process. The specific testing method is as follows:

[0134] The speedup ratio formula can be used , determine the acceleration ratio of the above scheme. It is measured that when the interpolation multiplier is 2, the acceleration ratio of the above scheme can reach 8-12 times, where C is the CPU single-core computing constant, P is the number of GPU stream processors, and D is the dimension of medical imaging data.

[0135] The surface roughness formula can be used , the surface roughness of the reconstructed medical image model of the above scheme was determined. After 300 smoothing iterations, the surface roughness ΔS of the above scheme was reduced to 20% of that of the related scheme.

[0136] The efficiency formula can be simplified by using patches , determine the patch simplification efficiency of the above scheme. It is measured that when the preset ratio is 0.3, the patch simplification efficiency η of the above scheme can reach 0.12, that is, while retaining 90% visual accuracy, the above scheme can reduce 30% of patches.

[0137] The test results obtained above, as well as the experimental results obtained by using the relevant scheme and the scheme of the embodiment of the present invention to reconstruct medical images (for example, see Figure 4 、 Figure 5 、 Figure 6 and Figure 7 , the experimental results of the medical imaging models obtained by reconstructing medical imaging data corresponding to different target tissues in the relevant schemes and the scheme of the embodiment of the present invention (this scheme) respectively) show that, compared with the interpolation time complexity of medical image reconstruction, (n is the volume data dimension, k is the computational complexity of the interpolation kernel). The above scheme fully utilizes the computing advantages of GPU to improve the processing speed and support batch processing of large-scale data. It also introduces automated image dimension, spatial information synchronization and automatic adaptation, CUDA-accelerated linear interpolation, enhanced grid generation algorithm including smoothing algorithm, automated processing pipeline, custom adjustable three-dimensional anisotropic morphological processing strategy, multi-stage nonlinear and linear filtering combination processing technology, grid simplification algorithm, flexible parameter configuration and batch processing capabilities. It can integrate the continuous medical image reconstruction process (extraction-smoothing-simplification-export) under the VTK framework. It has developed a highly modularized, core parameter centralized configuration, and parameter-controllable data processing framework and exception handling mechanism, which effectively solves the problems of low efficiency, high manual participation, insufficient flexibility, and poor mesh quality in medical image reconstruction, ensures the compatibility of GPU acceleration and abnormal input, achieves a quality-efficiency balance, and can increase the computational efficiency of the interpolation process in medical image reconstruction by 5-10 times. It supports real-time processing of 4K volume data, reduces 70% of redundant patches, and reduces the surface roughness of the reconstructed medical image model by 80%, significantly improving efficiency and model quality, and making it more adaptable, thereby achieving efficient, automated, highly robust, and flexible medical image reconstruction.

[0138] Figure 8This is a block diagram of the structure of a medical image reconstruction device provided by an embodiment of the present invention. The device is used to execute the medical image reconstruction method provided by any of the above embodiments. The device and the medical image reconstruction method of the above embodiments belong to the same inventive concept. For details not fully described in the embodiments of the medical image reconstruction device, please refer to the embodiments of the above medical image reconstruction method. Figure 8 The device may specifically include: a medical image data copying module 410, an interpolated data obtaining module 420, a data return module 430 and a medical image model reconstruction module 440.

[0139] Among them, the medical image data copy module 410 is used to copy the medical image data stored in the first memory of the central processing unit to the second memory of the graphics processing unit in response to the medical image reconstruction instruction; the interpolation data acquisition module 420 is used to execute the kernel function for implementing linear interpolation of the unified computing device architecture on the graphics processing unit, so as to use multiple threads started by the graphics processing unit to perform linear interpolation on the medical image data stored in the second memory in parallel to obtain interpolated data; the data return module 430 is used to return the interpolated data stored in the second memory to the first memory; the medical image model reconstruction module 440 is used to generate a first three-dimensional grid based on the interpolated data stored in the first memory, and reconstruct a medical image model corresponding to the medical image data based on the first three-dimensional grid.

[0140] Optionally, the device may also include: an interpolation multiple determination module, used to obtain the capacity of the second memory and determine the interpolation multiple based on the capacity; the interpolation data acquisition module 420 may include: an interpolation data acquisition submodule, used to use multiple threads started by the graphics processor to perform linear interpolation under the interpolation multiple on the medical imaging data stored in the second memory in parallel to obtain interpolated data.

[0141] Optionally, the medical image model reconstruction module 440 may include: a corrosion operation execution submodule, which is used to obtain the topological structure of the data to be corroded based on the interpolated data stored in the first memory, and perform a corrosion operation on the data to be corroded; a first three-dimensional grid generation submodule, which is used to check the topological structure during the corrosion process, and if the topological structure changes, stop executing the corrosion operation and generate a first three-dimensional grid based on the currently obtained corroded data.

[0142] Optionally, based on the above-mentioned device, the corrosion operation execution submodule may include: a quantity obtaining unit, used to mark the connected areas in the data to be corroded and obtain the number of connected areas, wherein the number is used to characterize the topological structure of the data to be corroded; the first three-dimensional grid generation submodule may include: a corrosion operation stopping unit, used to check the number during the corrosion process and stop executing the corrosion operation if the number changes.

[0143] Optionally, based on the above-mentioned device, the corrosion operation is performed based on the structural element, and the device may further include the following modules to create the structural element: a corrosion radius acquisition module, used to obtain the corrosion radius corresponding to the corrosion operation; a center position acquisition module, used to obtain the diameter of the structural element according to the corrosion radius, and obtain the center position according to the diameter of the structural element; a distance determination module, used to generate a second three-dimensional grid according to the diameter of the structural element, and determine the distance between each grid point in the second three-dimensional grid and the center position; a structural element creation module, used to obtain a spherical mask corresponding to the second three-dimensional grid according to the numerical relationship between each distance and the corrosion radius, and create a structural element based on the spherical mask.

[0144] Optionally, the medical image model reconstruction module 440 may include: a sorting result obtaining submodule, used to obtain the curvature of each vertex in the first three-dimensional mesh, and obtain the edge folding cost based on the curvature, so as to sort the edges in the first three-dimensional mesh according to the edge folding cost corresponding to each vertex to obtain a sorting result; a medical image model reconstruction submodule, used to simplify the first three-dimensional mesh in turn based on each edge in the sorting result, and when the simplification ratio of the first three-dimensional mesh reaches a preset ratio, reconstruct a medical image model corresponding to the medical image data based on the currently obtained simplified three-dimensional mesh.

[0145] Optionally, the medical image data copy module 410 may include: a first medical image data copy submodule, used to copy multiple medical image data stored in the first memory of the central processing unit to the second memory of the graphics processing unit respectively to achieve parallel reconstruction of multiple medical image data.

[0146] Optionally, the medical image data copy module 410 may include: an availability check submodule, used to check the availability of the graphics processor; and a second medical image data copy submodule, used to copy the medical image data stored in the first memory of the central processing unit to the second memory of the graphics processor when the graphics processor is available.

[0147] Optionally, the medical image data is stored in a medical image file, and the medical image data copy module 410 may include: an extension name acquisition submodule, used to obtain the extension name of the medical image file stored in the first memory of the central processing unit; a third medical image data copy submodule, used to read the medical image data in the medical image file based on a reader corresponding to the extension name, and copy the medical image data to the second memory of the graphics processing unit.

[0148] The medical image reconstruction device provided by an embodiment of the present invention, through a medical image data copying module, responds to a medical image reconstruction instruction, copies the medical image data stored in the first memory of the central processing unit to the second memory of the graphics processing unit, so that the graphics processing unit can perform linear interpolation on the medical image data; through an interpolation data acquisition module, for a kernel function for implementing linear interpolation of a unified computing device architecture, executes a kernel function on the graphics processing unit, so as to utilize multiple threads launched by the graphics processing unit to perform linear interpolation on the medical image data stored in the second memory in parallel to obtain interpolated data, and can efficiently perform linear interpolation on the medical image data in parallel through multiple threads launched by the graphics processing unit; through a data return module, the interpolation data stored in the second memory is returned to the first memory; through a medical image model reconstruction module, a first three-dimensional grid is generated based on the interpolation data stored in the first memory, and a medical image model corresponding to the medical image data is reconstructed based on the first three-dimensional grid to achieve reconstruction of the medical image model. The above-mentioned device copies the medical image data to the second memory of the graphics processor, and then uses multiple threads started by the graphics processor to efficiently and in parallel perform linear interpolation on the medical image data, which can improve the efficiency of interpolation of the medical image data. Moreover, the entire medical image reconstruction process does not need to rely on manual labor and can be carried out automatically, thereby improving the efficiency of medical image reconstruction.

[0149] The medical image reconstruction device provided by the embodiment of the present invention can execute the medical image reconstruction method provided by any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0150] It is worth noting that in the embodiment of the above-mentioned medical image reconstruction device, the various units and modules included are only divided according to functional logic, but are not limited to the above-mentioned division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention.

[0151] Figure 9A schematic diagram of an electronic device 10 that can be used to implement an embodiment of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices (such as helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or claimed herein.

[0152] like Figure 9 As shown, electronic device 10 includes at least one processor 11 and memory, such as read-only memory (ROM) 12 and random access memory (RAM) 13, communicatively connected to at least one processor 11. The memory stores computer programs executable by the at least one processor. Processor 11 can perform various appropriate actions and processes based on the computer programs stored in ROM 12 or loaded from storage unit 18 into RAM 13. RAM 13 can also store various programs and data required for the operation of electronic device 10. Processor 11, ROM 12, and RAM 13 are interconnected via bus 14. An input / output (I / O) interface 15 is also connected to bus 14.

[0153] Multiple components in the electronic device 10 are connected to the I / O interface 15, including an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a magnetic disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0154] Processor 11 can be any general-purpose and / or specialized processing component with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various specialized artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any other suitable processor, controller, microcontroller, etc. Processor 11 executes the various methods and processes described above, such as the medical image reconstruction method.

[0155] In particular, according to an embodiment of the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present invention includes a computer program product that includes a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network via the communication unit 19, or installed from the storage unit 18, or installed from the ROM 12. When the computer program is executed by the processor 11, the above-mentioned functions defined in the method of the embodiment of the present invention are performed.

[0156] In some embodiments, the medical image reconstruction method can be implemented as a computer program tangibly embodied in a computer-readable storage medium, such as the storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the medical image reconstruction method described above can be performed. Alternatively, in other embodiments, the processor 11 can be configured to perform the medical image reconstruction method in any other suitable manner (e.g., via firmware).

[0157] Various implementations of the systems and techniques described above can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on a chip (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include being implemented in one or more computer programs that are executable and / or interpreted on a programmable system that includes at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0158] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when the computer program is executed by the processor, the functions / operations specified in the flowcharts and / or block diagrams are implemented. The computer program may be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0159] In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, device, or apparatus. A computer-readable storage medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or apparatus, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media may include an electrical connection based on one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0160] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device that has: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0161] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0162] A computing system may include clients and servers. The clients and servers are typically remote from each other and typically interact via a communication network. This client-server relationship arises through computer programs running on the respective computers, creating a client-server relationship. The server may be a cloud server, also known as a cloud computing server or cloud host. This server is a hosting product within the cloud computing service ecosystem that addresses the management difficulties and limited scalability of traditional physical hosting and VPS services.

[0163] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in the present invention can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solution of the present invention can be achieved. This is not limited herein.

[0164] The above specific embodiments do not limit the scope of protection of the present invention. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention are intended to be included within the scope of protection of the present invention.

Claims

1. A medical image reconstruction method, characterized in that: include: In response to the medical image reconstruction instruction, copying the medical image data stored in the first memory of the central processing unit to the second memory of the graphics processing unit; executing a kernel function for implementing linear interpolation for a unified computing device architecture on the graphics processor to perform linear interpolation on the medical image data stored in the second memory in parallel using multiple threads launched by the graphics processor to obtain interpolated data; Returning the interpolated data stored in the second memory to the first memory; A first three-dimensional grid is generated based on the interpolated data stored in the first memory, and a medical image model corresponding to the medical image data is reconstructed based on the first three-dimensional grid.

2. The method according to claim 1, characterized in that Also includes: Obtaining a capacity of the second memory, and determining an interpolation multiple according to the capacity; The method of performing linear interpolation on the medical image data stored in the second memory in parallel by using the multiple threads started by the graphics processor to obtain interpolated data includes: The medical image data stored in the second memory are linearly interpolated at the interpolation multiple in parallel using a plurality of threads started by the graphics processor to obtain interpolated data.

3. The method according to claim 1, characterized in that Generating a first three-dimensional grid based on the interpolated data stored in the first memory includes: For data to be eroded obtained based on the interpolated data stored in the first memory, obtaining a topological structure of the data to be eroded, and performing an erosion operation on the data to be eroded; The topological structure is checked during the erosion process, and if the topological structure changes, the erosion operation is stopped, and a first three-dimensional mesh is generated based on the currently obtained eroded data.

4. The method according to claim 3, characterized in that The obtaining of the topological structure of the data to be eroded includes: Marking connected regions in the data to be eroded, and obtaining the number of the connected regions, wherein the number is used to characterize the topological structure of the data to be eroded; The checking of the topology during the corrosion process and stopping the corrosion operation if the topology changes include: The number is checked during the etching process, and if the number changes, the etching operation is stopped.

5. The method according to claim 3, characterized in that The erosion operation is performed based on a structuring element, which is created as follows: Obtaining a corrosion radius corresponding to the corrosion operation; Obtaining a structure element diameter according to the erosion radius, and obtaining a center position according to the structure element diameter; generating a second three-dimensional grid according to the diameter of the structural element, and determining a distance between each grid point in the second three-dimensional grid and the center position; According to the numerical relationship between each of the distances and the erosion radius, a spherical mask corresponding to the second three-dimensional grid is obtained, and the structural element is created based on the spherical mask.

6. The method according to claim 1, characterized in that The reconstructing a medical image model corresponding to the medical image data based on the first three-dimensional grid includes: For each vertex in the first three-dimensional mesh, obtaining a curvature of the vertex, and obtaining an edge collapse cost based on the curvature, so as to sort the edges in the first three-dimensional mesh according to the edge collapse costs corresponding to the vertices, thereby obtaining a sorting result; Based on each edge in the sorting result, the first three-dimensional grid is simplified in turn, and when the simplification ratio of the first three-dimensional grid reaches a preset ratio, a medical image model corresponding to the medical image data is reconstructed based on the currently obtained simplified three-dimensional grid.

7. The method according to claim 1, characterized in that The step of copying the medical image data stored in the first memory of the central processing unit to the second memory of the graphics processing unit includes: The plurality of medical image data stored in the first memory of the central processing unit are copied to the second memory of the graphics processing unit respectively, so as to realize parallel reconstruction of the plurality of medical image data.

8. The method according to claim 1, characterized in that The step of copying the medical image data stored in the first memory of the central processing unit to the second memory of the graphics processing unit includes: Check the availability of graphics processor; When the graphics processor is available, the medical image data stored in the first memory of the central processing unit is copied to the second memory of the graphics processor.

9. The method according to claim 1, characterized in that The medical image data is stored in a medical image file, and the step of copying the medical image data stored in the first memory of the central processing unit to the second memory of the graphics processing unit includes: Obtaining the extension of the medical image file stored in the first memory of the central processing unit; Based on the reader corresponding to the extension, the medical image data in the medical image file is read, and the medical image data is copied to the second memory of the graphics processor.

10. An electronic device, characterized in that: include: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor to enable the at least one processor to perform the medical image reconstruction method according to any one of claims 1 to 9.

Images