A method and system for adaptive digital reconstruction of radiological images based on genetic evolution
Through the adaptive digital reconstruction radiation imaging method based on genetic evolution, the problems of digital radiation projection algorithm in dealing with radiation dose, optical density differences and tissue composite imaging in complex medical imaging scenarios were solved, and high-quality DRR images were generated, showing more Clear soft tissue and organ structure improves the efficiency and accuracy of medical imaging conversion.
Patent Information
- Application Number
- CN202411433642.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-15
- Publication Date
- 2024-12-06
- Estimated Expiration
- 2044-10-15
AI Technical Summary
When existing digital radiation projection algorithms face complex medical imaging scenarios, it is difficult to effectively deal with different radiation doses, tissue optical density differences and tissue composite imaging, resulting in low image quality and inability to accurately display human soft tissues and organs.
Adaptive digital reconstruction radiographic imaging method based on genetic evolution is used to adjust the transparency function, filter operator, image fusion and histogram equalization parameters, and use spatially weighted multi-scale similarity as the fitness function to dynamically adjust the genetic evolution parameters to generate High-quality DRR imagery.
It improves the quality of composite imaging of different tissues, can display human soft tissues and organs more clearly, enhances the perception of complex textures and structures, and provides a more efficient and accurate medical image conversion solution.
Smart Images

Figure CN118941671B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of image processing, and in particular to a method and system for adaptive digital reconstruction of radiological images based on genetic evolution. Background Art
[0002] In the field of medical imaging, computed tomography (CT) is a commonly used imaging technology that produces a large number of tomographic images by performing cross-sectional scans of the human body. It provides detailed three-dimensional images for detecting and diagnosing various diseases and lesions. However, although CT imaging is widely used in clinical practice, there are some problems, such as: patients are repeatedly exposed to X-rays, the radiation dose exceeds the standard; and the problem of not being able to image in real time. On the other hand, X-rays have the characteristics of rapid imaging and relatively low radiation dose, and are widely used in clinical rapid screening and preliminary diagnosis. Compared with CT and X-ray technology, DRR overcomes the limitations of high radiation doses for real-time imaging and accurate positioning of specific anatomical structures. DRR is widely used in surgical navigation and radiotherapy planning, which helps doctors more accurately locate the target area and formulate treatment plans during surgery, thereby improving the accuracy of surgery and treatment effects.
[0003] The main challenges of current digital radiographic projection algorithms when faced with complex medical imaging scenarios are:
[0004] (1) Processing of different radiation doses. The radiation doses received by patients in different medical imaging scenarios vary greatly, and the dose changes caused by different scanning modes and parameter settings need to be considered. The higher the radiation dose, the more significant the absorption and attenuation of radiation by water, which affects the HU value in the CT data (the standardized value of tissue density in CT images).
[0005] (2) Processing the difference in optical density of different tissues. Different tissues in the human body have obvious differences in optical density, which puts strict requirements on the DRR algorithm. The algorithm needs to consider the absorption and scattering characteristics of the tissue and model the optical parameters of different tissues in detail. It involves complex physical models and multi-parameter optimization to ensure that the generated image is not distorted while fully expressing the anatomical structure.
[0006] (3) Composite imaging of different tissues. Due to the overlap of multiple tissues, the algorithm must be able to resolve the confusion between different tissues, such as dealing with large density differences and blurred boundaries. Summary of the invention
[0007] In order to at least overcome the above-mentioned deficiencies in the prior art, the purpose of the present application is to provide a method and system for adaptive digital reconstruction of radiological images based on genetic evolution.
[0008] In a first aspect, an embodiment of the present application provides a method for adaptive digital reconstruction of radiological images based on genetic evolution, comprising:
[0009] Acquire CT data and construct a voxel matrix corresponding to the CT data;
[0010] Initializing a transparency function and a color function, and generating the voxel matrix into a first DRR image according to the transparency function and the color function;
[0011] Initializing genetic evolution parameters and gene parameters, and enhancing the first DRR image to form a second DRR image according to the filter gene, image fusion gene and histogram equalization gene in the gene parameters; the gene parameters include transparency gene, color gene, filter gene, image fusion gene and histogram equalization gene;
[0012] Calculate and update the gene parameters according to the genetic evolution parameters, update the transparency function according to the transparency gene in the updated gene parameters, and update the color function according to the color gene in the updated gene parameters;
[0013] Generate a new first DRR image according to the updated transparency function and the color function, and enhance the new first DRR image according to the updated filter gene, image fusion gene and histogram equalization gene in the gene parameters to form a new second DRR image;
[0014] Repeat the updating of the gene parameters, the transparency function and the color function to generate the first DRR image and the second DRR image until the fitness of the second DRR image meets the preset requirements or the number of updates reaches a threshold, output the second DRR image as the final DRR image.
[0015] Furthermore, the transparency function adopts the following formula:
[0016]
[0017] The color function adopts the following formula:
[0018]
[0019] In the formula, k 1. k 2. k 3 and k 4 is the transparency function coefficient, l 1. l 2. l 3 and l 4 is the color function coefficient, x is the voxel grayscale value in the voxel matrix.
[0020] Further, updating the transparency function according to the updated transparency gene in the gene parameters, and updating the color function according to the updated color gene in the gene parameters include:
[0021] The transparency function coefficient in the transparency function is updated according to the updated transparency gene in the gene parameter, and the color function coefficient in the color function is updated according to the updated color gene in the gene parameter.
[0022] Furthermore, the generation of the first DRR image includes:
[0023] The color value obtained by observing the following voxel through the current voxel is calculated according to the following formula:
[0024]
[0025] In the formula, represents the current voxel, Indicates the transparency of the current voxel, represents the color of the current voxel, Indicates the color of the following voxels.
[0026] Furthermore, the generation of the second DRR image includes:
[0027] The first DRR image is edge enhanced by the filter gene to form an image edge image:
[0028] Image edge = f (F ori , G filter )
[0029] In the formula, f is Marginal function, F ori is the first DRR image, G filter For filtering genes, Image edge is the edge image, when G filter =1 f Using Gaussian operator; when G filter =2, f Using Sobel operator; when G filter =3, f Using Laplace operator;
[0030] The image edge image and the first DRR image are fused by the image fusion gene to generate a fused image:
[0031] Image fusion = G fusion ×Imageedge +(1-G fusion )×F ori
[0032] In the formula, Image fusion is the fused image, G fusion For image fusion genes;
[0033] The fused image is subjected to histogram enhancement processing by using the histogram equalization gene:
[0034] Image enhance = f e (Image fusion , G Hist )
[0035] In the formula, f e is the histogram enhancement function, Image enhance is the second DRR image, G Hist is the histogram equalization gene; when G Hist =1, f e The histogram equalization function is turned on; when G Hist =0, Image enhance =Image fusion .
[0036] Further, calculating and updating the gene parameter according to the genetic evolution parameter includes:
[0037] The dynamic adjustment factor for each iteration is calculated according to the following formula:
[0038]
[0039] Where SWMSS is the spatially weighted multi-scale similarity, N is the population size, and i is the i-th individual;
[0040] The next generation gene crossover rate is calculated according to the following formula:
[0041]
[0042] In the formula, is the lower threshold, is the upper threshold, is the previous generation crossover rate, is the next generation crossover rate, is the generation-by-generation adjustment rate, is the growth coefficient, is the attenuation coefficient;
[0043] The next generation gene mutation rate is calculated according to the following formula:
[0044]
[0045] In the formula, is the mutation rate of the previous generation, is the mutation rate of the next generation;
[0046] The probability of each individual being selected is calculated according to the following formula:
[0047]
[0048] Where P(i) is the probability of individual i being selected;
[0049] Gene crossover calculation is performed according to the following formula:
[0050]
[0051] The gene variation is calculated according to the following formula:
[0052]
[0053] Where P child is the offspring produced, and p1, p2 and p3 are the parents.
[0054] Furthermore, the spatial weighted multi-scale similarity is calculated according to the following formula:
[0055]
[0056]
[0057]
[0058]
[0059]
[0060]
[0061]
[0062]
[0063] In the formula, For single scale The spatial similarity between image X and image Y is: is the weight coefficient for each scale, For image X in the corresponding The mean brightness of the scaled image, For image Y in the corresponding The mean brightness of the scaled image, For image X in the corresponding The standard deviation of image brightness at scale, For image Y in the corresponding The standard deviation of image brightness at scale, and is the stability constant, It is in The spatial weight function on the scale, For The mean gradient of image X on the scale, For The gradient mean of image Y on the scale, abs is the absolute value function, k is the gradient parameter, size_row is the number of row pixels in the image, size_col is the number of column pixels in the image, For The gradient sum of image X on the scale, For images at position The gradient value of is the lateral gradient, is the vertical gradient, Indicates that the image is The gray value at .
[0064] Furthermore, the fitness meets the preset requirements including:
[0065] When there is a second DRR image of the previous round, the spatial weighted multi-scale similarity between the second DRR image of the current round and the second DRR image of the previous round is calculated, and when the spatial weighted multi-scale similarity is greater than a preset value, it is determined that the fitness meets the preset requirement;
[0066] When there is no second DRR image of the previous round, the spatial weighted multi-scale similarity of the second DRR image of the current round and the first DRR image of the current round is calculated, and when the spatial weighted multi-scale similarity is greater than a preset value, it is determined that the fitness meets the preset requirements.
[0067] In a second aspect, the present application also provides an adaptive digital reconstruction radiological image system based on genetic evolution, comprising:
[0068] An acquisition unit is configured to acquire CT data and construct a voxel matrix corresponding to the CT data;
[0069] An image unit, configured to initialize a transparency function and a color function, and generate the voxel matrix into a first DRR image according to the transparency function and the color function;
[0070] an enhancement unit, configured to initialize genetic evolution parameters and gene parameters, and enhance the first DRR image to form a second DRR image according to filter genes, image fusion genes and histogram equalization genes in the gene parameters; the gene parameters include transparency genes, color genes, filter genes, image fusion genes and histogram equalization genes;
[0071] an updating unit, configured to calculate and update the gene parameters according to the genetic evolution parameters, update the transparency function according to the transparency gene in the updated gene parameters, and update the color function according to the color gene in the updated gene parameters;
[0072] A generating unit is configured to generate a new first DRR image according to the updated transparency function and the color function, and enhance the new first DRR image to form a new second DRR image according to the updated filter gene, image fusion gene and histogram equalization gene in the gene parameters;
[0073] The iterative unit is configured to repeatedly update the genetic parameters, the transparency function and the color function and generate a first DRR image and a second DRR image until the fitness of the second DRR image meets a preset requirement or the number of updates reaches a threshold, and output the second DRR image as the final DRR image.
[0074] Furthermore, the iteration unit is further configured to:
[0075] When there is a second DRR image of the previous round, the spatial weighted multi-scale similarity between the second DRR image of the current round and the second DRR image of the previous round is calculated, and when the spatial weighted multi-scale similarity is greater than a preset value, it is determined that the fitness meets the preset requirement;
[0076] When there is no second DRR image of the previous round, the spatial weighted multi-scale similarity of the second DRR image of the current round and the first DRR image of the current round is calculated, and when the spatial weighted multi-scale similarity is greater than a preset value, it is determined that the fitness meets the preset requirements.
[0077] Compared with the prior art, the present invention has the following advantages and beneficial effects:
[0078] The present invention can effectively adjust the transparency function parameters, filter operator parameters, image fusion parameters, and the attenuation coefficient of water at different doses to adapt to different anatomical structures and scanning conditions through the optimization strategy of genetic evolution, thereby improving the quality of composite imaging of different tissues and showing more powerful performance in virtual X-ray films. This innovative background technology provides a more efficient and accurate solution for the conversion of CT images to DRR in the field of medical imaging. The present invention adopts a nonlinear cubic function to fit the transparency function and color function, which solves the problem of X-ray density difference processing of different tissues, that is, the absorption of X-rays by tissues; these function parameters are adaptively generated by genetic evolution, so that the projected DRR can more clearly display human soft tissues, organs, and bones. The present invention adopts spatial weighted multi-scale similarity (SWMSS) as the fitness function of genetic evolution. This function can capture the similarity of images more comprehensively and accurately by considering structural information at different scales; at the same time, the multi-scale framework of SWMSS helps to capture the local and global structures of the image at the same time, and improves the perception of complex textures and structures; multi-scale similarity enables the algorithm to adaptively select the appropriate scale and has better adaptability. The present invention dynamically adjusts the factors Adaptively adjust the selection algorithm, crossover calculation, and mutation algorithm of multiple parents. Enhance the global search ability of the algorithm, and obtain X-ray (DR images) that are closer to reality, providing more information for diagnosis, positioning, and navigation surgery. Generate G through genetic evolution filter Gene, G fusion Gene and G Hist Genes,respectively act on the edge enhancement module, image fusion module, and image equalization module to achieve image enhancement. BRIEF DESCRIPTION OF THE DRAWINGS
[0079] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of this application, and do not constitute a limitation of the embodiments of the present invention. In the drawings:
[0080] Figure 1 This is a schematic diagram of the steps of the method of the embodiment of the present application;
[0081] Figure 2 This is a schematic diagram of the method flow of the embodiment of the present application;
[0082] Figure 3 This is a schematic diagram of image enhancement according to an embodiment of the present application. DETAILED DESCRIPTION
[0083] To make the purpose, technical scheme and advantages of the embodiment of the present application clearer, the technical scheme in the embodiment of the present application will be clearly and completely described below in conjunction with the drawings in the embodiment of the present application. It should be understood that the drawings in the present application only serve the purpose of explanation and description and are not used to limit the scope of protection of the present application. In addition, it should be understood that the schematic drawings are not drawn in real proportion. The flowchart used in this application shows the operations implemented according to some embodiments of the embodiment of the present application. It should be understood that the operations of the flowchart can be implemented out of sequence, and the steps without logical context can be reversed in order or implemented simultaneously. In addition, those skilled in the art can add one or more other operations to the flowchart under the guidance of the content of the present application, or remove one or more operations from the flowchart.
[0084] In addition, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application described and shown in the drawings here can be arranged and designed in various configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present application.
[0085] Please refer to Figure 1 , which is a flow chart of a method for adaptive digital reconstruction of radiological images based on genetic evolution provided by an embodiment of the present invention. Furthermore, the method for adaptive digital reconstruction of radiological images based on genetic evolution may specifically include the contents described in the following steps S1 to S6.
[0086] S1: Acquire CT data and construct a voxel matrix corresponding to the CT data;
[0087] S2: Initializing a transparency function and a color function, and generating the voxel matrix into a first DRR image according to the transparency function and the color function;
[0088] S3: Initializing genetic evolution parameters and gene parameters, and enhancing the first DRR image to form a second DRR image according to the filter gene, image fusion gene and histogram equalization gene in the gene parameters; the gene parameters include transparency gene, color gene, filter gene, image fusion gene and histogram equalization gene;
[0089] S4: updating the gene parameters according to the genetic evolution parameter calculation, updating the transparency function according to the transparency gene in the updated gene parameters, and updating the color function according to the color gene in the updated gene parameters;
[0090] S5: generating a new first DRR image according to the updated transparency function and the color function, and enhancing the new first DRR image according to the updated filter gene, image fusion gene and histogram equalization gene in the gene parameters to form a new second DRR image;
[0091] S6: Repeat updating the gene parameters, the transparency function and the color function to generate the first DRR image and the second DRR image until the fitness of the second DRR image meets the preset requirements or the number of updates reaches a threshold, outputting the second DRR image as the final DRR image.
[0092] When implementing the embodiments of this application, please refer to Figure 2 , showing a schematic diagram of the process of an embodiment of the present application, in which it is necessary to obtain CT data and construct a voxel matrix, which belongs to the prior art and is not limited in the embodiments of the present application. In the process of generating the first DRR image, the basic principle is that the virtual X-ray film imaging adopts the principle of point light source ray projection. A virtual X-ray emission source is constructed in three-dimensional space. The X-rays emitted by the virtual light source penetrate the voxels of the CT data. Each beam of X-rays will encounter attenuation and absorption in the process of passing through the CT data, and then be projected onto the imaging plane to form a virtual X-ray image. In the process of initializing the ray projection renderer, a random method is used to initialize the HU value of the voxel at each sampling point on the ray path. The corresponding transparency function , color function Initial values, initialize camera position, camera focus, light source position, etc.
[0093] In the embodiment of the present application, when initializing the genetic evolution parameters and gene parameters, the population size N of the genetic evolution and the initial crossover rate are initialized. , initial mutation rate , fitness function end threshold FIT THRESHOLD , the 0th generation population Pop (0) Transparency gene G opaticy , color gene G color , filter gene G filter , image fusion gene G fusion , histogram equalization gene G Hist In the process of updating gene parameters, the embodiment of the present application uses a genetic evolution algorithm with multiple parents and adaptive crossover mutation to find the transparency function , color function The optimal solution of the parameters of the image enhancement algorithm is used to obtain the optimal value of the objective function. The combination of multiple individuals in the genetic evolution algorithm is called a single independent population. The initial population should be randomly selected to the solution space to ensure that a better solution can be found. Set the population size N, the initial crossover rate , initial mutation rate , fitness function end threshold FIT THRESHOLD The initial solution obtained by random initialization is the 0th generation population Pop (0) , the i-th individual in each generation is denoted by (N≥i≥0, MAX_ITER≥t≥0), N is the population size, MAX_ITER is the maximum number of iterations, and t is the current number of iterations. It is a set of solutions for the digital reconstruction of radiological images algorithm. Each individual is composed of multiple genes, and each gene corresponds to a parameter that needs to be fitted. In the present embodiment, the gene Ind is divided into five parts: the transparency gene G opaticy , color gene G color , filter gene G filter , image fusion gene G fusion , histogram equalization gene G Hist . Transparency gene G opaticy include , used to solve the transparency function k 1, k 2, k 3. k 4 coefficients; color gene G color include , used to solve the color function l 1, l 2, l 3. l 4 coefficients; filter gene G filter The possible values are 1, 2, and 3; image fusion gene G fusion Value range 0~1; histogram equalization gene G Hist The value is 0 or 1. The genetic evolution algorithm selects some individuals and eliminates others in each cycle iteration. The criterion for determining whether an individual is selected or eliminated is the fitness value of each individual.
[0094] In an embodiment of the present application, the updated transparency function and color function can generate a new first DRR image from the voxel matrix and enhance it through filtering genes, image fusion genes and histogram equalization genes. After a certain number of iterations or when the fitness of the second DRR image meets the preset requirements, it means that this influence can more accurately display the required content, and the final DRR image is output. The embodiment of the present application solves the problem that the existing DRR technology can only process tissues with large density differences such as bones and blood vessels, and has limitations in presenting soft tissues and other non-skeletal tissues, such as areas with different radiation doses and large changes in optical density, and multi-tissue composite imaging. The present invention can better display soft tissue features, realize composite imaging of bone tissue and soft tissue, and is closer to real X-rays (DR images), and can provide more information in certain diagnostic, positioning and navigation surgeries.
[0095] Furthermore, the transparency function adopts the following formula:
[0096]
[0097] The color function adopts the following formula:
[0098]
[0099] In the formula, k 1. k 2. k 3 and k 4 is the transparency function coefficient, l 1. l 2. l 3 and l 4 is the color function coefficient, x is the voxel grayscale value in the voxel matrix.
[0100] When the embodiment of the present application is implemented, the light projection renderer calculation includes the absorption and scattering process of X-rays in the object tissue, that is, simulating the propagation path and absorption of X-rays in different tissues. and color functions simulate the absorption of X-rays by tissues. So assuming , and has a nonlinear relationship with the CT value. Specifically, the constant function is air, and the transparency is 0; other CT values such as alveolar nodules, lung tissue, fat, organs, bone tissue and dense bone tissue have a cubic polynomial function nonlinear relationship with the transparency function and color function. Determine the transparency of different tissues and determine the mapping relationship between HU value and color. The color function is determined by a cubic polynomial function, as shown above, which can better reflect the nonlinear absorption characteristics of X-rays of different tissues. The color function is determined by a cubic polynomial function, as shown above.
[0101] Further, updating the transparency function according to the updated transparency gene in the gene parameters, and updating the color function according to the updated color gene in the gene parameters include:
[0102] The transparency function coefficient in the transparency function is updated according to the updated transparency gene in the gene parameter, and the color function coefficient in the color function is updated according to the updated color gene in the gene parameter.
[0103] Furthermore, the generation of the first DRR image includes:
[0104] The color value obtained by observing the following voxel through the current voxel is calculated according to the following formula:
[0105]
[0106] In the formula, represents the current voxel, Indicates the transparency of the current voxel, represents the color of the current voxel, Indicates the color of the following voxels.
[0107] When the embodiment of the present application is implemented, the transparency in the light projection model represents the ability of light to penetrate an object. Light penetrating a voxel will cause a change in the wavelength ratio. If it passes through multiple voxels, this change is cumulative. Therefore, the light projection model generates DRR by essentially mixing the color of the transparent voxel with the color of the subsequent voxel. The light projection model performs sampling and synthesis of the mixing technology from the front to the back, and the transparency and color can be calculated by the corresponding transparency function and color function.
[0108] Furthermore, the generation of the second DRR image includes:
[0109] The first DRR image is edge enhanced by the filter gene to form an image edge image:
[0110] Image edge = f (F ori , G filter )
[0111] In the formula, f is Marginal function, F ori is the first DRR image, G filter For filtering genes, Image edgeis the edge image, when G filter =1 f Using Gaussian operator; when G filter =2, f Using Sobel operator; when G filter =3, f Using Laplace operator;
[0112] The image edge image and the first DRR image are fused by the image fusion gene to generate a fused image:
[0113] Image fusion = G fusion ×Image edge +(1-G fusion )×F ori
[0114] In the formula, Image fusion is the fused image, G fusion For image fusion genes;
[0115] The fused image is subjected to histogram enhancement processing by using the histogram equalization gene:
[0116] Image enhance = f e (Image fusion , G Hist )
[0117] In the formula, f e is the histogram enhancement function, Image enhance is the second DRR image, G Hist is the histogram equalization gene; when G Hist =1, f e The histogram equalization function is turned on; when G Hist =0, Image enhance =Image fusion .
[0118] When implementing the embodiments of this application, please refer to Figure 3 In image processing, the purpose of using image enhancement algorithms is to improve the quality of images, enhance features in images, or improve visualization effects. In order to adapt to different doses and different sampling devices, the present invention uses a genetic algorithm to select different image enhancement strategies, thereby improving the similarity with real DR. The present invention adopts an image enhancement algorithm. This automatic selection method helps to improve the universality and adaptability of image enhancement, making the algorithm more versatile.
[0119] Image enhancement includes filtering operators to enhance edges, and histogram equalization to adjust the contrast of images. In experiments, it was found that some images generated by the ray projection method had blurred edges and low contrast. Edges contain important information in the field of medical images, such as tissue boundaries. Enhancing edges through filtering operators can make edges clearer and more prominent, which helps to more accurately identify and analyze structures in images and increase the similarity with real images. For the problem of low contrast, a histogram equalization strategy is adopted for the image to redistribute the grayscale of the image, which can stretch the brightness range and make the details in the image more prominent. The specific process of image enhancement is as follows: Figure 2 As shown, G was generated through genetic evolution filter Gene, G fusion Gene and G Hist The genes act on the edge enhancement module, image fusion module and image equalization module respectively, and output the enhanced DRR image.
[0120] Further, calculating and updating the gene parameter according to the genetic evolution parameter includes:
[0121] The dynamic adjustment factor for each iteration is calculated according to the following formula:
[0122]
[0123] Where SWMSS is the spatially weighted multi-scale similarity, N is the population size, and i is the i-th individual;
[0124] The next generation gene crossover rate is calculated according to the following formula:
[0125]
[0126] In the formula, is the lower threshold, is the upper threshold, is the previous generation crossover rate, is the next generation crossover rate, is the generation-by-generation adjustment rate, is the growth coefficient, is the attenuation coefficient;
[0127] The next generation gene mutation rate is calculated according to the following formula:
[0128]
[0129] In the formula, is the mutation rate of the previous generation, is the mutation rate of the next generation;
[0130] The probability of each individual being selected is calculated according to the following formula:
[0131]
[0132] Where P(i) is the probability of individual i being selected;
[0133] Gene crossover calculation is performed according to the following formula:
[0134]
[0135] The gene variation is calculated according to the following formula:
[0136]
[0137] Where P child is the offspring produced, and p1, p2 and p3 are the parents.
[0138] Furthermore, the spatial weighted multi-scale similarity is calculated according to the following formula:
[0139]
[0140]
[0141]
[0142]
[0143]
[0144]
[0145]
[0146]
[0147] In the formula, For single scale The spatial similarity between image X and image Y is: is the weight coefficient for each scale, For image X in the corresponding The mean brightness of the scaled image, For image Y in the corresponding The mean brightness of the scaled image, For image X in the corresponding The standard deviation of image brightness at scale, For image Y in the corresponding The standard deviation of image brightness at scale, and is the stability constant, It is in The spatial weight function on the scale, For The mean gradient of image X on the scale, For The gradient mean of image Y on the scale, abs is the absolute value function, k is the gradient parameter, size_row is the number of row pixels in the image, size_col is the number of column pixels in the image, For The gradient sum of image X on the scale, For images at position The gradient value of is the lateral gradient, is the vertical gradient, Indicates that the image is The gray value at .
[0148] When the embodiment of the present application is implemented, a specific gene parameter update scheme is provided, and the fitness value of the individual is calculated according to the fitness function, and the spatially weighted multi-scale similarity (SWMSS) is used as the fitness function. The optimization problem is a maximum optimization problem, so for individuals, the larger the fitness value, the better, indicating that the group of solutions is closer to the optimal solution, that is, closer to the real X-ray image. In addition, in order to generate new solutions and expand the solution space, crossover and mutation operations must be performed in each iteration. Crossover refers to the exchange of some genes between two individuals to form two new individuals, and mutation refers to the change of some genes of individuals. The algorithm uses adaptive crossover and mutation calculation to accelerate the search of the genetic evolution solution space, and dynamically calculates the crossover rate and mutation rate of each round of algorithm iteration according to the spatially weighted multi-scale similarity (SWMSS) fitness function. At the same time, a multi-parent strategy is adopted to enhance the global search ability of the algorithm and improve the convergence speed of the algorithm. Compared with the traditional dual-parent algorithm, multi-parent effectively improves the diversity of the search algorithm in the solution space by introducing additional parents, preventing it from falling into the local optimal solution. In addition, multiple parents also have stronger local search capabilities, which helps to accelerate the convergence speed of the algorithm and enable it to find qualified solutions faster. Among them, the genetic evolution algorithm uses the spatially weighted multi-scale similarity (SWMSS) fitness function, which combines the advantages of spatial weights and multi-scale analysis. Through spatial weights, this method takes into account the spatial relationship between pixels and uses gradient calculations to more accurately locate edges in images and capture texture information. At the same time, multi-scale analysis enables the algorithm to comprehensively and robustly evaluate image similarity, consider structural information at different scales, and improve the perception of local and global structures. This comprehensive use of spatial weights and multi-scale methods not only improves the sensitivity to complex textures and structures, but also makes the algorithm more adaptable to images with different input features and has better robustness.
[0149] Furthermore, the fitness meets the preset requirements including:
[0150] When there is a second DRR image of the previous round, the spatial weighted multi-scale similarity between the second DRR image of the current round and the second DRR image of the previous round is calculated, and when the spatial weighted multi-scale similarity is greater than a preset value, it is determined that the fitness meets the preset requirement;
[0151] When there is no second DRR image of the previous round, the spatial weighted multi-scale similarity of the second DRR image of the current round and the first DRR image of the current round is calculated, and when the spatial weighted multi-scale similarity is greater than a preset value, it is determined that the fitness meets the preset requirements.
[0152] Based on the same inventive concept, the embodiment of the present application also provides an adaptive digital reconstruction radiological image system based on genetic evolution, comprising:
[0153] An acquisition unit is configured to acquire CT data and construct a voxel matrix corresponding to the CT data;
[0154] An image unit, configured to initialize a transparency function and a color function, and generate the voxel matrix into a first DRR image according to the transparency function and the color function;
[0155] an enhancement unit, configured to initialize genetic evolution parameters and gene parameters, and enhance the first DRR image to form a second DRR image according to filter genes, image fusion genes and histogram equalization genes in the gene parameters; the gene parameters include transparency genes, color genes, filter genes, image fusion genes and histogram equalization genes;
[0156] an updating unit, configured to calculate and update the gene parameters according to the genetic evolution parameters, update the transparency function according to the transparency gene in the updated gene parameters, and update the color function according to the color gene in the updated gene parameters;
[0157] A generating unit is configured to generate a new first DRR image according to the updated transparency function and the color function, and enhance the new first DRR image to form a new second DRR image according to the updated filter gene, image fusion gene and histogram equalization gene in the gene parameters;
[0158] The iterative unit is configured to repeatedly update the genetic parameters, the transparency function and the color function and generate a first DRR image and a second DRR image until the fitness of the second DRR image meets a preset requirement or the number of updates reaches a threshold, and output the second DRR image as the final DRR image.
[0159] Furthermore, the iteration unit is further configured to:
[0160] When there is a second DRR image of the previous round, the spatial weighted multi-scale similarity between the second DRR image of the current round and the second DRR image of the previous round is calculated, and when the spatial weighted multi-scale similarity is greater than a preset value, it is determined that the fitness meets the preset requirement;
[0161] When there is no second DRR image of the previous round, the spatial weighted multi-scale similarity of the second DRR image of the current round and the first DRR image of the current round is calculated, and when the spatial weighted multi-scale similarity is greater than a preset value, it is determined that the fitness meets the preset requirements.
[0162] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.
[0163] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, or it can be an electrical, mechanical or other form of connection.
[0164] The units described as separate components may or may not be physically separated. As units, it is obvious that a person of ordinary skill in the art can realize that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.
[0165] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.
[0166] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a grid device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.
[0167] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.
Claims
1. A method for adaptive digital reconstruction of radiological images based on genetic evolution, characterized in that: include: Acquire CT data and construct a voxel matrix corresponding to the CT data; Initializing a transparency function and a color function, and generating the voxel matrix into a first DRR image according to the transparency function and the color function; Initializing genetic evolution parameters and gene parameters, and enhancing the first DRR image to form a second DRR image according to the filter gene, image fusion gene and histogram equalization gene in the gene parameters; the gene parameters include transparency gene, color gene, filter gene, image fusion gene and histogram equalization gene; Calculate and update the gene parameters according to the genetic evolution parameters, update the transparency function according to the transparency gene in the updated gene parameters, and update the color function according to the color gene in the updated gene parameters; Generate a new first DRR image according to the updated transparency function and the color function, and enhance the new first DRR image according to the updated filter gene, image fusion gene and histogram equalization gene in the gene parameters to form a new second DRR image; Repeatingly updating the gene parameters, the transparency function and the color function to generate a first DRR image and a second DRR image until the fitness of the second DRR image meets a preset requirement or the number of updates reaches a threshold, outputting the second DRR image as a final DRR image; The transparency function adopts the following formula: ; The color function adopts the following formula: ; In the formula, k 1. k 2. k 3 and k 4 is the transparency function coefficient, l 1. l 2. l 3 and l 4 is the color function coefficient, x is the voxel gray value in the voxel matrix; The generation of the first DRR image includes: The color value obtained by observing the following voxel through the current voxel is calculated according to the following formula: ; In the formula, represents the current voxel, Indicates the transparency of the current voxel, represents the color of the current voxel, Indicates the color of the following voxels.
2. The method for adaptive digital reconstruction of radiological images based on genetic evolution according to claim 1, characterized in that: Updating the transparency function according to the updated transparency gene in the gene parameters, and updating the color function according to the updated color gene in the gene parameters include: The transparency function coefficient in the transparency function is updated according to the updated transparency gene in the gene parameter, and the color function coefficient in the color function is updated according to the updated color gene in the gene parameter.
3. The method for adaptive digital reconstruction of radiological images based on genetic evolution according to claim 1, characterized in that: The generation of the second DRR image includes: The first DRR image is edge enhanced by the filter gene to form an image edge image: Image edge = f (F ori ,G filter ) In the formula, f is Marginal function, F ori is the first DRR image, G filter For filtering genes, Image edge is the edge image, when G filter =1 f Using Gaussian operator; when G filter =2, f Using Sobel operator; when G filter =3, f Using Laplace operator; The image edge image and the first DRR image are fused by the image fusion gene to generate a fused image: Image fusion = G fusion ×Image edge +(1-G fusion )×F ori In the formula, Image fusion is the fused image, G fusion For image fusion genes; The fused image is subjected to histogram enhancement processing by the histogram equalization gene to form a second DRR image: Image enhance = f e (Image fusion ,G Hist ) In the formula, f e is the histogram enhancement function, Image enhance is the second DRR image, G Hist is the histogram equalization gene; when G Hist =1, f e The histogram equalization function is turned on; when G Hist =0, Image enhance =Image fusion .
4. The method for adaptive digital reconstruction of radiological images based on genetic evolution according to claim 1, characterized in that: Calculating and updating the gene parameters according to the genetic evolution parameters includes: The dynamic adjustment factor for each iteration is calculated according to the following formula: ; Where SWMSS is the spatially weighted multi-scale similarity, N is the population size, and i is the i-th individual; The next generation gene crossover rate is calculated according to the following formula: ; In the formula, is the lower threshold, is the upper threshold, is the previous generation crossover rate, is the next generation crossover rate, is the generation-by-generation adjustment rate, is the growth coefficient, is the attenuation coefficient; The next generation gene mutation rate is calculated according to the following formula: ; In the formula, is the mutation rate of the previous generation, is the mutation rate of the next generation; The probability of each individual being selected is calculated according to the following formula: ; Where P(i) is the probability of individual i being selected; Gene crossover calculation is performed according to the following formula: ; The gene variation is calculated according to the following formula: ; Where P child is the offspring produced, and p1, p2 and p3 are the parents.
5. The method for adaptive digital reconstruction of radiological images based on genetic evolution according to claim 4, characterized in that: The spatially weighted multi-scale similarity is calculated according to the following formula: ; ; ; ; ; ; ; ; In the formula, For single scale The spatial similarity between image X and image Y is: is the weight coefficient for each scale, For image X in the corresponding The mean brightness of the scaled image, For image Y in the corresponding The mean brightness of the scaled image, For image X in the corresponding The standard deviation of image brightness at scale, For image Y in the corresponding The standard deviation of image brightness at scale, and is the stability constant, It is in The spatial weight function on the scale, For The mean gradient of image X on the scale, For The gradient mean of image Y on the scale, abs is the absolute value function, k is the gradient parameter, size_row is the number of row pixels in the image, size_col is the number of column pixels in the image, For The gradient of image X at the scale, For images at position The gradient value of is the lateral gradient, is the vertical gradient, Indicates that the image is The gray value at .
6. The method for adaptive digital reconstruction of radiological images based on genetic evolution according to claim 1, characterized in that: The fitness meets the preset requirements including: When there is a second DRR image of the previous round, the spatial weighted multi-scale similarity between the second DRR image of the current round and the second DRR image of the previous round is calculated, and when the spatial weighted multi-scale similarity is greater than a preset value, it is determined that the fitness meets the preset requirement; When there is no second DRR image of the previous round, the spatial weighted multi-scale similarity of the second DRR image of the current round and the first DRR image of the current round is calculated, and when the spatial weighted multi-scale similarity is greater than a preset value, it is determined that the fitness meets the preset requirements.
7. An adaptive digital reconstruction radiological imaging system based on genetic evolution using the method according to any one of claims 1 to 6, characterized in that: include: An acquisition unit is configured to acquire CT data and construct a voxel matrix corresponding to the CT data; An image unit, configured to initialize a transparency function and a color function, and generate the voxel matrix into a first DRR image according to the transparency function and the color function; an enhancement unit, configured to initialize genetic evolution parameters and gene parameters, and enhance the first DRR image to form a second DRR image according to filter genes, image fusion genes and histogram equalization genes in the gene parameters; the gene parameters include transparency genes, color genes, filter genes, image fusion genes and histogram equalization genes; an updating unit, configured to calculate and update the gene parameters according to the genetic evolution parameters, update the transparency function according to the transparency gene in the updated gene parameters, and update the color function according to the color gene in the updated gene parameters; A generating unit is configured to generate a new first DRR image according to the updated transparency function and the color function, and enhance the new first DRR image to form a new second DRR image according to the updated filter gene, image fusion gene and histogram equalization gene in the gene parameters; The iterative unit is configured to repeatedly update the genetic parameters, the transparency function and the color function and generate a first DRR image and a second DRR image until the fitness of the second DRR image meets a preset requirement or the number of updates reaches a threshold, and output the second DRR image as the final DRR image.
8. The adaptive digital reconstruction radiological image system based on genetic evolution according to claim 7, characterized in that: The iteration unit is further configured to: When there is a second DRR image of the previous round, the spatial weighted multi-scale similarity between the second DRR image of the current round and the second DRR image of the previous round is calculated, and when the spatial weighted multi-scale similarity is greater than a preset value, it is determined that the fitness meets the preset requirement; When there is no second DRR image of the previous round, the spatial weighted multi-scale similarity of the second DRR image of the current round and the first DRR image of the current round is calculated, and when the spatial weighted multi-scale similarity is greater than a preset value, it is determined that the fitness meets the preset requirements.
