Self-adaptive enhancement method and system for low-contrast region of medical image

Abstract

The invention discloses a medical image low-contrast region adaptive enhancement method and system, and relates to the technical field of medical image processing, and the method comprises the steps: receiving a target medical image set; carrying out deformation measurement, carrying out image registration according to a measurement result, and generating a first registration image and a second registration image; performing multi-level fusion on the first registration image and the second registration image to generate a fused medical image; carrying out local contrast calculation under a sliding window, and identifying and fusing a low-contrast region; and performing multi-feature analysis on the fused low-contrast region to generate a fused enhanced image. According to the invention, the technical problem of low quality and contrast of the medical image caused by difficulty in effectively identifying and enhancing the low-contrast area when the multi-modal medical image is processed in the prior art is solved, and the self-adaptive enhancement of the low-contrast area in the multi-modal medical image is realized. And the quality and the contrast ratio of the medical image are obviously improved.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image processing, and in particular to a method and system for adaptively enhancing low-contrast areas of medical images. Background Art

[0002] In medical diagnosis, technologies such as CT, MRI, and ultrasound imaging are important means of obtaining information about the human body. However, due to the influence of imaging principles, equipment performance, human tissue characteristics, and imaging environment, low-contrast areas often appear in medical images. For example, some tissues in CT have small differences in X-ray absorption, resulting in similar grayscale values, making it difficult to distinguish between structures and lesions; the soft tissue signal intensity of MRI is similar, which increases the difficulty of identification; ultrasound imaging also has poor contrast due to sound wave propagation problems. Existing medical image processing technology has shortcomings when dealing with low-contrast areas. Traditional enhancement methods such as histogram equalization are prone to amplify noise and are not very targeted to local areas. Local enhancement algorithms are difficult to adapt to complex image data of different modalities. When multimodal images are fused, how to identify and enhance low-contrast areas remains a difficult problem.

[0003] The existing technology has the technical problem of difficulty in effectively identifying and enhancing low-contrast areas when processing multimodal medical images, resulting in low quality and contrast of medical images. Summary of the invention

[0004] The present application provides a method and system for adaptively enhancing low-contrast areas of medical images, which is used to solve the technical problem in the prior art that it is difficult to effectively identify and enhance low-contrast areas when processing multimodal medical images, resulting in low quality and contrast of medical images.

[0005] In view of the above problems, the present application provides a method and system for adaptively enhancing low-contrast areas of medical images.

[0006] In a first aspect of the present application, a method for adaptively enhancing a low-contrast region of a medical image is provided, the method comprising: Receive a target medical image set, wherein the target medical image set includes at least a first medical image corresponding to a first modality and a second medical image corresponding to a second modality; perform deformation measurement on the first medical image and the second medical image, perform image registration according to the measurement results, and generate a first registered image and a second registered image, wherein the image registration includes rigid registration or elastic registration; perform multi-level fusion of the first registered image and the second registered image through multi-scale fusion to generate a fused medical image; perform local contrast calculation under a sliding window on the fused medical image to identify a fused low-contrast area; perform multi-feature analysis of grayscale distribution dynamics, grayscale gradient and frequency domain features on the fused low-contrast area, and based on the multi-feature analysis results, call a pre-trained adaptive enhancement module to perform joint enhancement processing of multiple contrast enhancement models to generate a fused enhanced image.

[0007] A second aspect of the present application provides a system for adaptively enhancing low-contrast areas of medical images, the system comprising: A medical image set acquisition module is used to receive a target medical image set, wherein the target medical image set includes at least a first medical image corresponding to a first modality and a second medical image corresponding to a second modality; a registration image generation module is used to perform deformation measurement on the first medical image and the second medical image, perform image registration according to the measurement results, and generate a first registered image and a second registered image, wherein the image registration includes rigid registration or elastic registration; a fused medical image generation module is used to perform multi-level fusion of the first registered image and the second registered image through multi-scale fusion to generate a fused medical image; a fused low-contrast area identification module is used to perform local contrast calculation under a sliding window on the fused medical image to identify a fused low-contrast area; a fused enhanced image generation module is used to perform multi-feature analysis of grayscale distribution dynamics, grayscale gradient and frequency domain features on the fused low-contrast area, and based on the multi-feature analysis results, call a pre-trained adaptive enhancement module to perform joint enhancement processing of multiple contrast enhancement models to generate a fused enhanced image.

[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages: Receive a target medical image set; perform deformation measurement on the first medical image and the second medical image, perform image registration based on the measurement results, and generate a first registered image and a second registered image; perform multi-level fusion of the first registered image and the second registered image through multi-scale fusion to generate a fused medical image; perform local contrast calculation under a sliding window on the fused medical image to identify fused low-contrast areas; perform multi-feature analysis on the fused low-contrast areas, and based on the multi-feature analysis results, call a pre-trained adaptive enhancement module to perform joint enhancement processing of multiple contrast enhancement models to generate a fused enhanced image. The method achieves the technical effect of realizing adaptive enhancement of low-contrast areas in multimodal medical images and significantly improving the quality and contrast of medical images. 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 A schematic diagram of a process flow of a method for adaptively enhancing low-contrast areas of medical images provided in an embodiment of the present application; Figure 2 A schematic diagram of the structure of a system for adaptively enhancing low-contrast areas of medical images provided in an embodiment of the present application.

[0011] Explanation of the reference numerals: medical image set acquisition module 10 , registration image generation module 20 , fusion medical image generation module 30 , fusion low-contrast region identification module 40 , fusion enhanced image generation module 50 . DETAILED DESCRIPTION

[0012] The present application provides a method and system for adaptively enhancing low-contrast areas of medical images, which is used to solve the technical problem in the prior art that it is difficult to effectively identify and enhance low-contrast areas when processing multimodal medical images, resulting in low quality and contrast of medical images.

[0013] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0014] Embodiment 1, as Figure 1As shown, the present application provides a method for adaptively enhancing low-contrast areas of medical images, the method comprising: Step S100: receiving a target medical image set, wherein the target medical image set includes at least a first medical image corresponding to a first modality and a second medical image corresponding to a second modality.

[0015] Specifically, the received target medical image set includes at least two medical images of different modalities, namely, a first medical image corresponding to a first modality and a second medical image corresponding to a second modality, such as a first medical image under a CT modality and a second medical image under a PET modality. These images are from the same user at the same location, and they carry different types of medical information. For example, CT images are good at presenting anatomical structures, while PET images highlight functional and metabolically active areas. The differences and complementarities between the two provide a rich data foundation for subsequent processing. These images are the core data source for subsequent operations such as image registration, fusion, and low-contrast area enhancement.

[0016] Step S200: performing deformation measurement on the first medical image and the second medical image, and performing image registration according to the measurement results to generate a first registered image and a second registered image, wherein the image registration includes rigid registration or elastic registration.

[0017] Specifically, for the first medical image and the second medical image, deformation measurement must be performed first. The specific method is to calculate the mutual information coefficient between the two, which can reflect the similarity and information correlation between the two images. When calculating the mutual information coefficient, the grayscale histograms of the first and second medical images are first counted to obtain the frequency of each grayscale value; then the joint grayscale histogram of the two images is counted to determine the frequency of simultaneous occurrence of specific grayscale value pairs; finally, according to the formula, the joint grayscale histogram frequency is divided by the product of the grayscale histogram frequencies of the two images, and the logarithm is taken and then accumulated to calculate the mutual information coefficient. The larger the coefficient, the higher the similarity of the two images. Based on this mutual information coefficient, deformation degree analysis is performed. The appropriate registration method is screened by comparing it with the pre-set mutual information coefficient threshold obtained by training historical rigid registration image samples. If the mutual information coefficient meets the rigid registration deformation condition constraint, the rigid registration is selected, that is, the initial translation alignment is performed using the center of gravity of the first and second medical images, and then the translation and rotation parameters are continuously adjusted with the help of the gradient descent method to maximize the mutual information between the two images; if it does not meet the condition, the elastic registration is used, and the images are first preliminarily aligned using the rigid registration, and then the deformation field model is constructed based on B-splines, and the deformation field is calculated using the gradient descent method to minimize the difference between the first medical image and the second medical image. After the above rigid registration or elastic registration process, the first registered image and the second registered image are finally generated, which lays the foundation for the subsequent multi-scale fusion and enhancement of the low-contrast area of ​​the medical image, ensuring that the information in the images of different modalities can be accurately corresponded and integrated.

[0018] Step S300: performing multi-level fusion on the first registered image and the second registered image through multi-scale fusion to generate a fused medical image.

[0019] Specifically, the two registered images are fused at multiple levels by multi-scale fusion technology to generate a fused medical image. First, the first registered image and the second registered image are pyramid decomposed respectively to construct a Gaussian pyramid, which decomposes the image into multiple levels of different resolutions, each of which contains different degrees of detailed information of the image. Then, at each resolution level of the Gaussian pyramid, the maximum value of the pixel points is selected, and the more representative pixel values ​​are selected from the pixels of the corresponding levels of the two images to generate a fused image corresponding to each resolution level, which preliminarily integrates the features of the two images. Then, starting from the bottom fused image, the fusion results of each layer of the image are synthesized with the previous level image layer by layer to construct a Laplacian pyramid image, which further refines and optimizes the fusion effect. Finally, based on the Laplacian pyramid image, all the details of the low-resolution layers are gradually superimposed on the image of the highest resolution layer, and the information at different resolutions is completely combined to reconstruct a fused medical image, realizing the comprehensive integration of medical image information of different modalities, and providing richer and more accurate image data for subsequent analysis and processing.

[0020] Step S400: performing local contrast calculation under a sliding window on the fused medical image to identify fused low-contrast areas.

[0021] Specifically, in order to identify the low-contrast area in the fused medical image, a sliding window method is used to calculate the local contrast. First, a sliding window of preset size is constructed. The size of the window is pre-set based on a large amount of experimental data and the characteristics of the medical image to ensure that the low-contrast area can be effectively detected. Then, let this sliding window slide pixel by pixel on the fused medical image. At each sliding position, the contrast index of the pixels in the window is calculated. It is calculated based on factors such as the gray value difference of the pixels in the window, and it can quantify the contrast of the image area in the window. Finally, the calculated pixel contrast index is compared with the preset contrast index, and the image blocks corresponding to the sliding window whose pixel contrast index is less than or equal to the preset contrast index are extracted. These extracted image blocks constitute the fused low-contrast area, laying the foundation for the subsequent targeted enhancement of the contrast of these areas.

[0022] Step S500: Perform multi-feature analysis of grayscale distribution dynamics, grayscale gradient and frequency domain features on the fused low-contrast area. Based on the multi-feature analysis results, call a pre-trained adaptive enhancement module to perform joint enhancement processing of multiple contrast enhancement models to generate a fused enhanced image.

[0023] Specifically, multi-feature analysis of grayscale distribution dynamics, grayscale gradient and frequency domain features is carried out on these areas. Grayscale distribution dynamic analysis can understand the distribution and change law of pixel grayscale values, grayscale gradient analysis can determine the intensity and direction of grayscale changes in the image, and frequency domain feature analysis helps to grasp the characteristics of the image under different frequency components. Based on historical data statistics, a set of preset low-contrast factors is determined in advance, and multiple groups of contrast enhancement samples including original medical image samples, enhanced medical image samples and corresponding multi-feature analysis samples are collected. Multiple contrast enhancement models are constructed and combined into a model network. Through these samples, the model network is trained for block fusion enhancement to obtain a pre-trained adaptive enhancement module. Based on the results of multi-feature analysis, this module is called to allow multiple contrast enhancement models to work together to perform joint enhancement processing on the fused low-contrast area. These models improve the display effect of low-contrast areas from different angles according to their respective characteristics, and finally generate fused enhanced images to improve the overall visibility of the image.

[0024] In a possible implementation, step S100 further includes: Step S110: the first modality and the second modality are different medical imaging modalities, and the first medical image and the second medical image are medical images of the same user at the same location under different medical imaging modalities.

[0025] Specifically, the first modality and the second modality must be different medical imaging modalities, such as the common combination of CT (computed tomography) and PET (positron emission tomography). CT can clearly present the details of the human body's anatomical structure, while PET is good at highlighting functional and metabolically active areas; there are also CT and MRI (magnetic resonance imaging). MRI has unique advantages in soft tissue imaging and can provide complementary information to CT. The first medical image and the second medical image are both from the same user at the same location. This condition ensures that the images reflect the same body part or area of ​​interest. Imaging the same part using different modalities can obtain medical information from multiple angles, providing rich and comprehensive data support for subsequent image fusion, low-contrast area enhancement, and accurate diagnosis of diseases.

[0026] In a possible implementation, step S200 further includes: Step S210: Calculate mutual information between the first medical image and the second medical image to generate a mutual information coefficient.

[0027] Step S220: performing deformation degree analysis based on the mutual information coefficient, screening rigid registration and elastic registration according to the analysis result, and determining a target registration scheme.

[0028] Step S230: registering the first medical image and the second medical image based on the target registration scheme to generate the first registered image and the second registered image.

[0029] Specifically, the first medical image and the second medical image are digitized and discretized to convert them into digital images based on a pixel matrix. For the first medical image, its pixel matrix is ​​traversed to count the frequency n of each gray value i appearing i , and combined with the total number of pixels N in the image, calculate the probability distribution of each gray value , thus constructing the grayscale histogram of the first medical image; similarly, perform the same operation on the second medical image to obtain its grayscale value probability distribution q(j). Then, synchronously traverse the pixel matrices of the two images and count the frequency n of the occurrence of pixel pairs with grayscale values ​​i in the first medical image and j in the second medical image. ij , and calculate its joint probability distribution , and then construct a joint grayscale histogram. Using the mutual information calculation formula (in Represent the first and second medical images respectively, are the gray levels of the first and second medical images respectively), and the mutual information coefficient is calculated.

[0030] Based on the generated mutual information coefficient, the deformation degree is analyzed and the rigid registration deformation condition constraint is constructed. This constraint is the mutual information coefficient threshold obtained by training the historical rigid registration image samples. The calculated mutual information coefficient is compared with the threshold. If the mutual information coefficient meets the rigid registration deformation condition constraint, it means that the deformation degree of the two images is relatively small and rigid registration is suitable; if it does not meet the rigid registration deformation condition constraint, it means that the image deformation is relatively complex and elastic registration is required. The target registration scheme is determined to ensure the accuracy and effectiveness of subsequent registration operations.

[0031] The first medical image and the second medical image are registered according to the determined target registration scheme. If rigid registration is selected, the center of gravity of the image will be used for initial translation alignment, and then the translation and rotation parameters will be adjusted with the help of the gradient descent method to maximize the mutual information between the two images; if elastic registration is used, rigid registration will be used for preliminary alignment first, and then a deformation field model will be constructed based on B-splines, and the deformation field will be calculated by the gradient descent method to minimize the difference between the images. After this registration process, the first registered image and the second registered image are successfully generated, laying a solid foundation for the subsequent multi-scale fusion and enhancement of low-contrast areas of medical images.

[0032] In a possible implementation, step S220 further includes: Step S221: constructing a rigid registration deformation condition constraint, wherein the rigid registration deformation condition constraint is a mutual information coefficient threshold trained by historical rigid registration image samples.

[0033] Step S222: Determine whether the mutual information coefficient satisfies the rigid registration deformation condition constraint.

[0034] Step S223: If yes, generate the target registration solution by rigid registration.

[0035] Step S224: If not, generate the target registration solution by elastic registration.

[0036] Specifically, in order to construct the rigid registration deformation condition constraint, the support vector machine (SVM) algorithm is used to determine the mutual information coefficient threshold based on the historical rigid registration image samples. First, a large number of historical rigid registration image samples in different scenes are collected, and the mutual information coefficient is calculated for each sample. These samples are divided into two categories: suitable for rigid registration and unsuitable for rigid registration, and are assigned labels respectively. Then the mutual information coefficient is used as a feature and the sample label is input into the SVM model as the target variable. SVM will find an optimal hyperplane in the feature space to separate the two types of samples as clearly as possible. During the training process, by adjusting the parameters of the model, such as the penalty factor and the kernel function parameters, the model performance is evaluated using methods such as cross-validation, and the model is continuously optimized to achieve the best classification effect. After the training is completed, a specific mutual information coefficient value can be determined at or near the intersection of the hyperplane and the feature axis, which is used as the threshold of the rigid registration deformation condition constraint. Subsequent new image registration tasks can use this threshold to determine whether rigid registration is suitable.

[0037] The mutual information coefficient of the two images is accurately calculated, which reflects the correlation and overlap of the information between the two images. After that, the calculated mutual information coefficient is compared with the mutual information coefficient threshold in the previously constructed rigid registration deformation condition constraint. If the calculated mutual information coefficient is greater than or equal to the threshold, it means that the two images meet the rigid registration deformation condition constraint under the current conditions, which means that the feature correspondence between them is relatively stable, and it is more likely to achieve accurate registration by using rigid registration; on the contrary, if the calculated mutual information coefficient is less than the threshold, it means that the rigid registration deformation condition constraint is not met. At this time, the feature correspondence between the images is more complex, and rigid registration may not achieve the ideal registration effect. It is necessary to consider using other more flexible registration methods such as elastic registration.

[0038] If the mutual information coefficient satisfies the rigid registration deformation condition constraint, that is, the calculated mutual information coefficient is greater than or equal to the pre-constructed mutual information coefficient threshold, it indicates that the feature correspondence between the first medical image and the second medical image is relatively stable and suitable for rigid registration. At this time, the target registration scheme is generated by rigid registration. The rigid registration process first uses the center of gravity of the first medical image and the second medical image to initialize the translation alignment to ensure that the two are initially matched in approximate positions. Then, with the help of the gradient descent method, the translation and rotation parameters are continuously adjusted. In this process, the mutual information between the images is continuously calculated, and the registration parameters are gradually optimized with the goal of maximizing the mutual information until the combination of translation and rotation parameters that maximizes the mutual information between the two images is found, thereby achieving precise alignment of the images in space and completing the generation of the target registration scheme.

[0039] If the mutual information coefficient does not meet the rigid registration deformation condition constraint, that is, the calculated mutual information coefficient is less than the pre-constructed mutual information coefficient threshold, it means that the feature correspondence between the images is relatively complex, and the rigid registration may not achieve the ideal registration effect. At this time, it is necessary to generate the target registration scheme by elastic registration. However, before performing elastic registration, rigid registration must be performed first. First, the first medical image and the second medical image are preliminarily aligned using rigid registration, and the two images are made close in approximate position through translation and rotation operations. After that, a deformation field model is constructed based on B-splines, which can describe the local deformation of the image. The deformation field is then calculated by the gradient descent method, and the parameters of the deformation field are continuously adjusted during the calculation process to minimize the difference between the two images, thereby achieving a more refined image registration and finally generating a target registration scheme.

[0040] In a possible implementation, step S223 further includes: Step S2231: The rigid registration uses the center of gravity of the first medical image and the second medical image to initialize the translation alignment, and uses the gradient descent method to maximize the mutual information between the first medical image and the second medical image by adjusting the translation and rotation parameters.

[0041] Step S2232: The elastic registration uses rigid registration to preliminarily align the first medical image and the second medical image, constructs a deformation field model based on B-splines, and calculates the deformation field by gradient descent method to minimize the difference between the first medical image and the second medical image.

[0042] Specifically, when performing rigid registration to generate a target registration scheme, the first step is to calculate the center of gravity of the first medical image and the second medical image. By analyzing the pixel distribution and other information of the two images respectively, their respective center of gravity positions are determined. Then, based on the two center of gravity positions, the images are initialized for translation operation, so that the first medical image and the second medical image are preliminarily aligned in space, so that the positional relationship between the two is closer. After that, the gradient descent method is used to further optimize the registration effect. The gradient descent method will continuously adjust the translation and rotation parameters of the image, and after each parameter adjustment, the mutual information between the first medical image and the second medical image is calculated. The mutual information can reflect the correlation and information overlap between the two images. With the goal of maximizing the mutual information, the parameters are continuously iterated and adjusted until a set of translation and rotation parameters is found so that the mutual information between the two images reaches the maximum value. At this time, the rigid registration process is completed, and a more accurate registration result is obtained.

[0043] When performing elastic registration, the first medical image and the second medical image are first preliminarily aligned using rigid registration (including translation and rotation operations). The purpose of this step is to reduce the amount of calculation for subsequent non-rigid registration, so that the two images are close in overall position first, laying the foundation for subsequent more refined registration. After completing the preliminary alignment, the deformation field model is constructed based on B-splines. B-splines (basis spline function) is a commonly used mathematical model that describes the local deformation of an image through a set of control points. These control points can be flexibly distributed in the image as needed. Each control point has a corresponding weight, which together determine the deformation mode of the local area of ​​the image. B-splines has good smoothness and local controllability, and can adapt well to complex shape changes in medical images. For example, in medical images, the shape of organs may be locally distorted and deformed due to physiological conditions, diseases and other factors. The B-splines deformation field model can accurately capture and describe these changes. After the deformation field model is constructed, the deformation field is calculated by the gradient descent method. Gradient descent is an optimization algorithm that measures the similarity between the first medical image and the second medical image with an energy function, such as mean square error, mutual information, structural similarity index, etc. During the calculation process, the parameters of the deformation field are continuously adjusted, and the parameters are updated along the direction that makes the energy function value decrease the fastest (i.e., the opposite direction of the gradient), so that the difference between the two images is gradually minimized. In each iteration, a new deformation field is calculated based on the current deformation field parameters. This deformation field is applied to the target image to gradually align it with the reference image. In order to smoothly apply the deformation field to the image, interpolation methods such as cubic interpolation or B-spline interpolation are usually used. Through interpolation, continuous deformation can be generated between control points to ensure the smoothness and naturalness of image deformation. In the entire elastic registration process, multiple iterations and optimizations are required to gradually adjust the deformation field until the error of alignment between the first medical image and the second medical image reaches the minimum, thereby achieving accurate elastic registration.

[0044] In a possible implementation, step S300 further includes: Step S310: performing pyramid decomposition on the first registered image and the second registered image respectively to construct a Gaussian pyramid.

[0045] Step S320: performing maximum value selection on the pixel points of each resolution level in the Gaussian pyramid to generate a fused image corresponding to each resolution level.

[0046] Step S330: Based on the fused images corresponding to each resolution level, starting from the fused image of the bottom layer, the fusion results of each layer of images are synthesized with the image of the previous level layer layer by layer to generate a high-resolution image, until the highest resolution layer is reached, and a Laplacian pyramid image is constructed.

[0047] Step S340: Based on the Laplacian pyramid image, all the details of the low-resolution layers are gradually superimposed on the image of the highest-resolution layer to reconstruct the fused medical image.

[0048] Specifically, pyramid decomposition is performed on the first registered image and the second registered image respectively to construct a Gaussian pyramid. Gaussian pyramid is an image pyramid obtained by Gaussian filtering and downsampling, which can present image information of different scales. In specific operations, for the first registered image, a Gaussian filter is first applied to it, which performs weighted averaging on each pixel and its neighborhood in the image, plays a role in smoothing the image and reducing noise interference in the image. Then, the image after Gaussian filtering is downsampled, usually by reducing the size of the image by half, such as from the original width and height of 2n pixels to width and height of n pixels, thereby obtaining a first-level lower resolution image, which retains the global structure of the first registered image. Then, the above Gaussian filtering and downsampling operations are repeated on the image with lower resolution at this level, and an image with lower resolution at the next level is obtained, and this is continuously iterated until the preset number of pyramid layers is reached. For the second registered image, the same method is used for processing, and finally a complete Gaussian pyramid is constructed for the first registered image and the second registered image, respectively, laying the foundation for subsequent image fusion operations.

[0049] For each resolution level image in the Gaussian pyramid, a pixel-based fusion operation will be performed. For the first registered image and the second registered image at the same resolution level, the pixels at their corresponding positions are compared. At each pixel position, the pixel value of the first registered image at that position is taken out, and the pixel value of the second registered image at the same position is also taken out, and then the two pixel values ​​are compared. The pixel value with the larger value is selected as the pixel value of the position after fusion. In this way, by performing the above maximum value selection operation on all pixels at the same resolution level, a new image can be obtained, and this new image is the fused image corresponding to the resolution level. For each resolution level of the Gaussian pyramid, the above maximum value selection operation on the pixels is repeated, and finally a fused image corresponding to each resolution level is generated. These fused images retain the more significant feature information in the first registered image and the second registered image as much as possible at their respective resolution levels, preparing for the subsequent further synthesis of high-resolution fused images.

[0050] After obtaining the fused image corresponding to each resolution level, the bottom layer refers to the layer with the highest resolution in the Gaussian pyramid, and the fused image of this layer is used as the starting point. The construction of the Laplacian pyramid is to better preserve the detailed information of the image so as to eventually reconstruct a high-quality fused medical image. The fused image of the bottom layer (highest resolution) is used as the basis. Then, the fused image of the previous level (slightly lower resolution) is taken, and the image of this level is upsampled, that is, the size of the image is enlarged so that its resolution is close to that of the bottom image. Upsampling usually makes the image blurry, and at this time, it is necessary to combine the information of the bottom image to restore the details. By performing a difference operation between the upsampled image and the bottom image, the detailed information between the image of this layer and the bottom image is obtained, that is, a layer of the Laplacian image. Then, the obtained Laplacian image of this layer is synthesized with the bottom image to obtain a new image. After that, the fused image of the next level (lower resolution) is taken, and the above-mentioned upsampling, difference calculation, and synthesis operations are repeated. The fusion results of each layer of the image are continuously synthesized with the image of the previous level to gradually restore the details of the image. As the image information of the low-resolution layer is processed upwards layer by layer, it is continuously fused with the image information of the high-resolution layer until the highest resolution layer is reached (that is, the top layer of the Gaussian pyramid, the layer with the lowest resolution). Through this layer-by-layer synthesis process, the Laplacian pyramid image is finally constructed. Each layer of the Laplacian pyramid contains detailed information of images at different resolutions, providing an important data foundation for the subsequent reconstruction of fused medical images based on these detailed information.

[0051] The Laplacian pyramid image contains image detail information at different resolution levels, which is the key to reconstructing the fused medical image. At this time, the image of the highest resolution layer in the Laplacian pyramid is used as the basis, because the image of this layer retains the finest part of the fused image. Starting from the next lower resolution layer next to it, the detail information contained in the Laplacian pyramid image of this layer is gradually superimposed on the image of the highest resolution layer. These detail information represents the difference between the images of adjacent resolution levels. Through superposition, the information lost due to downsampling is gradually restored, making the image richer and more complete. After superimposing the details of each lower resolution layer, continue to perform the same operation on the next lower resolution layer, and so on, until all the details of the low resolution layers are superimposed on the image of the highest resolution layer. In this process, the details of the image are continuously enriched, and the features such as texture and edge are gradually clear, and finally a fused medical image is reconstructed. The fused medical image integrates the features of the first registered image and the second registered image at different resolutions, providing a comprehensive and clear image basis for medical diagnosis and analysis.

[0052] In a possible implementation, step S500 further includes: Step S510: determining a preset low contrast factor set, wherein the preset low contrast factor set is based on historical data statistics.

[0053] Step S520: collecting multiple groups of contrast-enhanced samples based on the preset low-contrast factor set, wherein any group of contrast-enhanced samples includes original medical image samples and enhanced medical image samples, and corresponding multi-feature analysis samples.

[0054] Step S530: construct multiple model networks of the multiple contrast enhancement models, perform block fusion enhancement training on the multiple model networks with the multiple groups of contrast enhancement samples, generate the multiple contrast enhancement models, and construct the adaptive enhancement module.

[0055] Step S540: calling the adaptive enhancement module, and the multiple contrast enhancement models perform enhancement processing on the fused low-contrast area in the fused medical image based on the multi-feature analysis results to generate the fused enhanced image.

[0056] Specifically, a preset low-contrast factor set is determined. In order to accurately identify and process low-contrast areas in medical images, statistical analysis based on a large amount of historical medical image data is required. By sorting out and summarizing various relevant factors that appear in low-contrast areas in these historical data, such as parameter differences of different imaging devices, different physiological states of patients, and the impact of disease types on image contrast, a set of factors that may cause low contrast is determined. This set is the preset low-contrast factor set.

[0057] Multiple sets of contrast enhancement samples are collected based on a preset set of low-contrast factors. In actual operation, qualified samples are screened out from a large number of medical images according to different factors in the set. Each set of contrast enhancement samples includes original medical image samples, enhanced medical image samples after professional enhancement processing, and corresponding multi-feature analysis samples obtained after multi-feature analysis of the original medical image samples such as grayscale distribution dynamics, grayscale gradient and frequency domain features. These samples provide a rich data foundation for the subsequent training of contrast enhancement models.

[0058] Based on four existing models, multiple model networks of multiple contrast enhancement models are constructed, specifically covering local enhancement based on CLAHE, Retinex multi-scale enhancement (using the multi-scale Retinex (MSR) method to enhance local contrast), generative adversarial network (GAN) enhancement, and deep learning super-resolution reconstruction. First, according to the principles and structural characteristics of these models, the corresponding model networks are built respectively. For the local enhancement model network based on CLAHE, it can effectively improve the contrast of the local area of ​​the image; the Retinex multi-scale enhancement model network uses the MSR method to enhance the local contrast at multiple scales; the generative adversarial network (GAN) enhancement model network uses the adversarial training mechanism of the generator and the discriminator to optimize the image contrast; the deep learning super-resolution reconstruction model network focuses on improving the resolution of the image to improve the contrast. Afterwards, the model networks are trained for block fusion enhancement using multiple sets of contrast enhancement samples collected. The samples are input into each model network in blocks, and the parameters of the model are continuously adjusted during the training process. At the same time, the fusion strategy between the model networks is considered to make them cooperate with each other and complement each other's advantages. After multiple rounds of training and optimization, multiple contrast enhancement models with excellent performance are generated. Finally, these models are integrated together to construct an adaptive enhancement module for subsequent enhancement processing of low-contrast areas in fused medical images.

[0059] The adaptive enhancement module is called to enhance the fused low-contrast area in the fused medical image. After obtaining the fused medical image, a multi-feature analysis is first performed on it to obtain information such as the grayscale distribution dynamics, grayscale gradient, and frequency domain characteristics of the low-contrast area. These multi-feature analysis results are input into the adaptive enhancement module. The multiple contrast enhancement models in the module will collaboratively enhance the fused low-contrast area based on their learned knowledge and methods. Each model optimizes the low-contrast area from a different angle. Some models may focus on enhancing edge contrast, while others may be better at adjusting grayscale distribution. Through their combined action, a fused enhanced image is finally generated, which significantly improves the low-contrast area in the fused medical image.

[0060] In a possible implementation, step S530 further includes: Step S531: Using the original medical image samples and multi-feature analysis samples in any group of contrast enhanced samples as training inputs, and using the enhanced medical image samples as output supervision truth values, iteratively adjust the training parameters of the multiple model networks, wherein the training parameters include learning parameters corresponding to the multiple model networks, respectively, and enhanced fusion weights of the multiple model networks, and train the multiple model networks until convergence to generate the adaptive enhancement module.

[0061] Specifically, in order to construct the adaptive enhancement module, it is necessary to iteratively adjust the training parameters of multiple model networks. Select any set of contrast-enhanced samples, and use the original medical image samples and multi-feature analysis samples as training inputs, while the enhanced medical image samples are used as output supervisory truth values. During the training process, for multiple model networks, the training parameters that need to be adjusted include two parts: one is the learning parameters corresponding to each model network, which control the model's ability to capture and understand the input data features during the learning process; the other is the enhanced fusion weights of multiple model networks, which determine the size of the role played by each model network in the final fusion result. After the training starts, multiple model networks process the input data according to the current training parameters and generate their own output results. These output results are compared with the enhanced medical image samples as supervisory truth values, and the loss function value is calculated to measure the difference between the model output and the true result. Subsequently, based on the loss function value, the optimization algorithm (stochastic gradient descent method) is used to adjust the training parameters. This process will be repeated continuously, and each iteration will bring the model network closer to the optimal state. As the iterations continue, the loss function value will gradually decrease. When the loss function value decreases to a minimum and no longer changes significantly in subsequent iterations, it indicates that multiple model networks have been trained to converge. At this point, the learning parameters and enhanced fusion weights of each model network have reached a relatively stable and optimal state. Finally, these trained model networks are integrated together to generate an adaptive enhancement module, which can effectively enhance the low-contrast areas in the image based on the input medical image and its multi-feature analysis results.

[0062] In a possible implementation, step S400 further includes: Step S410: construct a preset sliding window.

[0063] Step S420: Slide the preset sliding window on the fused medical image to calculate the pixel contrast index under each sliding window.

[0064] Step S430: extracting the image block corresponding to the sliding window with a pixel contrast index or a preset contrast index to generate the fused low-contrast area.

[0065] Specifically, a preset sliding window is constructed. When constructing, the characteristics of medical images and the needs of subsequent calculations should be fully considered. First, the shape of the window is determined. The rectangular window is a common choice due to its regularity and computational convenience. It can easily scan the image row by row and column by column. Then the window size is set, which requires a comprehensive balance of many factors. If the window size is too small, although the calculation speed is fast, it may not cover enough pixel information to accurately reflect the contrast characteristics of the local area, and it is easy to miss the low-contrast area; if the window size is too large, although more information can be obtained, the amount of calculation will increase significantly, and the surrounding normal contrast area may also be included in the judgment range of the low-contrast area, resulting in misjudgment. Generally speaking, the size of common tissues and lesions in medical images is referred to, combined with multiple experiments and experience to determine the appropriate side length or radius. For example, for lung CT fusion images, after a lot of image analysis and testing, a rectangular window with a side length of 10-20 pixels may be selected. In addition, appropriate adjustments can be made based on the resolution of the image. A slightly larger window can be used for high-resolution images, while a smaller window can be used for low-resolution images to ensure that the window can effectively cover the target area without losing key information, thereby constructing a preset sliding window that meets the subsequent low-contrast area recognition needs.

[0066] After the preset sliding window is constructed, the preset sliding window is slid on the fused medical image. The sliding window method is used here. The window moves on the image pixel by pixel or in a certain step size to cover different areas of the image. For each area covered by the sliding window, its pixel contrast index needs to be calculated. The pixel contrast index here is calculated using the Local Contrast Measurement (LCM). According to the formula ,in Represents the pixel value in the local window, is a decimal, used to avoid the situation where the denominator is zero. Through this formula, the LCM value under each sliding window is calculated. This LCM value reflects the degree of difference in pixel values ​​within the sliding window, that is, the contrast. A larger LCM value indicates that the pixel contrast in the window is higher, and the image details and features are relatively obvious; while a smaller LCM value indicates that the pixel contrast in the window is lower, and there may be a low-contrast area. Subsequently, this LCM value can be compared with the preset threshold (LCM<0.1) to determine whether the area corresponding to the sliding window is a low-contrast area, providing data support for the subsequent accurate extraction of low-contrast areas.

[0067] Compare the calculated pixel contrast index with the preset contrast index (LCM<0.1). For those sliding windows whose pixel contrast index is less than or equal to the preset contrast index, extract the corresponding image blocks. These extracted image blocks show relatively low contrast characteristics in the fused medical image. By integrating all the image blocks that meet the conditions, a fused low-contrast area is generated. This area contains those parts of the image with insufficient contrast and unclear details, which provides clear goals and scope for subsequent targeted enhancement processing of these areas (such as enhancement through adaptive enhancement modules), which helps to improve the overall quality of medical images.

[0068] Embodiment 2 is based on the same inventive concept as the method for adaptively enhancing low-contrast areas of medical images in the aforementioned embodiment. Figure 2 As shown, the present application provides a system for adaptively enhancing low-contrast areas of medical images, and the system and method embodiments in the embodiments of the present application are based on the same inventive concept. The system includes: The medical image set acquisition module 10 is used to receive a target medical image set, wherein the target medical image set at least includes a first medical image corresponding to a first modality and a second medical image corresponding to a second modality.

[0069] The registration image generation module 20 is used to perform deformation measurement on the first medical image and the second medical image, perform image registration according to the measurement results, and generate a first registered image and a second registered image, wherein the image registration includes rigid registration or elastic registration.

[0070] The fused medical image generating module 30 is used to perform multi-level fusion on the first registered image and the second registered image through multi-scale fusion to generate a fused medical image.

[0071] The fused low-contrast region identification module 40 is used to perform local contrast calculation under a sliding window on the fused medical image to identify the fused low-contrast region.

[0072] The fused enhanced image generation module 50 is used to perform multi-feature analysis of the grayscale distribution dynamics, grayscale gradient and frequency domain characteristics of the fused low-contrast area. Based on the multi-feature analysis results, the pre-trained adaptive enhancement module is called to perform joint enhancement processing of multiple contrast enhancement models to generate a fused enhanced image.

[0073] Furthermore, the system is also used to implement the following functions: The first modality and the second modality are different medical imaging modalities, and the first medical image and the second medical image are medical images of the same user at the same location under different medical imaging modalities.

[0074] Furthermore, the system is also used to implement the following functions: Calculate the mutual information of the first medical image and the second medical image to generate a mutual information coefficient; perform deformation degree analysis based on the mutual information coefficient, screen rigid registration and elastic registration according to the analysis results, and determine a target registration scheme; register the first medical image and the second medical image based on the target registration scheme to generate the first registered image and the second registered image.

[0075] Furthermore, the system is also used to implement the following functions: Construct a rigid registration deformation condition constraint, wherein the rigid registration deformation condition constraint is a mutual information coefficient threshold trained by historical rigid registration image samples; determine whether the mutual information coefficient satisfies the rigid registration deformation condition constraint; if so, generate the target registration scheme by rigid registration; if not, generate the target registration scheme by elastic registration.

[0076] Furthermore, the system is also used to implement the following functions: The rigid registration uses the center of gravity of the first medical image and the second medical image to initialize the translation alignment, and uses the gradient descent method to maximize the mutual information between the first medical image and the second medical image by adjusting the translation and rotation parameters; the elastic registration uses the rigid registration to first perform a preliminary alignment on the first medical image and the second medical image, constructs a deformation field model based on B-splines, and calculates the deformation field by the gradient descent method, so that the difference between the first medical image and the second medical image is minimized.

[0077] Furthermore, the system is also used to implement the following functions: The first registered image and the second registered image are respectively subjected to pyramid decomposition to construct a Gaussian pyramid; the maximum value of the pixel points at each resolution level in the Gaussian pyramid is selected to generate a fused image corresponding to each resolution level; based on the fused image corresponding to each resolution level, starting from the bottom fused image, the fusion result of each layer of image is synthesized with the previous level image layer by layer to generate a high-resolution image until the highest resolution layer is reached to construct a Laplacian pyramid image; based on the Laplacian pyramid image, all the details of the low-resolution layers are gradually superimposed on the image of the highest resolution layer to reconstruct the fused medical image.

[0078] Furthermore, the system is also used to implement the following functions: Determine a preset low-contrast factor set, wherein the preset low-contrast factor set is based on historical data statistics; collect multiple groups of contrast enhancement samples based on the preset low-contrast factor set, wherein any group of contrast enhancement samples includes original medical image samples and enhanced medical image samples, and corresponding multi-feature analysis samples; construct multiple model networks of the multiple contrast enhancement models, perform block fusion enhancement training on the multiple model networks with the multiple groups of contrast enhancement samples, generate the multiple contrast enhancement models, and construct the adaptive enhancement module; call the adaptive enhancement module, and the multiple contrast enhancement models perform enhancement processing on the fused low-contrast area in the fused medical image based on the multi-feature analysis results to generate the fused enhanced image.

[0079] Furthermore, the system is also used to implement the following functions: The original medical image samples and multi-feature analysis samples in any group of contrast enhanced samples are used as training inputs, and the enhanced medical image samples are used as output supervision truth values ​​to iteratively adjust the training parameters of the multiple model networks, wherein the training parameters include learning parameters corresponding to the multiple model networks respectively, and enhanced fusion weights of the multiple model networks. The multiple model networks are trained until convergence to generate the adaptive enhancement module.

[0080] Furthermore, the system is also used to implement the following functions: Construct a preset sliding window; slide the preset sliding window on the fused medical image to calculate the pixel contrast index under each sliding window; extract the image block corresponding to the sliding window with a pixel contrast index or a pixel contrast index equal to the preset contrast index to generate the fused low-contrast area.

[0081] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0082] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.

[0083] This specification and drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A method for adaptively enhancing low-contrast areas of medical images, characterized in that: include: Receiving a target medical image set, wherein the target medical image set includes at least a first medical image corresponding to a first modality and a second medical image corresponding to a second modality; Performing deformation measurement on the first medical image and the second medical image, and performing image registration according to the measurement results to generate a first registered image and a second registered image, wherein the image registration includes rigid registration or elastic registration; Performing multi-level fusion on the first registered image and the second registered image through multi-scale fusion to generate a fused medical image; Performing local contrast calculation under a sliding window on the fused medical image to identify a fused low-contrast area; A multi-feature analysis of grayscale distribution dynamics, grayscale gradient and frequency domain features is performed on the fused low-contrast area. Based on the multi-feature analysis results, a pre-trained adaptive enhancement module is called to perform joint enhancement processing of multiple contrast enhancement models to generate a fused enhanced image.

2. The method for adaptively enhancing low-contrast areas of medical images according to claim 1, characterized in that: The first modality and the second modality are different medical imaging modalities, and the first medical image and the second medical image are medical images of the same user at the same location under different medical imaging modalities.

3. The method for adaptively enhancing low-contrast areas of medical images according to claim 1, characterized in that: Performing deformation measurement on the first medical image and the second medical image, performing image registration according to the measurement results, and generating a first registered image and a second registered image, including: performing mutual information calculation on the first medical image and the second medical image to generate a mutual information coefficient; Perform deformation degree analysis based on the mutual information coefficient, screen rigid registration and elastic registration according to the analysis result, and determine the target registration scheme; The first medical image and the second medical image are registered based on the target registration scheme to generate the first registered image and the second registered image.

4. The method for adaptively enhancing low-contrast areas of medical images as claimed in claim 3, characterized in that: The deformation degree is analyzed based on the mutual information coefficient, and the rigid registration and elastic registration are screened according to the analysis result to determine the target registration scheme, including: Constructing a rigid registration deformation condition constraint, wherein the rigid registration deformation condition constraint is a mutual information coefficient threshold trained by historical rigid registration image samples; Determining whether the mutual information coefficient satisfies the rigid registration deformation condition constraint; If so, generating the target registration solution by rigid registration; If not, the target registration solution is generated by elastic registration.

5. The method for adaptively enhancing low-contrast areas of medical images as claimed in claim 4, characterized in that: The rigid registration uses the first medical image and the second medical image to initialize the translation alignment with the center of gravity, and uses the gradient descent method to maximize the mutual information between the first medical image and the second medical image by adjusting the translation and rotation parameters; The elastic registration uses rigid registration to preliminarily align the first medical image and the second medical image, constructs a deformation field model based on B-splines, and calculates the deformation field by gradient descent method to minimize the difference between the first medical image and the second medical image.

6. The method for adaptively enhancing low-contrast areas of medical images according to claim 1, characterized in that: The method comprises: performing multi-level fusion on the first registered image and the second registered image through multi-scale fusion to generate a fused medical image, including: Performing pyramid decomposition on the first registered image and the second registered image respectively to construct a Gaussian pyramid; Selecting the maximum value of the pixel points at each resolution level in the Gaussian pyramid to generate a fused image corresponding to each resolution level; Based on the fused images corresponding to each resolution level, starting from the fused image at the bottom layer, the fusion results of each layer of images are synthesized with the image at the previous level layer layer by layer to generate a high-resolution image until the highest resolution layer is reached, thereby constructing a Laplacian pyramid image; Based on the Laplacian pyramid image, all the details of the low-resolution layers are gradually superimposed on the image of the highest-resolution layer to reconstruct the fused medical image.

7. The method for adaptively enhancing low-contrast areas of medical images according to claim 1, characterized in that: The fused low-contrast area is subjected to a multi-feature analysis of grayscale distribution dynamics, grayscale gradient and frequency domain features. Based on the multi-feature analysis results, a pre-trained adaptive enhancement module is called to perform joint enhancement processing of multiple contrast enhancement models to generate a fused enhanced image, including: Determining a preset low contrast factor set, wherein the preset low contrast factor set is based on historical data statistics; Collecting multiple groups of contrast-enhanced samples based on the preset low-contrast factor set, wherein any group of contrast-enhanced samples includes original medical image samples and enhanced medical image samples, and corresponding multi-feature analysis samples; Constructing multiple model networks of the multiple contrast enhancement models, performing block fusion enhancement training on the multiple model networks with the multiple groups of contrast enhancement samples, generating the multiple contrast enhancement models, and constructing the adaptive enhancement module; The adaptive enhancement module is called, and the multiple contrast enhancement models perform enhancement processing on the fused low-contrast area in the fused medical image based on the multi-feature analysis result to generate the fused enhanced image.

8. The method for adaptively enhancing low-contrast areas of medical images according to claim 7, characterized in that: Constructing the adaptive enhancement module includes: The original medical image samples and multi-feature analysis samples in any group of contrast enhanced samples are used as training inputs, and the enhanced medical image samples are used as output supervision truth values ​​to iteratively adjust the training parameters of the multiple model networks, wherein the training parameters include learning parameters corresponding to the multiple model networks respectively, and enhanced fusion weights of the multiple model networks. The multiple model networks are trained until convergence to generate the adaptive enhancement module.

9. The method for adaptively enhancing low-contrast areas of medical images as claimed in claim 1, characterized in that: Performing local contrast calculation under a sliding window on the fused medical image to identify a fused low-contrast area includes: Construct a preset sliding window; Sliding the preset sliding window on the fused medical image, and calculating a pixel contrast index under each sliding window; The pixel contrast index or the image block corresponding to the sliding window equal to the preset contrast index is extracted to generate the fused low-contrast area.

10. Medical image low contrast area adaptive enhancement system, characterized in that: include: A medical image set acquisition module, configured to receive a target medical image set, wherein the target medical image set includes at least a first medical image corresponding to a first modality and a second medical image corresponding to a second modality; a registration image generation module, configured to perform deformation measurement on the first medical image and the second medical image, perform image registration according to the measurement result, and generate a first registered image and a second registered image, wherein the image registration includes rigid registration or elastic registration; a fused medical image generation module, configured to perform multi-level fusion of the first registered image and the second registered image through multi-scale fusion to generate a fused medical image; A fusion low-contrast region identification module is used to perform local contrast calculation under a sliding window on the fused medical image to identify a fusion low-contrast region; The fused enhanced image generation module is used to perform multi-feature analysis of the grayscale distribution dynamics, grayscale gradient and frequency domain characteristics of the fused low-contrast area. Based on the multi-feature analysis results, the pre-trained adaptive enhancement module is called to perform joint enhancement processing of multiple contrast enhancement models to generate a fused enhanced image.

Images