A heart color ultrasound image processing method based on enhancement processing

Abstract

The application discloses a kind of based on enhancement processing heart color ultrasound image processing method, belong to image enhancement field, to improve the structural definition and tissue identification capability of heart color ultrasound image, the method first acquires original heart color ultrasound image and constructs dataset, then calculates pixel local energy and combines energy difference driving strategy and structure adjustment factor to establish signal excitation guide attention module, subsequently introduce disturbance generation mechanism and local response difference modeling construct local disturbance divergence guide mapping module, and optimize image features by nonlinear projection, then combining response residual, local information entropy and key adjustment factor to construct reversible response entropy coding module, control the enhancement intensity of different regions, integrate each module to construct heart color ultrasound image enhancement model, finally input heart color ultrasound image into model and output high-quality, detail-rich and with high diagnostic readability heart color ultrasound enhancement image.

Description

Technical Field

[0001] This invention belongs to the field of image enhancement, and specifically relates to a method for processing cardiac ultrasound images based on enhancement processing. Background Technology

[0002] Currently, echocardiogram images are important medical images in clinical diagnosis and auxiliary analysis of cardiac lesions. Their quality affects the accuracy of doctors in recognizing tissue boundaries and lesion areas. Traditional echocardiogram image enhancement methods mostly rely on image processing techniques with fixed rules, such as contrast enhancement, edge sharpening, and filtering and denoising. Under complex tissue structures and low signal-to-noise ratio conditions, these methods often suffer from edge blurring, loss of detail, and noise residue, limiting the stability and accuracy of the enhancement results in intelligent diagnosis. Therefore, there is an urgent need to propose processing methods that can combine tissue response characteristics and detail expression to improve the structural clarity and tissue recognition ability of low-quality echocardiogram images.

[0003] Publication No. CN117379098A discloses a cardiac ultrasound image enhancement system that intelligently determines the scanning area and acquisition quality by analyzing real-time sound wave attenuation values. After image acquisition, it performs automatic contrast and grayscale adjustment, automatic target area identification, and signal preprocessing to ensure image quality meets the requirements of subsequent enhancement processing, thereby improving the overall visibility of cardiac structures and image clarity. Publication No. CN112914610A discloses a deep learning-based contrast-enhanced echocardiogram ventricular wall thickness automatic analysis system. It first performs image quality identification and tissue pre-annotation, then evaluates feasibility and performs enhancement processing by the model, and finally automatically measures ventricular wall thickness, enhancing the ability to identify edges and distinguish structural details in the image. These methods have significant improvement effects on overall readability, edge clarity, and identification of key structures in cardiac color Doppler ultrasound scenarios.

[0004] However, existing technologies still have shortcomings when processing color Doppler ultrasound images with blurred boundaries and significant noise interference; the methods mentioned above fail to simultaneously take into account local response differences, perturbation characteristic modeling, and information entropy adjustment strategies, and ignore the impact of detailed regions on the overall quality during enhancement; in complex backgrounds, existing technologies are prone to artifacts, loss of detail, and inconsistent enhancement results, making it difficult to meet the clinical requirements for high-quality image enhancement. Summary of the Invention

[0005] This invention provides a cardiac ultrasound image processing method based on enhancement processing. It aims to construct a signal excitation-guided attention module, a local perturbation divergence-guided mapping module, and a reversible response entropy encoding module, and combine regional structure modeling, perturbation propagation mechanism, and information entropy enhancement strategy to achieve saliency modeling and detail enhancement of key tissue regions in cardiac ultrasound images, thereby improving the accuracy of image recognition, the clarity of boundary representation, and the robustness of the overall processing model.

[0006] This invention aims to propose a color Doppler ultrasound image enhancement model and provide a cardiac color Doppler ultrasound image processing method based on enhancement processing, including the following steps:

[0007] S1. Collect raw echocardiogram image data and generate an echocardiogram image dataset containing the target tissue region;

[0008] S2. Introduce a feature matrix and construct a signal-excited guided attention module. The echocardiogram image data is structurally enhanced through the signal-excited guided attention module to generate an enhanced echocardiogram image.

[0009] S3. Introduce local perturbation response difference calculation, construct local perturbation divergence guided mapping module, and process the enhanced cardiac ultrasound image through the local perturbation divergence guided mapping module to generate an optimized cardiac ultrasound image feature matrix.

[0010] S4. Design a response residual and local information entropy calculation mechanism, introduce key adjustment factors, construct a reversible response entropy encoding module, and enhance the optimized cardiac ultrasound image through the reversible response entropy encoding module to generate a high-quality enhanced cardiac ultrasound image.

[0011] S5. Integrating the signal excitation-guided attention module, the local perturbation divergence-guided mapping module, and the reversible response entropy coding module, a cardiac color Doppler ultrasound image enhancement processing model is constructed.

[0012] S6. Input the echocardiogram image into the echocardiogram image enhancement processing model for image enhancement, and output a high-quality enhanced echocardiogram image.

[0013] Preferably, in S2, constructing the signal-induced guided attention module specifically includes the following steps:

[0014] Step S21: Define the signal excitation intensity function Used to calculate the input feature matrix Each position The energy excitation intensity, mathematically modeled as follows:

[0015] ;

[0016] in, The total number of spatial locations, For the number of channels, For the first Position No. The eigenvalues ​​of the channel, As a signal excitation factor, it represents the location. The local energy response provides the basic input for subsequent connection matrix calculations.

[0017] Step S22: Introduce an energy difference-driven connection weight function. The mathematical model used to describe the degree of similarity connectivity between any two locations and to suppress weakly correlated connections with large energy differences is as follows:

[0018] ;

[0019] in, For temperature parameters, The energy difference threshold, This matrix is ​​an indicator function. Sparsity control was achieved through exponential decay and threshold screening, which resulted in strong connections between pixels with similar energy while irrelevant regions were weakened.

[0020] Step S23: Define the structural response adjustment factor This is used to enhance the connection weights between high-energy locations and suppress connections in low-energy regions, thereby improving the expressive power of structural features. The mathematical model is as follows:

[0021] ;

[0022] in, Energy excitation vector The maximum value, This is a non-linear adjustment index. This adjustment factor... Normalization of the energy product amplifies the contribution of high-response regions to network feature enhancement while reducing the impact of background noise regions.

[0023] Step S24: Connection matrix driven by energy difference With structural response adjustment factor matrix The combination of these elements constructs a structure-guided connection mapping matrix. It is used to reconstruct global feature relationships and enhance signal consistency. The mathematical model is as follows:

[0024] in, It integrates the dual effects of sparse connectivity and energy regulation, ensuring that highly correlated structural information is enhanced and providing guiding mapping for subsequent feature enhancement.

[0025] Step S25: Use the structure to guide the connection mapping matrix For the input feature matrix Perform mapping operations to obtain the output feature matrix. To achieve global enhancement of structural features and suppression of noise interference, the mathematical model is as follows: ;

[0026] in, The enhanced feature matrix serves as the input to the subsequent Local Perturbation Divergence Guided Mapping (LPDI) module, effectively improving the recognition and enhancement effect of structural regions in color ultrasound images.

[0027] Preferably, the signal excitation-guided attention module constructed through S2, in S21, by defining a signal energy excitation factor, can accurately calculate the energy response of each pixel position in the echocardiogram image, thereby effectively capturing the structural features and the correlation between regions in the echocardiogram image, avoiding the shortcomings of traditional methods that cannot dynamically capture local information, and improving the accuracy of structural enhancement; S22, by introducing an energy difference-driven connection weight function, can adaptively adjust the connection strength between each position according to the energy excitation difference of pixels, improving the flexibility and responsiveness of the echocardiogram image model; S23, by combining a local structural response adjustment factor, further strengthens the connection between high-energy regions, reduces the influence of low-energy regions on the model, and enhances the signal transmission effect in the echocardiogram image; S24, the nonlinear correction factor further optimizes the weights of the connection matrix, enabling the model to achieve more accurate structural enhancement in the echocardiogram image; Overall, this module significantly improves the accuracy and robustness of echocardiogram image structural enhancement through signal energy excitation and adaptive structural adjustment mechanisms.

[0028] Preferably, in S3, constructing the local perturbation divergence guided mapping module specifically includes the following steps:

[0029] Step S31: Excite the guiding feature matrix of the signal output from S2. Perturbation generation is performed to obtain the perturbation response matrix. This is done by introducing small perturbations at each location. Two perturbation versions were obtained. and Its mathematical model is:

[0030] ;

[0031] ;

[0032] in It is the perturbation amplitude, which is a fixed constant and will not be updated during training; The element-level sign function is used to introduce positive and negative perturbations, thereby generating two perturbation versions to compare and calculate the response differences at various locations in the image.

[0033] Step S32, for each channel Define the variance of the disturbance response difference. Its mathematical model is:

[0034] ;

[0035] in, Indicates channel Sensitivity to disturbances, i.e., its degree of instability. and Positions At the The perturbed version of the channel, this divergence reflects the channel's... The sensitivity to perturbations is indicated by a larger divergence value, which suggests a stronger response of that channel to the perturbation and poorer stability in the image region. This calculation allows us to assess the stability of each channel's response to perturbations, reflecting the local instability of the image.

[0036] Step S33: To achieve perturbation-guided mapping, construct the structure guidance matrix. Its mathematical model is:

[0037] ;

[0038] in, Indicates position and location The structural correlation between them, based on their differences in perturbation response. and The calculations show that this matrix effectively enhances regions with strong structural consistency by guiding signal flow by reflecting differences in disturbance response.

[0039] Step S34: Use the constructed structure guiding matrix For the input feature matrix The mapping is performed to obtain the final enhanced feature matrix. Its mathematical model is:

[0040] ;

[0041] The structural guidance matrix is ​​used here. By adjusting the flow of image information according to the differences in perturbation response, the transmission of information flow is optimized and the structural expression of more stable regions in the image is enhanced.

[0042] Step S35: Use the tanh function to perform a nonlinear projection transformation on the calculation results to obtain the final feature matrix, the mathematical model of which is:

[0043] ;

[0044] The tanh function enhances the dynamic range of the perturbation response region, ensuring that image details are enhanced. To maintain the amplitude of the original signal and ensure the integrity of the original information, the final output retains the characteristics after perturbation guidance while ensuring the consistency of image information.

[0045] Preferably, through the S3 local perturbation divergence guided mapping module, in S31, by defining a perturbation generation mechanism, the perturbation response at each pixel location in the image can be accurately calculated, thereby effectively capturing the stability differences of the cardiac ultrasound image region, avoiding the shortcomings of traditional methods that cannot dynamically adapt to local perturbations, and improving the accuracy of structure enhancement; in S32, by introducing perturbation response difference variance, the response sensitivity of each channel under perturbation can be accurately evaluated, thereby adjusting the propagation intensity of cardiac ultrasound image features according to the stability of the region, improving the adaptability and flexibility of the model; in S33, by constructing a structure guidance matrix, combined with perturbation response differences, the connection between regions with strong structural consistency is further enhanced, reducing the interference of regions with weak perturbation response to the model, and optimizing the information flow transmission effect; in S34, by using the structure guidance matrix to perform weighted mapping of input features, the signal transmission is effectively guided, ensuring that the features of stable regions in the image are enhanced; in S35, the nonlinear correction factor further optimizes the weight of the feature matrix, enabling the model to achieve more accurate feature enhancement in cardiac ultrasound images; overall, this module uses a local perturbation difference guidance mechanism and a structural response adjustment mechanism.

[0046] Preferably, in S4, constructing the reversible response entropy encoding module specifically includes the following steps:

[0047] Step S41: To accurately assess the reliability of the image structure, a locally reconstructed estimated echocardiogram image is first constructed to estimate the reconstructed value of each pixel in its neighborhood. This reconstruction process uses median filtering to avoid noise interference and the influence of local inconsistencies. The mathematical model is as follows:

[0048] ;

[0049] in In pixels The 3x3 neighborhood centered on the center, The set of neighboring pixels after removing the center pixel. This step, by removing local anomalies, provides a more stable basis for subsequent response calculations, resulting in a reconstructed estimate.

[0050] Step S42: Calculate the response residual between each pixel location and its reconstructed value. It is used to reflect the differences and instabilities of image regions, and its mathematical model is as follows:

[0051] ;

[0052] in, In response to the residual, a large residual value indicates that there is a significant change in the image content of the pixel area, while a small residual value indicates that the area is relatively smooth and stable.

[0053] Step S43: To measure the amount of information in the neighborhood of each pixel, define local entropy. This value reflects the complexity and information density of the image region, and its mathematical model is as follows:

[0054] ;

[0055] in, This represents the normalized probability distribution of pixel values ​​within the neighborhood. For neighborhood set, This represents the local entropy value of the pixel region. This entropy value can characterize the complexity of local regions of the image. A high entropy value indicates that the region has rich information, while a low entropy value indicates that the region has less information.

[0056] Step S44: Combine the response residual and local information entropy to calculate the response entropy score for each pixel. This score characterizes which information region the pixel belongs to, and its mathematical model is as follows:

[0057] ;

[0058] in, It is a very small constant used to prevent division by zero errors, and is used for response entropy scoring. This value is used to distinguish between high-information and low-information regions of an image. It is larger for low-information regions and smaller for high-information regions.

[0059] Step S45: Construct key regulatory factors This method dynamically adjusts the intensity of image enhancement based on the reflectance intensity and structural complexity of color Doppler ultrasound images. The adjustment factor is based on changes in reflectance intensity and structural feature reconstruction errors, ensuring that low-information areas are not over-enhanced during image enhancement, while high-information areas receive appropriate enhancement. Its mathematical model is as follows:

[0060] ;

[0061] in, Adjust the intensity as a global benchmark to control the overall strength of the enhancement; Indicates position The intensity of ultrasound reflection at a location reflects the reflection characteristics of tissue in the image; and These are the mean and standard deviation of the ultrasound reflectance intensity map, respectively, representing the global distribution of reflectance intensity in the image; It is a very small constant used to prevent division by zero errors; Indicates position The structural reconstruction error at that location reflects the complexity of the details in that area.

[0062] In this embodiment, the structural reconstruction error The mathematical model is:

[0063] ;

[0064] The average structural reconstruction error represents the overall structural complexity of the image. In this embodiment, the average structural reconstruction error... The mathematical model is:

[0065] ...

[0066] Step S46: Construct the final dehazing mask. This is used to control the difference in enhancement intensity between structural regions and low-reflectivity regions during image enhancement. The mask is achieved by combining response entropy scoring... and regulatory factors An adjustable enhancement factor based on local image information is generated to enhance high-reflectivity areas and suppress over-enhancement of low-reflectivity areas. Its mathematical model is as follows:

[0067] ;

[0068] in, The adjustment factor dynamically controls the image enhancement intensity, which depends on the ultrasonic reflection intensity and structural complexity. See step S45 for detailed calculation. The response entropy score measures the degree of combination between the response residual and local information entropy of each pixel, indicating whether the pixel belongs to a high-information region. The specific calculation is shown in step S44; mask. For high-information regions, response entropy scoring Smaller, regulating factor Larger, therefore mask It will maintain a small value to reduce over-enhancement of the image; for low-reflectivity areas, the response entropy score... Larger, regulating factor Smaller, therefore mask It will increase and enhance the detail and structural representation of these areas; the mask obtained through calculation. It is used as an adjustment factor for image enhancement, automatically adjusting the enhancement intensity of different regions to ensure that the structural information in the image is appropriately enhanced, while avoiding excessive enhancement of low-reflection areas.

[0069] Step S47: Use the response entropy coding mask The input image is enhanced to obtain the enhanced image. The goal of this step is to automatically adjust the enhancement intensity based on the different reflectivity and structural complexity of image regions, in order to highlight key information areas in the image while avoiding over-enhancing low-reflectivity areas. The mathematical model is as follows:

[0070] ;

[0071] in, These are the raw pixel values ​​of the input image; The mask value is calculated based on the response entropy score and the adjustment factor; These are the enhanced image pixel values; for highly reflective areas, due to Smaller pixels will result in relatively stable pixel values ​​in the enhanced image, avoiding over-enhancement; for low-reflectivity areas, Larger areas will have enhanced details, contrast, and structural clarity in the enhanced image; in this step, the image is weighted and enhanced to obtain the final enhanced image. The image has been enhanced with emphasis on structural regions and suppression of background regions using a response entropy coding mask. It plays a key role in enabling the enhancement process to automatically adjust according to the characteristics of the image content, thereby improving the image quality and recognizability.

[0072] Preferably, through the S4 reversible response entropy encoding module, in S41, by defining a local reconstruction estimation mechanism, the reconstruction error of each pixel position in the echocardiogram image can be accurately calculated, thereby effectively assessing the detail complexity of the image region and avoiding the shortcomings of traditional methods that cannot dynamically adapt to local changes, thus improving the accuracy of echocardiogram image enhancement; in S42, by calculating the response residual, the degree of change in each region of the echocardiogram image can be accurately assessed, thereby targeting the enhancement of structural detail regions and suppressing the over-enhancement of low-information regions; in S43, by introducing the calculation of local information entropy, the complexity of local regions in the echocardiogram image is accurately measured, providing a valid basis for subsequent echocardiogram image enhancement, enabling the model to better adapt to the diversity and complexity of echocardiogram images; in S44, by combining response entropy scoring with response residual, high-information regions and low-information regions in the echocardiogram image can be effectively distinguished. The system employs a three-tiered approach: S45, S46, S47, and S48. S48 introduces an illumination-component-driven adjustment factor to dynamically adjust the enhancement intensity based on the image's illumination information and structural complexity. This allows the cardiac ultrasound image enhancement process to better adapt to the characteristics of different regions, ensuring both local adaptability and global consistency of the enhancement effect. S48 uses a response entropy coding mask to adaptively adjust the enhancement intensity according to the response characteristics of different regions in the cardiac ultrasound image, further optimizing the structural representation and detail of the image, ensuring the preservation of key information and appropriate suppression of low-reflection areas. S49 finally enhances the cardiac ultrasound image using the response entropy coding mask, precisely adjusting the enhancement degree of different regions to obtain the enhanced cardiac ultrasound image, improving its quality and recognizability. Overall, this module, through the joint optimization of response residuals, local information entropy, and adjustment factors, ensures detail preservation and noise suppression during the cardiac ultrasound image enhancement process, improving the structural clarity and contrast of the cardiac ultrasound image.

[0073] Preferably, in S5, constructing the cardiac ultrasound image enhancement processing model specifically includes the following steps:

[0074] Step S51: First, the input echocardiogram image is preprocessed using a formatting method to extract its features and construct a feature matrix. This matrix is ​​used for subsequent image enhancement processing to ensure that the image data meets the model's input requirements.

[0075] Step S52: By constructing a signal excitation-guided attention module, the structural information in the echocardiogram image is enhanced based on the energy distribution of the image. This module constructs a connectivity matrix by calculating the signal excitation intensity at each location in the image and adjusts the connectivity strength through a response adjustment factor, thereby enhancing the recognizability of key information regions in the echocardiogram image.

[0076] Step S53: The local perturbation divergence-guided mapping module calculates the response difference of each region in the image through a perturbation generation mechanism, and adjusts the feature transfer intensity of the cardiac ultrasound image according to the perturbation response difference. By introducing a perturbation response difference matrix, the structural consistency of the cardiac ultrasound image is optimized, the details of the target region are enhanced, and the interference of low-reflection areas on the enhancement results is reduced.

[0077] In step S54, the reversible response entropy encoding module constructs an enhancement mask and ultimately outputs a high-quality enhanced echocardiogram image by calculating the response residual and local information entropy. The module evaluates the detail regions of the echocardiogram image based on the response residual and information entropy score, and dynamically adjusts the enhancement intensity of the echocardiogram image using a modulating factor, thereby optimizing the quality and structural representation of the echocardiogram image.

[0078] In summary, compared with existing technologies, the beneficial effects of this invention, by employing this technical solution, are as follows: The signal-excitation guided attention module successfully extracts important structural information from echocardiogram images by performing energy excitation calculations on the images; the local perturbation divergence-guided mapping module effectively adjusts the propagation intensity of echocardiogram image features by calculating perturbation response differences, thereby optimizing the enhancement effect on structural regions; and the reversible response entropy encoding module dynamically adjusts the enhancement intensity of echocardiogram images by combining response residuals and local information entropy, resulting in precise enhancement of image details and high-information regions. Overall, this model, through the fusion design of structure guidance, response difference adjustment, and information entropy-driven approaches, achieves high-quality enhancement of echocardiogram images, making target regions more prominent and details clearer, significantly improving the recognizability and diagnostic value of echocardiogram images. Attached Figure Description

[0079] Figure 1 A step-by-step diagram of a cardiac ultrasound image processing method based on enhancement processing.

[0080] Figure 2 This is a structural diagram of the signal-induced attention-guided module.

[0081] Figure 3 This is a structural diagram of the local perturbation divergence guided mapping module.

[0082] Figure 4 This is a diagram of the entropy encoding module for reversible response.

[0083] Figure 5 This is a diagram showing the overall structure of the color Doppler ultrasound image enhancement processing model.

[0084] Figure 6 Comparison of color Doppler ultrasound images before and after enhancement using a color Doppler ultrasound image enhancement processing model. Detailed Implementation

[0085] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0086] Please see the appendix Figure 1 -Appendix Figure 6 This invention provides a method for processing cardiac color Doppler ultrasound images based on enhancement processing.

[0087] As attached Figure 1 As shown in the flowchart, this invention proposes a method for processing cardiac ultrasound images based on enhancement processing. The specific implementation method includes the following steps:

[0088] S1. Acquire color Doppler ultrasound image data and generate a raw dataset of cardiac color Doppler ultrasound images containing the target tissue region.

[0089] Furthermore, as shown in the appendix Figure 1 The acquisition of cardiac ultrasound image data described in S1 specifically involves maintaining the stability of the acquisition device and controlling the imaging parameters, using a high-resolution ultrasound device to perform a non-destructive scan of the target area, and generating a cardiac ultrasound image dataset of 500 images. The dataset is in PNG format, and each image is uniformly preprocessed to a size of 512×512 pixels.

[0090] S2. Introduce a feature matrix and construct a signal-excited guided attention module. The echocardiogram image is structurally enhanced through the signal-excited guided attention module to generate an enhanced echocardiogram image.

[0091] Furthermore, as shown in the appendix Figure 1 The construction of the signal-induced attention module described in S2 is illustrated in the attached figure. Figure 2 As shown, the specific implementation of the module includes the following steps:

[0092] Further, in step S21, define the signal excitation intensity function. Used to calculate the input feature matrix Each position The energy excitation intensity, mathematically modeled as follows:

[0093] ;

[0094] The input echocardiogram image is 512×512 pixels in size and uses 3 channels. Since L is the total number of spatial locations, L = 512 × 512 = 262144 (that is, the total number of pixels in the image). Since C is the number of channels, C=3 (meaning that the color ultrasound image uses three channels). For the first Position No. The eigenvalues ​​of the channel, As a signal excitation factor, it represents the location. The local energy response provides the basic input for subsequent connection matrix calculations.

[0095] Further, in step S22, an energy difference-driven connection weight function is introduced. The mathematical model used to describe the degree of similarity connectivity between any two locations and to suppress weakly correlated connections with large energy differences is as follows:

[0096] ;

[0097] in, The temperature parameter is initially set to 0.1 and gradually decreased based on convergence. The energy difference threshold, the threshold Used to control which pixels establish connections, and to set thresholds. =2.0 means that two pixels are considered correlated only when the difference in their signal excitation is less than 2. This matrix is ​​an indicator function. Sparsity control was achieved through exponential decay and threshold screening, which resulted in strong connections between pixels with similar energy while irrelevant regions were weakened.

[0098] Further, in step S23, the structural response adjustment factor is defined. This is used to enhance the connection weights between high-energy locations and suppress connections in low-energy regions, thereby improving the expressive power of structural features. The mathematical model is as follows:

[0099] ;

[0100] in, Energy excitation vector The maximum value, The non-linear adjustment index is used here. The setting is 0.5 to avoid overly strong adjustment; this adjustment factor... Normalization of the energy product amplifies the contribution of high-response regions to network feature enhancement while reducing the impact of background noise regions.

[0101] Furthermore, step S24, the connection matrix driven by energy difference With structural response adjustment factor matrix The combination of these elements constructs a structure-guided connection mapping matrix. It is used to reconstruct global feature relationships and enhance signal consistency. The mathematical model is as follows:

[0102] ;

[0103] in, Integrating the dual effects of sparse connectivity and energy regulation, this structure-guided connectivity mapping matrix ensures that highly correlated structural information is enhanced and provides guiding mapping for subsequent feature enhancement. This technique is used to enhance connectivity in highly correlated regions and suppress connectivity in low-correlation regions in echocardiogram images, thereby achieving the reconstruction and enhancement of echocardiogram image features.

[0104] Further, in step S25, the structure is used to guide the connection mapping matrix. For the input feature matrix Perform mapping operations to obtain the output feature matrix. To achieve global enhancement of structural features and suppression of noise interference, the mathematical model is as follows:

[0105] ;

[0106] in, The enhanced feature matrix serves as the input to the subsequent Local Perturbation Divergence Guided Mapping (LPDI) module, effectively improving the recognition and enhancement effect of structural regions in color ultrasound images.

[0107] S3. By introducing the calculation of local perturbation response differences, a local perturbation divergence guided mapping module is constructed. The enhanced cardiac ultrasound image is processed by the local perturbation divergence guided mapping module to generate an optimized cardiac ultrasound image feature matrix.

[0108] Furthermore, as shown in the appendix Figure 3 The local perturbation divergence guided mapping module described in S3 is constructed using the following steps: Figure 3 As shown, the specific implementation of the module includes the following steps:

[0109] Furthermore, in step S31, the signal output from S2 is used to excite and guide the feature matrix. Perturbation generation is performed to obtain the perturbation response matrix. This is done by introducing small perturbations at each location. Two perturbation versions were obtained. and Its mathematical model is:

[0110] ;

[0111] ;

[0112] in This is the perturbation amplitude, a fixed constant, set to 0.05 here, and will not be updated during training; The element-level sign function is used to introduce positive and negative perturbations, thereby generating two perturbation versions to compare and calculate the response differences at various locations in the image.

[0113] Further, in step S32, for each channel Define the variance of the disturbance response difference. Its mathematical model is:

[0114] ;

[0115] in, Indicates channel Sensitivity to disturbances, i.e., its degree of instability. and Positions At the The perturbed versions of the channels, these matrices, generated in step S31, represent the versions of the cardiac ultrasound image feature matrices after adding and reducing perturbations. This divergence reflects the channel... The sensitivity to perturbations is indicated by a larger divergence value, suggesting a stronger response of that channel to perturbations and poorer stability in the cardiac ultrasound image region. This calculation allows us to assess the response stability of each channel under perturbations, reflecting the local instability of the cardiac ultrasound image.

[0116] Furthermore, in step S33, to achieve perturbation-guided mapping, we construct a structural guidance matrix. Its mathematical model is:

[0117] ;

[0118] in, Indicates position and location The structural correlation between them, based on their differences in perturbation response. and The calculations show that this matrix effectively enhances regions with strong structural consistency by guiding signal flow by reflecting differences in disturbance response.

[0119] Further, step S34 uses the constructed structure guiding matrix. For the input feature matrix The mapping process involves weighting the response features of each pixel to obtain the final enhanced feature matrix. Its mathematical model is:

[0120] ;

[0121] Among them, through the structural guiding matrix By adjusting the flow of image information according to the differences in perturbation response, the transmission of information flow is optimized and the structural expression of more stable regions in the image is enhanced.

[0122] Further, in step S35, the calculation result is subjected to a nonlinear projection transformation using the tanh function to obtain the final feature matrix, the mathematical model of which is:

[0123] ;

[0124] The tanh function enhances the dynamic range of the perturbation response region, ensuring that the details of the image are enhanced. When applied to the enhanced feature matrix, it can enhance the discernibility of information by performing nonlinear transformations on the features. To maintain the amplitude of the original signal and ensure the integrity of the original information, the final output retains the characteristics after perturbation guidance while ensuring the consistency of image information.

[0125] S4. Design a response residual and local information entropy calculation mechanism, introduce an adjustment factor, construct a reversible response entropy encoding module, and enhance the optimized cardiac ultrasound image through the reversible response entropy encoding module to generate a high-quality enhanced cardiac ultrasound image.

[0126] Furthermore, as shown in the appendix Figure 4 The reversible response entropy coding module described in S4 is constructed using the following steps: Figure 4 As shown, the specific implementation of the module includes the following steps:

[0127] Furthermore, in step S41, to accurately assess the reliability of the image structure, a local reconstruction estimation of the cardiac ultrasound image is first constructed to estimate the reconstruction value of each pixel in its neighborhood. This reconstruction process uses median filtering to avoid the effects of noise interference and local inconsistencies. Its mathematical model is as follows:

[0128] in In pixels The 3x3 neighborhood centered on the center, To obtain the set of neighboring pixels after removing the center pixel, a fixed 3×3 window is used for median filtering. This means the filtering operation considers the values ​​of each pixel and its eight neighboring pixels. This step, by removing local anomalies, provides a more stable basis for subsequent response calculations, resulting in a reconstructed estimate.

[0129] Further, in step S42, the response residual between each pixel location and its reconstructed value is calculated. It is used to reflect the differences and instabilities of image regions, and its mathematical model is as follows:

[0130] ;

[0131] in, In response to the residual, a large residual value indicates that there is a significant change in the image content of the pixel area, while a small residual value indicates that the area is relatively smooth and stable.

[0132] Further, in step S43, to measure the amount of information in the neighborhood of each pixel, local entropy is defined. This value reflects the complexity and information density of the image region, and its mathematical model is as follows: ;

[0133] in, This represents the normalized probability distribution of pixel values ​​within the neighborhood. The neighborhood set is a 3×3 window. This represents the local entropy value of the pixel region. This entropy value can characterize the complexity of local regions of the image. A high entropy value indicates that the region has rich information, while a low entropy value indicates that the region has less information.

[0134] Further, in step S44, the response residual and local information entropy are combined to calculate the response entropy score for each pixel. This score characterizes which information region the pixel belongs to, and its mathematical model is as follows:

[0135] ;

[0136] in, This is a very small constant used to prevent division by zero errors; it is set to 0.000001. (Response entropy score) This value is used to distinguish between high-information and low-information regions of an image. It is larger for low-information regions and smaller for high-information regions.

[0137] Further, in step S45, key regulatory factors are constructed. This method dynamically adjusts the intensity of image enhancement based on the reflectance intensity and structural complexity of color ultrasound images. The adjustment factor is based on changes in reflectance intensity and structural feature reconstruction errors, ensuring that low-information areas are not over-enhanced while high-information areas receive appropriate enhancement during the image enhancement process. Its mathematical model is as follows:

[0138] ;

[0139] in, Adjust the intensity as a global benchmark to control the overall strength of the enhancement; Indicates position The intensity of ultrasound reflection at a location reflects the reflection characteristics of tissue in the image; and These are the mean and standard deviation of the ultrasound reflectance intensity map, respectively, representing the global distribution of reflectance intensity in the image; It is a very small constant used to prevent division by zero errors; Indicates position The structural reconstruction error reflects the complexity of the details in that area. In this embodiment, the structural reconstruction error... The mathematical model is:

[0140] ;

[0141] The average structural reconstruction error represents the overall structural complexity of the image. In this embodiment, the average structural reconstruction error... The mathematical model is:

[0142] .

[0143] Further, in step S46, the final dehazing mask is constructed. It is used to control the difference in enhancement intensity between structural regions and low-reflection regions during the image enhancement process.

[0144] This mask combines the response entropy score. and regulatory factors An adjustable enhancement factor based on local image information is generated to enhance high-reflectivity areas and suppress over-enhancement of low-reflectivity areas. Its mathematical model is as follows:

[0145] ;

[0146] in, The adjustment factor dynamically controls the image enhancement intensity, which depends on the ultrasonic reflection intensity and structural complexity. See step S45 for detailed calculation. The response entropy score measures the degree of combination between the response residual and local information entropy of each pixel, indicating whether the pixel belongs to a high-information region. The specific calculation is shown in step S44; mask. For high-information regions, including edges, tissue interfaces, and blood vessels, response entropy scoring... Smaller, regulating factor Larger, therefore mask It will maintain a small value to reduce over-enhancement of the image; for low-reflectivity areas, the response entropy score... Larger, regulating factor Smaller, therefore mask It will increase and enhance the detail and structural representation of these areas; the mask obtained through calculation. This is used as a modulating factor for image enhancement, automatically adjusting the enhancement intensity of different regions to ensure that structural information in the image is appropriately enhanced while avoiding over-enhancement of low-reflectivity areas; the input weighted modulating factor and response entropy score Both are 2D arrays representing the feature matrices of the image.

[0147] Further, in step S47, a response entropy coding mask is used. The input image is enhanced to obtain the enhanced image. The goal of this step is to automatically adjust the enhancement intensity based on the different reflectivity and structural complexity of image regions, in order to highlight key information areas in the image while avoiding over-enhancing low-reflectivity areas. The mathematical model is as follows:

[0148] ;

[0149] in, The image pixel values ​​are those after enhancement. For high-reflectivity areas, the enhanced image maintains relatively stable pixel values ​​to avoid over-enhancement. For low-reflectivity areas, the enhanced image strengthens details, improving contrast and structural clarity. In this step, the image is weighted and enhanced to obtain the final enhanced image. The image has been enhanced with emphasis on structural regions and suppression of background regions using a response entropy-coded mask. It plays a key role in enabling the enhancement process to automatically adjust according to the characteristics of the image content, thereby improving the image quality and recognizability.

[0150] S5 integrates a signal excitation-guided attention module, a local perturbation divergence-guided mapping module, and a reversible response entropy coding module to construct a cardiac color Doppler ultrasound image enhancement processing model.

[0151] Furthermore, as shown in the appendix Figure 5 The model for enhancing and processing cardiac ultrasound images described in S5 is attached, with the specific construction steps as follows. Figure 5 As shown, the specific implementation of the module includes the following steps:

[0152] Further, in step S51, firstly, the input echocardiogram image is formatted and preprocessed to extract the features of the echocardiogram image and construct a feature matrix. This matrix is ​​used for subsequent echocardiogram image enhancement processing to ensure that the echocardiogram image data meets the model input requirements.

[0153] Further, in step S52, a signal-excitation-guided attention module is constructed to enhance the structural information in the echocardiogram image based on the energy distribution of the image. This module constructs a connectivity matrix by calculating the signal excitation intensity at each location in the echocardiogram image and adjusts the connectivity strength through a response adjustment factor, thereby enhancing the recognizability of key information regions in the echocardiogram image.

[0154] Furthermore, in step S53, the local perturbation divergence-guided mapping module calculates the response difference of each region in the image through a perturbation generation mechanism, and adjusts the feature transfer intensity of the cardiac ultrasound image based on the perturbation response difference. By introducing a perturbation response difference matrix, the structural consistency of the cardiac ultrasound image is optimized, the details of the target region are enhanced, and the interference of low-reflection regions on the enhancement results is reduced.

[0155] Furthermore, in step S54, the reversible response entropy encoding module constructs an enhancement mask and ultimately outputs an enhanced echocardiogram image by calculating the response residual and local information entropy. The module evaluates image detail regions based on the response residual and information entropy scores of the echocardiogram image region, and dynamically adjusts the enhancement intensity of the echocardiogram image by combining key adjustment factors, thereby optimizing the quality and structural representation of the echocardiogram image.

[0156] Furthermore, as shown in the appendix Figure 1 The cardiac ultrasound image enhancement model shown takes a cardiac ultrasound image as input and outputs an enhanced high-resolution cardiac ultrasound image, as shown in the attached figure. Figure 6 As shown, the specific implementation includes the following steps:

[0157] Furthermore, in step S6, the operating system platform used by the model is Linux, the language is Python 3.9.1, the processor is Jetson Xavier, the image processing library is OpenCV and PIL, the server hardware memory is 64G, and the dataset contains a total of 500 cardiac ultrasound images taken under different lighting conditions.

[0158] The above are merely preferred embodiments of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the inventive concept of the present invention, and these modifications and improvements all fall within the protection scope of the present invention.

Claims

1. A method for processing cardiac color Doppler ultrasound images based on enhancement processing, characterized in that, Includes the following steps: S1. Collect raw echocardiogram image data and generate an echocardiogram image dataset containing the target tissue region; S2. Introduce a feature matrix and construct a signal-excited guided attention module. The echocardiogram image data is structurally enhanced through the signal-excited guided attention module to generate an enhanced echocardiogram image. The signal-triggered attention module includes: S21. Define the signal excitation intensity function. Used to calculate the input feature matrix Each position The energy excitation intensity, mathematically modeled as follows: ； in, The total number of spatial locations, For the number of channels, For the first Position No. The eigenvalues ​​of the channel, As a signal excitation factor, it represents the location. The local energy response; S22. Introducing an energy difference-driven connection weighting function. The mathematical model describes the degree of similarity connectivity between any two locations and suppresses weakly correlated connections with large energy differences: ； in, For temperature parameters, The energy difference threshold, For indicator functions; S23. Define the structural response modulation factor This enhances the connection weights between high-energy locations and suppresses connections in low-energy regions, thereby improving the expressive power of structural features. The mathematical model is as follows: ； in, Energy excitation vector The maximum value, It is a non-linear adjustment index; S24, Connection matrix driven by energy difference With structural response adjustment factor matrix The combination of these elements constructs a structure-guided connection mapping matrix. The mathematical model is: ; S25 uses the structure to guide the connection mapping matrix For the input feature matrix Perform mapping operations to obtain the output feature matrix. ; S3. Introduce local perturbation response difference calculation, construct local perturbation divergence guided mapping module, and process the enhanced cardiac ultrasound image through the local perturbation divergence guided mapping module to generate an optimized cardiac ultrasound image feature matrix. The local perturbation divergence-guided mapping module includes: S31, Excitation and guidance feature matrix of the signal output from S2 Perturbation generation is performed to obtain the perturbation response matrix by introducing small perturbations at each location. Two perturbation versions were obtained. and Its mathematical model is: ; ； in The amplitude of the disturbance; Symbolic functions at the element level; S32, For each channel Define the variance of the disturbance response difference. Its mathematical model is: ； in, Indicates channel Sensitivity to disturbances and Positions At the The perturbation version of the channel; S301. Construct the structural guidance matrix Its mathematical model is: ； in, Indicates position and location The degree of structural correlation between them; S302, Using the constructed structure guidance matrix For the input feature matrix The mapping is performed to obtain the final enhanced feature matrix. Its mathematical model is: ； S303. Use the tanh function to perform a nonlinear projection transformation on the calculation results to obtain the final feature matrix. Its mathematical model is: ； in, To preserve the amplitude of the original signal; S4. Design a response residual and local information entropy calculation mechanism, introduce key adjustment factors, construct a reversible response entropy encoding module, and enhance the optimized cardiac ultrasound image through the reversible response entropy encoding module to generate a high-quality enhanced cardiac ultrasound image. The reversible response entropy encoding module includes: S41. Construct a locally reconstructed and estimated echocardiogram image. Median filtering is used to avoid the effects of noise interference and local inconsistencies. The mathematical model is as follows: ； in In pixels The neighborhood centered on, The set of neighboring pixels after removing the center pixel. These are the estimated values ​​after reconstruction; S42. Calculate the response residual between each pixel location and its reconstructed value. Its mathematical model is: ； S43. Define local entropy Its mathematical model is: ； in, This represents the normalized probability distribution of pixel values ​​within the neighborhood. For neighborhood set, This represents the local entropy value of the pixel region; S44. Combining the response residual and local information entropy, calculate the response entropy score for each pixel. Its mathematical model is: ； in, It is a very small constant; S401, Constructing Key Regulatory Factors Based on the changes in reflection intensity and the reconstruction error of structural features, its mathematical model is as follows: ； in, Adjust the intensity as a global benchmark. Indicates position Ultrasonic reflection intensity at the location, and These represent the mean and standard deviation of the ultrasound reflectance intensity map, respectively. It is a very small constant. Indicates position Structural reconstruction error at the location, The average structure reconstruction error of the entire image; S402, Construct the final defogging mask. This mask is achieved by combining the response entropy score. and regulatory factors A modulotable enhancement factor based on local image information is generated, and its mathematical model is as follows: ； in, As a regulating factor, For response entropy scoring; S403, Use response entropy coding mask The input image is enhanced to obtain the enhanced image. Its mathematical model is: ； in, These are the raw pixel values ​​of the input image; The mask value is calculated based on the response entropy score and the adjustment factor; These are the enhanced image pixel values; S5. An integrated signal-excitation guided attention module, a local perturbation divergence-guided mapping module, and a reversible response entropy coding module are used to construct a cardiac ultrasound image enhancement processing model, including: S51. Input the echocardiogram image dataset into the signal-guided attention module. Calculate the weight coefficient matrix based on the difference between local response intensity and global mean. Perform weighted processing on the input feature matrix to obtain the enhanced echocardiogram image. ; S52. Input the enhanced echocardiogram image into the local perturbation divergence guided mapping module. Introduce positive and negative perturbations of fixed amplitude at each pixel location, calculate the perturbation response difference, and standardize to generate a perturbation divergence weight matrix. Apply this weight matrix to feature optimization to obtain the optimized echocardiogram image feature matrix. ; S53. Optimize the feature matrix of the echocardiogram image. The reversible response entropy encoding module is input to calculate the pixel residual matrix and local information entropy. A key adjustment factor is introduced to correct the weights, completing the reversible encoding enhancement and obtaining a high-quality enhanced echocardiogram image. ; S6. Input the echocardiogram image into the echocardiogram image enhancement processing model for image enhancement, and output a high-quality enhanced echocardiogram image.

Images