MRI-based cardiac structure reconstruction system

Abstract

The present invention relates to the field of cardiac MRI image reconstruction technology, specifically to a cardiac structure reconstruction system based on MRI, the system comprising a soft threshold initial setting module, a threshold adaptive adjustment module, a control parameter optimization module, a cardiac structure detail analysis module, a key area marking module, an image clarity adjustment module, and a reconstruction efficiency optimization module. The present invention optimizes the reconstruction process through soft threshold adaptive adjustment and iterative control strategy, reduces the number of iterations and ensures image quality. At the same time, the dynamic adjustment of the soft threshold can accurately control the details of image processing according to the specific characteristics and noise level of the image, and improves the overall clarity of the reconstructed image. By combining differential geometry and topological data analysis technology, the complex geometric characteristics and spatial connectivity of the cardiac structure can be deeply analyzed, which not only improves the ability to accurately identify the cardiac structure, but also can capture subtle changes in the local curvature and morphology of the cardiac surface.

Description

Technical Field

[0001] The present invention relates to the technical field of cardiac MRI image reconstruction, and in particular to a cardiac structure reconstruction system based on MRI. Background Art

[0002] The field of cardiac MRI image reconstruction technology focuses on using magnetic resonance imaging (MRI) technology to obtain high-resolution images of the heart, and reconstructing the three-dimensional structure of the heart through image processing and computing technology. The field combines multiple technologies and methods such as medical imaging, computer vision, image processing, and machine learning to improve the accuracy of diagnosis of cardiac anatomical structure and functional abnormalities. At the same time, cardiac MRI image reconstruction technology can provide a more detailed and accurate view of the heart structure than traditional two-dimensional imaging technology, including the size and shape of each chamber of the heart and the details of the myocardial tissue, which is of great significance for evaluating heart disease, guiding clinical treatment plans, and monitoring treatment effects.

[0003] Among them, the MRI-based cardiac structure reconstruction system is a system that uses the cardiac images captured by MRI technology to reconstruct a three-dimensional model of the heart through software algorithms and calculation processes. It is used to provide more accurate and comprehensive cardiac structural information, assisting doctors to better analyze the anatomical structure and functional status of the heart, especially when diagnosing various heart diseases, heart malformations, and evaluating the situation before and after heart surgery. Through three-dimensional reconstruction, doctors can observe the cardiac structure from different angles and make more accurate diagnosis and treatment decisions.

[0004] Although the existing technology has made significant progress in cardiac MRI image reconstruction and achieved high-resolution views and three-dimensional reconstruction of cardiac structures, there are still some problems. For example, in the analysis of local curvature and morphological changes of the heart surface and the cavitation and connectivity of the heart structure, the subtle differences in cardiac tissue have not been fully revealed, affecting the diagnosis and evaluation of details. In the flexible adjustment of soft thresholds according to image features and noise levels and the precise control of the iteration process, although the number of iterations required for reconstruction can be effectively reduced, it will affect the image quality to a certain extent. Finally, when processing cardiac MRI images with high variability, it is easy to lead to inaccurate identification of key areas, affecting the quality and diagnostic value of the final reconstructed image. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings in the prior art and to propose a cardiac structure reconstruction system based on MRI.

[0006] In order to achieve the above-mentioned object, the present invention adopts the following technical scheme: a cardiac structure reconstruction system based on MRI, the system comprising a soft threshold initial setting module, a threshold adaptive adjustment module, a control parameter optimization module, a cardiac structure detail analysis module, a key area marking module, an image clarity adjustment module, and a reconstruction efficiency optimization module;

[0007] The soft threshold initial setting module analyzes the basic statistical data of the cardiac MRI image, calculates the average brightness and standard deviation of the image set, and selects the initial parameters of the soft threshold processing of the image based on the calculated data to generate an initial threshold parameter set;

[0008] The threshold adaptive adjustment module refines the initial threshold parameter set, adjusts the threshold according to the differences in the internal features and noise levels of the image, matches the detail retention requirements of various regions, and obtains an optimized threshold parameter set;

[0009] The control parameter optimization module uses the optimized threshold parameter set to set the starting conditions of the iterative process, including the number of iterations and the error threshold, and adjusts the control parameters according to the feedback information to establish an optimized control strategy;

[0010] The cardiac structure detail analysis module applies the established optimized control strategy to analyze the cardiac image, identify and describe the changes and features of the internal details of each region, and generate a structural detail description set;

[0011] The key region marking module analyzes and marks key parts of the cardiac structure, including regions that have an impact on diagnosis and treatment, based on the structural detail description set, and identifies and marks a set of cardiac key regions;

[0012] The image definition adjustment module adjusts the image resolution of the key area based on the heart key area set, highlights the key details, and optimizes the processing load of the overall image to obtain the image set after definition adjustment;

[0013] The reconstruction efficiency optimization module applies a parallel processing strategy to perform comprehensive data fusion analysis and generate a three-dimensional heart model based on an initial threshold parameter set, an optimized threshold parameter set, an optimized control strategy, a structural detail description set, a heart key area set and a clarity-adjusted image set.

[0014] The present invention has improvements in that the initial threshold parameter set includes an average brightness value of the image and a standard deviation value of the image, the optimized threshold parameter set includes brightness thresholds and contrast thresholds adjusted for a variety of image regions, the optimized control strategy includes an adjusted number of iterations, a set error tolerance, and a feedback adjustment mechanism, the structural detail description set includes texture features, morphological features, and color distribution of each heart region, the heart key region set includes a valve region, a coronary artery opening, and a ventricular wall thickness region, the clarity-adjusted image set includes images of key regions and optimized processing of the overall image, and the three-dimensional heart model includes geometric information of each chamber of the heart, texture information of the heart surface, and connectivity information of the heart's internal structure.

[0015] The present invention is improved in that the soft threshold initial setting module includes a basic data analysis submodule, a threshold parameter calculation submodule, and an initial parameter integration submodule;

[0016] The basic data analysis submodule calculates the average brightness and standard deviation of the image set based on the basic statistical data of the cardiac MRI images to generate a statistical parameter set;

[0017] The threshold parameter calculation submodule calculates the threshold of each image area based on the statistical parameter set, adjusts and matches multiple image characteristics, and obtains the adjusted threshold parameters;

[0018] The initial parameter integration submodule integrates the threshold settings of each image region based on the adjusted threshold parameters, optimizes the threshold parameters to match the detail requirements of various heart regions, and generates an initial threshold parameter set.

[0019] The present invention is improved in that the threshold adaptive adjustment module includes a regional feature analysis submodule, a noise level assessment submodule, and a threshold dynamic adjustment submodule;

[0020] The regional feature analysis submodule analyzes the brightness, contrast and texture features of multiple regions inside the cardiac MRI image based on the initial threshold parameter set to obtain regional feature analysis results;

[0021] The noise level assessment submodule assesses the noise level of each region based on the regional feature analysis result and refers to the local noise characteristics of the image to generate a noise level assessment result;

[0022] The threshold dynamic adjustment submodule dynamically adjusts the threshold of each area based on the noise level evaluation result, optimizes the balance between detail preservation and noise suppression, and constructs an optimized threshold parameter set.

[0023] The present invention is improved in that the control parameter optimization module includes an iteration condition setting submodule, a control parameter feedback adjustment submodule, and an optimization strategy construction submodule;

[0024] The iteration condition setting submodule evaluates the initial iteration effect in the image reconstruction process based on the optimized threshold parameter set, sets the number of iterations and the error threshold as the start condition, and generates the iteration condition configuration result;

[0025] The control parameter feedback adjustment submodule uses an iterative soft threshold adaptive optimization algorithm to monitor the current iteration progress and error reduction rate in real time according to the iteration condition configuration result. If the progress does not meet expectations and the error reduction does not meet the set standard, the iteration step size and error tolerance are automatically adjusted, and the dynamic error feedback mechanism is used to refine the control parameters to obtain the adjusted control parameters;

[0026] The iterative soft threshold adaptive optimization algorithm is based on Formula I:

[0027] Formula I;

[0028] Calculate parameter updates and generate adjusted control parameters;

[0029] The adaptive learning rate adjustment factor is defined by formula II:

[0030] Formula II;

[0031] in, are model parameters, is the updated model parameter value, is the model parameter value before updating, is the basic learning rate, The loss function parameter The gradient of is the adaptive learning rate adjustment factor, is the current iteration number, is the image feature difference, is the image quality indicator, is the weight coefficient, dynamically adjusted The calculation is determined by applying a genetic algorithm to pre-training data, wherein the pre-training data includes information on image feature differences, image quality indicators, and the number of iterations;

[0032] The optimization strategy construction submodule comprehensively adjusts the control parameters, analyzes the changes in efficiency and accuracy during the iteration process, optimizes the iteration strategy, shortens the reconstruction time while ensuring image quality, and constructs an optimized control strategy.

[0033] The present invention is improved in that the cardiac structure detail analysis module includes a regional internal detail analysis submodule, a detail feature recognition submodule, and a structure detail description integration submodule;

[0034] The regional internal detail analysis submodule performs detailed scanning on each region of the cardiac MRI image based on an optimized control strategy, analyzes brightness, contrast and texture features, reveals structural changes, and generates regional feature analysis results;

[0035] The detail feature recognition submodule applies differential geometry feature extraction and topological data analysis based on the regional feature analysis results to calculate the local curvature and average curvature of the heart surface, identify the cavities and connectivity in the structure, reveal the cardiac tissue structure and geometric features, and obtain detail feature recognition results;

[0036] The differential geometry feature extraction is performed according to Formula III and Formula IV:

[0037] Formula III;

[0038] Formula IV;

[0039] Calculate the local curvature and mean curvature of cardiac MRI images;

[0040] in, and are the two principal curvatures at the center of the cardiac MRI image, and is the curvature offset adjusted based on image features, and is the weight coefficient, which is determined by minimizing the difference between the cardiac structure reconstruction error and the current cardiac structure, is the weighting coefficient in the mean curvature calculation, is a fixed offset, is the local curvature, is the mean curvature;

[0041] The structural detail description integration submodule integrates the detail feature recognition results, records the structural and feature changes of the heart, including the features and mutual relationships of each region, and establishes a structural detail description set.

[0042] The present invention is improved in that the key area marking module includes a key part analysis submodule, a diagnosis and treatment impact area marking submodule, and a heart key area integration submodule;

[0043] The key part analysis submodule scans and identifies structures in the heart image, including ventricles and valves, based on the structure detail description set, analyzes the brightness, contrast and texture differences of the region, and obtains key structure analysis results;

[0044] The diagnosis and treatment impact region marking submodule evaluates the importance of each marked region to diagnosis and treatment based on the key structure analysis results, highlights the key regions using image processing technology, and generates a diagnosis and treatment impact region marking set;

[0045] The cardiac key region integration submodule integrates the diagnosis and treatment-affected region marker set, constructs a complete cardiac key region view, highlights the location and features of each marker region, and establishes a cardiac key region set.

[0046] The present invention is improved in that the image definition adjustment module includes a key area resolution adjustment submodule, a detail highlighting processing submodule, and an image integration submodule;

[0047] The key area resolution adjustment submodule analyzes the requirements of each key area based on the heart key area set, performs resolution optimization operations on the marked key areas, and generates resolution improvement results;

[0048] The detail highlighting processing submodule uses the resolution enhancement result to enhance the key details in the image again, and iteratively highlights the key structural features by adjusting the contrast and sharpening to obtain a detail enhanced image;

[0049] The image integration submodule integrates detail-enhanced images, fuses all processed key area images with the initial MRI image, uniformly adjusts the clarity and visual effects of the overall image set, and establishes an image set with adjusted clarity.

[0050] The present invention is improved in that the reconstruction efficiency optimization module includes a parallel processing submodule, a data fusion analysis submodule, and a three-dimensional heart model generation submodule;

[0051] The parallel processing submodule deploys a parallel computing strategy based on the initial threshold parameter set, the optimized threshold parameter set, the optimized control strategy, the structural detail description set, the cardiac key area set and the clarity-adjusted image set, and synchronously analyzes and processes the data set to generate a parallel processing data set;

[0052] The data fusion analysis submodule utilizes parallel processing of data sets to comprehensively analyze the information in each data set, optimizes the accuracy and reliability of the model by integrating key structural details and image quality adjustment results, and obtains fusion analysis results;

[0053] The three-dimensional heart model generation submodule converts the fused data into a model based on the fusion analysis results using three-dimensional reconstruction technology to reflect the structural and functional characteristics of the heart and establish a three-dimensional heart model.

[0054] Compared with the prior art, the advantages and positive effects of the present invention are:

[0055] In the present invention, the reconstruction process is optimized through the adaptive adjustment of soft threshold and iterative control strategy, and the image quality is guaranteed while the number of iterations is reduced. At the same time, the dynamic adjustment of the soft threshold can accurately control the details of image processing according to the specific characteristics and noise level of the image, and improve the overall clarity of the reconstructed image. By combining differential geometry and topological data analysis technology, the complex geometric characteristics and spatial connectivity of the heart structure can be deeply analyzed, which not only improves the ability to accurately identify the heart structure, but also can capture the slight changes in the local curvature and morphology of the heart surface, providing richer and more accurate information for diagnosing heart disease. The key area marking and image clarity adjustment module can further improve the recognition accuracy and image quality of the key structures of the heart, and provide clinicians with a more accurate image basis to support more accurate diagnosis and treatment planning. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 The module diagram of the cardiac structure reconstruction system based on MRI proposed by the present invention;

[0057] Figure 2 The present invention provides a system framework diagram of a cardiac structure reconstruction system based on MRI;

[0058] Figure 3 A schematic diagram of a soft threshold initialization setting module in a cardiac structure reconstruction system based on MRI proposed by the present invention;

[0059] Figure 4 A schematic diagram of a threshold adaptive adjustment module in a cardiac structure reconstruction system based on MRI proposed by the present invention;

[0060] Figure 5 A schematic diagram of a control parameter optimization module in an MRI-based cardiac structure reconstruction system proposed in the present invention;

[0061] Figure 6 A schematic diagram of a cardiac structure detail analysis module in a cardiac structure reconstruction system based on MRI proposed by the present invention;

[0062] Figure 7 A schematic diagram of a key area marking module in a cardiac structure reconstruction system based on MRI proposed by the present invention;

[0063] Figure 8 A schematic diagram of an image clarity adjustment module in a cardiac structure reconstruction system based on MRI proposed by the present invention;

[0064] Fig. 9 This is a schematic diagram of a reconstruction efficiency optimization module in an MRI-based cardiac structure reconstruction system proposed in the present invention. DETAILED DESCRIPTION

[0065] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0066] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating positions or positional relationships, are based on the positions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined.

[0067] Example: See Figure 1 , the present invention provides a technical solution: a cardiac structure reconstruction system based on MRI, the system includes a soft threshold initial setting module, a threshold adaptive adjustment module, a control parameter optimization module, a cardiac structure detail analysis module, a key area marking module, an image clarity adjustment module, and a reconstruction efficiency optimization module;

[0068] The soft threshold initial setting module analyzes the basic statistical data of the cardiac MRI image, calculates the average brightness and standard deviation of the image set, and selects the initial parameters of the soft threshold processing of the image based on the calculation, thereby generating an initial threshold parameter set;

[0069] The threshold adaptive adjustment module refines the initial threshold parameter set, adjusts the threshold according to the differences in the internal features and noise levels of the image, matches the detail retention requirements of various regions, and obtains the optimized threshold parameter set;

[0070] The control parameter optimization module uses the optimized threshold parameter set to set the starting conditions of the iteration process, including the number of iterations and the error threshold, and adjusts the control parameters according to the feedback information to establish an optimized control strategy;

[0071] The cardiac structure detail analysis module applies the established optimized control strategy to analyze the cardiac image, identify and describe the changes and characteristics of the internal details of each region, and generate a structural detail description set;

[0072] The key area marking module analyzes and marks the key parts of the heart structure based on the structure detail description set, including the areas that have an impact on diagnosis and treatment, and identifies and marks the key area set of the heart;

[0073] The image clarity adjustment module adjusts the image resolution of the key area based on the heart key area set, highlights the key details, and optimizes the processing load of the overall image to obtain an image set after clarity adjustment;

[0074] The reconstruction efficiency optimization module applies a parallel processing strategy to perform comprehensive data fusion analysis and generate a three-dimensional heart model based on the initial threshold parameter set, the optimized threshold parameter set, the optimized control strategy, the structural detail description set, the cardiac key area set and the clarity-adjusted image set.

[0075] The initial threshold parameter set includes the average brightness value of the image and the standard deviation value of the image. The optimized threshold parameter set includes the brightness threshold and contrast threshold adjusted for various image areas. The optimized control strategy includes the adjusted number of iterations, the set error tolerance, and the feedback adjustment mechanism. The structural detail description set includes the texture features, morphological features, and color distribution of each heart area. The key heart area set includes the valve area, coronary artery opening, and ventricular wall thickness area. The clarity-adjusted image set includes key area images and optimized processing of the overall image. The three-dimensional heart model includes the geometric information of each chamber of the heart, the texture information of the heart surface, and the connectivity information of the internal structure of the heart.

[0076] See also Figure 2 and Figure 3 ,The soft threshold initial setting module includes a basic data analysis submodule, a threshold parameter calculation submodule, and an initial parameter integration submodule;

[0077] The basic data analysis submodule calculates the average brightness and standard deviation of the image set based on the basic statistical data of the cardiac MRI images. The specific process of generating the statistical parameter set is as follows:

[0078] The basic data analysis submodule is based on the basic statistical data of cardiac MRI images and uses Python's NumPy library to execute the mean() and std() functions to calculate the average brightness and standard deviation of the image set. The mean() function is used to calculate the average brightness of each pixel in the image set, and the std() function is used to evaluate the standard deviation of the brightness value distribution of the image set. This is used as the basic indicator for image quality assessment and a statistical parameter set is generated.

[0079] The threshold parameter calculation submodule calculates the threshold of each image area based on the statistical parameter set, adjusts and matches multiple image characteristics, and obtains the adjusted threshold parameters in the following specific process:

[0080] The threshold parameter calculation submodule is based on the statistical parameter set and adopts the threshold optimization method based on the gradient descent algorithm. It uses Python's Scikit-learn library to execute the GradientDescentOptimizer() function for threshold calculation. The function takes the average brightness value and the standard deviation value as input parameters, and gradually reduces the error between the predicted threshold and the actual threshold by iteratively adjusting the learning rate parameter alpha. It adjusts and matches multiple image characteristics to generate adjusted threshold parameters.

[0081] The initial parameter integration submodule integrates the threshold settings of each image region based on the adjusted threshold parameters, optimizes the threshold parameters to match the detail requirements of various heart regions, and generates the specific process of the initial threshold parameter set as follows:

[0082] The initial parameter integration submodule is based on the adjusted threshold parameters and adopts vectorized data processing technology. It uses Python's Pandas library to execute the DataFrame() function to integrate the threshold settings of each image area. The DataFrame() function creates a data frame object and stores the threshold parameters of different image areas as column vectors to achieve efficient data integration and access, optimize the threshold parameters to match the detailed requirements of various heart areas, and generate an initial threshold parameter set.

[0083] See also Figure 2 and Figure 4 ,The threshold adaptive adjustment module includes a regional feature analysis submodule, a noise level assessment submodule, and a threshold dynamic adjustment submodule;

[0084] The regional feature analysis submodule analyzes the brightness, contrast and texture features of multiple regions in the cardiac MRI image based on the initial threshold parameter set. The specific process of obtaining the regional feature analysis results is as follows:

[0085] The regional feature analysis submodule is based on the initial threshold parameter set and adopts image processing technology. It uses the cv2.calcHist() function in the OpenCV library to perform brightness and contrast analysis of multiple regions inside the cardiac MRI image. The brightness and contrast characteristics of each region are obtained by calculating the histogram distribution of the image. At the same time, the cv2.Laplacian() function is used to analyze the texture characteristics of the image. The sharpness of the texture is obtained by calculating the Laplacian operator of the image. The brightness, contrast and texture characteristics are combined to generate the regional feature analysis results.

[0086] The noise level assessment submodule assesses the noise level of each area based on the regional feature analysis results and the local noise characteristics of the image. The specific process of generating the noise level assessment results is as follows:

[0087] The noise level assessment submodule is based on the regional feature analysis results and adopts local variance estimation technology. It uses the np.var() function of the NumPy library to calculate the variance of the image data in each region. The variance value reflects the degree of fluctuation of the image signal. Regions with larger fluctuations are considered to have higher noise levels. By comparing the variance values ​​of different regions, the noise level of each region is evaluated to generate a noise level assessment result.

[0088] The threshold dynamic adjustment submodule dynamically adjusts the threshold of each area based on the noise level evaluation results to optimize the balance between detail preservation and noise suppression. The specific process of constructing the optimized threshold parameter set is as follows:

[0089] The threshold dynamic adjustment submodule adopts an adaptive threshold adjustment algorithm based on the noise level assessment results, and uses MATLAB's imadjust() function to dynamically adjust the threshold of each area. The imadjust() function adjusts the grayscale value range of the image according to the input image data and noise assessment results. This method optimizes the balance between detail retention and noise suppression. The adjustment operation adjusts the brightness and contrast thresholds according to the noise assessment results to generate an optimized threshold parameter set.

[0090] See also Figure 2 and Figure 5 ,The control parameter optimization module includes an iterative condition setting submodule, a control parameter feedback adjustment submodule, and an optimization strategy building submodule;

[0091] The iterative condition setting submodule evaluates the initial iteration effect in the image reconstruction process based on the optimized threshold parameter set, sets the number of iterations and the error threshold as the starting conditions, and generates the specific process of the iterative condition configuration result as follows:

[0092] The iterative condition setting submodule is based on the optimized threshold parameter set and adopts numerical analysis methods. It uses the optimize.minimize() function in Python's SciPy library to evaluate the initial iteration effect in the image reconstruction process. The optimize.minimize() function minimizes the error by defining an objective function. The objective function is calculated based on the difference between the reconstructed image and the original image. The function also sets the number of iterations and the error threshold as constraints of the optimization process to generate an iterative condition configuration result.

[0093] The control parameter feedback adjustment submodule uses the iterative soft threshold adaptive optimization algorithm to monitor the current iteration progress and error reduction rate in real time according to the iteration condition configuration results. If the progress does not meet expectations and the error reduction does not meet the set standard, the iteration step size and error tolerance are automatically adjusted, and the dynamic error feedback mechanism is used to refine the control parameters. The specific process of obtaining the adjusted control parameters is as follows:

[0094] The control parameter feedback adjustment submodule configures the results according to the iteration conditions and uses the tf.train.GradientDescentOptimizer() function of the TensorFlow framework to implement the iterative soft threshold adaptive optimization algorithm. The tf.train.GradientDescentOptimizer() function adjusts the learning rate parameters to monitor the current iteration progress and error reduction rate in real time, automatically adjusts the algorithm's iteration step size and error tolerance, implements a dynamic error feedback mechanism, refines the control parameters, and generates adjusted control parameters.

[0095] The iterative soft threshold adaptive optimization algorithm is based on Formula I:

[0096] Formula I;

[0097] Calculate parameter updates and generate adjusted control parameters;

[0098] The adaptive learning rate adjustment factor is defined by formula II:

[0099] Formula II;

[0100] in, are model parameters, representing the variables that need to be optimized, such as weights and bias terms in the image reconstruction model. is the updated model parameter value, is the model parameter value before updating, is the basic learning rate, which controls the step size of each parameter update. The loss function parameter The gradient indicates the direction and magnitude of the change of the loss function under the current parameters. is the adaptive learning rate adjustment factor, is the current iteration number, is the image feature difference, is the image quality indicator, is the weight coefficient, dynamically adjusted The calculation is determined by applying a genetic algorithm on the pre-training data, and a loss function is defined to evaluate the performance of the model under different weight coefficients, and the coefficients are iteratively optimized to minimize the loss function. The pre-training data includes information on the feature difference of the image, image quality indicators, and the number of iterations;

[0101] The execution process is as follows:

[0102] S1: Before training, first initialize the model parameters Parameters include weights and bias terms in the image reconstruction model, and setting the base learning rate The initial value of

[0103] S2: Collect pre-training data for evaluating model performance and adjusting learning rate during training;

[0104] S3: For each training instance, calculate the current model parameters The loss function under Gradient , the gradient indicates the direction and magnitude of change of the loss function under the current parameters;

[0105] S4: According to the current number of iterations , image feature difference and image quality indicators , dynamically calculate the learning rate adjustment factor , and the final factor is calculated by formula II, where is the base of natural logarithms;

[0106] S5: Use the calculated dynamic learning rate factor , and the gradient of the loss function , update the model parameters through formula I. Reduce the value of the loss function to improve the performance of the model;

[0107] S6: Repeat steps S3 to S5 for each training sample or each group of batch processed samples until a predetermined number of iterations is reached or the model performance meets a predetermined termination condition, such as a decrease in the loss function is less than a certain threshold.

[0108] The optimization strategy construction submodule comprehensively adjusts the control parameters, analyzes the changes in efficiency and accuracy during the iteration process, optimizes the iteration strategy, shortens the reconstruction time while ensuring image quality, and the specific process of constructing the optimized control strategy is as follows:

[0109] The optimization strategy construction sub-module comprehensively adjusts the control parameters, adopts simulation technology, and uses MATLAB's Simulink environment to analyze the efficiency and accuracy changes of the iterative process. Simulink constructs a dynamic system model of image reconstruction, simulates the iterative process under different control parameters, and evaluates the impact of parameter adjustment on reconstruction time and image quality. According to the simulation results, the iterative strategy is optimized to generate the optimized control strategy.

[0110] See also Figure 2 and Figure 6 ,The cardiac structure detail analysis module includes a regional internal detail analysis submodule, a detail feature recognition submodule, and a structure detail description integration submodule;

[0111] Based on the optimized control strategy, the regional internal detail analysis submodule performs detailed scanning on each region of the cardiac MRI image, analyzes the brightness, contrast and texture features, reveals structural changes, and generates the specific process of regional feature analysis results as follows:

[0112] The regional internal detail analysis submodule is based on an optimized control strategy and adopts image enhancement technology. It uses the ImageEnhance module in Python's PIL library to perform a detailed scan of each area of ​​the cardiac MRI image. The ImageEnhance module enhances image details by adjusting brightness, contrast, and sharpness to make the structural changes in the image more obvious. At the same time, the ImageFilter module is used to apply edge detection filters to highlight the texture features of the image. The enhanced brightness, contrast, and texture features are combined to generate regional feature analysis results.

[0113] The detail feature recognition submodule uses differential geometry feature extraction and topological data analysis based on the regional feature analysis results to calculate the local curvature and average curvature of the heart surface, identify the voids and connectivity in the structure, reveal the cardiac tissue structure and geometric features, and obtain the specific process of detail feature recognition results as follows:

[0114] Based on the results of regional feature analysis, the detail feature recognition submodule adopts computational geometry algorithms and topological principles, and uses the ndimage module in Python's SciPy library to perform differential geometry feature extraction and topological data analysis. The ndimage module evaluates the local curvature and average curvature of the heart surface by calculating the high-order derivatives of the image data, and uses SciPy's spatial module to analyze the spatial relationship of the image, identify the voids and connectivity in the structure, and reveal the cardiac tissue structure and geometric features to generate detail feature recognition results.

[0115] Differential geometry feature extraction, according to formula III and formula IV:

[0116] Formula III;

[0117] Formula IV;

[0118] Calculate the local curvature and mean curvature of cardiac MRI images;

[0119] in, and The two principal curvatures at the center of the cardiac MRI image are used to describe the geometric shape of the local area of ​​the image. In cardiac structure analysis, it helps to reveal the concave and convex characteristics of the heart surface. and is the curvature offset adjusted based on image features. In order to adjust the original curvature value according to the local features of the image and Make fine adjustments to increase sensitivity and accuracy to detail. and is the weight coefficient, which is determined by minimizing the difference between the cardiac structure reconstruction error and the current cardiac structure, and is used to adjust and The weight coefficient is introduced to consider that the importance of local structure description may be different in different image regions, so as to achieve the best image interpretation effect. is the weighting factor in the mean curvature calculation, balancing and The effect of the mean curvature on the calculation of the curvature. The contribution ratio of the two principal curvatures can be more finely controlled during the calculation process to adapt to the characteristics of different heart structures. is a fixed offset used in the adjustment of the average curvature. This enables the algorithm to adapt to the quality differences of different cardiac structures and MRI images, increasing the flexibility and adaptability of the calculation. is the local curvature, is the mean curvature.

[0120] The execution process is as follows:

[0121] S1: Preprocess the cardiac MRI images, including denoising, normalizing brightness and contrast, and extracting the basic features of the cardiac MRI images;

[0122] S2: Use the second-order derivative of the image to calculate the two principal curvatures at each pixel and ;

[0123] S3: Determine the weight coefficients using a set of training data by minimizing the difference between the reconstructed model and the actual heart structure and and The training data includes a set of cardiac MRI images, accurate cardiac structure annotations corresponding to the images, geometric feature descriptors of the images, and clinical parameters of cardiac structures;

[0124] S4: Weight coefficient determined according to formula III and And the local feature adjustment amount, calculate the improved local curvature of each pixel. Calculate the improved average curvature according to formula IV;

[0125] S5: Determine the fixed offset , based on the statistical data in the training set, a global adjustment term is introduced when calculating the mean curvature to improve the adaptability of the model between different image qualities and cardiac structural variations;

[0126] S6: Perform 3D reconstruction of the cardiac structure using the improved local curvature and mean curvature information. Further optimize the reconstruction results based on the difference between the reconstruction results and the actual data. Ensure that the final reconstructed heart model is close to the real structure.

[0127] The structural detail description integration submodule integrates the detailed feature recognition results and records the structural and characteristic changes of the heart, including the characteristics and mutual relationships of each area. The specific process of establishing the structural detail description set is as follows:

[0128] The structural detail description integration submodule integrates the detail feature recognition results, adopts data fusion technology, and uses Python's Pandas library to create a DataFrame object to record the structural and feature changes of the heart. The DataFrame object stores the characteristics and relationships of each area as column data, providing an efficient structured data expression to support subsequent data analysis and visualization operations, and to establish a structural detail description set.

[0129] See also Figure 2 and Figure 7 ,The key area marking module includes the key part analysis submodule, the diagnosis and treatment affected area marking submodule, and the heart key area integration submodule;

[0130] The key parts analysis submodule scans and identifies structures in the heart image, including ventricles and valves, based on the structure detail description set, and analyzes the brightness, contrast, and texture differences of the region. The specific process of obtaining the key structure analysis results is as follows:

[0131] The key parts analysis submodule is based on a set of structural detail descriptions and uses feature recognition technology. It uses the cv2.matchTemplate() function in Python's OpenCV library for template matching. The cv2.matchTemplate() function identifies structures such as ventricles and valves by searching for areas that match predefined templates in cardiac MRI images, and combines it with the cv2.calcHist() function to calculate the brightness, contrast, and texture histograms of the matching areas, evaluate the differences in image characteristics of the regions, and generate key structure analysis results.

[0132] The diagnosis and treatment impact region marking submodule evaluates the importance of each marked region to diagnosis and treatment based on the key structure analysis results, and uses image processing technology to highlight the key regions. The specific process of generating the diagnosis and treatment impact region marking set is as follows:

[0133] The diagnosis and treatment impact area marking submodule adopts image annotation technology based on the key structure analysis results, and uses the patches module in Python's matplotlib library to add graphic annotations. The patches module highlights the area by drawing color-coded rectangular or circular marks on the identified cardiac structure area, and combines the numpy library to perform feature enhancement processing on the image area, thereby enhancing the visualization of key areas, emphasizing the diagnostic and treatment importance of each marked area, and generating a diagnosis and treatment impact area marking set.

[0134] The cardiac key region integration submodule integrates the diagnosis and treatment of the affected region marker set, builds a complete cardiac key region view, and highlights the location and characteristics of each marker region. The specific process of establishing the cardiac key region set is as follows:

[0135] The cardiac key area integration submodule integrates the marker set of the affected areas of comprehensive diagnosis and treatment, adopts image synthesis technology, and uses the Image.blend() function in Python's PIL library to fuse the original cardiac MRI image with the marker image. The Image.blend() function sets appropriate transparency parameters to maintain the details of the original image while highlighting the markers of the key areas. In this way, a complete cardiac key area view is constructed and a cardiac key area set is established.

[0136] See also Figure 2 and Figure 8 ,The image clarity adjustment module includes a key area resolution adjustment submodule, a detail highlighting processing submodule, and an image integration submodule;

[0137] The key area resolution adjustment submodule analyzes the requirements of each key area based on the heart key area set, performs resolution optimization operations on the marked key areas, and generates the specific process of resolution improvement results as follows:

[0138] The key area resolution adjustment submodule is based on the heart key area set and adopts super-resolution reconstruction technology. It uses the cv2.resize() function in Python's OpenCV library to optimize the resolution of the marked key areas. The cv2.resize() function increases the resolution of the image through the interpolation algorithm and upsamples the image by using the bicubic interpolation method, thereby improving the resolution of the image in specific areas, ensuring that the details are more clearly visible and generating resolution improvement results.

[0139] The detail highlighting submodule uses the resolution enhancement result to enhance the key details in the image again. By adjusting the contrast and sharpening, the key structural features are iteratively highlighted. The specific process of obtaining the detail enhanced image is as follows:

[0140] The detail highlighting processing submodule uses the resolution enhancement result and the image enhancement algorithm to enhance the key details in the image again. The contrast is adjusted through Python's ImageEnhance module and the cv2.filter2D() function is used for sharpening. The cv2.filter2D() function defines a high-pass filter to enhance the image edges, making the key structural features in the image more prominent. The image details are iteratively enhanced to obtain a detail-enhanced image.

[0141] The image integration submodule integrates the detail enhanced images, fuses all processed key area images with the initial MRI images, uniformly adjusts the clarity and visual effects of the overall image set, and establishes the image set after clarity adjustment. The specific process is as follows:

[0142] The image integration submodule integrates detail-enhanced images and uses image fusion technology to fuse all processed key area images with the initial MRI image. The Python PIL library is used for image fusion operations, and the Image.blend() function is used to proportionally mix the two images pixel by pixel to uniformly adjust the clarity and visual effects of the entire image set and establish an image set with adjusted clarity.

[0143] See also Figure 2 and Fig. 9 ,The reconstruction efficiency optimization module includes a parallel processing submodule, a data fusion analysis submodule, and a 3D heart model generation submodule;

[0144] The parallel processing submodule deploys parallel computing strategies based on the initial threshold parameter set, the optimized threshold parameter set, the optimized control strategy, the structural detail description set, the cardiac key area set, and the clarity-adjusted image set, and performs synchronous analysis and processing on the data set. The specific process of generating the parallel processing data set is as follows:

[0145] The parallel processing submodule adopts a parallel computing framework based on the initial threshold parameter set, optimized threshold parameter set, optimized control strategy, structural detail description set, cardiac key area set and clarity-adjusted image set, and uses Python's multiprocessing library to deploy parallel computing strategies. The multiprocessing library creates multiple processes to allow synchronous analysis and processing of data sets to be executed in parallel on different cores, thereby accelerating data processing speed, reducing overall execution time, and generating parallel processing data sets.

[0146] The data fusion analysis submodule uses parallel processing of data sets to comprehensively analyze the information in each data set. By integrating key structural details and image quality adjustment results, the accuracy and reliability of the model are optimized. The specific process of obtaining the fusion analysis results is as follows:

[0147] The data fusion analysis submodule utilizes parallel processing of data sets, adopts data fusion algorithm, and uses the np.concatenate() function in Python's NumPy library to comprehensively analyze the information in each data set. The np.concatenate() function merges arrays from different data sources, integrates key structural details and image quality adjustment results, improves the accuracy and reliability of the model, ensures the integrity and accuracy of the final 3D reconstructed model, and obtains fusion analysis results.

[0148] The 3D heart model generation submodule uses 3D reconstruction technology to transform the fused data into a model based on the fusion analysis results to reflect the structural and functional characteristics of the heart. The specific process of establishing a 3D heart model is as follows:

[0149] The 3D heart model generation submodule applies 3D reconstruction technology based on the fusion analysis results and uses the VTK library to convert the fused data into a model. The VTK library reads the processed data set and uses voxelization and surface rendering technology to generate a 3D geometric model of the heart. The model accurately reflects the structural and functional characteristics of the heart, supports further medical analysis and diagnosis processes, and establishes a 3D heart model.

[0150] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. MRI-based cardiac structure reconstruction system, characterized by: The system includes a soft threshold initial setting module, a threshold adaptive adjustment module, a control parameter optimization module, a cardiac structure detail analysis module, a key area marking module, an image clarity adjustment module, and a reconstruction efficiency optimization module; The soft threshold initial setting module analyzes the basic statistical data of the cardiac MRI image, calculates the average brightness and standard deviation of the image set, and selects the initial parameters of the soft threshold processing of the image based on the calculated data to generate an initial threshold parameter set; The threshold adaptive adjustment module refines the initial threshold parameter set, adjusts the threshold according to the differences in the internal features and noise levels of the image, matches the detail retention requirements of various regions, and obtains an optimized threshold parameter set; The control parameter optimization module uses the optimized threshold parameter set to set the starting conditions of the iterative process, including the number of iterations and the error threshold, and adjusts the control parameters according to the feedback information to establish an optimized control strategy; The cardiac structure detail analysis module applies the established optimized control strategy to analyze the cardiac image, identify and describe the changes and features of the internal details of each region, and generate a structural detail description set; The key region marking module analyzes and marks key parts of the cardiac structure, including regions that have an impact on diagnosis and treatment, based on the structural detail description set, and identifies and marks a set of cardiac key regions; The image definition adjustment module adjusts the image resolution of the key area based on the heart key area set, highlights the key details, and optimizes the processing load of the overall image to obtain the image set after definition adjustment; The reconstruction efficiency optimization module applies a parallel processing strategy to perform comprehensive data fusion analysis and generate a three-dimensional heart model based on an initial threshold parameter set, an optimized threshold parameter set, an optimized control strategy, a structural detail description set, a heart key area set and a clarity-adjusted image set.

2. The MRI-based cardiac structure reconstruction system according to claim 1, characterized in that: The initial threshold parameter set includes the average brightness value of the image and the standard deviation value of the image, the optimized threshold parameter set includes the brightness threshold and contrast threshold adjusted for multiple image areas, the optimized control strategy includes the adjusted number of iterations, the set error tolerance, and the feedback adjustment mechanism, the structural detail description set includes the texture features, morphological features, and color distribution of each heart area, the heart key area set includes the valve area, the coronary artery opening, and the ventricular wall thickness area, the clarity-adjusted image set includes the key area image and the optimization processing of the overall image, and the three-dimensional heart model includes the geometric information of each chamber of the heart, the texture information of the heart surface, and the connectivity information of the internal structure of the heart.

3. The MRI-based cardiac structure reconstruction system according to claim 1, characterized in that: The soft threshold initial setting module includes a basic data analysis submodule, a threshold parameter calculation submodule, and an initial parameter integration submodule; The basic data analysis submodule calculates the average brightness and standard deviation of the image set based on the basic statistical data of the cardiac MRI images to generate a statistical parameter set; The threshold parameter calculation submodule calculates the threshold of each image area based on the statistical parameter set, adjusts and matches multiple image characteristics, and obtains the adjusted threshold parameters; The initial parameter integration submodule integrates the threshold settings of each image region based on the adjusted threshold parameters, optimizes the threshold parameters to match the detail requirements of various heart regions, and generates an initial threshold parameter set.

4. The MRI-based cardiac structure reconstruction system according to claim 1, characterized in that: The threshold adaptive adjustment module includes a regional feature analysis submodule, a noise level assessment submodule, and a threshold dynamic adjustment submodule; The regional feature analysis submodule analyzes the brightness, contrast and texture features of multiple regions inside the cardiac MRI image based on the initial threshold parameter set to obtain regional feature analysis results; The noise level assessment submodule assesses the noise level of each region based on the regional feature analysis result and refers to the local noise characteristics of the image to generate a noise level assessment result; The threshold dynamic adjustment submodule dynamically adjusts the threshold of each region based on the noise level evaluation result, optimizes the balance between detail preservation and noise suppression, and constructs an optimized threshold parameter set.

5. The cardiac structure reconstruction system based on MRI according to claim 1, characterized in that: The control parameter optimization module includes an iteration condition setting submodule, a control parameter feedback adjustment submodule, and an optimization strategy construction submodule; The iteration condition setting submodule evaluates the initial iteration effect in the image reconstruction process based on the optimized threshold parameter set, sets the number of iterations and the error threshold as the start condition, and generates the iteration condition configuration result; The control parameter feedback adjustment submodule uses an iterative soft threshold adaptive optimization algorithm to monitor the current iteration progress and error reduction rate in real time according to the iteration condition configuration result. If the progress does not meet expectations and the error reduction does not meet the set standard, the iteration step size and error tolerance are automatically adjusted, and the dynamic error feedback mechanism is used to refine the control parameters to obtain the adjusted control parameters; The iterative soft threshold adaptive optimization algorithm is according to Formula I: Formula I; Calculate parameter updates and generate adjusted control parameters; The adaptive learning rate adjustment factor is defined by formula II: Formula II; in, are model parameters, is the updated model parameter value, is the model parameter value before updating, is the basic learning rate, The loss function parameter The gradient of is the adaptive learning rate adjustment factor, is the current iteration number, is the image feature difference, is the image quality indicator, is the weight coefficient, dynamically adjusted The calculation is determined by applying a genetic algorithm to pre-training data, wherein the pre-training data includes information on image feature differences, image quality indicators, and the number of iterations; The optimization strategy construction submodule comprehensively adjusts the control parameters, analyzes the changes in efficiency and accuracy during the iteration process, optimizes the iteration strategy, shortens the reconstruction time while ensuring image quality, and constructs an optimized control strategy.

6. The MRI-based cardiac structure reconstruction system according to claim 1, characterized in that: The cardiac structure detail analysis module includes a regional internal detail analysis submodule, a detail feature recognition submodule, and a structure detail description integration submodule; The regional internal detail analysis submodule performs detailed scanning on each region of the cardiac MRI image based on an optimized control strategy, analyzes brightness, contrast and texture features, reveals structural changes, and generates regional feature analysis results; The detail feature recognition submodule applies differential geometry feature extraction and topological data analysis based on the regional feature analysis results to calculate the local curvature and average curvature of the heart surface, identify the cavities and connectivity in the structure, reveal the cardiac tissue structure and geometric features, and obtain detail feature recognition results; The differential geometry feature extraction is performed according to Formula III and Formula IV: Formula III; Formula IV; Calculate the local curvature and mean curvature of cardiac MRI images; in, and are the two principal curvatures at the center of the cardiac MRI image, and is the curvature offset adjusted based on image features, and is the weight coefficient, which is determined by minimizing the difference between the cardiac structure reconstruction error and the current cardiac structure, is the weighting coefficient in the mean curvature calculation, is a fixed offset, is the local curvature, is the mean curvature; The structural detail description integration submodule integrates the detail feature recognition results, records the structural and feature changes of the heart, including the features and mutual relationships of each region, and establishes a structural detail description set.

7. The MRI-based cardiac structure reconstruction system according to claim 1, characterized in that: The key area marking module includes a key part analysis submodule, a diagnosis and treatment affected area marking submodule, and a heart key area integration submodule; The key part analysis submodule scans and identifies structures in the heart image, including ventricles and valves, based on the structure detail description set, analyzes the brightness, contrast and texture differences of the region, and obtains key structure analysis results; The diagnosis and treatment impact region marking submodule evaluates the importance of each marked region to diagnosis and treatment based on the key structure analysis results, highlights the key regions using image processing technology, and generates a diagnosis and treatment impact region marking set; The cardiac key region integration submodule integrates the diagnosis and treatment-affected region marker set, constructs a complete cardiac key region view, highlights the location and features of each marker region, and establishes a cardiac key region set.

8. The MRI-based cardiac structure reconstruction system according to claim 1, characterized in that: The image definition adjustment module includes a key area resolution adjustment submodule, a detail highlighting processing submodule, and an image integration submodule; The key area resolution adjustment submodule analyzes the requirements of each key area based on the heart key area set, performs resolution optimization operations on the marked key areas, and generates resolution improvement results; The detail highlighting processing submodule uses the resolution enhancement result to enhance the key details in the image again, and iteratively highlights the key structural features by adjusting the contrast and sharpening to obtain a detail enhanced image; The image integration submodule integrates detail-enhanced images, fuses all processed key area images with the initial MRI image, uniformly adjusts the clarity and visual effects of the overall image set, and establishes an image set with adjusted clarity.

9. The MRI-based cardiac structure reconstruction system according to claim 1, characterized in that: The reconstruction efficiency optimization module includes a parallel processing submodule, a data fusion analysis submodule, and a three-dimensional heart model generation submodule; The parallel processing submodule deploys a parallel computing strategy based on the initial threshold parameter set, the optimized threshold parameter set, the optimized control strategy, the structural detail description set, the cardiac key area set and the clarity-adjusted image set, and synchronously analyzes and processes the data set to generate a parallel processing data set; The data fusion analysis submodule utilizes parallel processing of data sets to comprehensively analyze the information in each data set, optimizes the accuracy and reliability of the model by integrating key structural details and image quality adjustment results, and obtains fusion analysis results; The three-dimensional heart model generation submodule converts the fused data into a model based on the fusion analysis results using three-dimensional reconstruction technology to reflect the structural and functional characteristics of the heart and establish a three-dimensional heart model.

Images