Nuclear magnetic image scanning data reconstruction method and system

Abstract

The invention discloses a nuclear magnetic image scanning data reconstruction method and system, and belongs to the field of medical imaging. According to the method, the problem of poor quality of the reconstructed image in the prior art is solved, the motion artifacts can be corrected by adopting a navigation echo method and a motion compensation method in the scanning process of the MRI scanning equipment, so that the quality of the reconstructed image is improved, and the image quality is improved. Nuclear magnetic image scanning data are obtained and then subjected to denoising and enhancement and data standardization and normalization, the reliability can be improved, the quality of reconstructed images is further improved, the nuclear magnetic image scanning data are reconstructed through the neural network model, quality changes of the reconstructed images under different MRI scanning devices or different scanning parameters are analyzed, and the reconstruction efficiency is improved. According to the method, the effect of the reconstructed image is verified, the reconstructed image is subjected to image segmentation and image registration processing, and the quality of the reconstructed image is evaluated through quantitative and qualitative methods, so that the quality of the reconstructed image can be evaluated more comprehensively and more accurately.

Description

Technical Field

[0001] The present invention relates to the field of medical imaging technology, and in particular to a method and system for reconstructing nuclear magnetic resonance image scanning data. Background Art

[0002] Magnetic Resonance Imaging (MRI) is a non-invasive medical imaging technology that uses strong magnetic fields and harmless radio waves to generate detailed images of the human body. However, in the process of MRI scanning, data acquisition and image reconstruction are crucial steps that can directly affect image quality and diagnostic accuracy. During data acquisition, different devices or different scanning parameters will lead to different reconstructed image quality. In addition, due to the uncertainty of the physiological movement of the research object, motion artifacts are prone to occur during data acquisition, thereby reducing the quality of the reconstructed image; therefore, it does not meet the existing needs. In this regard, we propose a method and system for reconstructing nuclear magnetic image scanning data. Summary of the invention

[0003] The purpose of the present invention is to provide a method and system for reconstructing nuclear magnetic image scanning data. During the scanning process of an MRI scanning device, a navigation echo method and a motion compensation method are used to correct motion artifacts caused by the physiological activities of a research object, thereby improving the quality of the reconstructed image. After obtaining the nuclear magnetic image scanning data of the research object, denoising and enhancement and data standardization and normalization are performed, which can improve the reliability of the nuclear magnetic image scanning data, thereby further improving the quality of the reconstructed image. Then, the nuclear magnetic image scanning data is reconstructed using a neural network model, and the quality changes of the reconstructed images under different MRI scanning devices or different scanning parameters are analyzed based on the reconstructed images to verify the effect of the reconstructed images. Finally, the reconstructed images are segmented and registered, and the quality of the reconstructed images is evaluated by quantitative and qualitative methods, thereby solving the problems raised in the above-mentioned background technology.

[0004] To achieve the above object, the present invention provides the following technical solution: a method for reconstructing nuclear magnetic resonance image scanning data, comprising the following steps: Set the scanning parameters of the MRI scanning equipment, including repetition time, echo time and matrix size; Placing the research object on the MRI scanning device, starting the MRI scanning device to scan and record the data during the imaging process, and obtaining the nuclear magnetic resonance image scanning data of the research object; Process the acquired nuclear magnetic resonance image scanning data, including denoising and enhancement and data standardization and normalization; Construct a neural network model and perform training and testing, and use the trained neural network model to reconstruct the MRI scan data; The reconstructed images are further processed, including segmentation and registration, and the quality of the reconstructed images is evaluated by quantitative and qualitative methods.

[0005] Furthermore, the quality of the reconstructed image is evaluated by quantitative and qualitative methods, including: The reconstructed image is scaled according to a plurality of preset sizes to obtain a scaled image corresponding to each preset size, feature extraction is performed on all scaled images to obtain image features, and the image features are unified to obtain standard features; Based on the standard features of the reconstructed image and the scaled image, the quality value of the reconstructed image is calculated according to the following formula ; Where n represents the number of scaled images, m represents the number of features, represents the jth standard eigenvalue in the reconstructed image, represents the jth standard eigenvalue in the i-th scaled image, represents the scaling weight value of the i-th scaled image; Determining whether the quality value of the reconstructed image is greater than a preset quality value; If so, determining that the clarity of the reconstructed image meets the quality requirements; Otherwise, determining that the clarity of the reconstructed image does not meet the quality requirement; After determining that the clarity of the reconstructed image meets the quality requirements, an initial reconstructed image before registration is obtained, and a cross-correlation coefficient between the reconstructed image and the initial reconstructed image is calculated according to the following formula; in, represents the cross-correlation coefficient between the reconstructed image and the initial reconstructed image, represents the covariance between the reconstructed image and the initial reconstructed image, represents the variance of the reconstructed image, represents the variance of the initial reconstructed image, represents the expected variance of the reconstructed image based on the initial reconstructed image, represents the expected variance of the initial reconstructed image based on the reconstructed image; Determining whether the cross-correlation coefficient between the reconstructed image and the initial reconstructed image meets a preset coefficient requirement; If so, determining that the artifacts of the reconstructed image meet the quality requirements; Otherwise, it is determined that the artifacts of the reconstructed image do not meet the quality requirement.

[0006] Furthermore, during the scanning process of the MRI scanning device, corresponding correction methods are used to correct motion artifacts in view of artifact problems caused by physiological activities of the research subject. The correction methods include navigation echo methods and motion compensation methods.

[0007] Furthermore, the correction method is specifically as follows: Navigation echo method, used for: During the scanning process of the MRI scanning device, the image artifacts caused by respiratory physiological activities are tracked and corrected, and the diaphragm movement is tracked using 2D radio frequency pulses. Data is collected when the diaphragm movement is within the set acceptable range. Motion compensation methods for: By describing the movement of each small block in adjacent frames of the nuclear magnetic resonance image scanning data, each small block of the previous frame is moved to the corresponding position of the current frame, thereby reducing the spatial redundancy and motion artifacts in the nuclear magnetic resonance image scanning data sequence.

[0008] Furthermore, the denoising and enhancement are specifically as follows: Denoising: Denoising algorithms are used to denoise the MRI scan data to remove noise and retain image details. Denoising algorithms include wavelet transform or non-local mean filtering; Enhancement processing: Enhancement processing methods are used to enhance the quality, contrast and clarity of nuclear magnetic resonance image scanning data. Enhancement processing methods include spatial domain enhancement methods and frequency domain enhancement methods.

[0009] Furthermore, the data standardization and normalization are specifically as follows: The data format is unified for: Unify the scanning data formats of nuclear magnetic resonance images obtained by different MRI scanning devices or different scanning parameters; Data normalization for: Standardize the MRI scan data to eliminate the differences between different MRI scan data; Data normalization, used to: Map the MRI scan data to a specific range, which is [0,1] or [-1,1].

[0010] Furthermore, a neural network model is constructed and trained and tested, which specifically includes the following processes: Collecting magnetic resonance imaging scan data of 10 research subjects to obtain a first data set, and dividing the first data set into a training set and a test set; By using different MRI scanning devices or adjusting scanning parameters, magnetic resonance imaging scan data of another 10 research subjects are collected to obtain a second data set, and the second data set is divided into a training set and a test set; Construct a neural network model, and use the training set and test set to train and test the neural network model respectively; The neural network model is trained and tested using the training set and the test set divided by the first data set, and the trained neural network model is used to reconstruct the nuclear magnetic resonance image scanning data to obtain a first reconstructed image; The neural network model is trained and tested using the training set and the test set divided by the second data set, and the trained neural network model is used to reconstruct the nuclear magnetic resonance image scanning data to obtain a second reconstructed image; The first reconstructed image is compared with the second reconstructed image, and the robustness of the neural network model is verified by comparing and analyzing the quality changes of the reconstructed images under different MRI scanning devices or different scanning parameters; After verification, the neural network model is deployed to reconstruct the MRI image scanning data.

[0011] Furthermore, the reconstructed image is segmented and registered, specifically: Image segmentation: Image segmentation is used to divide the reconstructed image into non-overlapping and characteristic regions; Image registration: Image registration is used to perform geometric transformation on the reconstructed image so that the reconstructed image can be aligned in space.

[0012] Furthermore, geometric transformation of the reconstructed image by image registration includes: Initializing the interpolation processing of the reconstructed image according to a preset interpolation method to obtain an interpolation pixel position of the reconstructed image, and determining the pixel grayscale value of the interpolation pixel position based on the pixel values ​​of the pixels around the interpolation pixel position and the interpolation direction to obtain an interpolated image; Based on feature point matching, the reconstructed image is deformed to the space where the interpolated image is located to obtain an initial registration image; Extracting image feature information from the initial registered image, and selecting a feature point set consisting of main feature points based on the image feature information as a registration basis, and re-registering the initial registered image based on the registration basis to obtain a second registered image; Extracting mutual information between the second registered images, and establishing a mutual information matrix based on the image sequence and information attributes of the second registered images; Based on the difference between the horizontal correlation relationship between the mutual information matrices and the preset horizontal correlation relationship, a horizontal optimization value is obtained, and based on the difference between the vertical correlation relationship between the mutual information matrices and the preset vertical correlation relationship, a vertical optimization value is obtained; Acquire initial optimization parameters of the optimization algorithm, and adjust the initial optimization parameters based on the lateral optimization value and the longitudinal optimization value to obtain target optimization parameters; Acquire an initial transformation relationship from the reconstructed image to the second registered image, and optimize the initial transformation relationship based on the target optimization parameter to obtain a target transformation relationship; The reconstructed image is geometrically transformed based on the target transformation relationship to obtain a target registration image. A nuclear magnetic image scanning data reconstruction system, used for a nuclear magnetic image scanning data reconstruction method, comprising: Data acquisition unit for: The scanning parameters of the MRI scanning device are set, and the data of the imaging process is scanned and recorded using the MRI scanning device. During the scanning process, the navigation echo method and motion compensation method are used to correct the motion artifacts, and finally the nuclear magnetic resonance image scanning data of the research object is obtained.

[0013] Data processing unit for: De-noise and enhance the acquired nuclear magnetic resonance image scanning data and perform data standardization and normalization; An image reconstruction unit, for: Construct a neural network model and conduct training and testing. Use the trained neural network model to reconstruct the nuclear magnetic resonance image scanning data, and analyze the quality changes of the reconstructed images under different MRI scanning devices or different scanning parameters based on the reconstructed images. Reconstruction image processing unit for: The reconstructed image is subjected to image segmentation and image registration processing, and the quality of the reconstructed image is evaluated by quantitative and qualitative methods.

[0014] Compared with the prior art, the present invention has the following beneficial effects: The present invention can correct motion artifacts by adopting a navigation echo method and a motion compensation method during the scanning process of an MRI scanning device, can solve the artifact problem caused by the physiological activities of the research object, thereby effectively reducing the interference of motion artifacts on the image, and improving the quality of the reconstructed image. After the nuclear magnetic image scanning data of the research object is obtained, it is processed, and the reliability of the nuclear magnetic image scanning data can be improved through denoising and enhancement and data standardization and normalization, thereby further improving the quality of the reconstructed image. Then, a neural network model is constructed and trained and tested, and the nuclear magnetic image scanning data is reconstructed using the neural network model. The quality changes of the reconstructed images under different MRI scanning devices or different scanning parameters are analyzed according to the reconstructed images to verify the effect of the reconstructed images. Finally, the reconstructed images are segmented and image registration is performed, which can improve the accuracy and efficiency of the reconstructed image analysis. The quality of the reconstructed images is evaluated by quantitative and qualitative methods, which can more comprehensively and accurately evaluate the quality of the reconstructed images. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a flow chart of the method for reconstructing nuclear magnetic resonance image scanning data of the present invention; Figure 2 It is a structural diagram of the nuclear magnetic resonance image scanning data reconstruction system of the present invention. DETAILED DESCRIPTION

[0016] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0017] In order to solve the technical problem that the quality of reconstructed images varies due to different devices or different scanning parameters during data acquisition, and that motion artifacts are prone to occur during data acquisition due to the uncertainty of the physiological movement of the research object, thereby reducing the quality of reconstructed images, please refer to Figure 1-Figure 2 , this embodiment provides the following technical solutions: The method for reconstructing nuclear magnetic resonance image scanning data comprises the following steps: Set the scanning parameters of the MRI scanning equipment, including repetition time, echo time and matrix size; Placing the research object on the MRI scanning device, starting the MRI scanning device to scan and record the data during the imaging process, and obtaining the nuclear magnetic resonance image scanning data of the research object; Process the acquired nuclear magnetic resonance image scanning data, including denoising and enhancement and data standardization and normalization; Construct a neural network model and perform training and testing, and use the trained neural network model to reconstruct the MRI scan data; Further processing of the reconstructed image, including segmentation and registration, and evaluation of the quality of the reconstructed image by quantitative and qualitative methods; Among them, the quantitative evaluation method mainly measures the image quality by calculating various indicators, and the qualitative evaluation method mainly evaluates the image quality through human visual perception. The quantitative evaluation method can provide objective data support, but it cannot completely simulate the complex perception process of the human eye. Although the qualitative evaluation method is subjective, it can capture some visual artistry or perceptual pros and cons that cannot be discovered by quantitative methods. Therefore, combining quantitative and qualitative evaluation methods can more comprehensively and accurately evaluate the quality of the reconstructed image.

[0018] In one embodiment, the quality of the reconstructed image is evaluated by quantitative and qualitative methods, including: The reconstructed image is scaled according to a plurality of preset sizes to obtain a scaled image corresponding to each preset size, feature extraction is performed on all scaled images to obtain image features, and the image features are unified to obtain standard features; Based on the standard features of the reconstructed image and the scaled image, the quality value of the reconstructed image is calculated according to the following formula ; Where n represents the number of scaled images, m represents the number of features, represents the jth standard eigenvalue in the reconstructed image, represents the jth standard eigenvalue in the i-th scaled image, represents the scaling weight value of the i-th scaled image; Determining whether the quality value of the reconstructed image is greater than a preset quality value; If so, determining that the clarity of the reconstructed image meets the quality requirements; Otherwise, determining that the clarity of the reconstructed image does not meet the quality requirement; After determining that the clarity of the reconstructed image meets the quality requirements, an initial reconstructed image before registration is obtained, and a cross-correlation coefficient between the reconstructed image and the initial reconstructed image is calculated according to the following formula; in, represents the cross-correlation coefficient between the reconstructed image and the initial reconstructed image, represents the covariance between the reconstructed image and the initial reconstructed image, represents the variance of the reconstructed image, represents the variance of the initial reconstructed image, represents the expected variance of the reconstructed image based on the initial reconstructed image, represents the expected variance of the initial reconstructed image based on the reconstructed image; Determining whether the cross-correlation coefficient between the reconstructed image and the initial reconstructed image meets a preset coefficient requirement; If so, determining that the artifacts of the reconstructed image meet the quality requirements; Otherwise, it is determined that the artifacts of the reconstructed image do not meet the quality requirement.

[0019] In this embodiment, the scaling weight value of the scaled image is determined according to the scaling ratio.

[0020] The beneficial effect of the above design is that by quantitatively evaluating the image quality from two aspects: image clarity and artifacts, a more comprehensive and accurate evaluation of the quality of the reconstructed image can be achieved, which is of great significance for improving the quality and diagnostic value of the reconstructed image. During the scanning process of the MRI scanning device, the motion artifacts are corrected by using corresponding correction methods, including navigation echo method and motion compensation method, in view of the artifacts caused by the physiological activities of the research subjects. The correction method is specifically as follows: Navigation echo method, used for: During the scanning process of the MRI scanning device, the image artifacts caused by respiratory physiological activities are tracked and corrected, and the diaphragm movement is tracked using 2D radio frequency pulses. Data is collected when the diaphragm movement is within the set acceptable range. Motion compensation methods for: By describing the movement of each small block in adjacent frames of the nuclear magnetic resonance image scanning data, each small block of the previous frame is moved to the corresponding position of the current frame, thereby reducing the spatial redundancy and motion artifacts in the nuclear magnetic resonance image scanning data sequence.

[0021] The technical effect of the above content is: during the scanning process of the MRI scanning equipment, due to the artifact problem caused by the physiological activities of the research subject, the corresponding correction method can be used to correct the motion artifact. The navigation echo method tracks the diaphragm movement through 2D radio frequency pulses, and collects data when the diaphragm movement is within the set acceptable range. The navigation echo method can effectively reduce the interference of motion artifacts on the image and improve the image quality. The motion compensation method describes the changes in adjacent frames, uses advanced prediction algorithms to synthesize the lost frames, and moves each small block of the previous frame to the corresponding position of the current frame to reduce motion artifacts. The motion compensation method can reduce or eliminate motion artifacts and significantly improve the image quality. The navigation echo method and the motion compensation method can reduce or eliminate motion artifacts, which is of great significance for improving the quality of reconstructed images and diagnostic value. ‌ Denoising and enhancement, specifically: Denoising: Denoising algorithms are used to denoise the MRI scan data to remove noise and retain image details. Denoising algorithms include wavelet transform or non-local mean filtering; Enhancement processing: Enhance the quality, contrast and clarity of nuclear magnetic resonance image scanning data using enhancement processing methods, including spatial domain enhancement methods and frequency domain enhancement methods; The spatial domain enhancement method directly processes the pixels of the image. Commonly used spatial domain enhancement methods include: ‌Histogram equalization‌: By changing the histogram distribution of the image, the pixel intensity distribution is made more uniform, thereby enhancing the contrast of the entire image; ‌Image Sharpening and Edge Enhancement‌: Enhances the edges in an image by applying a high-pass filter, making the boundaries clearer; ‌Spatial domain filtering‌: including smoothing filtering and sharpening filtering, used to remove image noise or enhance image details; The frequency domain enhancement method is achieved by modifying the spectrum of the image after Fourier transformation. Common frequency domain enhancement methods include: ‌Low-pass filtering‌: removes the high-frequency part of the image to remove noise; ‌High-pass filtering‌: Enhances the high-frequency part of the image to sharpen the image; Homomorphic filtering: combines the characteristics of low-pass and high-pass filtering to enhance the dynamic range of images.

[0022] Data standardization and normalization, specifically: The data format is unified for: Unify the scanning data formats of nuclear magnetic resonance images obtained by different MRI scanning devices or different scanning parameters; Data normalization for: Standardize the MRI scan data to eliminate the differences between different MRI scan data; Data normalization, used to: Map the MRI scan data to a specific range, which is [0,1] or [-1,1].

[0023] The technical effect of the above content is: after the nuclear magnetic resonance image scanning data is obtained, denoising and enhancement and data standardization and normalization will be performed. By using wavelet transform or non-local mean filtering, the noise in the nuclear magnetic resonance image scanning data can be removed and the image details can be retained. The spatial domain enhancement method and the frequency domain enhancement method can be used to enhance the quality, contrast and clarity of the nuclear magnetic resonance image scanning data. By denoising and enhancement, the quality, contrast and details of the nuclear magnetic resonance image scanning data can be improved. By unifying the data format, the format of nuclear magnetic resonance image scanning data obtained by different MRI scanning devices or different scanning parameters can be unified. By data standardization, the nuclear magnetic resonance image scanning data can be standardized to eliminate the differences between different nuclear magnetic resonance image scanning data. By data normalization, the nuclear magnetic resonance image scanning data can be mapped to a specific range. By standardizing and normalizing the nuclear magnetic resonance image scanning data, the data can have the same dimension and scale, thereby better reflecting the true distribution of the data. At the same time, the accuracy and convergence speed of the model can be improved, making the model training more stable and efficient.

[0024] Build a neural network model and perform training and testing, which includes the following processes: Collecting magnetic resonance imaging scan data of 10 research subjects to obtain a first data set, and dividing the first data set into a training set and a test set; By using different MRI scanning devices or adjusting scanning parameters, magnetic resonance imaging scan data of another 10 research subjects are collected to obtain a second data set, and the second data set is divided into a training set and a test set; Construct a neural network model, and use the training set and test set to train and test the neural network model respectively; The neural network model is trained and tested using the training set and the test set divided by the first data set, and the trained neural network model is used to reconstruct the nuclear magnetic resonance image scanning data to obtain a first reconstructed image; The neural network model is trained and tested using the training set and the test set divided by the second data set, and the trained neural network model is used to reconstruct the nuclear magnetic resonance image scanning data to obtain a second reconstructed image; The first reconstructed image is compared with the second reconstructed image, and the robustness of the neural network model is verified by comparing and analyzing the quality changes of the reconstructed images under different MRI scanning devices or different scanning parameters; After verification, the neural network model is deployed to reconstruct the MRI image scanning data.

[0025] The technical effect of the above content is: constructing a neural network model and performing training and testing, collecting nuclear magnetic resonance imaging (MRI) scan data of 20 research subjects in total, dividing the nuclear magnetic resonance imaging (MRI) scan data of 10 of the research subjects into a first data set, and dividing the first data set into a training set and a test set, and dividing the nuclear magnetic resonance imaging (MRI) scan data of the other 10 research subjects into a second data set, the second data set is nuclear magnetic resonance imaging (MRI) scan data obtained by using different MRI scanning devices or adjusting scanning parameters, the second data set is a control group of the first data set, and the second data set is also divided into a training set and a test set, firstly, using the training set and the test set divided from the first data set to train and test the neural network model, and then using the trained neural network model to train and test the nuclear magnetic resonance imaging (MRI) scan data. The first reconstructed image can be obtained by reconstructing the scanning data using the training set and the test set divided by the second data set. Then, the neural network model is trained and tested. The trained neural network model is used to reconstruct the nuclear magnetic resonance image scanning data to obtain the second reconstructed image. Finally, the first reconstructed image is compared with the second reconstructed image. The robustness of the neural network model is verified by comparing and analyzing the quality changes of the reconstructed images under different MRI scanning devices or different scanning parameters. Through the verification, the MRI scanning device can be reasonably selected and the MRI scanning parameters can be adjusted, so as to effectively improve the image quality. After verification, the neural network model is deployed to reconstruct the nuclear magnetic resonance image scanning data, so as to realize the reconstruction of the nuclear magnetic resonance image scanning data.

[0026] The reconstructed image is segmented and registered, specifically: Image segmentation: Image segmentation is used to divide the reconstructed image into non-overlapping and characteristic regions; Image registration: Image registration is used to perform geometric transformation on the reconstructed image so that the reconstructed image can be aligned in space.

[0027] In one embodiment, geometrically transforming the reconstructed image by image registration includes: Initializing the interpolation processing of the reconstructed image according to a preset interpolation method to obtain an interpolation pixel position of the reconstructed image, and determining the pixel grayscale value of the interpolation pixel position based on the pixel values ​​of the pixels around the interpolation pixel position and the interpolation direction to obtain an interpolated image; Based on feature point matching, the reconstructed image is deformed to the space where the interpolated image is located to obtain an initial registration image; Extracting image feature information from the initial registered image, and selecting a feature point set consisting of main feature points based on the image feature information as a registration basis, and re-registering the initial registered image based on the registration basis to obtain a second registered image; Extracting mutual information between the second registered images, and establishing a mutual information matrix based on the image sequence and information attributes of the second registered images; Based on the difference between the horizontal correlation relationship between the mutual information matrices and the preset horizontal correlation relationship, a horizontal optimization value is obtained, and based on the difference between the vertical correlation relationship between the mutual information matrices and the preset vertical correlation relationship, a vertical optimization value is obtained; Acquire initial optimization parameters of the optimization algorithm, and adjust the initial optimization parameters based on the lateral optimization value and the longitudinal optimization value to obtain target optimization parameters; Acquire an initial transformation relationship from the reconstructed image to the second registered image, and optimize the initial transformation relationship based on the target optimization parameter to obtain a target transformation relationship; The reconstructed image is geometrically transformed based on the target transformation relationship to obtain a target registration image.

[0028] In this embodiment, the average value of the pixel values ​​of the pixels around the interpolation pixel position corresponding to the interpolation direction is used as the pixel grayscale value of the interpolation pixel position.

[0029] In this embodiment, the reconstructed image is deformed to the space where the interpolated image is located to obtain an initial registration image to achieve a large-scale transformation of the reconstructed image, thereby providing an image basis for subsequent registration.

[0030] In this embodiment, the optimization algorithm is pre-selected or designed according to the optimization target.

[0031] In this embodiment, the second registered image implements feature-based registration between the initial registered images.

[0032] In this embodiment, the initial transformation relationship from the reconstructed image to the second registered image includes the transformation relationship from the reconstructed image to the initial registered image and then to the second registered image.

[0033] In this embodiment, the reconstructed image is geometrically transformed based on the target transformation relationship to obtain a target registered image to achieve accurate registration from two aspects: image features and association between images.

[0034] The beneficial effects of the above-mentioned design scheme are: by deforming the reconstructed image to the space where the interpolated image is located based on feature point matching, an initial registered image is obtained, the initial registered image is re-registered based on the registration basis to obtain a second registered image, an initial conversion relationship from the reconstructed image to the second registered image is obtained, the initial conversion relationship is optimized based on the target optimization parameters to obtain a target conversion relationship, the reconstructed image is geometrically transformed based on the target conversion relationship to obtain a target registered image, and the target registered image is accurately registered from two aspects of image features and the relationship between images, so that the reconstructed image can be aligned in space.

[0035] The technical effect of the above content is: through image segmentation, the reconstructed image can be divided into non-overlapping areas with their own characteristics. The image segmentation methods mainly include threshold-based segmentation, region-based segmentation, edge-based segmentation, etc. The threshold segmentation method divides the image into different grayscale intervals by selecting one or more thresholds, which is suitable for images with high contrast between the target and the background. The region segmentation rule divides pixels with similar attributes into the same region, while edge detection performs segmentation based on the sudden change of the boundary of objects in the image. The reconstructed image is geometrically transformed through image registration to achieve the purpose of spatial alignment of the reconstructed image. The image registration methods include methods based on feature point matching and methods based on image mutual information. Feature point matching performs registration by finding common feature points between images, while the mutual information method achieves registration by optimizing the mutual information between images. Segmenting and registering the reconstructed image can improve the quality and accuracy of the reconstructed image, thereby improving the accuracy and efficiency of the analysis of the reconstructed image. A nuclear magnetic image scanning data reconstruction system, used for a nuclear magnetic image scanning data reconstruction method, comprising: Data acquisition unit for: The scanning parameters of the MRI scanning device are set, and the data of the imaging process is scanned and recorded using the MRI scanning device. During the scanning process, the navigation echo method and motion compensation method are used to correct the motion artifacts, and finally the nuclear magnetic resonance image scanning data of the research object is obtained.

[0036] Data processing unit for: De-noise and enhance the acquired nuclear magnetic resonance image scanning data and perform data standardization and normalization; An image reconstruction unit, for: Construct a neural network model and conduct training and testing. Use the trained neural network model to reconstruct the nuclear magnetic resonance image scanning data, and analyze the quality changes of the reconstructed images under different MRI scanning devices or different scanning parameters based on the reconstructed images. Reconstruction image processing unit for: The reconstructed image is subjected to image segmentation and image registration processing, and the quality of the reconstructed image is evaluated by quantitative and qualitative methods.

[0037] The technical effect of the above content is: the nuclear magnetic image scanning data reconstruction system is composed of a data acquisition unit, a data processing unit, an image reconstruction unit and a reconstructed image processing unit, and the nuclear magnetic image scanning data reconstruction system is used to implement the above-mentioned nuclear magnetic image scanning data reconstruction method, the nuclear magnetic image scanning data of the research object can be acquired through the data acquisition unit, and during the scanning process, the motion artifacts can be corrected through the navigation echo method and the motion compensation method, so as to reduce or eliminate the motion artifacts, thereby improving the quality of the reconstructed image, and the acquired nuclear magnetic image scanning data can be denoised and enhanced and the data can be standardized and normalized through the data processing unit, so as to improve the reliability of the nuclear magnetic image scanning data, thereby further improving the quality of the reconstructed image, and then, through the image reconstruction unit A neural network model is constructed and trained and tested. The neural network model is used to reconstruct the nuclear magnetic resonance image scanning data. The quality changes of the reconstructed images under different MRI scanning devices or different scanning parameters are analyzed based on the reconstructed images. The MRI scanning equipment can be reasonably selected and the MRI scanning parameters can be adjusted, thereby effectively improving the image quality. After verification, the neural network model is deployed to reconstruct the nuclear magnetic resonance image scanning data, thereby realizing the reconstruction of the nuclear magnetic resonance image scanning data. Finally, the reconstructed image can be segmented and registered by the reconstructed image processing unit, which can improve the accuracy and efficiency of the reconstructed image analysis. The quality of the reconstructed image can be evaluated more comprehensively and accurately by quantitative and qualitative methods.

[0038] Working principle: During the scanning process of the MRI scanning device, the motion artifacts are reduced or eliminated through the navigation echo method and the motion compensation method, thereby improving the quality of the reconstructed image. After the nuclear magnetic image scanning data of the research object is obtained, it is processed. Through denoising and enhancement and data standardization and normalization, the reliability of the nuclear magnetic image scanning data can be improved, thereby further improving the quality of the reconstructed image. Then, a neural network model is constructed and trained and tested. The nuclear magnetic image scanning data is reconstructed using the neural network model, and the quality changes of the reconstructed images under different MRI scanning devices or different scanning parameters are analyzed according to the reconstructed images to verify the effect of the reconstructed image. The MRI scanning equipment can be reasonably selected and the MRI scanning parameters can be adjusted, thereby effectively improving the image quality. Finally, the reconstructed images are segmented and registered, which can improve the quality and accuracy of the reconstructed images, thereby improving the accuracy and efficiency of the reconstructed image analysis. The quality of the reconstructed images can be evaluated more comprehensively and accurately by quantitative and qualitative methods.

[0039] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0040] While the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that many changes, modifications, substitutions and variations can be made to the embodiments without departing from the principles and spirit of the invention.

Claims

1. A method for reconstructing nuclear magnetic resonance image scanning data, characterized in that: The following steps are involved: Set the scanning parameters of the MRI scanning equipment, including repetition time, echo time and matrix size; Placing the research object on the MRI scanning device, starting the MRI scanning device to scan and record the data during the imaging process, and obtaining the nuclear magnetic resonance image scanning data of the research object; Process the acquired nuclear magnetic resonance image scanning data, including denoising and enhancement and data standardization and normalization; Construct a neural network model and perform training and testing, and use the trained neural network model to reconstruct the MRI scan data; The reconstructed images are further processed, including segmentation and registration, and the quality of the reconstructed images is evaluated by quantitative and qualitative methods.

2. The method for reconstructing nuclear magnetic resonance image scanning data according to claim 1, characterized in that: The quality of the reconstructed images was evaluated by quantitative and qualitative methods, including: The reconstructed image is scaled according to a plurality of preset sizes to obtain a scaled image corresponding to each preset size, feature extraction is performed on all scaled images to obtain image features, and the image features are unified to obtain standard features; Based on the standard features of the reconstructed image and the scaled image, the quality value of the reconstructed image is calculated according to the following formula ; Where n represents the number of scaled images, m represents the number of features, represents the jth standard eigenvalue in the reconstructed image, represents the jth standard eigenvalue in the i-th scaled image, Indicates the scaling weight value of the i-th scaled image; determines whether the quality value of the reconstructed image is greater than a preset quality value; If so, determining that the clarity of the reconstructed image meets the quality requirements; Otherwise, determining that the clarity of the reconstructed image does not meet the quality requirement; After determining that the clarity of the reconstructed image meets the quality requirements, an initial reconstructed image before registration is obtained, and a cross-correlation coefficient between the reconstructed image and the initial reconstructed image is calculated according to the following formula; in, represents the cross-correlation coefficient between the reconstructed image and the initial reconstructed image, represents the covariance between the reconstructed image and the initial reconstructed image, represents the variance of the reconstructed image, represents the variance of the initial reconstructed image, represents the expected variance of the reconstructed image based on the initial reconstructed image, represents the expected variance of the initial reconstructed image based on the reconstructed image; Determining whether the cross-correlation coefficient between the reconstructed image and the initial reconstructed image meets a preset coefficient requirement; If so, determining that the artifacts of the reconstructed image meet the quality requirements; Otherwise, it is determined that the artifacts of the reconstructed image do not meet the quality requirement.

3. The method for reconstructing nuclear magnetic resonance image scanning data according to claim 1, characterized in that: During the scanning process of the MRI scanning device, corresponding correction methods are used to correct motion artifacts. The correction methods include navigation echo methods and motion compensation methods.

4. The method for reconstructing nuclear magnetic resonance image scanning data according to claim 3, characterized in that: The correction method is specifically as follows: Navigation echo method, used for: During the scanning process of the MRI scanning device, 2D radio frequency pulses are used to track the movement of the diaphragm, and data is collected when the diaphragm movement is within the set acceptable range; Motion compensation methods for: By describing the movement of each small block in adjacent frames of nuclear magnetic resonance image scanning data, each small block of the previous frame is moved to the corresponding position of the current frame, thereby reducing spatial redundancy and motion artifacts in the nuclear magnetic resonance image scanning data sequence.

5. The method for reconstructing nuclear magnetic resonance image scanning data according to claim 1, characterized in that: The denoising and enhancement are specifically as follows: Denoising: Denoising algorithms are used to denoise the MRI scan data to remove noise and retain image details. Denoising algorithms include wavelet transform or non-local mean filtering; Enhancement processing: Enhancement processing methods are used to enhance the quality, contrast and clarity of nuclear magnetic resonance image scanning data. Enhancement processing methods include spatial domain enhancement methods and frequency domain enhancement methods.

6. The method for reconstructing nuclear magnetic resonance image scanning data according to claim 1, characterized in that: The data standardization and normalization are specifically as follows: The data format is unified for: Unify the scanning data formats of nuclear magnetic resonance images obtained by different MRI scanning devices or different scanning parameters; Data normalization for: Standardize the MRI scan data to eliminate the differences between different MRI scan data; Data normalization, used to: Map the MRI scan data to a specific range, which is [0,1] or [-1,1].

7. The method for reconstructing nuclear magnetic resonance image scanning data according to claim 1, characterized in that: Build a neural network model and perform training and testing, which includes the following processes: Collecting magnetic resonance imaging scan data of 10 research subjects to obtain a first data set, and dividing the first data set into a training set and a test set; By using different MRI scanning devices or adjusting scanning parameters, magnetic resonance imaging scan data of another 10 research subjects are collected to obtain a second data set, and the second data set is divided into a training set and a test set; Construct a neural network model, and use the training set and test set to train and test the neural network model respectively; The neural network model is trained and tested using the training set and the test set divided by the first data set, and the trained neural network model is used to reconstruct the nuclear magnetic resonance image scanning data to obtain a first reconstructed image; The neural network model is trained and tested using the training set and the test set divided by the second data set, and the trained neural network model is used to reconstruct the nuclear magnetic resonance image scanning data to obtain a second reconstructed image; The first reconstructed image is compared with the second reconstructed image, and the robustness of the neural network model is verified by comparing and analyzing the quality changes of the reconstructed images under different MRI scanning devices or different scanning parameters; After the verification is completed, the MRI scanning equipment and scanning parameters are determined, and the neural network model trained using the MRI scanning equipment and scanning parameters is used to reconstruct the nuclear magnetic resonance image scanning data.

8. The method for reconstructing nuclear magnetic resonance image scanning data according to claim 1, characterized in that: The reconstructed image is segmented and registered, specifically: Image segmentation: Image segmentation is used to divide the reconstructed image into non-overlapping and characteristic regions; Image registration: Image registration is used to geometrically transform the reconstructed image.

9. The method for reconstructing nuclear magnetic resonance image scanning data according to claim 1, characterized in that: The geometric transformation of the reconstructed image through image registration includes: Initializing the interpolation processing of the reconstructed image according to a preset interpolation method to obtain an interpolation pixel position of the reconstructed image, and determining the pixel grayscale value of the interpolation pixel position based on the pixel values ​​of the pixels around the interpolation pixel position and the interpolation direction to obtain an interpolated image; Based on feature point matching, the reconstructed image is deformed to the space where the interpolated image is located to obtain an initial registration image; Extracting image feature information from the initial registered image, and selecting a feature point set consisting of main feature points based on the image feature information as a registration basis, and re-registering the initial registered image based on the registration basis to obtain a second registered image; Extracting mutual information between the second registered images, and establishing a mutual information matrix based on the image sequence and information attributes of the second registered images; Based on the difference between the horizontal correlation relationship between the mutual information matrices and the preset horizontal correlation relationship, a horizontal optimization value is obtained, and based on the difference between the vertical correlation relationship between the mutual information matrices and the preset vertical correlation relationship, a vertical optimization value is obtained; Acquire initial optimization parameters of the optimization algorithm, and adjust the initial optimization parameters based on the lateral optimization value and the longitudinal optimization value to obtain target optimization parameters; Acquire an initial transformation relationship from the reconstructed image to the second registered image, and optimize the initial transformation relationship based on the target optimization parameter to obtain a target transformation relationship; The reconstructed image is geometrically transformed based on the target transformation relationship to obtain a target registration image.

10. A nuclear magnetic image scanning data reconstruction system, used to implement the nuclear magnetic image scanning data reconstruction method according to any one of claims 1 to 8, characterized in that: include: Data acquisition unit for: Set the scanning parameters of the MRI scanning device, use the MRI scanning device to scan and record the data during the imaging process, and use the navigation echo method and motion compensation method to correct the motion artifacts during the scanning process, and finally obtain the nuclear magnetic resonance image scanning data of the research object. The data processing unit is used to: De-noise and enhance the acquired nuclear magnetic resonance image scanning data and perform data standardization and normalization; An image reconstruction unit, for: Construct a neural network model and conduct training and testing. Use the trained neural network model to reconstruct the nuclear magnetic resonance image scanning data, and analyze the quality changes of the reconstructed images under different MRI scanning devices or different scanning parameters based on the reconstructed images. Reconstruction image processing unit for: The reconstructed image is subjected to image segmentation and image registration processing, and the quality of the reconstructed image is evaluated by quantitative and qualitative methods.

Images