A magnetic resonance imaging image degradation method, device, and medium
By integrating artifact simulation algorithms based on the physical imaging principles of MRI, we can generate MRI degradation images that are more consistent with clinical practice. This solves the problem of the limited simulation dimensions of existing methods, provides rich training data, and improves the clinical applicability of the algorithm.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- XUANWU HOSPITAL OF CAPITAL UNIV OF MEDICAL SCI
- Filing Date
- 2026-04-24
- Publication Date
- 2026-07-24
AI Technical Summary
Existing MRI image degradation methods cannot accurately simulate the complex degradation patterns caused by multiple factors in MRI images, making it difficult for trained algorithms to effectively adapt to real-world low-quality clinical images.
By integrating multiple artifact simulation algorithms based on the physical imaging principles of MRI, a refined parameter control system is constructed to simulate various artifacts in MRI images, such as random noise, data acquisition and encoding distortion, physical interactions, physiological motion, and hardware systems, thereby generating degraded images that are more consistent with clinical reality.
It provides rich and reliable training and testing data, which significantly improves the clinical applicability of the algorithm and can reproduce common mixed and multi-dimensional degradation scenarios in clinical practice, breaking through the limitation of single simulation dimension.
Smart Images
Figure CN122453984A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of sensor testing technology, and in particular to a method, device and medium for magnetic resonance imaging image degradation. Background Technology
[0002] Magnetic Resonance Imaging (MRI), with its superior imaging capabilities of the brain, soft tissues, and other areas, has become an indispensable core tool in clinical diagnosis and medical scientific research. With the deepening penetration of artificial intelligence technology into the medical field, research on intelligent MRI image processing algorithms is receiving increasing attention.
[0003] However, the further development of these studies faces a core bottleneck: in clinical practice, severely degraded images collected by medical institutions due to patient movement, equipment limitations, physiological factors, etc., are usually deleted directly rather than archived because they cannot meet diagnostic needs and involve patient and hospital privacy. This results in a very small number of real low-quality samples that can be used for algorithm training or testing.
[0004] Currently, in the absence of real low-quality datasets, existing MRI image degradation methods can only simulate image degradation by artificially adding general noise such as Gaussian noise, salt-and-pepper noise, or deformation to high-quality MRI images. However, the degradation patterns of MRI images have significant unique characteristics, which are closely related to the principles of imaging physics. This makes methods derived from the field of natural image processing unable to accurately reproduce the unique, physically-based artifact features of MRI images. Furthermore, in clinical practice, the degradation of MRI image quality is often the result of multiple factors working together, and may simultaneously contain mixed degradation effects such as random noise, physiological motion artifacts, and equipment-related artifacts. Simply adding noise cannot reproduce such complex scenarios, ultimately making it difficult for trained algorithms to effectively adapt to real low-quality clinical images. Summary of the Invention
[0005] To address the technical problems existing in the background art, embodiments of this application provide a method, device, and medium for degrading magnetic resonance imaging (MRI) images. The method includes: acquiring an original MRI image and performing standardized preprocessing on the original MRI image to obtain a standardized image; simulating MRI artifacts on the standardized image according to the physical characteristics of MRI artifact formation and a preset artifact control parameter range to obtain a degraded MRI image; the MRI artifacts include at least random interference and noise artifacts, data acquisition and encoding distortion artifacts, physical interaction artifacts, physiological motion artifacts, and hardware system artifacts; and outputting the degraded MRI image for visualization.
[0006] In one example, based on the physical characteristics of magnetic resonance imaging artifact formation and a preset artifact control parameter table, random interference and noise artifact simulation are performed on the standardized image. Specifically, this includes: superimposing a first Gaussian white noise onto the real part of the standardized image and superimposing a second Gaussian white noise independent of the first Gaussian white noise onto the imaginary part of the standardized image; the mean of both the first and second Gaussian white noises is zero and their standard deviation is controlled by the noise intensity parameter in the preset artifact control parameter table; calculating the magnitude of the complex signal after superimposing the noise to generate a Rician noise magnetic resonance imaging image to simulate Rician noise artifacts.
[0007] In one example, the method further includes: superimposing normally distributed random noise onto the standardized image to generate a Gaussian noise magnetic resonance imaging image to simulate Gaussian noise artifacts; the standard deviation of the random noise is controlled by a Gaussian noise intensity parameter in a preset artifact control parameter table.
[0008] In one example, the method further includes: in the standardized image, selecting pixels using a preset random algorithm and setting the pixel values of the pixels to extreme values to generate a salt-and-pepper noise magnetic resonance imaging image to simulate salt-and-pepper noise artifacts; the proportion of pixels set to extreme values is controlled by a noise density parameter in a preset artifact control parameter table.
[0009] In one example, based on the physical characteristics of magnetic resonance imaging artifact formation and a preset artifact control parameter table, data acquisition and encoding distortion artifact simulation are performed on the standardized image. Specifically, this includes: converting the standardized image to K-space using Fourier transform; truncating the high-frequency components on the periphery of the K-space using a rectangular window function; the degree of truncation of the high-frequency components is controlled by the truncation intensity parameter in the preset artifact control parameter table; and reconstructing the truncated K-space high-frequency components using inverse Fourier transform to obtain a magnetic resonance imaging image including ringing fringes to simulate Gibbs artifacts.
[0010] In one example, the method further includes: calculating the width of the region in the phase encoding direction of the normalized image based on a preset effective field-of-view ratio parameter; periodically extending the normalized image in the phase encoding direction according to a preset extension period; and linearly superimposing the signal located outside one edge of the normalized image in the extended normalized image to the corresponding region of the opposite edge of the normalized image to simulate folding artifacts.
[0011] In one example, the method further includes: the mixing intensity of the linear superposition is controlled by a preset folding intensity parameter; converting the normalized image to K-space and attenuating the high-frequency components of the normalized image through a low-pass filter; the degree of high-frequency attenuation is controlled by a preset K-space cutoff frequency parameter; downsampling the normalized image with high-frequency attenuation and resampling the downsampled normalized image to its original size through an interpolation algorithm; the degree of spatial resolution reduction is controlled by a preset downsampling intensity parameter; adding Rician noise to the normalized image after downsampling and upsampling to simulate low-resolution artifacts; the level of the added Rician noise is controlled by a preset additional noise intensity parameter.
[0012] In one example, based on the physical characteristics of magnetic resonance imaging artifact formation and a preset artifact control parameter table, physical interaction artifact simulation is performed on the standardized image. Specifically, this includes: defining a sensitive region in the standardized image to simulate the difference in magnetic susceptibility of the material based on prior knowledge; generating a parameterized displacement field within the sensitive region based on preset distortion direction parameters and deformation intensity parameters; the parameterized displacement field defines the movement direction and distance of each pixel within the sensitive region; and mapping each pixel within the sensitive region to a new position according to the parameterized displacement field through bicubic interpolation resampling to synthesize a magnetic resonance imaging image including local geometric distortion, signal accumulation, and signal voids to simulate magnetic susceptibility artifacts.
[0013] In one example, the method further includes: presetting the position and influence radius of the metal artifact in the standardized image; generating a core attenuation field using a distance-based cubic smoothing function to attenuate the image signal within a region centered at the position and within the influence radius to simulate signal voids; creating a non-rigid vortex displacement field and constructing rotational and torsional components around the position using trigonometric functions; applying the vortex displacement field to the coordinates of the standardized image using bicubic interpolation resampling to synthesize an image containing stretching, compression, and vortex-like geometric deformations; generating directional ripple noise related to the polar angle and stripe noise composed of multiple frequency sine waves, and limiting the intensity and spatial distribution of the directional ripple noise and stripe noise within the influence radius; and adaptively mixing the signal-attenuated image, the geometrically deformed image, the generated directional ripple noise, and the stripe noise with the standardized image using multi-layer fusion technology, and performing a smooth transition at the boundaries to generate a synthetic image including the metal artifact to simulate the metal artifact.
[0014] In one example, the method further includes: generating a convolutional kernel for signal mixing based on preset mixing kernel size parameters and mixing intensity parameters; the types of the convolutional kernel include Gaussian kernels, uniform kernels, and edge-preserving kernels; in an adaptive mixing mode, calculating the gradient magnitude of the normalized image to identify tissue boundary regions, and enhancing the local mixing intensity in the identified boundary regions with a preset magnitude, and applying a preset mixing intensity in the internal regions of uniform tissue; for each pixel in the normalized image, weighting and summing the neighboring pixel values according to the weights of the convolutional kernel to obtain pixel values after simulating partial volume effects, generating an image including tissue boundary blurring and small structure distortion to simulate partial volume artifacts.
[0015] In one example, based on the physical characteristics of magnetic resonance imaging artifact formation and a preset artifact control parameter table, physiological motion artifact simulation is performed on the standardized image. Specifically, this includes: generating a complex phase perturbation field, the spatial variation of which is defined by preset phase encoding direction, physiological motion frequency, displacement amplitude, and global perturbation intensity parameters; converting the standardized image to K-space using a Fourier transform; performing complex multiplication on the generated complex phase perturbation field and the K-space data to simulate the modulation of the data phase by periodic motion; and performing an inverse Fourier transform on the modulated K-space data and calculating the modulus to obtain an image including artifacts arranged at equal intervals along the phase encoding direction, thus simulating periodic motion artifacts.
[0016] In one example, the method further includes: generating a random phase perturbation field, wherein the phase portion of the random phase perturbation field is generated by a random displacement field model, and the perturbation amplitude and the dominant spatial direction are controlled by corresponding preset parameters; converting the normalized image to K-space; performing a complex multiplication operation on the generated random phase perturbation field and the K-space data to simulate phase errors caused by random motion; performing an inverse Fourier transform on the K-space data after the complex multiplication operation and calculating the magnitude to obtain an image with a decreased signal-to-noise ratio to simulate non-periodic motion artifacts.
[0017] In one example, the method further includes: converting the standardized image to K-space; generating a complex phase perturbation field at a specific frequency position in K-space along the phase encoding direction according to preset ghosting number and interval parameters; generating an amplitude attenuation field; performing complex multiplication operations on the K-space data, the complex phase perturbation field, and the amplitude attenuation field to achieve joint modulation of the signal; and performing an inverse Fourier transform on the modulated K-space data to reconstruct an image of ghosting in the phase encoding direction to simulate motion artifacts.
[0018] In one example, based on the physical characteristics of magnetic resonance imaging artifact formation and a preset artifact control parameter table, hardware system-level artifact simulation is performed on the standardized image. Specifically, this includes: constructing a low-frequency smooth intensity modulation field based on a Gaussian vignetting function, wherein the overall dynamic range of the intensity modulation field is controlled by a preset field non-uniformity intensity parameter; and performing pixel-by-pixel multiplication operations between the generated intensity modulation field and the standardized image to generate an image including smooth brightness gradients to simulate bias field artifacts.
[0019] In one example, the method further includes: converting the standardized image to K-space using Fourier transform; setting interference pulses at paired specific locations in K-space according to preset mesh direction and spatial frequency parameters; the setting of the interference pulses follows the conjugate symmetry of K-space; performing inverse Fourier transform on the K-space data after setting the interference pulses to reconstruct an image including global, periodic alternating bright and dark stripes to simulate mesh artifacts; the direction and density of the stripes are controlled by the arrangement direction and interval of the interference pulses in K-space, the overall clarity of the stripes is controlled by preset stripe intensity parameters, and the relative intensity of the interference pulses is controlled by preset peak relative intensity parameters.
[0020] In one example, the process involves acquiring a raw magnetic resonance imaging (MRI) image and performing a standardized preprocessing step to obtain a standardized image. Specifically, this includes: acquiring a publicly available desensitized MRI image from a medical institution and identifying this desensitized MRI image as the raw MRI image; using a DICOM format parsing function to read the DICOM format data of the raw MRI image to obtain the pixel values and maximum pixel value; converting the pixel values and maximum pixel value from integer data types to double-precision floating-point data types; and normalizing the converted pixel values to a preset range to obtain the standardized image. This normalization is achieved by dividing each pixel value by the maximum pixel value.
[0021] On the other hand, embodiments of this application provide a magnetic resonance imaging image degradation device, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform any of the above-mentioned magnetic resonance imaging image degradation methods.
[0022] On the other hand, embodiments of this application provide a non-volatile computer storage medium for magnetic resonance imaging image degradation, which stores computer-executable instructions that can execute any of the above-mentioned magnetic resonance imaging image degradation methods.
[0023] The above-described technical solutions adopted in the embodiments of this application can achieve the following beneficial effects: This application effectively addresses the scarcity of real low-quality MRI images by integrating multiple artifact simulation algorithms based on the physical imaging principles of MRI and constructing a sophisticated parameter control system. This provides abundant and reliable training and testing data for the development and evaluation of algorithms for image quality assessment, restoration, and super-resolution. Furthermore, it abandons the general natural image noise model and directly simulates MRI-specific degradation patterns with clear physical causes, such as motion artifacts and magnetic susceptibility artifacts. This makes the generated degradation images more consistent with clinical reality and significantly improves the clinical applicability of algorithms trained on this data. Through independently adjustable multi-type artifact simulation modules, multiple degradation factors can be introduced individually or in combination to reproduce common mixed-type, multi-dimensional degradation scenarios in clinical practice, overcoming the limitation of existing methods that simulate only a single dimension. Attached Figure Description
[0024] To more clearly illustrate the technical solution of this application, some embodiments of this application will be described in detail below with reference to the accompanying drawings, in which: Figure 1 A schematic flowchart of a magnetic resonance imaging image degradation method provided in an embodiment of this application; Figure 2 A simulation process and parameter adjustment diagram for Rician noise artifacts provided in this application embodiment; Figure 3 A simulation of Gibbs artifacts and parameter adjustment diagram provided for embodiments of this application; Figure 4 A flowchart and parameter adjustment diagram for simulating fold artifacts provided in this application embodiment; Figure 5 A simulation of low-resolution artifacts and parameter adjustment diagram provided in this application embodiment; Figure 6 A simulation of magnetic susceptibility artifacts and parameter adjustment diagram provided for embodiments of this application; Figure 7 A flowchart and parameter adjustment diagram for simulating metal artifacts provided in this application embodiment; Figure 8 A flowchart and parameter adjustment diagram for simulating partial volumetric artifacts provided in this application embodiment; Figure 9 A flowchart and parameter adjustment diagram for simulating periodic motion artifacts provided in this application embodiment; Figure 10 A flowchart and parameter adjustment diagram for simulating non-periodic motion artifacts provided in this application embodiment; Figure 11 A simulation flow artifact process and parameter adjustment diagram provided in this application embodiment; Figure 12 A simulation bias field artifact process and parameter adjustment diagram provided for embodiments of this application; Figure 13 A flowchart and parameter adjustment diagram for simulating moiré artifacts provided in this application embodiment; Figure 14 A complete artifact simulation view provided for embodiments of this application; Figure 15 This is a schematic diagram of the structure of a magnetic resonance imaging image degradation device provided in an embodiment of this application. Detailed Implementation
[0025] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0026] Some embodiments of this application will now be described in detail with reference to the accompanying drawings.
[0027] Figure 1 This is a flowchart illustrating a magnetic resonance imaging image degradation method provided in an embodiment of this application. This method can be applied to different business domains. Certain input parameters or intermediate results in this process can be manually adjusted to help improve accuracy.
[0028] The analysis method involved in the embodiments of this application can be implemented by a terminal device or a server, and this application does not impose any special limitations on it. For ease of understanding and description, the following embodiments are all described in detail using a server as an example.
[0029] Based on this Figure 1 The process may include the following steps: S101: Acquire the original magnetic resonance imaging image and perform standardized preprocessing on the original magnetic resonance imaging image to obtain a standardized image.
[0030] In some embodiments of this application, the source of the original magnetic resonance imaging (MRI) image can be a publicly available research database that has been anonymized by a medical institution, or a high-quality DICOM format image exported from a local clinical image archiving and communication system that complies with ethical guidelines. Specifically, the system calls DICOM format parsing functions from libraries such as pydicom or SimpleITK to read the pixel array data of the selected image and the metadata stored in the file header, extracting the maximum pixel value of the image as the benchmark for subsequent normalization.
[0031] To ensure the precision of all subsequent mathematical operations and avoid rounding errors caused by integer operations, the system uniformly converts the read integer pixel data into double-precision floating-point numbers. Then, normalization processing is performed, that is, traversing each pixel in the image and dividing its floating-point value by the previously extracted maximum pixel value, thereby linearly mapping the intensity of all pixels to the closed interval [0, 1].
[0032] This series of operations constitutes a standardized preprocessing workflow. Its purpose is to transform raw MRI images from different scanning devices, different imaging sequences, and with different contrast and dynamic range into standardized images with consistent numerical scale and data type. This provides a stable and comparable input benchmark for various subsequent physical principle-based artifact simulation algorithms, ensuring the consistency and repeatability of simulation results.
[0033] S102: Based on the physical characteristics of magnetic resonance imaging artifact formation and a preset artifact control parameter table, perform magnetic resonance imaging artifact simulation on the standardized image to obtain a degraded magnetic resonance imaging image; the magnetic resonance imaging artifacts include at least random interference and noise artifacts, data acquisition and encoding distortion artifacts, physical interaction artifacts, physiological motion artifacts, and hardware system artifacts.
[0034] In some embodiments of this application, this step is the core of the invention, aiming to introduce various clinically common MRI-specific artifacts into standardized images with high fidelity through a series of algorithms based on physical principles.
[0035] Users first select the type of artifact to be simulated and set its intensity parameters from the preset artifact control parameter table through a graphical interface or configuration file. The system's built-in artifact simulation algorithm library is then invoked.
[0036] For random interference and noise-type artifacts, when simulating Rician noise, the algorithm first treats the normalized image as the real part of a complex signal and generates an independent, equally sized imaginary part with zero values, thus constructing a complex image. Next, two statistically independent Gaussian white noise fields with zero mean and standard deviation controlled by the noise intensity parameter L are generated and superimposed on the real and imaginary parts of the complex image, respectively. Subsequently, the magnitude of this noisy complex signal is calculated. This nonlinear operation converts the Gaussian noise in the real and imaginary parts into Rician noise in the final amplitude image, simulating the process by which MRI equipment generates clinically usable amplitude images from complex data.
[0037] When simulating Gaussian noise, a value that obeys the following rules is directly generated. A noise matrix with the same distribution and size as the image, where the standard deviation σ is controlled by the Gaussian noise intensity parameter and is added pixel-by-pixel to the normalized image.
[0038] When simulating salt-and-pepper noise, the algorithm generates a random matrix of the same size as the image, uniformly distributed in [0,1], and compares it with the noise density parameter prob. Random values less than [the specified value] are then selected. The pixel is set to 0, and the pixel value is greater than 0. The pixel is set to 1, while the other pixels remain unchanged.
[0039] For artifacts related to data acquisition and encoding distortion, when simulating Gibbs artifacts, the standardized image is transformed to the K-space (frequency domain) using a two-dimensional Fast Fourier Transform. Then, based on the user-defined truncation intensity parameter (e.g., 0 to 10), the proportion of high-frequency regions around the K-space that are zeroed is determined. For example, an intensity of 5 might mean zeroing out the outermost 25% of the high-frequency coefficients of the K-space matrix. Subsequently, an inverse Fourier transform is performed on the truncated K-space. After taking the modulus, alternating bright and dark ringing fringes caused by the lack of high-frequency information can be observed at the tissue boundary. The extension direction of the fringes can be specified as the X or Y direction using the phase encoding direction parameter.
[0040] When simulating folding artifacts, the algorithm primarily operates in the image domain. Based on the effective field of view (FOV) ratio, it calculates the width at which the image will be folded on both sides of a specified phase-encoding direction. Next, the original image is copied three times in this direction and stitched together to simulate an infinitely periodically extended signal source. Finally, the portions of the extended image extending beyond the effective field of view on the left and right sides are extracted and linearly superimposed (the mixing intensity is controlled by the folding intensity parameter) onto the corresponding regions on the right and left sides of the resulting image, thus creating a mirrored folding effect of the contralateral anatomical structure.
[0041] When simulating low-resolution artifacts, a composite strategy is adopted. First, a Gaussian low-pass filter is used in the K-space to attenuate high frequencies (the cutoff frequency is controlled by the K-space cutoff frequency) to simulate the scanning bandwidth limitation. Then, downsampling is performed in the spatial domain, and then upsampling back to the original size is performed through bilinear interpolation to simulate finite voxel resolution. Finally, additional Rician noise (the intensity is controlled by the additional noise intensity) can be selectively added to simulate the overall degradation effect of low signal-to-noise ratio scanning.
[0042] For physical interaction artifacts, when simulating magnetic susceptibility artifacts, users can interactively define sensitive regions on the image or automatically define them using preset templates. Within these regions, a smooth displacement field is generated based on the main distortion direction and distortion intensity parameters. This field defines the vector that each pixel needs to move. Then, through bicubic interpolation resampling, the pixels in the original sensitive region are "distorted" to new positions according to the displacement field. Signal highlights are generated at the points where the displacement converges, and signal holes are generated at the points where the displacement diverges, thus simulating the geometric distortion caused by magnetic field inhomogeneity.
[0043] In simulating metal artifacts, a "signal hole" template that smoothly decays outward from the center is first generated using a distance function based on preset artifact center coordinates and core influence radius. Simultaneously, a non-rigid vortex displacement field is generated to simulate tissue stretching and rotational deformation caused by severe magnetic field distortion due to metal. Furthermore, directional ripple noise and complex stripe noise are generated to simulate signal oscillations at the interface. Finally, through multi-layer fusion technology, signal attenuation, geometric distortion, and noise effects are adaptively blended into the original image to generate a metal artifact image containing a typical "black hole" and surrounding distorted stripes.
[0044] The simulation of partial volumetric artifacts employs a convolutional blending strategy. A convolutional kernel (such as a Gaussian kernel) is generated based on the blending kernel size and blending intensity. This kernel is then used to convolve the image, resulting in each pixel's value being a weighted average of the tissue signals within its neighborhood. This leads to blurring of fine structures and boundary diffusion. If the adaptive blending switch is enabled, the system first calculates the image gradient, increasing the blending intensity at tissue boundaries and decreasing it in uniform regions, making the simulation more consistent with anatomical reality.
[0045] For physiological motion artifacts, when simulating periodic motion artifacts (such as breathing and heartbeat), the algorithm directly introduces periodic phase perturbations into the K-space. Specifically, firstly, based on parameters such as the frequency of physiological motion, displacement amplitude, and perturbation intensity, a two-dimensional phase perturbation field that varies sinusoidally or cosinely along the phase encoding direction is generated. This complex phase perturbation field is then multiplied point-by-point with the K-space data of the original image, which is equivalent to applying a motion-related phase offset to each K-space line. After IFFT reconstruction, a series of equally spaced "ghost images" are generated along the phase encoding direction.
[0046] The simulation of non-periodic motion artifacts is similar, but the phase perturbation field is generated by a random number generator. The perturbation amplitude and spatial correlation are controlled by the perturbation intensity and dominant direction parameters, ultimately resulting in a smeared blur rather than a clear "ghost" in the overall image.
[0047] Simulation of flow artifacts (such as blood vessel or cerebrospinal fluid pulsation) is also based on K-space phase modulation, but an amplitude attenuation field is introduced. Specifically, the standardized image is converted to K-space; along the phase encoding direction, a complex phase perturbation field is generated at a specific frequency position in K-space according to preset ghosting number and interval parameters; further, an amplitude attenuation field is generated; complex multiplication is performed on the K-space data, the complex phase perturbation field, and the amplitude attenuation field to achieve joint modulation of the signal; finally, the modulated K-space data is subjected to inverse Fourier transform to reconstruct the image of ghosting in the phase encoding direction to simulate flow artifacts.
[0048] For hardware system artifacts, when simulating bias field artifacts, the algorithm generates a two-dimensional, slowly varying intensity modulation field based on a Gaussian vignetting function. The overall degree of non-uniformity is controlled by the field non-uniformity intensity parameter. Multiplying this modulation field pixel-by-pixel with a normalized image generates a brightness shadow that smoothly varies from the center to the edge on the image.
[0049] To simulate moiré artifacts, a standardized image is converted to K-space using a Fourier transform. In K-space, interference pulses are set at paired specific locations according to preset moiré direction and spatial frequency parameters. The setting of the interference pulses follows the conjugate symmetry of K-space. An inverse Fourier transform is performed on the K-space data after setting the interference pulses to reconstruct an image including global and periodic alternating bright and dark stripes to simulate moiré artifacts. The direction and density of the stripes are controlled by the arrangement direction and interval of the interference pulses in K-space, the overall clarity of the stripes is controlled by preset stripe intensity parameters, and the relative intensity of the interference pulses is controlled by preset peak relative intensity parameters.
[0050] S103: Output the degraded magnetic resonance imaging image to visualize the degraded magnetic resonance imaging image.
[0051] In some embodiments of this application, after completing all specified artifact simulation steps, the system outputs the generated degraded magnetic resonance imaging image data. The output format is diverse to meet different application scenarios: Firstly, the image data, along with its corresponding artifact type labels and all intensity parameter values, can be batch-saved to local hard disk or network storage in a structured format to construct a standardized degraded image dataset for deep learning model training and testing.
[0052] Secondly, through application programming interfaces or message queues, single or batch degraded images can be transmitted in real time to downstream image quality assessment systems, image restoration algorithm testing platforms, or digital twin simulation environments.
[0053] Third, real-time visualization is achieved through a graphical user interface integrated into the software tool of this method. In the graphical user interface, the original high-quality image and the degraded image are usually displayed side by side. Users can dynamically adjust the intensity parameters of any type of artifact using sliders and observe the image degradation process in real time.
[0054] In addition, the interface provides auxiliary analysis tools such as image contrast adjustment, pseudo-color mapping, local magnification, and profile signal intensity plotting to help researchers or engineers intuitively and quantitatively evaluate the visual characteristics and physical plausibility of the generated artifacts. Ultimately, these controllable, realistic, and precisely labeled degraded images can be directly used to drive and evaluate the development of next-generation intelligent processing algorithms for MRI image denoising, super-resolution, and artifact correction.
[0055] It should be noted that, although the embodiments in this application are based on... Figure 1 Steps S101 to S103 will be described sequentially, but this does not mean that steps S101 and S103 must be performed in a strict order. The reason this embodiment follows this order is... Figure 1 The order in which steps S101 to S103 are described is provided to facilitate understanding of the technical solutions of the embodiments of this application by those skilled in the art. In other words, in the embodiments of this application, the order of steps S101 to S103 can be appropriately adjusted according to actual needs.
[0056] pass Figure 1 This application utilizes a method that integrates multiple artifact simulation algorithms based on the physical imaging principles of MRI and constructs a refined parameter control system. This effectively compensates for the scarcity of real low-quality MRI images, providing rich and reliable training and testing data for the development and evaluation of algorithms for image quality assessment, restoration, and super-resolution. Furthermore, it abandons the general natural image noise model and directly simulates MRI-specific degradation patterns with clear physical causes, such as motion artifacts and magnetic susceptibility artifacts. This makes the generated degradation images more consistent with clinical reality and significantly improves the clinical applicability of algorithms trained on this data. Through independently adjustable multi-type artifact simulation modules, multiple degradation factors can be introduced individually or in combination to reproduce common mixed-type, multi-dimensional degradation scenarios in clinical practice, overcoming the limitation of existing methods that simulate only a single dimension.
[0057] Figure 2 This application provides a simulation of Rician noise artifacts and a parameter adjustment diagram.
[0058] Figure 3 This application provides a simulation of Gibbs artifact flow and parameter adjustment diagram.
[0059] Figure 4 This application provides a flowchart of a simulation of fold artifacts and parameter adjustment.
[0060] Figure 5 This application provides a simulation of low-resolution artifacts and a parameter adjustment diagram.
[0061] Figure 6 This application provides a simulation of magnetic susceptibility artifacts and a parameter adjustment diagram.
[0062] Figure 7 This document provides a flowchart and parameter adjustment diagram for simulating metal artifacts in an embodiment of this application.
[0063] Figure 8This application provides a flowchart and parameter adjustment diagram for simulating partial volumetric artifacts in an embodiment.
[0064] Figure 9 This application provides a flowchart and parameter adjustment diagram for simulating periodic motion artifacts in an embodiment.
[0065] Figure 10 This application provides a flowchart and parameter adjustment diagram for simulating non-periodic motion artifacts.
[0066] Figure 11 This application provides a simulation flow artifact process and parameter adjustment diagram.
[0067] Figure 12 This application provides a simulation of bias field artifacts and a parameter adjustment diagram.
[0068] Figure 13 This application provides a simulation process and parameter adjustment diagram for a mesh artifact.
[0069] Figure 14 This is a complete artifact simulation view provided for an embodiment of this application.
[0070] Figure 15 A schematic diagram of a magnetic resonance imaging image degradation device provided in this application embodiment includes: At least one processor; and, A memory that is communicatively connected to at least one processor; wherein, A magnetic resonance imaging image degradation method is provided in which the memory stores instructions executable by at least one processor, such that the at least one processor is able to perform any of the above-mentioned methods.
[0071] Some embodiments of this application provide a non-volatile computer storage medium for magnetic resonance imaging image degradation, which stores computer-executable instructions capable of executing any of the above-described magnetic resonance imaging image degradation methods.
[0072] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the device and medium embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the description of the method embodiments.
[0073] The devices and media provided in this application are one-to-one with the methods. Therefore, the devices and media also have similar beneficial technical effects as their corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the devices and media will not be repeated here.
[0074] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0075] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.
[0076] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.
[0077] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
[0078] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.
[0079] Memory may include non-persistent storage in computer-readable media, random access memory (RAM), and non-volatile memory such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.
[0080] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.
[0081] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.
[0082] The above are merely embodiments of this application and are not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, or improvements made within the technical principles of this application should fall within the protection scope of this application.
Claims
1. A method for degrading magnetic resonance imaging images, characterized in that, The method includes: The raw magnetic resonance imaging image is acquired, and the raw magnetic resonance imaging image is subjected to normalization preprocessing to obtain a normalized image; Based on the physical characteristics of magnetic resonance imaging artifact formation and a preset artifact control parameter table, the standardized image is simulated to obtain a degraded magnetic resonance imaging image; the magnetic resonance imaging artifacts include at least random interference and noise artifacts, data acquisition and encoding distortion artifacts, physical interaction artifacts, physiological motion artifacts, and hardware system artifacts. Output the degraded magnetic resonance imaging image to visualize it.
2. The method according to claim 1, characterized in that, Based on the physical characteristics of magnetic resonance imaging artifact formation and a preset artifact control parameter table, random interference and noise artifact simulation is performed on the standardized image, specifically including: For simulating Rician noise artifacts, a first Gaussian white noise is superimposed on the real part signal of the normalized image, and a second Gaussian white noise independent of the first Gaussian white noise is superimposed on the imaginary part signal of the normalized image; the mean of the first Gaussian white noise and the second Gaussian white noise are both zero and their standard deviation is controlled by the noise intensity parameter in the preset artifact control parameter table. The magnitude of the complex signal after superimposed noise is calculated, and a magnetic resonance imaging image of Rician noise is generated to simulate Rician noise artifacts. To simulate Gaussian noise artifacts, random noise following a normal distribution is superimposed on the standardized image to generate a magnetic resonance imaging image with Gaussian noise, thereby simulating Gaussian noise artifacts; the standard deviation of the random noise is controlled by the Gaussian noise intensity parameter in a preset artifact control parameter table. To simulate salt-and-pepper noise artifacts, in the standardized image, pixels are selected by a preset random algorithm and their pixel values are set to extreme values to generate a magnetic resonance imaging image with salt-and-pepper noise, thereby simulating salt-and-pepper noise artifacts; the proportion of pixels set to extreme values is controlled by the noise density parameter in the preset artifact control parameter table.
3. The method according to claim 1, characterized in that, Based on the physical characteristics of magnetic resonance imaging artifact formation and a preset artifact control parameter table, data acquisition and encoding distortion artifact simulation are performed on the standardized image, specifically including: To simulate Gibbs artifacts, the normalized image is transformed to K-space using Fourier transform; The high-frequency components in the outer periphery of the K-space are truncated using a rectangular window function; the degree of truncation of the high-frequency components is controlled by the truncation intensity parameter in a preset artifact control parameter table. The truncated high-frequency components of the K-space were reconstructed by inverse Fourier transform to obtain a magnetic resonance imaging image including ringing fringes, in order to simulate Gibbs artifacts. For the simulation of folding artifacts, the width of the region in the phase encoding direction in the standardized image is calculated based on the preset effective field-of-view ratio parameter. According to a preset extension period, the standardized image is periodically extended in the phase encoding direction; Signals located outside one edge of the extended normalized image are linearly superimposed onto the corresponding region of the opposite edge of the normalized image to simulate folding artifacts; the mixing intensity of the linear superposition is controlled by a preset folding intensity parameter. For the simulation of low-resolution artifacts, the normalized image is converted to K-space, and the high-frequency components of the normalized image are attenuated by a low-pass filter; the degree of high-frequency attenuation is controlled by a preset K-space cutoff frequency parameter. The normalized image with high-frequency attenuation is downsampled, and the downsampled normalized image is resampled to the original size using an interpolation algorithm; the degree of spatial resolution reduction is controlled by a preset downsampling intensity parameter. Rician noise is added to the normalized image after downsampling and upsampling to simulate low-resolution artifacts; the level of added Rician noise is controlled by a preset additional noise intensity parameter.
4. The method according to claim 1, characterized in that, Based on the physical characteristics of magnetic resonance imaging artifact formation and a preset artifact control parameter table, physical interaction artifact simulation is performed on the standardized image, specifically including: Regarding the simulation of magnetic susceptibility artifacts, in the standardized image, a sensitive region of the magnetic susceptibility difference of the simulated material is defined based on prior knowledge; Within the sensitive area, a parameterized displacement field is generated based on preset distortion direction parameters and deformation intensity parameters; the parameterized displacement field defines the movement direction and distance of each pixel within the sensitive area. By using bicubic interpolation resampling, each pixel in the sensitive area is mapped to a new position according to the parameterized displacement field to synthesize a magnetic resonance imaging image including local geometric distortion, signal accumulation and signal voids, in order to simulate magnetic susceptibility artifacts. Regarding the simulation of metal artifacts, the position and influence radius of the metal artifacts are preset in the standardized image; A core attenuation field is generated by a distance-based cubic smoothing function to attenuate the image signal within a region centered at the said location and with the said radius of influence, in order to simulate a signal hole. A non-rigid vortex displacement field is created, and rotational and torsional components around the said position are constructed using trigonometric functions; By using bicubic interpolation resampling, the vortex displacement field is applied to the coordinates of the standardized image to synthesize an image containing stretching, compression, and vortex-like geometric deformation. Generate directional ripple noise related to the polar angle, and stripe noise composed of multiple frequency sine waves superimposed, and limit the intensity and spatial distribution of the directional ripple noise and stripe noise within the influence radius; By using multi-layer fusion technology, the signal-attenuated image, the geometrically deformed image, the generated directional ripple noise and stripe noise are adaptively mixed with the normalized image, and a smooth transition is performed at the boundary to generate a synthetic image including metal artifacts to simulate metal artifacts. Regarding the simulation of partial volumetric artifacts, a convolution kernel for signal mixing is generated based on preset mixing kernel size parameters and mixing intensity parameters; the types of convolution kernels include Gaussian kernels, uniform kernels, and edge-preserving kernels; In adaptive blending mode, the gradient magnitude of the normalized image is calculated to identify tissue boundary regions, and the local blending intensity is enhanced by a preset magnitude in the identified boundary regions, while the preset blending intensity is applied in the internal regions of uniform tissue. For each pixel in the standardized image, the neighboring pixel values are weighted and summed according to the weight of the convolution kernel to obtain the pixel value after simulating partial volume effects, generating an image including tissue boundary blurring and small structure distortion to simulate partial volume artifacts.
5. The method according to claim 1, characterized in that, Based on the physical characteristics of magnetic resonance imaging artifact formation and a preset artifact control parameter table, physiological motion artifact simulation is performed on the standardized image, specifically including: The simulation of periodic motion artifacts generates a complex phase perturbation field. The spatial variation law of the complex phase perturbation field is defined by preset phase encoding direction, physiological motion frequency, displacement amplitude and global perturbation intensity parameters. The standardized image is transformed to K-space using Fourier transform; The generated complex phase perturbation field is multiplied by the K-space data to simulate the modulation of the data phase by periodic motion; The modulated K-space data is subjected to inverse Fourier transform and the magnitude is calculated to obtain an image including artifacts arranged at equal intervals along the phase encoding direction to simulate periodic motion artifacts. Regarding the simulation of non-periodic motion artifacts, a random phase perturbation field is generated. The phase part of the random phase perturbation field is generated by a random displacement field model, and the perturbation amplitude and the dominant spatial direction are controlled by corresponding preset parameters. Convert the standardized image to K-space; The generated random phase perturbation field is multiplied by the K-space data using a complex multiplication operation to simulate the phase error caused by random motion. The inverse Fourier transform of the K-space data after complex multiplication is performed and the modulus is calculated to obtain an image with a decreased signal-to-noise ratio, in order to simulate non-periodic motion artifacts. Regarding the simulation of motion artifacts, the normalized image is converted to K-space; Along the phase encoding direction, a complex phase perturbation field is generated at a specific frequency position in K space according to the preset number of ghosts and interval parameters; Generate an amplitude decay field; The K-space data, the complex phase perturbation field, and the amplitude attenuation field are subjected to complex multiplication to achieve joint modulation of the signal; An inverse Fourier transform is performed on the modulated K-space data to reconstruct an image of ghosting in the phase coding direction, in order to simulate motion artifacts.
6. The method according to claim 1, characterized in that, Based on the physical characteristics of magnetic resonance imaging artifact formation and a preset artifact control parameter table, hardware system-based artifact simulation is performed on the standardized image, specifically including: Simulated bias field artifacts: A low-frequency smooth intensity modulation field is constructed based on the Gaussian vignetting function, and the overall dynamic range of the intensity modulation field is controlled by a preset field non-uniformity intensity parameter; The generated intensity modulation field is multiplied pixel-by-pixel with the normalized image to generate an image with smooth brightness gradients to simulate bias field artifacts.
7. The method according to claim 6, characterized in that, The method further includes: Simulated mesh artifacts: The standardized image is transformed to K-space using Fourier transform; In the K-space, interference pulses are set at paired specific locations according to preset mesh direction and spatial frequency parameters; the setting of the interference pulses follows the conjugate symmetry of the K-space. An inverse Fourier transform is performed on the K-space data after setting the interference pulse to reconstruct an image including global and periodic alternating bright and dark stripes to simulate mesh artifacts. The direction and density of the stripes are controlled by the arrangement direction and interval of the interference pulse in K-space, the overall clarity of the stripes is controlled by a preset stripe intensity parameter, and the relative intensity of the interference pulse is controlled by a preset peak relative intensity parameter.
8. The method according to claim 1, characterized in that, The process of acquiring the raw magnetic resonance imaging (MRI) image and performing standardized preprocessing on the raw MRI image to obtain a standardized image specifically includes: Obtain publicly available desensitized magnetic resonance imaging (MRI) images from medical institutions and identify the desensitized MRI images as the original MRI images. The DICOM format parsing function is used to read the DICOM format data of the original magnetic resonance imaging image to obtain the pixel value and maximum pixel value of the original magnetic resonance imaging image. Convert the pixel value and the maximum pixel value from integer data type to double-precision floating-point data type; The converted pixel values are normalized to a preset range to obtain a standardized image; the normalization is achieved by dividing each pixel value by the maximum pixel value.
9. A magnetic resonance imaging image degradation device, characterized in that, include: At least one processor; as well as, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform a magnetic resonance imaging image degradation method according to any one of claims 1-8.
10. A magnetic resonance imaging image degradation storage medium, storing computer-executable instructions, characterized in that, The computer-executable instructions are capable of executing a magnetic resonance imaging image degradation method according to any one of claims 1-8.