Image reconstruction method and system
By introducing non-convex regularization methods and interactive systems in MRI image reconstruction technology, the deviation problem of traditional methods when processing incomplete data is solved, the accuracy and quality of image reconstruction are significantly improved, and the practical application needs such as clinical diagnosis are met.
Patent Information
- Application Number
- CN202510253230.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-05
- Publication Date
- 2025-06-17
AI Technical Summary
The traditional MRI image reconstruction method has deviation problems when processing incomplete data, resulting in a reduced accuracy of the reconstruction image and cannot meet the needs of practical applications such as clinical diagnosis.
A MRI image reconstruction technology based on non-convex regularity is proposed, including an image reconstruction method in online scanning mode and offline reconstruction mode. By designing an interactive system for MRI image reconstruction, it provides data cutting slide bars and a variety of reconstruction algorithms, improving the user-friendliness of image reconstruction and improving the algorithm.
Through non-convex regularization technology and interactive system, the accuracy and quality of image reconstruction are significantly improved, and the features of the original image can be restored more accurately, meeting the practical application needs such as clinical diagnosis.
Smart Images
Figure CN120163892A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of image processing, and in particular to an image reconstruction method and system. Background Art
[0002] Image reconstruction technology is crucial in many fields such as medical imaging and remote sensing. Taking magnetic resonance imaging (MRI) in medical imaging as an example, traditional MRI imaging has the problem of long acquisition time, which not only makes patients uncomfortable, but also may cause image artifacts due to patient movement, affecting the accuracy of diagnosis. At the same time, during the data acquisition process, due to limitations of factors such as equipment performance and acquisition environment, the acquired raw data is often incomplete, which poses a great challenge to image reconstruction. In existing image reconstruction methods, such as traditional sparse convex regularization models, when dealing with such incomplete data, they often exhibit bias problems, resulting in reduced accuracy of reconstructed images, which cannot meet the needs of practical applications such as clinical diagnosis. In practical problems such as image deblurring and MRI reconstruction, how to recover high-quality images from noisy observation data has always been a difficult problem that the industry needs to solve urgently. Summary of the invention
[0003] The following is a summary of the subject matter described in detail herein. This summary is not intended to limit the scope of the claims.
[0004] The embodiments of the present application provide an image reconstruction method and system to improve user friendliness and facilitate algorithm improvement and research.
[0005] In a first aspect, the embodiments of the present application provide an image reconstruction method and system, comprising the following steps:
[0006] S1. Propose an MRI image reconstruction technique based on non-convex regularization.
[0007] S2. Designed an interactive system for MRI image reconstruction.
[0008] In some embodiments, S1. proposes a MRI image reconstruction technique based on non-convex regularization. Image reconstruction method in online scanning mode:
[0009] In the online scanning mode, obtain the data cutting mode determined by the user. Since the acquisition of the original data is not completed at this time, usually only the time-based data cutting mode is available. Obtain the original data parameters (i.e., the planned acquisition time) according to this mode, and generate a data cutting slider. The user sets the start value and the end value on the slider. When setting, it is necessary to follow the rule that if the maximum value N on the data cutting slider is greater than the minimum cutting amount Min, the available range of the start value is (0, N - Min); if the maximum value N is not greater than the minimum cutting amount Min, the start value is defaulted to 0. After setting, collect the original data, load the original data under the data cutting slider according to the data generation time, split the original data according to the start value and the end value, and use the split data as the reconstructed data for image reconstruction.
[0010] In some embodiments, S1: Propose an MRI image reconstruction technique based on non-convex regularization. The image reconstruction method in the offline reconstruction mode:
[0011] In the offline reconstruction mode, the acquisition of the original data has been completed, and both the time-based and count-based data cutting modes can be selected. After obtaining the data cutting mode determined by the user, obtain the original data parameters according to this mode. If the time-based data cutting mode is selected, the parameter is the acquisition time of the collected original data; if the count-based data cutting mode is selected, the parameter is the acquisition amount of the collected original data. After generating the data cutting slider, the user sets the start value and the end value, also following specific setting rules. Then, load the original data according to the data cutting mode selected by the user, split the original data according to the start value and the end value, and use the split data as the reconstructed data for image reconstruction. In the time-based data cutting mode, load and split the data according to the data generation time; in the count-based data cutting mode, load and split the data in the increasing order of the acquisition amount.
[0012] In the second aspect, in some embodiments, S2: Design an interactive system for MRI image reconstruction. It includes:
[0013] System startup and interface display: When performing MRI image reconstruction work, the operator opens the MATLAB software and runs the GUI.m file, and then enters the interactive system interface for the MRI image reconstruction technique based on non-convex regularization and its application. The interface is simple and clear, and has function buttons such as "MRI image reconstruction algorithm" and "basic image processing".
[0014] Image input operation: Click the "MRI Reconstruction Algorithm" button to enter the interactive system for MRI image reconstruction algorithms. At this time, the system interface presents multiple functional modules such as image input, sampling mode, image reconstruction algorithm, and algorithm comparison. Then click the "Open" button, and the system pops up a file selection window, supporting multiple image formats, such as common.bmp,.png,.mat, etc. The operator selects a.mat format image of a brain MRI, and this image is successfully loaded into the system, and its relevant information is displayed in the "Original Image Waveform Diagram" area, providing a basis for subsequent processing.
[0015] Sampling mode selection: In the sampling mode module, the system provides three options: random mode, Cartesian mode, and radial ray mode. The operator selects the radial ray mode according to actual needs, and this selection takes effect immediately in the system. The system prepares to perform corresponding sampling operations on the input image according to the characteristics of the selected mode, which will affect the subsequent image reconstruction effect.
[0016] Residual multiple setting: To more intuitively observe the image reconstruction effect, the operator sets the residual multiple to 5 in the system. The residual multiple is used to control the display degree of the difference between the reconstructed image and the original image. A higher multiple can more clearly show the difference between the two, facilitating the operator to analyze the reconstruction effect.
[0017] Reconstruction algorithm selection and result display: The image reconstruction algorithm module of the system provides six algorithms, namely traditional TV, MCTV, SCADTV, At anTV, CauchyTV, and MCauchyTV. The operator first selects the traditional TV algorithm and clicks to run, and the system quickly starts the calculation program. After a period of operation, the restored image and the residual image are displayed in the "Reconstructed Image Waveform Diagram" and "Residual Image" areas, and at the same time, the relevant indicators of this algorithm, such as PSNR (Peak Signal-to-Noise Ratio), RE (Relative Error), and SSIM (Structural Similarity), etc. are displayed in the "Evaluation Criteria" area, facilitating the operator to evaluate the reconstruction effect. Then, the operator selects the MCauchyTV algorithm to run. After waiting for a while, the obtained reconstructed image is significantly better than the traditional TV model in terms of clarity and detail restoration. Comparing the evaluation criteria of the two, the MCauchyTV algorithm has a higher PSNR value, a lower RE value, and an SSIM value closer to 1, indicating that its reconstruction effect is better and it can more accurately restore the characteristics of the original image.
[0018] Algorithm comparison function: To comprehensively understand the performance differences of different algorithms, the operator clicks the "Algorithm Comparison" button. The system generates a chart through numerical analysis to compare the feasibility of six models from indicators such as RE, PSNR, and SSIM, clearly showing the advantages of the MCauchyTV model in these indicators. At the same time, the system also demonstrates the effectiveness of the CauchyTV model and the MCauchyTV model from the visualization aspect. By comparing the residual images, it is obvious that the residual of the images reconstructed by these two models is less and the image quality is higher, providing a strong basis for the operator to select the appropriate reconstruction algorithm. Summary of the Invention
[0020] The accompanying drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the technical solutions of the present invention, and do not constitute a limitation to the technical solutions of the present invention.
[0021] Figure 1 It is a schematic flowchart of a method for image reconstruction according to the present invention.
[0022] Figure 2 It is a schematic diagram of an image reconstruction system according to the present invention. Detailed Embodiments
[0023] In order to make the objectives, technical solutions and advantages of the present invention clearer, the following further details the present invention in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.
[0024] It should be noted that although the functional modules are divided in the device schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order from the module division in the device or the flowchart. Terms such as "first", "second", etc. in the specification, claims or the above-mentioned drawings are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence.
[0025] In a first aspect, embodiments of the present application provide an image reconstruction method and system, including the following steps:
[0026] S1. MRI image reconstruction technology based on non-convex regularization.
[0027] S2. MRI image reconstruction interactive system.
[0028] In some embodiments, the MRI image reconstruction technology based on non-convex regularization proposed in S1. Image reconstruction method in the online scanning mode:
[0029] In the online scanning mode, obtain the data cutting mode determined by the user. Since the acquisition of the original data is not completed at this time, usually only the time-based data cutting mode is available. According to this mode, obtain the original data parameters (i.e., the planned acquisition time) and generate a data cutting slider. The user sets the start value and the end value on the slider, and when setting, it is necessary to follow the rule that if the maximum value N on the data cutting slider is greater than the minimum cutting amount Min, the available range of the start value is (0, N - Min); if the maximum value N is not greater than the minimum cutting amount Min, the start value is defaulted to 0. After the setting is completed, collect the original data, load the original data under the data cutting slider according to the data occurrence time, split the original data according to the start value and the end value, and use the split data as the reconstructed data for image reconstruction.
[0030] In some embodiments, S1: Propose an MRI image reconstruction technique based on non-convex regularization. The image reconstruction method in the offline reconstruction mode:
[0031] In the offline reconstruction mode, the acquisition of the original data has been completed, and both the time-based and count-based data cutting modes can be selected. After obtaining the data cutting mode determined by the user, obtain the original data parameters according to this mode. If the time-based data cutting mode is selected, the parameter is the acquisition time of the collected original data; if the count-based data cutting mode is selected, the parameter is the acquisition amount of the collected original data. After generating the data cutting slider, the user sets the start value and the end value, also following specific setting rules. Then, load the original data according to the data cutting mode selected by the user, split the original data according to the start value and the end value, and use the split data as the reconstructed data for image reconstruction. In the time-based data cutting mode, load and split the data according to the data occurrence time; in the count-based data cutting mode, load and split the data in the increasing order of the acquisition amount.
[0032] In the second aspect, in some embodiments, S2: Design an interactive system for MRI image reconstruction. It includes:
[0033] System startup and interface display: When performing MRI image reconstruction work, the operator opens the MATLAB software and runs the GUI.m file, and then enters the interactive system interface for the MRI image reconstruction technique based on non-convex regularization and its application. This interface is simple and clear, and is provided with function buttons such as "MRI image reconstruction algorithm" and "basic image processing".
[0034] Image input operation: Click the "MRI Reconstruction Algorithm" button to enter the interactive system for MRI image reconstruction algorithm. At this time, the system interface presents multiple functional modules such as image input, sampling mode, image reconstruction algorithm, and algorithm comparison. Then click the "Open" button, and the system pops up a file selection window, supporting multiple image formats, such as common.bmp,.png,.mat, etc. The operator selects a.mat format image of a brain MRI, and this image is successfully loaded into the system, and its relevant information is displayed in the "Original Image Waveform" area, providing a basis for subsequent processing.
[0035] Sampling mode selection: In the sampling mode module, the system provides three options: random mode, Cartesian mode, and radial ray mode. The operator selects the radial ray mode according to actual needs, and this selection takes effect immediately in the system. The system prepares to perform corresponding sampling operations on the input image according to the characteristics of the selected mode, which will affect the subsequent image reconstruction effect.
[0036] Residual multiple setting: To more intuitively observe the image reconstruction effect, the operator sets the residual multiple to 5 in the system. The residual multiple is used to control the display degree of the difference between the reconstructed image and the original image. A higher multiple can more clearly show the difference between the two, facilitating the operator to analyze the reconstruction effect.
[0037] Reconstruction algorithm selection and result display: The image reconstruction algorithm module of the system provides six algorithms, namely traditional TV, MCTV, SCADTV, At anTV, CauchyTV, and MCauchyTV. The operator first selects the traditional TV algorithm and clicks to run, and the system quickly starts the calculation program. After a period of operation, the restored image and the residual image are displayed in the "Reconstructed Image Waveform" and "Residual Image" areas, and at the same time, the relevant indicators of this algorithm, such as PSNR (Peak Signal-to-Noise Ratio), RE (Relative Error), and SSIM (Structural Similarity), etc. are displayed in the "Evaluation Criteria" area, facilitating the operator to evaluate the reconstruction effect. Then, the operator selects the MCauchyTV algorithm to run. After waiting for a while, the obtained reconstructed image is significantly better than the traditional TV model in terms of clarity and detail restoration. Comparing the evaluation criteria of the two, the MCauchyTV algorithm has a higher PSNR value, a lower RE value, and an SSIM value closer to 1, indicating that its reconstruction effect is better and it can more accurately restore the characteristics of the original image.
[0038] Algorithm comparison function: To comprehensively understand the performance differences of different algorithms, the operator clicks the "Algorithm Comparison" button. The system generates charts through numerical analysis, compares the feasibility of six models from indicators such as RE, PSNR, and SSIM, and clearly shows the advantages of the MCauchyTV model in these indicators. At the same time, the system also demonstrates the effectiveness of the CauchyTV model and the MCauchyTV model from the visualization aspect. By comparing the residual images, it is obvious that the images reconstructed by these two models have less residual and higher image quality, providing a strong basis for the operator to select the appropriate reconstruction algorithm.
Claims
1. An image reconstruction method and system, characterized in that: The following steps are involved: S1. Propose an MRI image reconstruction technique based on non-convex regularization. S2. Designed an interactive system for MRI image reconstruction.
2. The image reconstruction method and system according to claim 1, characterized in that: S1. Propose a MRI image reconstruction technique based on non-convex regularization. Image reconstruction method in online scanning mode: In the online scanning mode, the data cutting mode determined by the user is obtained. Since the original data acquisition is not completed at this time, usually only the time-based data cutting mode is available. According to this mode, the original data parameters (i.e., the planned acquisition time) are obtained and a data cutting slider is generated. The user sets the start value and end value on the slider. When setting, the start value can be set in the range of (0, N-Min) if the maximum value N on the data cutting slider is greater than the minimum cutting amount Min. If the maximum value N is not greater than the minimum cutting amount Min, the start value defaults to 0. After the settings are completed, collect the original data, load the original data under the data cutting slider according to the data occurrence time, segment the original data according to the start value and end value, and use the segmented data as the reconstruction data for image reconstruction.
3. The image reconstruction method and system according to claim 1, characterized in that: S1. Propose an MRI image reconstruction technique based on non-convex regularization. Image reconstruction method in offline reconstruction mode: In the offline reconstruction mode, the acquisition of raw data is completed, and both time-based and count-based data segmentation modes are available. After obtaining the data cutting mode determined by the user, the original data parameters are obtained according to the mode. If the time-based data cutting mode is selected, the parameters are the collection time of the collected original data; If the count-based data cutting mode is selected, the parameter is the amount of raw data collected. After the data cutting slider is generated, the user sets the start value and the end value, which also follow the specific setting rules. Then, the original data is loaded according to the data cutting mode selected by the user, the original data is segmented according to the start value and the end value, and the segmented data is used as the reconstruction data for image reconstruction. In the time-based data segmentation mode, data is loaded and segmented according to the time when the data occurs; in the count-based data segmentation mode, data is loaded and segmented in increasing amounts.
4. The image reconstruction method and system according to claim 1, characterized in that: S2. Designed an interactive system for MRI image reconstruction. System startup and interface display: When performing MRI image reconstruction, the operator opens the MATLAB software, runs the GUI.m file, and then enters the interactive system interface of MRI image reconstruction technology and application based on non-convex regularization. The interface is concise and clear, with function buttons such as "MRI image reconstruction algorithm" and "basic image processing". Image input operation: Click the "MRI reconstruction algorithm" button to enter the MRI image reconstruction algorithm interactive system. At this time, the system interface presents multiple functional modules such as image input, sampling mode, image reconstruction algorithm, and algorithm comparison. Then click the "Open" button, and the system pops up a file selection window, which supports a variety of image formats, such as common .bmp, .png, .mat, etc. The operator selects a .mat format image of brain MRI, which is successfully loaded into the system, and its related information is displayed in the "Original Image Waveform" area, providing a basis for subsequent processing. Sampling mode selection: In the sampling mode module, the system provides three options: random mode, Cartesian mode and radial ray mode. The operator selects the radial ray mode according to actual needs, and the selection takes effect immediately in the system. The system prepares to perform corresponding sampling operations on the input image according to the characteristics of the selected mode, which will affect the subsequent image reconstruction effect. Residual multiple setting: In order to observe the image reconstruction effect more intuitively, the operator sets the residual multiple in the system to 5. The residual multiple is used to control the display degree of the difference between the reconstructed image and the original image. A higher multiple can more clearly show the difference between the two, making it easier for the operator to analyze the reconstruction effect. Reconstruction algorithm selection and result display: The system's image reconstruction algorithm module provides six algorithms, including traditional TV, MCTV, SCADTV, AtanTV, CauchyTV and MCauchyTV. The operator first selects the traditional TV algorithm and clicks to run, and the system quickly starts the calculation program. After a period of calculation, the restored image and residual image are displayed in the "Reconstructed Waveform" and "Residual Image" areas. At the same time, the relevant indicators of the algorithm are displayed in the "Evaluation Criteria" area, such as PSNR (peak signal-to-noise ratio), RE (relative error) and SSIM (structural similarity) values, which are convenient for operators to evaluate the reconstruction effect. After that, the operator selects the MCauchyTV algorithm to run. After waiting for a while, the reconstructed image obtained is significantly better than the traditional TV model in clarity and detail restoration. Comparing the evaluation criteria of the two, the MCauchyTV algorithm has a higher PSNR value, a lower RE value, and an SSIM value closer to 1, indicating that its reconstruction effect is better and can more accurately restore the characteristics of the original image. Algorithm comparison function: In order to more comprehensively understand the performance differences of different algorithms, the operator clicks the "Algorithm Comparison" button. The system generates charts through numerical analysis, comparing the feasibility of the six models from the perspective of RE, PSNR, and SSIM, and clearly showing the advantages of the MCauchyTV model in these indicators. At the same time, the system also demonstrates the effectiveness of the CauchyTV model and the MCauchyTV model from a visual perspective. By comparing the residual images, it is obvious that the images reconstructed by the two models have fewer residuals and higher image quality, providing a strong basis for operators to choose appropriate reconstruction algorithms.