Magnetic resonance high-quality rapid imaging method capable of reducing visual field

Abstract

The invention relates to the technical field of magnetic resonance imaging, and discloses a magnetic resonance high-quality rapid imaging method capable of reducing a visual field. The method comprises the following steps of pulse and sequence configuration, local region excitation and data acquisition, image reconstruction, regional denoising optimization and software integration. According to the magnetic resonance high-quality rapid imaging method capable of reducing the visual field, bidirectional spatial selective excitation in the frequency and phase encoding direction is realized through a single two-dimensional radio frequency pulse, a circular or elliptical region of interest can be adapted without multi-pulse matching, and meanwhile, the magnetic resonance imaging efficiency is improved by combining small FOV undersampling and an EPI sequence. The phase coding number and the k space sampling interval are optimized along with the reduction of the visual field, so that the imaging time is greatly shortened, and the relaxation attenuation blurring and partial resonance image distortion artifacts of the EPI sequence are relieved; and regional non-local mean denoising is matched.

Description

Technical Field

[0001] This invention relates to the field of magnetic resonance imaging technology, specifically to a high-quality, rapid magnetic resonance imaging method with a reduced field of view. Background Technology

[0002] Magnetic Resonance Imaging (MRI) is a non-invasive medical imaging technique based on the nuclear magnetic resonance effect. With its advantages such as no ionizing radiation, excellent soft tissue contrast, and the ability to perform tomographic imaging in any direction, it has been widely used in clinical diagnosis and scientific research. Its core principle is to collect raw k-space data and reconstruct images by interacting with hydrogen nuclei in human tissues through magnetic fields and radio frequency pulses, so as to present information about the internal structure and lesions of the human body.

[0003] In MRI imaging, there are key contradictions and constraints among field of view (FOV), imaging time, spatial resolution, and signal-to-noise ratio (SNR). Traditional full FOV imaging requires dense sampling in k-space according to the Nyquist sampling theorem to avoid aliasing artifacts, resulting in a large amount of data acquisition and excessively long imaging time. At the same time, spatial resolution and SNR are negatively correlated, that is, the higher the resolution and the smaller the pixel size, the lower the SNR. However, improving the SNR often requires increasing the number of scans or extending the acquisition time, which further exacerbates the conflict between various performance indicators.

[0004] To address the aforementioned issues, existing technologies are mainly divided into two categories: one category employs undersampling combined with reconstruction algorithms (such as parallel imaging and low-rank reconstruction) to shorten imaging time, but the reconstructed image is still the full field of view (FOV), failing to fully utilize the clinical diagnostic need to focus only on local organs (such as the heart, prostate, and hindbrain), and there is still room for improvement in data acquisition efficiency; the other category achieves local imaging through external volume suppression (OVS) technology or multi-pulse localization excitation, but external volume suppression technology is easily affected by B1 field inhomogeneity, resulting in incomplete signal suppression and relaxation recovery problems; multi-pulse localization excitation technology (such as PRESS and STEAM) has drawbacks such as limited minimum echo time, excessive specific absorption rate (SAR), and decreased signal-to-noise ratio, and it can usually only achieve unidirectional selection, making it difficult to flexibly adapt to regions of interest of different shapes and locations.

[0005] In addition, fast imaging sequences, such as echo planar imaging (EPI), can shorten the acquisition time, but the long readout time can easily lead to signal relaxation and attenuation, causing image blurring, and the partial resonance effect can cause spatial misalignment artifacts. Traditional denoising algorithms, such as non-mean filtering (NLM), have problems such as high computational complexity, fixed parameters, and "noise halo" easily generated at high contrast edges, and cannot adapt to the differences in image features of different regions in local imaging.

[0006] Therefore, there is an urgent need for a local magnetic resonance imaging method that can balance imaging speed, resolution, and signal-to-noise ratio, and effectively suppress artifacts, in order to meet the clinical demand for high-quality and rapid local imaging. Summary of the Invention

[0007] (a) Technical problems to be solved To address the shortcomings of existing technologies, this invention provides a high-quality, rapid magnetic resonance imaging method with a narrow field of view. It features bidirectional spatial selective excitation using a single two-dimensional pulse, small FOV imaging that balances rapid acquisition with low artifacts, regional denoising to adapt to local imaging features, and real-time parameter matching to the region of interest. This solves the problems of existing local imaging techniques, such as the need for multi-pulse coordination / single-directional selection, long full FOV imaging time and prominent artifacts, fixed parameters in traditional denoising algorithms leading to loss of detail or incomplete denoising, and inability to flexibly adapt to the imaging needs of local organs at different locations / sizes.

[0008] (II) Technical Solution To achieve the aforementioned objectives of "bidirectional spatially selective excitation using a single two-dimensional pulse, small FOV imaging balancing rapid acquisition and low artifacts, regional denoising to adapt to local imaging features, and real-time parameter matching to the region of interest," this invention provides the following technical solution: a high-quality, rapid magnetic resonance imaging method with a narrowed field of view, comprising the following steps: Step 1: Pulse and Sequence Configuration: Design a two-dimensional radio frequency excitation pulse, which applies an oscillating gradient in both the frequency encoding direction and the phase encoding direction to achieve spatial selectivity in both directions, replacing the traditional one-dimensional radio frequency excitation pulse; embed the two-dimensional radio frequency excitation pulse into a planar echo imaging (EPI) sequence to form a local excitation imaging sequence; Step 2: Local Region Excitation and Data Acquisition: The user-preset region of interest (ROI) is excited by the two-dimensional radio frequency (RF) excitation pulse. The ROI can be adjusted to achieve a circular or elliptical shape according to the oscillation gradient parameters. Based on the position coordinates and size parameters of the ROI input by the user, the phase shift, bandwidth, and gradient amplitude of the two-dimensional RF excitation pulse are adjusted. Based on the local excitation imaging sequence, the number of phase codes required at the same resolution can be reduced. Furthermore, undersampling can be used to acquire k-space data corresponding to the ROI in the phase coding direction. The number of phase coding lines acquired is less than the number of phase coding lines required for full-field imaging. Step 3: Image Reconstruction: Reconstruct the k-space data using a parallel imaging algorithm or a low-rank reconstruction algorithm to generate an initial image; Step 4: Regional Denoising Optimization: Based on the theoretical excitation contour of the two-dimensional radio frequency excitation pulse, the initial image is divided into a central region, an edge transition region, and a peripheral non-excitation region. Different search window sizes, neighborhood window sizes, and similarity weight coefficients are configured for the central region, the edge transition region, and the peripheral non-excitation region, respectively. A non-local mean denoising algorithm is used to calculate a weighted average based on the structural similarity between pixel blocks, and denoising processing is performed on each region to output the final imaging image.

[0009] Preferably, in step 1, the repetition time of the planar echo imaging (EPI) sequence ( The signal acquisition time within a single execution cycle of the sequence is determined based on the type of human tissue corresponding to the region of interest. The formula is: = ; in, Indicates the repetition time. Indicates the number of the first phase code. Indicates the number of the second phase codes; the and Determined based on the size of the region of interest, and The value is less than the number of first phase codes corresponding to full-field imaging. In 3D imaging, The value can be set according to the specific imaging resolution requirements and the target area range.

[0010] Preferably, in step 2, when acquiring k-space data using a two-dimensional pulse local excitation method, the sampling interval of k-space is... The size increases as the field of view (FOV) corresponding to the region of interest shrinks, reducing the number of acquisitions. It can also be further combined with undersampling, where the phase coding parameters are determined together based on the size of the region of interest and the parameters of the planar echo imaging sequence.

[0011] Preferably, in step 3, the parallel imaging algorithm is mainly the GRAPPA algorithm; the low-rank reconstruction algorithm uses the linear correlation between local data in the k-space of magnetic resonance to construct a low-rank matrix, and obtains the rank constraint of the matrix through singular value decomposition (SVD) for iterative reconstruction. At the same time, the low-rank reconstruction algorithm has a certain denoising capability, which can improve the image signal-to-noise ratio to a certain extent.

[0012] Preferably, in step 4, the search window size configured for the peripheral non-excitation region is 21×21 pixels to 31×31 pixels, and the neighborhood window size is 7×7 pixels to 9×9 pixels; the search window size configured for the central region is 9×9 pixels to 15×15 pixels, and the neighborhood window size is 3×3 pixels to 5×5 pixels. At the same time, by estimating the standard deviation of image noise and determining the NLM smoothing parameter h accordingly, the denoising intensity is adaptively adjusted.

[0013] Preferably, in step 4, before performing denoising on the central region, the local gradient values ​​of each pixel block in the central region are calculated first, and the central region is divided into smooth sub-regions and texture sub-regions according to the local gradient values. Based on the complexity of different regions, a region-related modulation factor is introduced into the exponential decay term of the non-local mean weight, so that the smooth region enhances noise suppression, while the texture region adopts a more conservative weight decay to avoid over-smoothing, thereby achieving a balance between noise suppression and structure preservation.

[0014] Preferably, the method also includes a software integration step, which integrates the corresponding algorithm modules for pulse and sequence configuration, local region excitation and data acquisition, image reconstruction, and regional denoising optimization into the magnetic resonance scanning platform. The scanning platform allows users to input the location coordinates, size parameters, and imaging resolution requirements of the region of interest through an interactive interface.

[0015] Preferably, in step 1, when the two-dimensional radio frequency excitation pulse is combined with the planar echo imaging (EPI) sequence, a pre-scan is used in the EPI sequence for phase correction to eliminate Nyquist artifacts.

[0016] Preferably, in step 2, the size range of the region of interest is 40mm×40mm—150mm×150mm, and the number of phase coding lines when acquiring k-space data is 1 / 3 to 1 / 2 of the number of phase coding lines required for full-field imaging.

[0017] (III) Beneficial Effects Compared with existing technologies, this invention provides a high-quality, rapid magnetic resonance imaging method with a reduced field of view, which has the following advantages: 1. This high-quality, rapid magnetic resonance imaging method with a reduced field of view achieves bidirectional spatially selective excitation of frequency and phase encoding directions through a single two-dimensional radio frequency pulse. It can adapt to circular or elliptical regions of interest without the need for multiple pulses. At the same time, by combining small FOV undersampling and EPI sequences, the number of phase encoding lines and k-space sampling intervals are optimized as the field of view is reduced. This significantly shortens the imaging time (the number of phase encoding lines is only 1 / 3 to 1 / 2 of that of the full field of view) and alleviates the relaxation decay blur and off-resonance misalignment artifacts of EPI sequences. Combined with regional nonlocal mean denoising (differentiated configuration of windows and weights according to the excitation contour, with further subdivision of smooth / textured sub-regions in the central area), it effectively solves the problem of detail loss or incomplete noise reduction caused by fixed denoising parameters in traditional methods, achieving a synergistic improvement in imaging speed, resolution, and signal-to-noise ratio.

[0018] 2. This high-quality, rapid magnetic resonance imaging method with a narrow field of view supports real-time adjustment of pulse phase shift, bandwidth, and gradient amplitude based on the size of the region of interest. Combined with a software-integrated interactive interface, it can flexibly adapt to the imaging needs of different local organs such as the posterior brain and prostate. The choice of reconstruction algorithm, such as parallel imaging (GRAPPA) or low-rank reconstruction (limited images support the construction of low-rank matrices), and the use of pre-scanning for phase correction of EPI sequences further ensure the imaging quality in small fields of view (40mm×40mm—150mm×150mm). It avoids the shortcomings of existing local imaging techniques, such as unidirectional selection, incomplete suppression, or SAR exceeding the limit, and provides a feasible path for high-quality, rapid local magnetic resonance imaging in clinical practice. Attached Figure Description

[0019] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which: Figure 1 This is a flowchart of the magnetic resonance imaging method for reducing the field of view according to the present invention; Figure 2 This is an illustration of the imaging field of view and image resolution of the present invention; Figure 3 The image shown is an EPI imaging reconstruction image from this invention, demonstrating the removal of relevant artifacts and the reduction of distortion. Figure 4 This is a comparison of the final imaging results before and after denoising in the case of hindbrain imaging using local excitation and single-excitation-based SE-EPI sequences according to the present invention. Figure 5 This is a flowchart of the modules of the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] like Figure 1 As shown, the high-quality, rapid magnetic resonance imaging method with a narrowed field of view includes the following core steps: Step 1: Pulse and sequence configuration (corresponding) Figure 5 (Mid-pulse and sequence implementation module).

[0022] A two-dimensional radio frequency excitation pulse is designed, which simultaneously applies an oscillating gradient field in both the frequency encoding and phase encoding directions. This eliminates the need for multiple pulses and directly replaces the traditional one-dimensional radio frequency excitation pulse. The pulse parameters (phase offset, bandwidth, gradient amplitude) can be adjusted to adapt to regions of interest of different shapes. The two-dimensional radio frequency excitation pulse is then embedded into an echo-planar imaging (EPI) sequence to form a local excitation imaging sequence.

[0023] Key optimizations: Phase correction using pre-scanning in EPI sequences to eliminate Nyquist artifacts easily generated in traditional EPI sequences; repetition time of EPI sequences. Determine based on the type of human tissue corresponding to the region of interest (e.g., shortening heart tissue is required). To capture dynamic changes, prostate tissue can be appropriately prolonged. (Improve signal-to-noise ratio), acquisition time Still satisfied = ,in (Number of first phase codes) and The number of second phase codes is determined based on the size of the region of interest, and the total number is less than the number of phase codes corresponding to full-field imaging, thus reducing the amount of data collected from the source.

[0024] Step 2: Local Region Excitation and Data Acquisition (corresponding to) Figure 5 (Acquisition and Reconstruction Module).

[0025] Excitation control: The two-dimensional radio frequency excitation pulse designed in step 1 is used to excite the user-preset circular or elliptical region of interest (size range 40mm×40mm—150mm×150mm). The user can input the position coordinates and size parameters of the region of interest through the scanning interface, and adjust the phase offset, bandwidth and gradient amplitude of the pulse in real time to ensure that the excitation area and the region of interest are accurately matched.

[0026] Data Acquisition: Based on the local excitation imaging sequence, k-space data was acquired using an undersampling method along the phase encoding direction. The number of phase encoding lines acquired was 1 / 3 to 1 / 2 of the number required for full-field imaging. During the undersampling process, the sampling interval in k-space was... Increases as the field of view (FOV) decreases (e.g.) Figure 3 As shown, in the local excitation scenario (smaller than the full field of view scene), combined with the formula: ; in, f is the partial resonant quantity in Hertz (Hz), and BW is the sampling bandwidth. This represents the spatial misalignment of the image. Therefore... Increasing the size can significantly alleviate spatial misalignment artifacts caused by the partial resonance effect, while reducing the readout time of the EPI sequence and reducing image blurring caused by signal relaxation attenuation.

[0027] Step 3: Image Reconstruction (corresponding to...) Figure 5 (Acquisition and Reconstruction Module).

[0028] To reconstruct the undersampled k-space data, a parallel imaging algorithm or a low-rank reconstruction algorithm is applied to generate an initial image. The parallel imaging algorithm selected is the GRAPPA algorithm, which utilizes the linear correlation between the k-space data acquired by the multi-channel coils to fill the gaps in the undersampled data. The low-rank reconstruction algorithm utilizes the linear correlation between local k-space data in magnetic resonance imaging to construct a low-rank matrix, and then uses SVD to obtain the rank constraint of the matrix for iterative reconstruction. Simultaneously, this low-rank reconstruction algorithm has a certain denoising capability, which can improve the image signal-to-noise ratio to some extent. Both types of algorithms can be selected according to the specific imaging situation to ensure that there are no obvious aliasing artifacts after reconstruction of undersampled data (e.g., ...). Figure 3 As shown, locally excited reconstructed images have significantly fewer artifacts than fully excited images.

[0029] Step 4: Region-specific noise reduction optimization (corresponding) Figure 5 (China Signal-to-Noise Ratio Enhancement Module).

[0030] Region division: Based on the theoretical excitation profile of the two-dimensional radio frequency excitation pulse (e.g. Figure 4 As shown, the excitation regions are distributed in a circular pattern, dividing the initial image into a central region (uniform signal and rich details), an edge transition region (large signal gradient changes), and a peripheral non-excitation region (mainly noise). Differentiated parameter configuration: Peripheral non-excitation area: Configure a large search window (21×21 pixels - 31×31 pixels) and a neighboring window (7×7 pixels - 9×9 pixels) to enhance noise suppression; Central region: Configure small search windows (9×9 pixels - 15×15 pixels) and neighborhood windows (3×3 pixels - 5×5 pixels) to avoid loss of detail; further calculate the local gradient values ​​of pixel blocks in the central region, divide the region into smooth sub-regions (small gradient values) and textured sub-regions (large gradient values), and introduce a region-related modulation factor into the exponential decay term of the non-local mean weight according to the complexity of different regions. This enhances noise suppression in smooth regions, while using a more conservative weight decay in textured regions to avoid over-smoothing, thereby achieving a balance between noise suppression and structure preservation. Denoising Execution: A nonlocal mean denoising algorithm is employed, which calculates a weighted average based on the structural similarity between pixel blocks, performs denoising processing on each region, and outputs the final image (e.g., Figure 4 As shown, the noise in the denoised image is significantly reduced while the edge details are well preserved.

[0031] Step 5: Software Integration (corresponding to) Figure 5 (Software scanning module).

[0032] The aforementioned algorithm modules for pulse and sequence configuration, local region excitation and data acquisition, image reconstruction, and regional denoising optimization are integrated into the magnetic resonance scanning platform. The platform allows users to input the location coordinates, size parameters, and imaging resolution requirements of the region of interest through an interactive interface, enabling real-time parameter adjustment and automated imaging processes.

[0033] See Figure 5 As shown, the pulse and sequence implementation module: considering the performance of the magnetic resonance scanner hardware system, it implements two-dimensional excitation pulses through pulse design principles and can perform imaging, while ensuring that parameters can be adjusted according to the region of interest, implementing EPI fast imaging sequences and imaging testing, and achieving artifact removal, such as Nyquist artifacts.

[0034] Accelerated Acquisition and Reconstruction Module: Determines the accelerated acquisition algorithm and module to be adopted, calculates the phase coding parameters required for the sequence, realizes the complete reconstruction of undersampled data, and removes relevant image artifacts, such as aliasing artifacts.

[0035] Signal-to-noise ratio (SNR) enhancement module: Based on the implementation of the denoising algorithm and combined with the characteristics of magnetic resonance imaging, the denoising effect is optimized. At the same time, it solves the problems of the traditional NLM algorithm. Based on the excitation imaging contour information of the two-dimensional pulse, the image is divided into the central region, the edge transition region, and the peripheral non-excitation region. The weights are calculated through the structural similarity index, and different parameter strategies are set for different regions to improve the image SNR. This avoids the disadvantage of increasing the scanning time by averaging multiple times to improve the SNR.

[0036] The software scanning module integrates the method into the scanning platform, allowing users to select the location, size, and other imaging parameters of the region of interest to achieve high-quality imaging.

[0037] The following is in conjunction with the appendix Figure 1-5 As shown, taking "posterior brain magnetic resonance imaging" as an example, the implementation process of the present invention is explained in detail: Step 1: Pulse and sequence configuration: Design a two-dimensional radio frequency excitation pulse. The oscillation gradients in the frequency encoding direction and the phase encoding direction are not the same. Adjust the pulse phase by adjusting the relevant parameters of the radio frequency pulse to move the excitation position and ensure that the excitation area is moved to the back of the head and the excitation shape is elliptical (to better fit the shape of the back of the head). Two-dimensional radio frequency pulses are embedded into an EPI sequence, and DWI imaging is performed using a single-excitation-based SE-EPI sequence. Based on the posterior brain tissue, a time of 3500ms was set, and phase correction was performed using a pre-scan to eliminate Nyquist artifacts. Calculation of acquisition time: Full FOV in multi-layer imaging scenarios =128 (number of first phase codes), 12 layers, small FOV =64, and in the multi-layer acquisition process, multi-chip technology is used. A small FOV can increase the number of layers scanned in a single pass using multi-chip technology, thus effectively reducing the acquisition time; or a small FOV If consistent with full FOV, the final imaging resolution of the image can be improved while keeping the scan time constant.

[0038] Step 2: Local Region Excitation and Data Acquisition Users can locate the region of interest (posterior visual cortex) through the scanning platform's interactive interface, set the imaging FOV parameters, and the system will automatically adjust the frequency phase shift of the two-dimensional radio frequency pulse and the gradient amplitude in both directions. k-space data was acquired using an undersampling method. The number of phase encoding lines was set to half of the full field of view (128 lines) (64 lines). The FOV of the phase encoding direction was reduced by half. The k-space sampling interval was... It will increase by half, combined with the formula. If all other parameters remain the same, Increasing the size reduces the spatial misalignment by 50%.

[0039] Step 3: Image Reconstruction: The GRAPPA parallel imaging algorithm (acceleration factor 2) is used to reconstruct the undersampled k-space data to generate an initial image without obvious aliasing artifacts.

[0040] Step 4: Regional Denoising Optimization: Based on the two-dimensional radio frequency pulse excitation contour, the initial image is divided into the central region (strong imaging signal), the edge transition region, and the peripheral non-excitation region (the remaining region, mainly noise). Parameter configuration: outer region search window 25×25 pixels, neighboring window 8×8 pixels; center region search window 12×12 pixels, neighboring window 4×4 pixels; calculate the local gradient value of the pixel block in the center region, and divide the region according to the gradient for noise reduction; By performing non-local mean denoising, the signal-to-noise ratio of the final image is significantly improved, and image details are clearly visible. Furthermore, by using a block denoising algorithm, the problem of having to scan and average the image multiple times in DWI imaging (especially with high b values) is avoided, effectively preventing excessively long scan times. Alternatively, the in-plane resolution of the image can be improved without extending the scan time, facilitating detailed diagnosis.

[0041] Step 5: Software integration and verification: The above algorithm module is integrated into the 1.5T MRI scanning platform. After the user inputs the parameters of the region of interest, the system quickly completes the entire process from excitation to final imaging, and the imaging quality meets the requirements of clinical visual function MRI diagnosis.

[0042] The embodiments verify that the present invention, through "two-dimensional local excitation - small FOV undersampling - regional denoising", can optimize time, resolution and signal-to-noise ratio in imaging of different local organs and adapt to clinical needs.

[0043] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, 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 a process, method, article, or apparatus. Without further limitations, 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 said element.

[0044] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method of high quality fast imaging with reduced field of view using magnetic resonance, characterized in that, The method comprises the following steps: Step 1: Pulse and sequence configuration: design a two-dimensional radio frequency excitation pulse which simultaneously applies a wobble gradient in the frequency encoding direction and the phase encoding direction to achieve spatial selectivity in both directions, instead of a traditional one-dimensional radio frequency excitation pulse; embed the two-dimensional radio frequency excitation pulse into a planar echo imaging sequence to form a local excitation imaging sequence; Step 2: Local region excitation and data acquisition: excite a user-predefined region of interest through the two-dimensional radio frequency excitation pulse, and the region of interest can be circular or elliptical in shape according to the wobble gradient parameters; adjust the phase offset, bandwidth and gradient amplitude of the two-dimensional radio frequency excitation pulse according to the position coordinates and size parameters of the region of interest input by the user; based on the local excitation imaging sequence, the number of required phase encodings can be reduced under the same resolution, and the k-space data corresponding to the region of interest can be further acquired in the phase encoding direction in combination with an undersampling mode, and the number of acquired phase encoding lines is less than that required for full-field imaging; Step 3: Image reconstruction: apply a parallel imaging algorithm or a low-rank reconstruction algorithm to the k-space data for reconstruction to generate an initial image; Step 4: Regional denoising optimization: divide the initial image into a central region, an edge transition region and a peripheral non-excitation region according to the theoretical excitation profile of the two-dimensional radio frequency excitation pulse; configure different search window sizes, neighborhood window sizes and similarity weight coefficients in the central region, the edge transition region and the peripheral non-excitation region respectively; adopt a non-local mean denoising algorithm, calculate a weighted average value based on the structural similarity between pixel blocks, perform denoising processing on each region, and output a final imaging image.

2. The method of claim 1, wherein: In step 1, the repetition time of the planar echo imaging sequence is determined according to the human tissue type corresponding to the region of interest; the signal acquisition time within a single execution cycle of the sequence The formula is satisfied: = ； wherein, represents a repetition time, represents a first number of phase encodings, represents a second number of phase encodings; the and is determined according to the size of the region of interest, and the value of is less than the first number of phase encodings corresponding to full field of view imaging; when three-dimensional imaging is performed, the value of can be set according to specific imaging resolution requirements and target region range.

3. The method of claim 1, wherein: In step 2, when k-space data is acquired in a two-dimensional pulsed local excitation manner, the sampling interval of k-space The number of acquisitions is reduced as the field of view corresponding to the region of interest is reduced, and further combined with under-sampling, the phase encoding parameter of which is determined according to the size of the region of interest and the parameters of the planar echo imaging sequence.

4. The method of claim 1, wherein: In step 3, the parallel imaging algorithm is mainly GRAPPA algorithm; the low-rank reconstruction algorithm utilizes the linear correlation between local data in the magnetic resonance k-space to construct a low-rank matrix, and obtains the rank constraint of the matrix through singular value decomposition to perform iterative reconstruction, and the low-rank reconstruction algorithm has a certain denoising ability, which can improve the image signal-to-noise ratio to a certain extent.

5. The method of claim 1, wherein: In step 4, the search window size of the peripheral non-excitation region is configured to be 21x21 pixels-31x31 pixels, and the neighborhood window size is configured to be 7x7 pixels-9x9 pixels; the search window size of the central region is configured to be 9x9 pixels-15x15 pixels, and the neighborhood window size is configured to be 3x3 pixels-5x5 pixels; at the same time, the NLM smoothing parameter h is determined according to the estimated image noise standard deviation, to realize adaptive adjustment of the denoising strength.

6. The method of claim 1, wherein: In step 4, before performing denoising processing on the central region, the local gradient values of each pixel block in the central region are calculated, and the central region is divided into a smooth sub-region and a texture sub-region according to the local gradient values; according to the complexity of different regions, a region-related modulation factor is introduced in the exponential decay term of the non-local mean weight, so that the smooth region enhances noise suppression, while the texture region adopts a more conservative weight decay to avoid excessive smoothing, thereby achieving a balance between noise suppression and structure preservation.

7. The method of claim 1, wherein: Also included is a software integration step, which integrates the pulse and sequence configuration, local region excitation and data acquisition, image reconstruction, and sub-region denoising optimization corresponding algorithm modules into a magnetic resonance scanning platform that supports users to input the position coordinates, size parameters, and imaging resolution requirements of the region of interest through an interactive interface.

8. The method of claim 1, wherein: In step 1, when the two-dimensional radio frequency excitation pulse is combined with the planar echo imaging sequence, the pre-scan is used for phase correction in the planar echo imaging sequence to eliminate the Nyquist artifact.

9. The method of claim 1, wherein: In step 2, the size range of the region of interest is 40mm×40mm-150mm×150mm, and the number of phase encoding lines when collecting k-space data is 1 / 3-1 / 2 of the number of phase encoding lines required for full field of view imaging.

Images