Water-fat separation magnetic resonance imaging method, medium and equipment

Abstract

The invention relates to a water-fat separation magnetic resonance imaging method, a medium and equipment. The water-fat separation magnetic resonance imaging method comprises the steps that S1, water-fat in-phase and out-phase images are collected, and a wrapping phase diagram is obtained through calculation; s2, calculating and obtaining an unwrapped phase diagram based on a phase unwrapping method of pixel clustering and local curved surface fitting; and S3, completing water-fat separation according to the real phase obtained by unwrapping. According to the water-fat separation magnetic resonance imaging method, a two-dimensional phase imaging algorithm based on pixel clustering and local curved surface fitting is adopted for phase unwrapping, a potential real phase can still be obtained robustly under the conditions of serious noise, rapid phase change or non-connected regions, the water-fat separation accuracy is improved, and the water-fat separation efficiency is improved. The water-fat exchange phenomenon can be effectively avoided, the accurate water image and fat image are generated, the clinical diagnosis requirement is met, and a new technical scheme is provided for phase correlation magnetic resonance imaging application.

Description

Technical Field

[0001] This invention belongs to the field of cardiac magnetic resonance imaging, and specifically relates to water-lipid separation magnetic resonance imaging methods, media, and equipment. Background Technology

[0002] Fat is an important component of human tissue, mainly distributed in the abdominal cavity, subcutaneous tissue, and between muscle fibers, and plays an important physiological role. In magnetic resonance imaging (MRI), fat has a short T1 and a long T2, appearing as a high signal on T1- and T2-weighted images. Due to these characteristics, fat can mask lesions and reduce the effectiveness of contrast-enhanced scans in clinical MRI examinations, thus interfering with diagnosis. Furthermore, there is a clinical need to quantitatively measure fat content, such as for assessing the degree of fatty liver.

[0003] Currently, the mainstream fat suppression techniques in clinical practice include chemical shift-selective presaturation (CFS), short-tau inversion recovery (STIR), and the Dixon technique. While CFS offers high selectivity and ease of use, it is highly dependent on field strength and requires high uniformity of the main magnetic field and radiofrequency field. STIR, compared to CFS, is less dependent on the main magnetic field and radiofrequency field, but suffers from lower image signal-to-noise ratio, lower selectivity for signal suppression, longer scan time, and incompatibility with contrast-enhanced imaging. The first two methods only achieve fat suppression, while the Dixon method, as a water-lipid separation technique, can quantify fat while simultaneously suppressing it.

[0004] The Dixon method, proposed by William Thomas Dixon in 1984, utilizes the chemical shift difference between water and fat protons. By adjusting sequence parameters, it acquires images with an angle of 0° and π between water and fat, respectively, resulting in a water-fat in-phase image and a water-fat out-of-phase image. These two images are then used to calculate the water and fat images. The original two-point Dixon technique suffers from phase errors due to inhomogeneities in the main magnetic field, leading to water-fat misconversion in the obtained water and fat images. To overcome this defect, in 1991, Glover and Schneider proposed the three-point Dixon method, which adds a third measurement (-π, 0, π) to the original two measurements. This additional information is used to correct the phase error. However, because the three-point Dixon method requires three measurements, the scanning time is longer, and it is susceptible to motion artifacts. In reality, three-point Dixon data contains redundant information, and phase error correction can be achieved using only two-point Dixon data. Therefore, in 1996, Thomas E. Skinner proposed the extended two-point Dixon method. Compared to the three-point Dixon method, the extended two-point Dixon method can reduce data acquisition time and has been more widely used in practice. However, this method also has certain drawbacks. Near the water-fat boundary, where the proportions of water and fat are similar, the signal-to-noise ratio is low, leading to magnetic field inhomogeneity and inaccurate phase error correction.

[0005] The key to the Dixon method lies in correcting phase errors caused by inhomogeneity of the main magnetic field. Currently developed phase error correction methods mainly include phase unwrapping, region-growing, and regional iterative phasor extraction. These methods have all driven the development of Dixon technology. In recent years, the combination of multi-echo Dixon sequences and parallel coil acquisition technology has further improved scanning speed. Moreover, because it is based on a multi-peak fat model, it has also further improved the accuracy of quantitative fat measurement and has been successfully applied to precise quantitative fat studies in the liver.

[0006] However, the Dixon technique still has the following problems: (1) Excessive phase entanglement caused by magnetic field inhomogeneity: The Dixon method relies on the difference in chemical shifts between water and fat protons, and is quite sensitive to the inhomogeneity of the main magnetic field, which may lead to a large phase error and affect the separation effect of water and fat. (2) "Sawtooth" phenomenon at the water-fat boundary: During the phase correction process of the two-point Dixon method, near the water-fat boundary, the proportions of water and fat are similar, and the signal-to-noise ratio is low, which leads to inaccurate phase error correction caused by magnetic field inhomogeneity, resulting in the "sawtooth" phenomenon. Therefore, improvements are still needed. Summary of the Invention

[0007] To address the shortcomings of the prior art, this invention aims to provide a method, medium, and device for water-lipid separation magnetic resonance imaging based on Dixon sequences.

[0008] A first aspect of the present invention provides a water-lipid separation magnetic resonance imaging method, comprising the following steps: S1 acquires in-phase and out-of-phase images of the water-lipid mixture and calculates the entanglement phase diagram.

[0009] S2 uses a phase unwinding method based on pixel clustering and local surface fitting to calculate the unwound phase map; S3 completes the separation of water and lipids based on the true phase obtained from untangling.

[0010] This method employs a two-dimensional phase imaging algorithm based on pixel clustering and local surface fitting for phase unwrapping. It can robustly obtain the potential true phase even in the presence of severe noise, rapid phase changes, or disconnected regions, thereby improving the accuracy of water-fat separation.

[0011] Preferably, in step S2, in-phase and out-of-phase images of water and lipids are acquired based on FSE-Dixon and GRE-Dixon sequences.

[0012] Preferably, step S2 includes the following steps: S21 constructs a multi-resolution image pyramid based on the acquired in-phase and out-of-phase images; S22 From the lowest resolution layer to the highest resolution layer of the multi-resolution image pyramid, a layered phase unwrapping strategy is adopted, and phase unwrapping is performed by combining pixel clustering and the CLOSE algorithm for local surface fitting. The unwrapping results of the low-resolution layer are used to guide the phase unwrapping of the high-resolution layer.

[0013] Preferably, the construction of the multi-resolution image pyramid in step S21 includes downsampling and upsampling operations, both using bilinear interpolation with a scaling factor of 0.5, and the multi-resolution image pyramid has 3 layers.

[0014] Preferably, the specific process of layered phase unwrapping in step S22 is as follows: (a) For the phase to be unwrapped in the current resolution layer, the initial phase unwrapping is performed using the upsampled phase map from the unwrapping results of the low resolution layer; (b) Input the initial phase unwrapping result into the CLOSE algorithm, and divide the pixels into easily unwrapped blocks and difficult-to-unwrap residual pixels according to the local change threshold of the true phase and the preset minimum block size; (c) Using the region-growing local polynomial surface fitting method, the easily unwound blocks are unwound intra-block, unwound inter-block, and the difficult-to-unwound residual pixels are unwound in sequence.

[0015] More preferably, the formula for calculating the initial phase unwrapping in step (a) is: ,in ; This is the initial phase unwrapping result. Let n be the k-th layer of phase to be unwrapped, and n be an integer offset. For the upsampled phase, the round() function is used to calculate the nearest integer.

[0016] More preferably, the minimum block size preset in step (b) is adjusted according to the resolution layer. For example, in one embodiment, the lowest resolution layer is 16 pixels, the intermediate resolution layer is 32 pixels, and the highest resolution layer is 64 pixels.

[0017] Preferably, the calculation formula for water-lipid separation is: , ; in, These are in-phase images. It is an inverse phase image. The true phase after untangling is represented by W, which is the water signal, and F is the grease signal.

[0018] A second aspect of the present invention also provides a computer-readable storage medium storing computer instructions for causing a processor to perform phase unwrapping and water-oil separation imaging using the method described above.

[0019] A third aspect of the present invention also provides a magnetic resonance imaging device that uses the above-described method for phase unwrapping and water-fat separation imaging.

[0020] The beneficial effects of this invention are as follows: (1) The water-lipid separation magnetic resonance imaging method of the present invention uses a two-dimensional phase imaging algorithm based on pixel clustering and local surface fitting to unwrap the phase. It can still robustly obtain the potential true phase even in the presence of severe noise, rapid phase change or disconnected regions, improve the accuracy of water-lipid separation, effectively avoid water-lipid exchange phenomenon, generate accurate water and lipid images, meet the needs of clinical diagnosis, and provide a new technical solution for phase-related MRI applications.

[0021] (2) Furthermore, the solution of the present invention generates a smoother phase map through the noise suppression effect of the multi-resolution pyramid structure, which reduces the difficulty of phase unwrapping. It can obtain unwrapping results without obvious wrapping residue in scenarios with disconnected regions, different signal-to-noise ratios and different phase change levels. Compared with traditional methods (such as B0-NICEbd, ROMEO, GraphCut, CLOSE), it has stronger anti-interference ability.

[0022] (3) Furthermore, the present invention optimizes the CLOSE algorithm, reduces the sensitivity to the selection of clustering threshold, eliminates the need to manually adjust the optimal threshold according to different application scenarios, and can maintain a low error rate even in extreme threshold cases.

[0023] (4) The computational efficiency of the present invention is moderate and can meet the efficiency requirements of clinical applications. Detailed Implementation

[0024] To enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be further described in detail below with reference to embodiments.

[0025] Dixon's water-lipid separation technology utilizes the difference in chemical shifts between water and fat protons to separate the signals of these two tissues. However, due to magnetic field inhomogeneities, significant phase errors can occur. Appropriate phase error correction can improve the accuracy of water-lipid separation.

[0026] This invention provides a water-lipid separation magnetic resonance imaging method, comprising the following steps: S1 acquires in-phase and out-of-phase images of the water-lipid mixture and calculates the entanglement phase diagram.

[0027] S2 uses a phase unwinding method based on pixel clustering and local surface fitting to calculate the unwound phase map; S3 completes the separation of water and lipids based on the true phase obtained from untangling.

[0028] The method of the present invention uses a two-dimensional phase imaging algorithm based on pixel clustering and local surface fitting for phase unwrapping, which can robustly obtain the potential true phase even in the presence of severe noise, rapid phase changes or disconnected regions, thereby improving the accuracy of water-fat separation.

[0029] In a preferred embodiment, step S1 of the present invention uses FSE-Dixon and GRE-Dixon sequences to acquire in-phase and out-of-phase images of water and lipid tissue, and calculates the entanglement phase map. The scanning parameters used—repetition time, echo time, number of layers, FOV, and echo train length—are all clinically common parameters. The main scanning sites are the hip joint, ankle joint, cervical spine, and lumbar spine, which are clinically used for fat suppression techniques, to obtain scanning data.

[0030] In some specific embodiments, the following parameters may be used: Sagittal ankle joint: TR / TE = 2060 / 45.6 ms, slice thickness = 3.0 mm, FOV = 190 × 190 mm 2 Acquisition matrix = 280×210, echo chain length = 7, bandwidth = 80 kHz; Sagittal lumbar spine: TR / TE = 2180 / 91.2 ms, slice thickness = 4.0 mm, FOV = 280 × 280 mm 2 Acquisition matrix = 280×180, echo chain length = 15, bandwidth = 80 kHz; Axial pelvis: TR / TE=3300 / 51 ms, slice thickness=4.0 mm, FOV=340×340 mm 2 Acquisition matrix = 320×238, echo train length = 9, bandwidth = 100 kHz; Axial view of the knee joint: TR / TE = 1900 / 43.8 ms, slice thickness = 4.0 mm, FOV = 180 × 180 mm 2 Acquisition matrix = 280×170, echo chain length = 5, bandwidth = 80 kHz.

[0031] Step S2: Based on pixel clustering and local surface fitting, a phase unwinding method is used to calculate the unwound phase map. Specifically, pixels are first clustered into blocks that are easy to unwind and residual pixels that are difficult to unwind. Then, a local polynomial surface fitting method based on region growth is used to sequentially perform intra-block, inter-block, and residual pixel phase unwinding, and finally obtain the unwound phase map.

[0032] In a preferred embodiment, step S2 includes the following steps: S21 constructs a multi-resolution image pyramid based on the acquired in-phase and out-of-phase images; S22 From the lowest resolution layer to the highest resolution layer of the multi-resolution image pyramid, a layered phase unwrapping strategy is adopted, and phase unwrapping is performed by combining pixel clustering and the CLOSE algorithm for local surface fitting. The unwrapping results of the low-resolution layer are used to guide the phase unwrapping of the high-resolution layer.

[0033] Employing a multi-resolution pyramid structure can suppress noise, generate a smoother phase map, and reduce the difficulty of phase unwrapping. It can obtain unwrapping results without obvious wrapping residue in scenarios with disconnected regions, different signal-to-noise ratios, and different phase change levels. Compared with traditional methods (such as B0-NICEbd, ROMEO, Graph Cut, and CLOSE), it has stronger anti-interference ability.

[0034] The local surface fitting CLOSE algorithm is used for hierarchical phase unwinding. Compared with the traditional CLOSE algorithm, it has significantly reduced sensitivity to the selection of clustering threshold. There is no need to manually adjust the optimal threshold according to different application scenarios. Even in extreme cases where the threshold is set to π / 10 or π / 5, it can still maintain a low error rate. Moreover, the water-lipid separation accuracy is high, which can effectively avoid water-lipid exchange phenomenon and generate accurate water and lipid images to meet the needs of clinical diagnosis.

[0035] In a preferred embodiment, the construction of the multi-resolution image pyramid in step S21 includes downsampling and upsampling operations, both using bilinear interpolation with a scaling factor of 0.5. The multi-resolution image pyramid has three layers, providing moderate computational efficiency that meets the efficiency requirements of clinical applications while ensuring processing quality. In a specific embodiment, using a three-layer pyramid structure, processing a 280×210 image takes only about 3 seconds (hardware configuration: ASUS, AMD Ryzen 9 7940HX, 16 GB RAM).

[0036] In one specific embodiment, the downsampling operation involves generating low-resolution layers of the pyramid step-by-step through bilinear interpolation downsampling based on the acquired in-phase and out-of-phase images. The downsampling scaling factor is 0.5, i.e., the k-th layer image. It is obtained by downsampling the image of layer k-1, and the expression is: , where k=2,...,K, and K is the number of pyramid levels (e.g., K=3).

[0037] Upsampling operation: High-resolution layers of the pyramid are generated through bilinear interpolation upsampling. The scaling factor for upsampling is also 0.5, meaning that the (k-1)th layer image is obtained by upsampling the kth layer image. The expression is: , where k=K,...,2.

[0038] Phase unwrapping is generally performed progressively from the lowest resolution layer (Level K-1) to the highest resolution layer (Level 1) of the pyramid. The unwrapping results of the low-resolution layers are used to guide the phase unwrapping of the high-resolution layers. In a preferred embodiment, the specific process of layered phase unwrapping in step S22 is as follows: (a) For the phase to be unwrapped in the current resolution layer, the initial phase unwrapping is performed using the upsampled phase map from the unwrapping results of the low resolution layer; Because the interpolation operations during downsampling and upsampling slightly alter the original data and introduce additional errors, the upsampled phase cannot be directly used as input to the CLOSE algorithm. In a preferred embodiment, the initial phase unwrapping calculation formula in step (a) is: ,in ; This is the initial phase unwrapping result. Let n be the k-th layer of phase to be unwrapped, and n be an integer offset. The round() function is used to calculate the nearest integer to the expression within parentheses, representing the upsampled phase from the unwrapping result of the low-resolution layer.

[0039] (b) Input the initial phase unwrapping result into the CLOSE algorithm, and divide the pixels into easily unwrapped blocks and difficult-to-unwrap residual pixels according to the local change threshold of the true phase and the preset minimum block size; This step involves the CLOSE algorithm to finely detangle pixels. The CLOSE algorithm first categorizes pixels into easily tangled blocks and difficult-to-detangle residual pixels based on a local change threshold of the true phase. Simultaneously, small blocks with fewer pixels than a preset minimum block size are also classified as residual pixels. Different minimum block sizes are set for different resolution layers. In a preferred embodiment, the preset minimum block size in step (b) is adjusted according to the resolution layer. For example, in one specific embodiment, the lowest resolution layer has 16 pixels, the intermediate resolution layer has 32 pixels, and the highest resolution layer has 64 pixels.

[0040] (c) The Local Polynomial Surface Fitting (LPSF) method is used to sequentially untangle easily untangled blocks intra-block untangled, inter-block untangled, and untangled difficult-to-untangle residual pixels.

[0041] In a preferred embodiment, the calculation formula for water-fat separation in step S3 is: , ; in, These are in-phase images. It is an inverse phase image. The true phase after untangling is represented by W, which is the water signal, and F is the grease signal.

[0042] As can be seen, this invention provides a phase unwrapping method based on multi-resolution pixel clustering and local surface fitting. By integrating the multi-resolution strategy with the local surface fitting CLOSE algorithm, the dependence of the CLOSE algorithm on the clustering threshold is reduced, the robustness and accuracy of Dixon MRI phase unwrapping are improved, and the water-lipid separation effect is optimized, providing a new technical solution for phase-correlated MRI applications.

[0043] Embodiments of the present invention also provide a computer-readable storage medium storing computer instructions for causing a processor to execute the aforementioned phase unwrapping and water-oil separation imaging.

[0044] Embodiments of the present invention also provide a magnetic resonance imaging device that utilizes the methods mentioned in the embodiments of the present invention for phase unwrapping and water-fat separation imaging.

[0045] This invention is not limited to the applications listed in the specification and embodiments. It can be applied to various fields suitable for this invention. Without departing from the spirit and essence of this invention, those skilled in the art can easily make other modifications and variations, but these corresponding modifications and variations should all fall within the protection scope claimed by this invention.

[0046] The above description is only a part of the embodiments of the present invention, and is not intended to limit the implementation and protection scope of the present invention. Those skilled in the art should realize that any equivalent substitutions and obvious changes made based on the content of the present invention specification should be included within the protection scope of the present invention.

Claims

1. A water-fat separation magnetic resonance imaging method, characterized in that, Includes the following steps: S1 acquires in-phase and out-of-phase images of the water-lipid mixture and calculates the entanglement phase map. S2 uses a phase unwinding method based on pixel clustering and local surface fitting to calculate the unwound phase map; S3 completes the separation of water and lipids based on the true phase obtained from untangling.

2. The method as described in claim 1, characterized in that, In step S2, in-phase and out-of-phase images of water and lipids are acquired based on FSE-Dixon and GRE-Dixon sequences.

3. The method as described in claim 1, characterized in that, Step S2 includes the following steps: S21 constructs a multi-resolution image pyramid based on the acquired in-phase and out-of-phase images; S22 From the lowest resolution layer to the highest resolution layer of the multi-resolution image pyramid, a layered phase unwrapping strategy is adopted, and phase unwrapping is performed by combining pixel clustering and the CLOSE algorithm for local surface fitting. The unwrapping results of the low-resolution layer are used to guide the phase unwrapping of the high-resolution layer.

4. The method according to claim 3, characterized in that, The construction of the multi-resolution image pyramid in step S21 includes downsampling and upsampling operations, both using bilinear interpolation with a scaling factor of 0.

5. The multi-resolution image pyramid has 3 layers.

5. The method according to claim 3, characterized in that, Step S22, layered phase unwrapping, includes: (a) For the phase to be unwrapped in the current resolution layer, the initial phase unwrapping is performed using the upsampled phase map from the unwrapping results of the low resolution layer; (b) Input the initial phase unwrapping result into the CLOSE algorithm, and divide the pixels into easily unwrapped blocks and difficult-to-unwrap residual pixels according to the local change threshold of the true phase and the preset minimum block size; (c) Using the region-growing local polynomial surface fitting method, the easily unwound blocks are unwound intra-block, unwound inter-block, and the difficult-to-unwound residual pixels are unwound in sequence.

6. The method according to claim 5, characterized in that, The formula for calculating the initial phase unwrapping in step (a) is: ,in ; This is the initial phase unwrapping result. Let n be the k-th layer of phase to be unwrapped, and n be an integer offset. For the upsampled phase, the round() function is used to calculate the nearest integer.

7. The method according to claim 5, characterized in that, The minimum block size preset in step (b) is adjusted according to the resolution layer.

8. The method according to claim 1, characterized in that, In step S3, the calculation formula for water-fat separation is: in, These are in-phase images. It is an inverse phase image. The true phase after untangling is represented by W, which is the water signal, and F is the grease signal.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause a processor to execute the method of any one of claims 1-8 for phase unwrapping and water-lipid separation imaging.

10. A magnetic resonance imaging device, characterized in that, Phase unwrapping and water-lipid separation imaging are performed using the method described in any one of claims 1-8.

Images