CN109188327B - Magnetic resonance image fast reconstruction method based on tensor product complex small compact framework - Google Patents
Magnetic resonance image fast reconstruction method based on tensor product complex small compact framework Download PDFInfo
- Publication number
- CN109188327B CN109188327B CN201811124319.XA CN201811124319A CN109188327B CN 109188327 B CN109188327 B CN 109188327B CN 201811124319 A CN201811124319 A CN 201811124319A CN 109188327 B CN109188327 B CN 109188327B
- Authority
- CN
- China
- Prior art keywords
- magnetic resonance
- image
- resonance image
- tensor product
- matrix
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R33/00—Arrangements or instruments for measuring magnetic variables
- G01R33/20—Arrangements or instruments for measuring magnetic variables involving magnetic resonance
- G01R33/44—Arrangements or instruments for measuring magnetic variables involving magnetic resonance using nuclear magnetic resonance [NMR]
- G01R33/48—NMR imaging systems
- G01R33/54—Signal processing systems, e.g. using pulse sequences ; Generation or control of pulse sequences; Operator console
- G01R33/56—Image enhancement or correction, e.g. subtraction or averaging techniques, e.g. improvement of signal-to-noise ratio and resolution
- G01R33/561—Image enhancement or correction, e.g. subtraction or averaging techniques, e.g. improvement of signal-to-noise ratio and resolution by reduction of the scanning time, i.e. fast acquiring systems, e.g. using echo-planar pulse sequences
- G01R33/5611—Parallel magnetic resonance imaging, e.g. sensitivity encoding [SENSE], simultaneous acquisition of spatial harmonics [SMASH], unaliasing by Fourier encoding of the overlaps using the temporal dimension [UNFOLD], k-t-broad-use linear acquisition speed-up technique [k-t-BLAST], k-t-SENSE
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/05—Detecting, measuring or recording for diagnosis by means of electric currents or magnetic fields; Measuring using microwaves or radio waves
- A61B5/055—Detecting, measuring or recording for diagnosis by means of electric currents or magnetic fields; Measuring using microwaves or radio waves involving electronic [EMR] or nuclear [NMR] magnetic resonance, e.g. magnetic resonance imaging
Landscapes
- Health & Medical Sciences (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- General Health & Medical Sciences (AREA)
- Radiology & Medical Imaging (AREA)
- Engineering & Computer Science (AREA)
- High Energy & Nuclear Physics (AREA)
- Signal Processing (AREA)
- Medical Informatics (AREA)
- Condensed Matter Physics & Semiconductors (AREA)
- Biophysics (AREA)
- Pathology (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- General Physics & Mathematics (AREA)
- Molecular Biology (AREA)
- Surgery (AREA)
- Animal Behavior & Ethology (AREA)
- Public Health (AREA)
- Veterinary Medicine (AREA)
- Magnetic Resonance Imaging Apparatus (AREA)
Abstract
The invention discloses a magnetic resonance image fast reconstruction method based on a tensor product complex small compact frame, and provides a new research method for magnetic resonance fast imaging. According to the invention, the K space data is subjected to undersampling by utilizing a Cartesian sampling track mode, so that the scanning speed of the equipment is greatly improved; the image is decomposed from multiple directions based on sparse transformation of a tensor product complex small compact frame, so that the precision of magnetic resonance imaging is improved; the convex optimization problem in the image reconstruction is solved by using a projection fast iterative soft threshold algorithm, so that the reconstruction speed of the magnetic resonance image is increased; in the projection fast iterative soft threshold algorithm, regularization parameters are calculated in a self-adaptive mode by adopting a bivariate contraction-based method, so that the process of blindly selecting the parameters is omitted, and the reconstruction quality of the image is improved.
Description
Technical Field
The invention belongs to the technical field of magnetic resonance imaging, and particularly relates to a magnetic resonance image fast reconstruction method based on a tensor product complex wavelet compact framework.
Background
Magnetic Resonance Imaging (MRI) has the characteristics of no ionizing radiation, multi-angle Imaging, no damage to human tissues and the like, so that the MRI becomes an important detection means in medical clinic and medical scientific research; however, MRI has the disadvantages of slow imaging speed and low time resolution of dynamic MRI; meanwhile, a single patient has long scanning time and high cost, and further popularization of MRI is limited. In order to increase the scanning speed and improve the scanning efficiency, Parallel Imaging (PI) technology and compressive sensing theory have become a research hotspot in recent years.
The parallel magnetic resonance imaging technology adopts a multi-channel sampling coil to simultaneously receive magnetic resonance signals, and reduces the number of phase encoding times required when compared with the use of a single coil, thereby reducing the signal acquisition time and accelerating the imaging speed. Currently, two methods are widely used to reconstruct images from acquired partially parallel data, one of which uses coil sensitivity information, such as sense (sensitivity encoding); another approach uses correlations between multi-channel K-space data, such as GRAPPA (Generalized Auto-calibrating Parallel acquisition). However, the acceleration factor of parallel magnetic resonance is still limited due to the performance limitations of the severely ill-conditioned system matrix and reconstruction algorithms.
The compressed sensing theory holds that as long as the signals are sparse, or sparse in a certain transform domain, we can break through the limitation of Nyquist sampling rate and use less data to recover a large amount of signals.
Disclosure of Invention
The invention aims to overcome the defects of the method and provides a magnetic resonance image fast reconstruction method based on a tensor product complex small compact frame, the method takes the tensor product complex small compact frame as sparse transformation, and decomposes signals from multiple directions, thereby improving the precision of magnetic resonance imaging; the image reconstruction problem is solved by using a projection fast iterative soft threshold algorithm, so that the speed of magnetic resonance imaging reconstruction is improved; and the regularization parameters are calculated in a self-adaptive manner based on a bivariate shrinkage method, so that the reconstruction quality of the image is further improved.
The method comprises the following specific steps:
1) and performing undersampling on the magnetic resonance K space data by utilizing a Cartesian sampling track to obtain the K space undersampled data.
2) And performing zero filling on the K space undersampled data, and then obtaining an initial magnetic resonance image through Fourier inverse transformation.
3) Taking the initial magnetic resonance image as an input image, carrying out sparse decomposition by using a tensor product complex wavelet compact frame, and obtaining a decomposition coefficient expression after decomposition as follows:
wherein α is the decomposition coefficient, x is the input image, and T is the transposed symbol;a complex wavelet tightening framework is a tensor product;
4) and (3) performing minimum solution on the objective function of the magnetic resonance image by adopting a projection fast iterative soft threshold algorithm, wherein the optimization problem is expressed as follows:
the N-dimensional complex vector x and the M-dimensional complex vector y respectively represent a magnetic resonance image to be reconstructed and K-space undersampled data, the N × N complex matrix F represents a discrete Fourier transform matrix, the M × N real matrix U represents an undersampled matrix when the K-space data are undersampled, and the M × N real matrix U represents an undersampled matrix when the K-space data are undersampled<N; λ is the regularization parameter, | ·| non-woven phosphor2Is a two-norm operation, | | ·| non-conducting phosphor1Is a norm operation.
The iteration equation is as follows:
μ=xk+F*UT(y-UFxk)
wherein x iskRepresenting the reconstructed image at the k-th iteration; c. mu and sk+1Are all intermediate transition vectors, F*And UTRespectively representing discrete Fourier transformsTransposing of transform and sampling matrices;is a soft threshold function for shrinking the wavelet coefficients; t is tkStep size for the k-th iteration.
And when the iteration times reach a set value, outputting the reconstructed magnetic resonance image.
Further, each layer α in decomposition coefficient αqThe corresponding λ is evaluated using a bivariate contraction function, as follows:
wherein, αqIs the current scaling factor, αpIs αqCoefficient of the parent scale, σαIs within the current sub-band (current layer) q at α q3 × 3 window centered prediction variance, noise varianceUsing a median estimator to estimate:
α thereinhIs the coefficient of the high-frequency sub-band in the first layer decomposition, and the mean is the operator for taking the median.
The invention discloses a magnetic resonance image fast reconstruction method based on a tensor product complex small compact framework, and provides a new research method for magnetic resonance fast imaging. According to the invention, the K space data is subjected to undersampling by utilizing the Cartesian sampling track, so that the scanning speed of the equipment is greatly improved; the image is decomposed from multiple directions based on sparse transformation of a tensor product complex small compact frame, so that the precision of magnetic resonance imaging is improved; the convex optimization problem in the image reconstruction is solved by using a projection fast iterative soft threshold algorithm, so that the reconstruction speed of the magnetic resonance image is increased; and in the projection fast iterative soft threshold algorithm, regularization parameters are calculated in a self-adaptive mode by adopting a bivariate contraction-based method, so that the reconstruction quality of the image is further improved.
Drawings
FIG. 1 is a schematic flow chart of the present invention.
Fig. 2 is a diagram of a cartesian sampling trajectory pattern.
Fig. 3(a) is a shoulder magnetic resonance image.
Fig. 3(b) is a chest magnetic resonance image.
Fig. 3(c) shows a brain magnetic resonance image.
FIG. 4(a) is a graph showing the peak signal-to-noise ratio PSNR of the magnetic resonance image of the brain compared with the currently popular D L MRI, PBDWS, PANO and GBRWT algorithms under different acceleration factors.
Fig. 4(b) is a diagram comparing the present invention with the currently popular D L MRI, PBDWS, PANO and GBRWT algorithms to measure SSIM on structural similarity of magnetic resonance images of the brain under different acceleration factors.
Detailed Description
In order to more specifically describe the present invention, the following detailed description is provided for the technical solution of the present invention with reference to the accompanying drawings and the specific embodiments.
As shown in fig. 1, the method for rapidly reconstructing a magnetic resonance image based on a tensor product complex wavelet compact framework of the present invention comprises the following specific steps:
1. undersampling magnetic resonance K space data based on a Cartesian sampling track shown in FIG. 2 to obtain K space undersampled data, wherein a white pixel point value is 1 and represents a sampling point; the black pixel point value is 0, indicating a non-sampling point.
2. And (4) performing zero filling on the K-space undersampled data (setting non-sampling point values as 0), and obtaining an initial magnetic resonance image through Fourier inversion.
3. Taking the initial magnetic resonance image as an input image, carrying out sparse decomposition by using a tensor product complex wavelet compact frame, and obtaining a decomposition coefficient expression after decomposition as follows:
wherein α isDecomposition coefficients (also called wavelet coefficients), x being the input image and T being the transposed symbol;for the tensor product complex small compact framework, the construction method adopts the prior art, such as Han B, Zhao Z.Tensor product complete frame with creating direction, which is published in 7 months in 2014 in the Journal on ImagingSciences.
4. And (3) minimizing and solving an objective function of the magnetic resonance image by adopting a projection fast iterative soft threshold algorithm (pFISTA), wherein the optimization problem is expressed as follows:
where the N-dimensional complex vector x and the M-dimensional complex vector y represent the magnetic resonance image to be reconstructed and the K-space undersampled data, respectively, the N × N complex matrix F represents a discrete fourier transform matrix, the M × N (M × N)<N) the real number matrix U represents an undersampling matrix when undersampling K space data; λ is the regularization parameter, | ·| non-woven phosphor2Is a two-norm operation, | | ·| non-conducting phosphor1Is a norm operation.
The iteration equation is as follows:
μ=xk+F*UT(y-UFxk)
wherein x iskRepresenting the reconstructed image at the k-th iteration; c. mu and sk+1Are all intermediate transition vectors, F*And UTRespectively representing inverse discrete Fourier transform and transposition of a sampling matrix;is a soft threshold function for shrinking the wavelet coefficients; t is tkStep size for the k-th iteration.
Wherein each layer α of decomposition coefficients α is evaluated using a bivariate contraction functionqThe corresponding λ is specifically as follows:
wherein, αqIs the current scaling factor, αpIs αqCoefficient of the parent scale, σαIs within the current sub-band (current layer) q at α q3 × 3 window centered prediction variance, noise varianceUsing a median estimator to estimate:
α thereinhIs the coefficient of the high-frequency sub-band in the first layer decomposition, and the mean is the operator for taking the median.
And when the iteration times reach a set value, outputting the reconstructed magnetic resonance image.
In order to verify the reconstruction performance of the magnetic resonance image fast reconstruction method (TPCTF-BS) based on the tensor product complex wavelet compact framework, a simulation experiment is performed by using three fully sampled magnetic resonance images, as shown in fig. 3(a), 3(b), and 3(c), which are respectively the magnetic resonance image data of the shoulder, the chest, and the brain, fig. 2 is a cartesian sampling matrix with an acceleration factor of 4 (equivalent to a sampling rate of 25%), which is used to undersample data on K space, the method of the present invention is compared with the currently popular magnetic resonance reconstruction algorithm, which mainly includes D L MRI, PBDWS, PANO, and GBRWT, and the reconstruction effects of different algorithms are measured by using peak signal-to-noise ratio (signal-to-noise ratio), and Structural Similarity (SSIM) as objective evaluation indexes to verify the reconstruction performance of the method, two objective evaluation indexes are defined as:
wherein x is the reconstructed image,for images under full sampling, W represents images x andh denotes the images x andheight of (d), muxAndare x andthe average value of (a) of (b),andare x andthe variance of (a) is determined,is the covariance of the two images, C1And C2Is a constant used to maintain stability.
Tables 1 and 2 record that under-sampling 25% K-space data under Cartesian under-sampling matrix, and different algorithms are used for reconstructing PSNR and SSIM of a magnetic resonance image, so that the TPCTF-BS has the best reconstruction quality.
Table 3 shows the average reconstruction time of the magnetic resonance image under different algorithms, and it can be found that the TPCTF-BS of the present invention has the fastest reconstruction speed.
Fig. 4(a) and fig. 4(b) are PSNR and SSIM of magnetic resonance images reconstructed from magnetic resonance images of a brain under different acceleration factors, respectively, and it can be seen that as the acceleration factor increases, i.e. the data of a sampling K space decreases, the reconstruction quality of all algorithms decreases, but under the same acceleration factor, the TPCTF-BS of the present invention still has the highest PSNR and SSIM values, indicating that the method has applicability under any reconstruction condition.
TABLE 1
Image \ algorithm | DLMRI | PBDWS | PANO | GBRWT | TPCTF-BS |
Shoulder image | 35.69 | 37.49 | 37.37 | 36.64 | 37.69 |
Chest image | 26.36 | 26.33 | 26.69 | 26.03 | 28.50 |
Brain image | 28.55 | 30.84 | 29.35 | 28.92 | 31.33 |
TABLE 2
Image \ algorithm | DLMRI | PBDWS | PANO | GBRWT | TPCTF-BS |
Shoulder image | 0.9291 | 0.9433 | 0.9436 | 0.9376 | 0.9484 |
Chest image | 0.7390 | 0.8285 | 0.7829 | 0.7998 | 0.8620 |
Brain image | 0.8488 | 0.9079 | 0.8704 | 0.8890 | 0.9145 |
TABLE 3
Algorithm | DLMRI | PBDWS | PANO | GBRWT | TPCTF-BS |
Reconstruction time(s) | 208.7 | 92.6 | 54.6 | 95.6 | 45.8 |
The above-mentioned embodiments are provided to further explain the objects, technical solutions and advantages of the present invention in detail, and it should be understood that the above-mentioned embodiments are only examples of the present invention and are not intended to limit the scope of the present invention. It should be understood that any modifications, equivalents, improvements and the like, which come within the spirit and principle of the invention, may occur to those skilled in the art and are intended to be included within the scope of the invention.
Claims (2)
1. A magnetic resonance image fast reconstruction method based on a tensor product complex small compact frame is characterized in that: the method comprises the following specific steps:
1) performing undersampling on the magnetic resonance K space data by utilizing a Cartesian sampling track to obtain K space undersampled data;
2) performing zero filling on the K space undersampled data, and then obtaining an initial magnetic resonance image through Fourier inverse transformation;
3) taking the initial magnetic resonance image as an input image, carrying out sparse decomposition by using a tensor product complex wavelet compact frame, and obtaining a decomposition coefficient expression after decomposition as follows:
wherein α is the decomposition coefficient, x is the input image, and T is the transposed symbol;a complex wavelet tightening framework is a tensor product;
4) and (3) performing minimum solution on the objective function of the magnetic resonance image by adopting a projection fast iterative soft threshold algorithm, wherein the optimization problem is expressed as follows:
the N-dimensional complex vector x and the M-dimensional complex vector y respectively represent a magnetic resonance image to be reconstructed and K-space undersampled data, the N × N complex matrix F represents a discrete Fourier transform matrix, the M × N real matrix U represents an undersampled matrix when the K-space data are undersampled, and the M × N real matrix U represents an undersampled matrix when the K-space data are undersampled<N; λ is the regularization parameter, | ·| non-woven phosphor2Is a two-norm operation, | | ·| non-conducting phosphor1Is a norm operation;
the iteration equation is as follows:
μ=xk+F*UT(y-UFxk)
wherein x iskRepresenting the reconstructed image at the k-th iteration; c. mu and sk+1Are all intermediate transition vectors, F*And UTRespectively representing inverse discrete Fourier transform and transposition of a sampling matrix;is a soft threshold function for shrinking the wavelet coefficients; t is tkStep length of the k iteration;
and when the iteration times reach a set value, outputting the reconstructed magnetic resonance image.
2. The method for rapidly reconstructing magnetic resonance image based on tensor product complex small compact framework as claimed in claim 1, wherein each layer α in decomposition coefficient αqThe corresponding λ is evaluated using a bivariate contraction function, as follows:
wherein, αqIs the current scaling factor, αpIs αqCoefficient of the parent scale, σαIs within the current sub-band (current layer) q at αq3 × 3 window centered prediction variance, noise varianceUsing a median estimator to estimate:
α thereinhIs the coefficient of the high-frequency sub-band in the first layer decomposition, and the mean is the operator for taking the median.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811124319.XA CN109188327B (en) | 2018-09-26 | 2018-09-26 | Magnetic resonance image fast reconstruction method based on tensor product complex small compact framework |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811124319.XA CN109188327B (en) | 2018-09-26 | 2018-09-26 | Magnetic resonance image fast reconstruction method based on tensor product complex small compact framework |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109188327A CN109188327A (en) | 2019-01-11 |
CN109188327B true CN109188327B (en) | 2020-07-14 |
Family
ID=64907409
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811124319.XA Active CN109188327B (en) | 2018-09-26 | 2018-09-26 | Magnetic resonance image fast reconstruction method based on tensor product complex small compact framework |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109188327B (en) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US11553139B2 (en) | 2020-09-29 | 2023-01-10 | International Business Machines Corporation | Video frame synthesis using tensor neural networks |
CN114742720B (en) * | 2022-03-16 | 2023-03-14 | 哈尔滨工业大学(深圳) | Tensor completion method, tensor completion device, equipment and storage medium |
Family Cites Families (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR100755446B1 (en) * | 2006-06-28 | 2007-09-04 | 중앙대학교 산학협력단 | Apparatus for deforming nurbs surface using modified fem and method thereof |
CN102148987B (en) * | 2011-04-11 | 2012-12-12 | 西安电子科技大学 | Compressed sensing image reconstructing method based on prior model and 10 norms |
KR101811774B1 (en) * | 2016-03-15 | 2017-12-22 | (주)넥셀 | Apparatus and method for processing graphics |
CN106485764B (en) * | 2016-11-02 | 2019-11-22 | 中国科学技术大学 | The quick exact reconstruction methods of MRI image |
CN108510564A (en) * | 2018-03-29 | 2018-09-07 | 东北大学 | A kind of MR image reconstruction method based on compressed sensing |
-
2018
- 2018-09-26 CN CN201811124319.XA patent/CN109188327B/en active Active
Also Published As
Publication number | Publication date |
---|---|
CN109188327A (en) | 2019-01-11 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Schlemper et al. | A deep cascade of convolutional neural networks for dynamic MR image reconstruction | |
WO2018099321A1 (en) | Generalized tree sparse-based weighted nuclear norm magnetic resonance imaging reconstruction method | |
CN107991636B (en) | Rapid magnetic resonance image reconstruction method based on adaptive structure low-rank matrix | |
CN109615675B (en) | Image reconstruction method for multi-channel magnetic resonance imaging | |
CN108010094B (en) | Magnetic resonance image reconstruction method and device | |
CN112819949B (en) | Magnetic resonance fingerprint image reconstruction method based on structured low-rank matrix | |
CN111754598B (en) | Local space neighborhood parallel magnetic resonance imaging reconstruction method based on transformation learning | |
Liu et al. | A deep framework assembling principled modules for CS-MRI: unrolling perspective, convergence behaviors, and practical modeling | |
CN112991483B (en) | Non-local low-rank constraint self-calibration parallel magnetic resonance imaging reconstruction method | |
CN109188327B (en) | Magnetic resonance image fast reconstruction method based on tensor product complex small compact framework | |
CN109920017B (en) | Parallel magnetic resonance imaging reconstruction method of joint total variation Lp pseudo norm based on self-consistency of feature vector | |
CN116725515B (en) | Magnetic resonance rapid imaging method | |
Shastri et al. | Denoising generalized expectation-consistent approximation for MR image recovery | |
Zong et al. | Fast reconstruction of highly undersampled MR images using one and two dimensional principal component analysis | |
Yashtini et al. | Partially parallel MR image reconstruction using sensitivity encoding | |
CN114723644A (en) | Compressed sensing magnetic resonance image reconstruction method and device, storage medium and electronic equipment | |
Yu et al. | Universal generative modeling in dual domains for dynamic MRI | |
Yaman et al. | Comparison of neural network architectures for physics-driven deep learning MRI reconstruction | |
Duan et al. | Adaptive transform learning and joint sparsity based PLORAKS parallel magnetic resonance image reconstruction | |
Liu et al. | Hybrid regularization for compressed sensing MRI: Exploiting shearlet transform and group-sparsity total variation | |
CN108346167B (en) | MRI image reconstruction method based on simultaneous sparse coding under orthogonal dictionary | |
An et al. | A structural oriented training method for gan based fast compressed sensing mri | |
Chen et al. | Magnetic resonance image reconstruction via L0-norm minimization | |
CN117078785B (en) | Rapid non-Cartesian magnetic resonance intelligent imaging method | |
Yang et al. | Median filter based compressed sensing model with application to MR image reconstruction |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
EE01 | Entry into force of recordation of patent licensing contract |
Application publication date: 20190111 Assignee: WENCHENG YIDONG TECHNOLOGY Co.,Ltd. Assignor: ZHEJIANG SCI-TECH University Contract record no.: X2022330000458 Denomination of invention: Fast Reconstruction Method of Magnetic Resonance Image Based on Tensor Product Complex Wavelet Compact Frame Granted publication date: 20200714 License type: Common License Record date: 20220825 |
|
EE01 | Entry into force of recordation of patent licensing contract |