CN108872899B - Reconstructing device, method, magnetic resonance system, equipment and the medium of magnetic resonance image - Google Patents
Reconstructing device, method, magnetic resonance system, equipment and the medium of magnetic resonance image Download PDFInfo
- Publication number
- CN108872899B CN108872899B CN201810689647.8A CN201810689647A CN108872899B CN 108872899 B CN108872899 B CN 108872899B CN 201810689647 A CN201810689647 A CN 201810689647A CN 108872899 B CN108872899 B CN 108872899B
- Authority
- CN
- China
- Prior art keywords
- data
- reconstruction
- magnetic resonance
- module
- image
- 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
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
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Signal Processing (AREA)
- High Energy & Nuclear Physics (AREA)
- Condensed Matter Physics & Semiconductors (AREA)
- General Physics & Mathematics (AREA)
- Magnetic Resonance Imaging Apparatus (AREA)
Abstract
The embodiment of the invention discloses a kind of reconstructing device of magnetic resonance image, method, magnetic resonance system, equipment and media, the reconstructing device includes CPU, and the multiple GPU connecting with CPU, CPU include data processing module and post processing of image module, GPU includes rebuilding module；Data processing module is divided into multiple groups reconstruction data group for obtaining magnetic resonance equipment reconstruction data collected, and by reconstruction data, and every group of son is rebuild data component and is assigned to corresponding GPU；It rebuilds module and rebuilds data group progress image reconstruction for antithetical phrase to generate the magnetic resonance image of corresponding body layer；Post processing of image module, which is used to obtain, rebuilds the magnetic resonance image that module generates, and post-processes to magnetic resonance image.It is longer the time required to solving the problems, such as to carry out MR image reconstruction using the MR image reconstruction means of the prior art, reach the technical effect that the MR image reconstruction time is greatly lowered.
Description
Technical field
The present embodiments relate to technical field of medical equipment more particularly to a kind of reconstructing devices of magnetic resonance image, side
Method, magnetic resonance system, equipment and medium.
Background technique
Magnetic resonance is widely used in medical diagnosis, and in order to rapidly obtain clearly magnetic resonance image, people are from data
The various aspects such as acquisition, image reconstruction improve magnetic resonance system.Due to the appearance of parallel imaging technique, so that magnetic resonance
Data acquisition time greatly shortens.By taking head magnetic resonance image as an example, using the magnetic resonance number on parallel imaging technique acquisition head
According to can shorten to 3 minutes or so.Although parallel imaging technique substantially reduces the time of MR data acquisition, using existing
MR image reconstruction means to brain MR data carry out image reconstruction take around 40 minutes.When relative to acquisition
Between, the too long MR image reconstruction time greatly reduces the working efficiency of magnetic resonance system, and user experience is poor, is unfavorable for
The clinical expansion of mr imaging technique.
Summary of the invention
The embodiment of the invention provides a kind of reconstructing device of magnetic resonance image, method, magnetic resonance system, equipment and Jie
Matter, it is longer the time required to carrying out MR image reconstruction using the MR image reconstruction means of the prior art with solution to ask
Topic.
In a first aspect, the embodiment of the invention provides a kind of reconstructing device of magnetic resonance image, including CPU, and with institute
Multiple GPU of CPU communication connection are stated, the CPU includes data processing module and post processing of image module, and the GPU includes weight
Model block；
The data processing module is for obtaining magnetic resonance equipment reconstruction data collected, and by the reconstruction data
It is divided into multiple groups and rebuilds data group, and every group of son is rebuild into data component and is assigned to corresponding GPU, wherein the sub- reconstruction
Data group includes the reconstruction data of the default body number of plies；
The module of rebuilding is used to receive the son reconstruction data group that the data processing module is distributed, and to the son
It rebuilds data group and carries out image reconstruction to generate the magnetic resonance image of corresponding body layer；
Described image post-processing module is used to obtain the magnetic resonance image that the reconstruction module generates, and to the magnetic resonance
Image is post-processed, and the post-processing, which includes at least, exports the magnetic resonance image to magnetic resonance image display end.
Second aspect, the embodiment of the invention also provides a kind of method for reconstructing of magnetic resonance image, comprising:
Magnetic resonance equipment reconstruction data collected are obtained by the data processing mould of CPU, and by the reconstruction data
It is divided into multiple groups and rebuilds data group, and every group of son is rebuild into data component and is assigned to corresponding GPU, wherein every group of son is rebuild
Data group includes the reconstruction data of the default body number of plies；
The son that the data processing module is distributed is received by the reconstruction module of GPU and rebuilds data group, and to described
Son rebuilds data group and carries out image reconstruction to generate the magnetic resonance image of corresponding body layer；
The magnetic resonance image that the reconstruction module generates is obtained by the post processing of image module of CPU, and total to the magnetic
Vibration image is post-processed, and the post-processing, which includes at least, exports the magnetic resonance image to magnetic resonance image display end.
The third aspect, the embodiment of the invention also provides a kind of magnetic resonance systems, comprising:
Magnetic resonance data acquisition end, the magnetic resonance data acquisition end are used to acquire the reconstruction data of target object；
Reconstructing device, the reconstructing device are used to obtain magnetic resonance equipment by the data processing module of CPU collected
Data are rebuild, and the reconstruction data are divided into multiple groups and rebuild data group, and every group of son is rebuild into data component and is matched
To corresponding GPU, wherein every group of son rebuilds the reconstruction data that data group includes the default body number of plies；And the reconstruction mould for passing through GPU
Block receives the son that the data processing module distributed and rebuilds data group, and to the sub- reconstruction data group carry out image reconstruction with
Generate the magnetic resonance image of corresponding body layer；And the magnetic resonance figure that the reconstruction module generates is obtained by post processing of image module
Picture, and the magnetic resonance image is exported to magnetic resonance image display end；
Magnetic resonance image display end, the magnetic resonance image display end is for receiving and showing described image post-processing module
The magnetic resonance image of output.
Fourth aspect, the embodiment of the invention also provides a kind of computer equipment, the computer equipment includes:
One or more processors；
Storage device, for storing one or more programs；
When one or more of programs are executed by one or more of processors, so that one or more of processing
Device realizes the method for reconstructing of the magnetic resonance image as described in second aspect.
5th aspect, it is described the embodiment of the invention also provides a kind of storage medium comprising computer executable instructions
Computer executable instructions as computer processor when being executed for executing the weight of the magnetic resonance image as described in second aspect
Construction method.
The technical solution of the reconstructing device of magnetic resonance image provided in an embodiment of the present invention, including CPU, and it is logical with CPU
Believe that multiple GPU of connection, CPU include data processing module and post processing of image module, GPU includes rebuilding module；Data processing
Module is divided into multiple groups reconstruction data group for obtaining magnetic resonance equipment reconstruction data collected, and by reconstruction data,
And every group of son is rebuild into data component and is assigned to corresponding GPU, wherein sub- reconstruction data group includes the reconstruction number of the default body number of plies
According to；It rebuilds module processing module is distributed for receiving data son reconstruction data group and antithetical phrase reconstruction data group and carries out figure
The magnetic resonance image of corresponding body layer is generated as rebuilding；Post processing of image module, which is used to obtain, rebuilds the magnetic resonance figure that module generates
Picture, and magnetic resonance image is post-processed, post-processing includes at least and exports magnetic resonance image to magnetic resonance image display end.
By CPU control multiple GPU simultaneously to rebuild data carry out concurrent operation, the computation capability that GPU can be made full use of superpower,
The technological means for carrying out MR image reconstruction by CPU compared with the existing technology, is greatly improved the reconstruction of magnetic resonance image
Speed, and then the reconstruction time of magnetic resonance image is significantly reduced, be conducive to the clinic for promoting customer experience and mr techniques
It promotes.
Detailed description of the invention
To describe the technical solutions in the embodiments of the present invention more clearly, make required in being described below to embodiment
Attached drawing does one and simply introduces, it should be apparent that, drawings in the following description are some embodiments of the invention, for this
For the those of ordinary skill of field, without creative efforts, it can also be obtained according to these attached drawings others
Attached drawing.
Fig. 1 is the block diagram representation of the reconstructing device for the magnetic resonance image that the embodiment of the present invention one provides；
Fig. 2 is the block diagram representation of the reconstructing device for the another magnetic resonance image that the embodiment of the present invention one provides；
Fig. 3 is the magnetic resonance pulse signal schematic representation that the embodiment of the present invention one provides；
Fig. 4 is the k-space sample track schematic diagram that the embodiment of the present invention one provides；
Fig. 5 A is the aliased image schematic diagram that the embodiment of the present invention one provides；
Fig. 5 B is the original image schematic diagram that the embodiment of the present invention one provides；
Fig. 6 is the Gadgetron platform schematic diagram that the embodiment of the present invention one provides；
Fig. 7 is the reconstruction process schematic diagram for the reconstruction module that the embodiment of the present invention one provides；
Fig. 8 is the flow chart of the method for reconstructing of magnetic resonance image provided by Embodiment 2 of the present invention；
Fig. 9 is the magnetic resonance system schematic diagram that the embodiment of the present invention three provides；
Figure 10 is the structural schematic diagram for the computer equipment that the embodiment of the present invention four provides.
Specific embodiment
To make the object, technical solutions and advantages of the present invention clearer, hereinafter with reference to attached in the embodiment of the present invention
Figure, clearly and completely describes technical solution of the present invention by embodiment, it is clear that described embodiment is the present invention one
Section Example, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art are not doing
Every other embodiment obtained under the premise of creative work out, shall fall within the protection scope of the present invention.
Embodiment one
Fig. 1 is the block diagram representation of the reconstructing device for the magnetic resonance image that the embodiment of the present invention one provides.The present embodiment
Technical solution is suitable for the case where reconstruction data collected to magnetic resonance equipment carry out image reconstruction.As shown in Figure 1, the reconstruction
Device includes CPU 11, and the multiple GPU 12 communicated to connect with CPU 11, CPU 11 include data processing module 111 and figure
As post-processing module 112, GPU 12 includes rebuilding module 121；Data processing module 111 is acquired for obtaining magnetic resonance equipment
Reconstruction data, and data will be rebuild be divided into multiple groups and rebuild data group, and every group of son is rebuild into data component and is assigned to
Corresponding GPU 12, wherein sub- reconstruction data group includes the reconstruction data of the default body number of plies；Module 121 is rebuild for receiving number
Data group is rebuild according to the son that processing module 111 is distributed and antithetical phrase rebuilds data group and carries out image reconstruction to generate corresponding body
The magnetic resonance image of layer；Post processing of image module 112, which is used to obtain, rebuilds the magnetic resonance image that module 121 generates, and total to magnetic
Vibration image is post-processed, and post-processing includes at least and exports the magnetic resonance image to magnetic resonance image display end.
Since many magnetic resonance equipments are equipped with the Gadgetron platform of open source, reconstructing device can pass through Gadgetron platform
The magnetic resonance equipment reconstruction data collected that are attached thereto are obtained, it, can be with when reconstructing device connects more magnetic resonance equipments
The reconstruction data for more magnetic resonance equipments being attached thereto are obtained simultaneously.
The reconstruction data of one target object generally include the reconstruction data of many body layers, at the data of the present embodiment CPU
Module is managed after the reconstruction data for getting a target object, or when current reconstruction data to be allocated are single target pair
When the reconstruction data of elephant, data will be rebuild according to body layer coding and be divided into multiple groups reconstruction data group, so that every height rebuilds number
It include the reconstruction data of the default body number of plies according to group.Preferably, the coding that every height rebuilds data group institute occlusion body layer is continuous
Coding.
For the ease of rebuilding the division of data, CPU further includes monitoring module 113, for monitoring each GPU state, so that
Data processing module will rebuild data group according to each GPU state and be divided into multiple sub- reconstruction data groups, and antithetical phrase rebuilds data
Group is allocated.GPU state includes busy condition, idle state, malfunction and lock state.Illustratively, if rebuilding
Device includes two GPU, and when the two GPU are currently at idle state, CPU will then rebuild data and be divided into two groups of sons reconstructions
Then two groups of sons are rebuild data and are respectively sent to different GPU by data, by two GPU respectively to the received son weight of respective institute
It builds data group while carrying out image reconstruction.When reconstructing device includes three GPU, and only it is in idle condition there are two GPU, it is another
It is a to be in the lock state or when malfunction, it data will be rebuild is divided into two groups of sons and rebuild data, then rebuild this two groups of sons
Data send current two free time GPU respectively.
In order to further increase image reconstruction speed, GPU further includes memory application module 122, and GPU receives sub- reconstruction number
According to memory application module 122 is passed through after group to the CPU application global memory right of possession corporeal right, occupy so as to rebuild module in global memory
Sub- reconstruction data group received to institute carries out image reconstruction during power, to generate the magnetic resonance image of corresponding body layer, compared to existing
There is the every magnetic resonance image for having rebuild an individual layer of GPU to discharge a memory, global memory's right of possession corporeal right that this implementation uses can
Make GPU to received all body layers magnetic resonance image ability Free up Memory, by reduce memory application and release number,
Reduce GPU image reconstruction times.
In order to improve the reconstruction number in the convenience that reconstructing device uses, especially reconstructing device there are multiple target objects
According to when, need to solve the problems, such as rebuild data processing sequence, to meet various needs.Certainly, there are multiple targets in reconstructing device
The reconstruction data of object, usually because the reconstructing device connects more magnetic resonance equipments simultaneously, so existing simultaneously from multiple magnetic
Resonant device obtains the case where rebuilding data.In order to solve the problems, such as to rebuild data processing sequence, CPU reads each reconstruction data
Identification information, such as acquisition time, magnetic resonance equipment coding and the precedence information for rebuilding data of the reconstruction data.It is wherein excellent
First grade can be such as urgent by operator according to doctor's advice input.When CPU determines currently own according to the identification information for rebuilding data
When the priority of the reconstruction data of target object is identical, then the reconstruction data of target object are determined according to reconstruction data obtaining time
Reconstruction time；When acquisition time is identical, then the reconstruction data of target object are determined according to the preset order of magnetic resonance equipment
Reconstruction time.When CPU is determined there are when the reconstruction data of high priority target object, then according to the identification information for rebuilding data
Preferentially the reconstruction data are handled.
In order to improve each target object reconstruction data reconstruction speed, the present embodiment is preferably by target object
It rebuilds data and is divided into multiple sub- reconstruction data groups, and every height reconstruction data component is assigned in corresponding GPU.But it can manage
Solution, when the reconstruction data that current reconstructions data to be allocated are multiple target objects, and the data volume of each reconstruction data
When smaller, it can also be divided according to target object to data are rebuild, the reconstruction data distribution of as each target object
One GPU.
In order to further increase image reconstruction speed, the GPU of the present embodiment further includes Bandwidth control module 123, passes through band
The data transfer bandwidth and data transfer locations of per thread is arranged in wide control module 123, and to avoid adjacent thread, there are numbers
According to access span.Bandwidth control module 123 can control thread transmission bandwidth by design kernel function, avoid adjacent thread
Access data have span.Kernel function can be the form of x=a [b × i], wherein b%32 range be 0~32,0 bandwidth most
Height, 32 bandwidth are minimum, differ about 32 times.For example, realizing the addition of two matrixes, C=A+B with GPU；Design a kernel function
Add, i=threadId, c (i)=a (i)+b (i)；Here i is an ID of thread, number of threads is arranged by CPU, CPU will
Number of threads order is sent to GPU, and GPU prepares the thread of respective numbers according to number of threads order, and per thread is carried out this
A kernel function.The present embodiment identifies per thread by threadId.When matrix there are 10,10 lines need to be just set
Journey, i is actually 0 to 9, in this way, matrix just completes sum operation.The i=threadId of the present embodiment, c (i)=a (i)+b
(i), GPU can continuously obtain memory at this time.If i=threadId × 32, c (i)=a (i)+b (i) is then deposited for span
It takes.It is unable to satisfy access to merge, efficiency is poor.
In order to improve the reconstruction data acquisition speed of magnetic resonance image, the reconstruction data of the present embodiment are excellent by magnetic resonance equipment
Gated 3D Wave-CAIPI imaging technique (Controlled Aliasing in Parallel Imaging, abbreviation
CAIPI, the controllable aliasing parallel imaging technique of wave) it obtains.Wherein, 3D Wave-CAIPI imaging technique by modification phase and
Offset on frequency coding increases the changes in distribution of interlayer coil sensitivity with quick obtaining and rebuilds data.Specifically: in y ladder
Direction is spent (referring to the G in Fig. 3P) and z gradient direction (referring to the G in Fig. 3S) on plus a sinusoidal wave pulse, the two pulse
Phase by pi/2, as shown in Figure 3, wherein RFFor radio frequency direction, GRFor readout direction, ADC is sampled data.K spatial sampling rail
Mark is spiral shape, as shown in Figure 4.Corresponding relationship between the aliased image (Fig. 5 B) and original image (Fig. 5 A) of Wave-CAIPI can
It is indicated by following formula:
Wherein, Fx is the Fourier transform (direction x is readout direction here) in the direction x, Fx-1Be the direction x Fourier it is inverse
Transformation, PSF are space divergence estimations.Wave (x, y, z) corresponds to aliased image, the corresponding original image of m [x, y, z].
Image reconstruction is the process that original image is obtained according to aliased image.In order to improve the quality of magnetic resonance image, this reality
The reconstruction module for applying example rebuilds the received sub- reconstruction data group of institute by conjugate gradient least square method to generate magnetic and be total to
Shake image.It is minimum based on existing conjugate gradient although 3D Wave-CAIPI imaging technique can rapidly obtain reconstruction data
Square law is longer the time required to carrying out image reconstruction to reconstruction data acquired in the parallel imaging technique, and time-consuming is usually several
Ten minutes, it is difficult to carry out the popularization of clinical application.Using the conjugate gradient least square method (Conjugate of the present embodiment
Gradients Least Squares, abbreviation CGLS) combine more GPU concurrent operations to carry out image reconstruction, image can be greatly reduced
Reconstruction time.
As shown in fig. 6, Gadgetron platform includes customized data structure and some algorithm for reconstructing, and rebuilds and calculate
Reconstruction module corresponding to method is segmented into multiple separate units, the reading interface magnetic resonance imaging number of Gadgetron platform
According to, linear data structure is converted by n by non-sequence and ties up matrix, it is (single by one or more reconstruction by reconstruction module 121
Member 1211 forms) image reconstruction is carried out to it, then the image data after reconstruction is total to by the magnetic for writing the incoming magnetic resonance equipment of interface
Shake image display end.Tool box 13 is the set for rebuilding 121 callable unit of module, can be an independent algorithm for reconstructing,
It is also possible to junior unit such as a Fourier transform and inverse Fourier transform of algorithm for reconstructing.In actual use, according to reality
It needs to add in configuration file.
Rebuild module 12 reconstruction procedures include:
1, it pre-processes
Rebuilding data is the measurement data that directly exports of magnetic resonance equipment, rebuilds module by its preprocessing module to dividing
The son matched is rebuild the measurement data that data group is included and is pre-processed, and rebuilds data group, and updated son weight to update son
The data format for building data group is [readout phase slice channel], and wherein readout is to read data, and phase is
Phase, slice are the body number of plies, and channel is parallel acquisition coil number.
2, it reads when the data and the corresponding yz plane gradient data of the body layer of precursor layer and the body layer are corresponding
The coil number of coil sensitive data.
3, it is based on conjugate gradient least-squares algorithm, image weight is carried out to each column data when precursor layer according to formula 1
It builds, until the image reconstruction to all column datas when precursor layer is completed, to obtain the magnetic resonance image when precursor layer.
It is understood that pre-treatment step can also be completed by the data processing module of CPU, data processing module is complete
At after pre-treatment step again by rebuild data be divided into several sub- reconstruction data groups, at this point, the reconstruction module of GPU can directly into
Row above-mentioned steps 2 and step 3.
Wherein, it is based on conjugate gradient least square method, image weight is carried out to each column data when precursor layer according to formula 1
The detailed process built are as follows:
1, parameter vector b (vector that b is made of the column data when precursor layer, referring to the input of the CGLS of Fig. 7) is inputted.
2, it initializes: r0=b, s0=A (b, T), p1=s0,γ0=| | s0||2,x0=0.Wherein, T is that formula (1) is reverse
Marker bit is operated, i.e., is encoded to obtain original image process according to WAIPI by aliased image, is specifically shown in the operation of the second reconstruction unit.
Customized CUDA kernel function device_copy completes the mutual assignment between the vector identified by sparse matrix, improves just
Beginningization speed.
3, i=1,2,3 ... n, circulation execute following step a to step h.
a)、qi=A (pi, NT), NT is the positive process markup of corresponding formula (1), encodes to obtain according to WAIPI by original image
Aliased image process, referring specifically to the realization process of algorithm corresponding to the first reconstruction unit.
b)、αi=γi-1||qi||2。
c)、xi=xi-1+αipi。
d)、ri=ri-1-αiqi。
e)、si=A (ri, T), which is formula (1) inverse process, original image process is obtained by decoding by aliased image,
Referring specifically to the realization process of algorithm corresponding to the second reconstruction unit.
d)、γi-1=| | si||2。
g)、βi=γi/γi-1。
h)、pi+1=si+βipi。
4, judge whether to reach termination condition i > n or αipi< n0, if conditions are not met, 3 are thened follow the steps, if full
Foot, thens follow the steps 5.
5, x is exportedi, xiIt is the column of required original image.
Wherein, algorithm corresponding to the first reconstruction unit is Wave=A (w, rcv, PSFyz), and w is original digital image data, and rcv is
Coil sensitivities image, PSFyz are space divergence matrixes.This process is the positive process of formula 1, obtains aliasing by original image m
The process of image wave.It implements process are as follows:
1.1), I=w.*rcv.
1.2), I=fftshift (I), % Fourier variation, and using fs/2 as center right and left mutually changing after Fourier's variation
1.3), I=FxI。
1.4), I=fftshift (I).
1.5), I=PSFyz.*I.
1.6), I=fftshift (I)
1.7)、
1.8), I=fftshift (I)
1.9), wave=∑Rz RyWave, %wave merge in the direction YZ.
Second reconstruction unit is the inverse process of formula (1), obtains original image, the realization of corresponding algorithm by aliased image
Journey are as follows:
2.1), I=fftshift (inx) % to input matrix, that is, in matrix the direction x carry out fftshift, wherein the side x
To being matrix column direction.
2.2), I=Fx(I)。
2.3), I=fftshift (Ix) % to matrix I the direction x carry out fftshift.
2.4), because aliased image is in the sampling of the direction y and z, reconstruction will restore I=repmat (I, [1, Rz, Rz, 1]) %.
2.5), I=I.*conj (PSFyz).%conj seeks conjugation.
2.6), I=fftshift (Ix)。
2.7)、
2.8), I=fftshift (Ix)。
2.9), I=cutx (I) .*conj (rcv).%cutx is in the direction x central part interception image height.
2.10), m=∑coil Icoil.The fusion of % multi-coil.
Rebuild the step 3 and the first reconstruction unit, the vectorial addition in the second reconstruction unit, multiplication and modulus meter of module
It calculates and is completed using the library function in the library cuSparse.
For the first reconstruction unit and the second reconstruction unit, following CUDA kernel function is defined for parallel computation, is divided
Not are as follows:
G_fftshift gpu Parallel Implementation fftshift.
Two matrix dot product of g_dotMulti gpu Parallel Implementation and conjugation dot product.
The multiple coil coils of g_sum gpu Parallel Implementation merge.
The 1.10 of g_sumYZ gpu the first reconstruction unit of Parallel Implementation), i.e. wave merges in the direction YZ.
G_fft calls the library cufft, calculates Fourier transform, calls cufft_plan_many to realize multiple one-dimensional
Fft/ifft operation.
Since the library cufft does not support sparse format to store.The present embodiment creation storage transfer function MemTranser is carried out
Mutual conversion between sparse storage and normal storage format is used in the front-rear position of step a) and step e).
The whole flow process of conjugate gradient least square method, in addition to single argument is calculated with CPU (univariate data, between equipment
Passing time can ignore), vector operation GPU memory complete.So only starting to carry out with iteration ends in iteration
The data between CPU and GPU are needed to transmit when data input and output.Other calculating do not need data between CPU and GPU and pass
It passs, advantageously reduces image reconstruction times.
The comparing result of scheme and the prior art described in the present embodiment is as follows:
It is obtained by 7T magnetic resonance equipment (Siemens Company of Erlangen, Germany) and rebuilds data, the magnetic resonance equipment packet
32 channel head coils are included, data acquisition is carried out with 3D Wave-CAIPI imaging technique, to adopt multiple equal for the direction z and the direction y deficient
Be 3, when sampling, accelerator coefficient is related with owing to adopt multiple, when y owe to adopt digit be 3, z owe to adopt multiple be 3 when, then it is practical to accelerate
Coefficient is up to 9 times.Rebuild experiment porch are as follows: MATLABR2017a (MathWorks Inc., Natick, Massachusetts,
USA), CPU is Intel (R) Core (TM) i7-4770 double-core (3.4G Hz, 8M Cache), and memory is 8G；Operating system is
Microsoft Windows 7,64-bit；GPU is NVIDIA GTX770 (1536CUDA Cores, 1.2G Hz).For one
The head magnetic resonance image of adult, data visual field size FOV=240 × 240 × 192mm3, the resolution ratio in three directions are all
1mm completes image reconstruction based on CPU and takes around 40 minutes.And the present embodiment is based on CPU+2GPU and calculates completion image reconstruction
About 4 minutes time-consuming, time efficiency improves 10 times.
The technical solution of the reconstructing device of magnetic resonance image provided in an embodiment of the present invention, including CPU, and it is logical with CPU
Believe that multiple GPU of connection, CPU include data processing module and post processing of image module, GPU includes rebuilding module；Data processing
Module is divided into multiple groups reconstruction data group for obtaining magnetic resonance equipment reconstruction data collected, and by reconstruction data,
And every group of son is rebuild into data component and is assigned to corresponding GPU, wherein sub- reconstruction data group includes the reconstruction number of the default body number of plies
According to；It rebuilds module processing module is distributed for receiving data son reconstruction data group and antithetical phrase reconstruction data group and carries out figure
The magnetic resonance image of corresponding body layer is generated as rebuilding；Post processing of image module, which is used to obtain, rebuilds the magnetic resonance figure that module generates
Picture, and magnetic resonance image is post-processed, post-processing includes at least and exports magnetic resonance image to magnetic resonance image display end.
By CPU control multiple GPU simultaneously to rebuild data carry out concurrent operation, the computation capability that GPU can be made full use of superpower,
The technological means for carrying out MR image reconstruction by CPU compared with the existing technology, is greatly improved the reconstruction of magnetic resonance image
Speed, and then the reconstruction time of magnetic resonance image is significantly reduced, be conducive to the clinic for promoting customer experience and mr techniques
It promotes.
Embodiment two
Fig. 8 is the flow chart of the method for reconstructing of magnetic resonance image provided by Embodiment 2 of the present invention.The technology of the present embodiment
Scheme is suitable for the case where carrying out MR image reconstruction by multiple GPU.This method can be by provided in an embodiment of the present invention
The reconstructing device of magnetic resonance image executes.Software or hardware realization can be selected in this method, and configures and use in the processor, has
Body includes the following steps:
S101, magnetic resonance equipment reconstruction data collected are obtained by the data processing module of CPU, and number will be rebuild
Data group is rebuild according to multiple groups is divided into, and every group of son is rebuild into data component and is assigned to corresponding GPU, wherein every group of sub- weight
Build the reconstruction data that data group includes the default body number of plies.
Since many magnetic resonance equipments are equipped with the Gadgetron platform of open source, reconstructing device can pass through Gadgetron platform
The magnetic resonance equipment reconstruction data collected that are attached thereto are obtained, it, can be with when reconstructing device connects more magnetic resonance equipments
The reconstruction data for more magnetic resonance equipments being attached thereto are obtained simultaneously.
The reconstruction data of one target object generally include the reconstruction data of many body layers, at the data of the present embodiment CPU
Module is managed after the reconstruction data for getting a target object, current reconstruction data to be allocated are single target pair in other words
When the reconstruction data of elephant, data will be rebuild according to body layer coding and be divided into multiple groups reconstruction data group, so that every height rebuilds number
It include the reconstruction data of the default body number of plies according to group.Preferably, the coding that every height rebuilds data group institute occlusion body layer is continuous
Coding.
For the ease of rebuilding the division of data, CPU further includes monitoring module, for monitoring each GPU state, so that data
Processing module will rebuild data group according to each GPU state and be divided into multiple sub- reconstruction data groups, and antithetical phrase rebuild data group into
Row distribution.GPU state includes busy, idle, failure and locking.Illustratively, if reconstructing device includes two GPU, when this
When two GPU are currently at idle state, CPU will then rebuild data and be divided into two groups of sons reconstruction data, then by two groups of sub- weights
It builds data and is respectively sent to different GPU, image is carried out simultaneously to the respective received sub- reconstruction data group of institute respectively by two GPU
It rebuilds.When reconstructing device includes three GPU, and only there are two being in idle condition, when another is in locking or malfunction,
Data will be rebuild and be divided into two groups of sons reconstruction data, this two groups of sons are then rebuild into data and send current two free time GPU respectively.
In order to improve the reconstruction number in the convenience that reconstructing device uses, especially reconstructing device there are multiple target objects
According to when, need to solve the problems, such as rebuild data processing sequence, to meet various needs.Certainly, there are multiple targets in reconstructing device
The reconstruction data of object, usually because the reconstructing device connects more magnetic resonance equipments simultaneously, so existing simultaneously from multiple magnetic
Resonant device obtains the case where rebuilding data.In order to solve the problems, such as to rebuild data processing sequence, CPU reads each reconstruction data
Identification information, such as acquisition time, magnetic resonance equipment coding and the precedence information for rebuilding data of the reconstruction data.It is wherein excellent
First grade can be such as urgent by operator according to doctor's advice input.When CPU determines currently own according to the identification information for rebuilding data
When the priority of the reconstruction data of target object is identical, then the reconstruction data of target object are determined according to reconstruction data obtaining time
Reconstruction time；When acquisition time is identical, then the reconstruction data of target object are determined according to the preset order of magnetic resonance equipment
Reconstruction time.When CPU is determined there are when the reconstruction data of high priority target object, then according to the identification information for rebuilding data
Preferentially the reconstruction data are handled.
In order to improve each target object reconstruction data reconstruction speed, the present embodiment is preferably by target object
It rebuilds data and is divided into multiple sub- reconstruction data groups, and every height reconstruction data component is assigned in corresponding GPU.But it can manage
Solution, when the reconstruction data that current reconstructions data to be allocated are multiple target objects, and the data volume of each reconstruction data
When smaller, it can also be divided according to target object to data are rebuild, the reconstruction data distribution of as each target object
One GPU.
In order to further increase image reconstruction speed, the GPU of the present embodiment further includes Bandwidth control module 123, passes through band
The data transfer bandwidth and data transfer locations of per thread is arranged in wide control module 123, and avoiding adjacent thread, there are data
Access span.Bandwidth control module 123 can control thread transmission bandwidth by design kernel function, and adjacent thread is avoided to deposit
Access is according to there is span, and kernel function can be the form of x=a [b × i], and b%32 range is 0~32,0 bandwidth highest, 32 bandwidth
It is minimum, differ about 32 times.For example, realizing the addition of two matrixes, C=A+B with GPU；Design a kernel function add:i=
threadId；C (i)=a (i)+b (i)；Here i is an ID of thread, and the quantity of thread is arranged by CPU, and CPU is by thread
The order of quantity is sent to GPU, and GPU prepares the thread of respective numbers according to the order of number of threads, and per thread is carried out this
A kernel function.Per thread is identified by threadId.When matrix there are 10,10 threads need to be just set, and i is actually
It is 1 to 10, in this way, matrix just completes sum operation.The i=threadId of the present embodiment, c (i)=a (i)+b (i), at this time
GPU can continuously obtain memory.If i=threadId × 32, c (i)=a (i)+b (i) is then span access.It can not expire
Foot access merges, and efficiency is poor.
S102, the son that the data processing module distributed is received by the reconstruction module of GPU rebuild data group and right
Son rebuilds data group and carries out image reconstruction to generate the magnetic resonance image of corresponding body layer.
Before or after GPU receives sub- reconstruction data group, by memory application module to the CPU application global memory right of possession corporeal right, with
Make to rebuild module sub- reconstruction data group progress image reconstruction received to institute during global memory's right of possession corporeal right, to generate corresponding body
The magnetic resonance image of layer, the magnetic resonance image for having rebuild an individual layer every compared to existing GPU discharge a memory, this implementation
Global memory's right of possession corporeal right of use can make GPU to received all body layers magnetic resonance image ability Free up Memory, by subtracting
The application of few memory and release number, reduce GPU image reconstruction times.
S103, the magnetic resonance image that the reconstruction module generates is obtained by the post processing of image module of CPU, and to described
Magnetic resonance image is post-processed, and the post-processing, which includes at least to export the magnetic resonance image to magnetic resonance image, to be shown
End.
Image reconstruction is then the process that original image is obtained according to aliased image.In order to improve the quality of magnetic resonance image, this
The reconstruction module of embodiment rebuilds to generate magnetic the received sub- reconstruction data group of institute by conjugate gradient least square method
Resonance image.Although 3D Wave-CAIPI imaging technique can rapidly obtain reconstruction data, it is existing based on this parallel at
The image reconstruction times of reconstruction data as acquired in technology are but very long, and time-consuming is difficult to be faced usually in dozens of minutes
The popularization of bed application.The present embodiment combines aforementioned more GPU concurrent operations to carry out image reconstruction using conjugate gradient least square method,
The specific reconstruction procedures of its algorithm are referring to described in previous embodiment, and it will not be described here for the present embodiment.
Post processing of image module, which is used to obtain, rebuilds the magnetic resonance image that module generates, and carries out figure to magnetic resonance image
As post-processing, for example magnetic resonance image is exported to magnetic resonance image display end and is shown, and magnetic resonance image is passed through
Network transmission is to doctor workstation.
The technical solution of MR image reconstruction method provided in an embodiment of the present invention, passes through the data processing module of CPU
Magnetic resonance equipment reconstruction data collected are obtained, and data will be rebuild and be divided into multiple groups reconstruction data group, and will be every
Group rebuilds data component and is assigned to corresponding GPU, wherein every group of son rebuilds the reconstruction data that data group includes the default body number of plies；
The son that data processing module is distributed, which is received, by the reconstruction module of GPU rebuilds data group and antithetical phrase reconstruction data group progress
Image reconstruction is to generate the magnetic resonance image of corresponding body layer；It is obtained by the post processing of image module of CPU and rebuilds what module generated
Magnetic resonance image, and magnetic resonance image is post-processed, post-processing includes at least and exports magnetic resonance image to magnetic resonance figure
As display end.Multiple GPU are controlled by CPU and carry out concurrent operation to rebuilding data simultaneously, can make full use of superpower parallel of GPU
Computing capability is carried out the technological means of MR image reconstruction by CPU compared with the existing technology, is greatly improved magnetic resonance figure
The reconstruction speed of picture, and then the reconstruction time of magnetic resonance image is significantly reduced, be conducive to promote customer experience and magnetic resonance skill
The clinical expansion of art.
Embodiment three
Fig. 9 is the magnetic resonance system schematic diagram that the embodiment of the present invention three provides.As shown in figure 9, the magnetic resonance system includes:
Magnetic resonance data acquisition end 2, the magnetic resonance data acquisition end 2 are used to acquire the reconstruction data of target object；
Reconstructing device 1, the reconstructing device are used to obtain magnetic resonance equipment by the data processing module of CPU collected
Data are rebuild, and the reconstruction data are divided into multiple groups and rebuild data group, and every group of son is rebuild into data component and is matched
To corresponding GPU, wherein every group of son rebuilds the reconstruction data that data group includes the default body number of plies；And the reconstruction mould for passing through GPU
Block receives the son that the data processing module distributed and rebuilds data group, and to the sub- reconstruction data group carry out image reconstruction with
Generate the magnetic resonance image of corresponding body layer；And the magnetic resonance figure that the reconstruction module generates is obtained by post processing of image module
Picture, and the magnetic resonance image is exported to magnetic resonance image display end；
Magnetic resonance image display end 3, the magnetic resonance image display end 3 is for receiving and showing that described image post-processes mould
The magnetic resonance image of block output.
The technical solution of magnetic resonance system provided in an embodiment of the present invention, including magnetic resonance data acquisition end, magnetic resonance number
It is used to acquire the reconstruction data of target object according to collection terminal；Reconstructing device, reconstructing device are used for the data processing module by CPU
Magnetic resonance equipment reconstruction data collected are obtained, and data will be rebuild and be divided into multiple groups reconstruction data group, and will be every
Group rebuilds data component and is assigned to corresponding GPU, wherein every group of son rebuilds the reconstruction data that data group includes the default body number of plies；
And the son that data processing module is distributed is received by the reconstruction module of GPU and rebuilds data group, and a data group is rebuild to institute's
Image reconstruction is carried out to generate the magnetic resonance image of corresponding body layer；And it is obtained by post processing of image module and rebuilds module generation
Magnetic resonance image, and magnetic resonance image is exported to magnetic resonance image display end；Magnetic resonance image display end is for receiving simultaneously
Show the magnetic resonance image of post processing of image module output.Multiple GPU are controlled by CPU simultaneously to transport reconstruction data parallel
It calculates, the computation capability that GPU can be made full use of superpower, MR image reconstruction is carried out by CPU compared with the existing technology
Technological means is greatly improved the reconstruction speed of magnetic resonance image, and then significantly reduces the reconstruction time of magnetic resonance image, favorably
In the clinical expansion for promoting customer experience and mr techniques.
Example IV
Figure 10 is the structural schematic diagram for the computer equipment that the embodiment of the present invention four provides, as shown in Figure 10, the computer
Equipment includes processor 401, memory 402, input unit 403 and output device 404；Processor 401 in computer equipment
Quantity can be one or more, in Figure 10 by taking a processor 401 as an example；Processor 401, storage in computer equipment
Device 402, input unit 403 and output device 404 can be connected by bus or other modes, to be connected by bus in Figure 10
It is connected in example.
Memory 402 is used as a kind of computer readable storage medium, can be used for storing software program, journey can be performed in computer
Sequence and module, if the corresponding program instruction/module of the method for reconstructing of the magnetic resonance image in the embodiment of the present invention is (for example, number
According to processing module 111, post processing of image module 112 and rebuild module 121).Processor 401 is stored in memory by operation
Software program, instruction and module in 402 are realized thereby executing the various function application and data processing of equipment
The method for reconstructing for the magnetic resonance image stated.
Memory 402 can mainly include storing program area and storage data area, wherein storing program area can store operation system
Application program needed for system, at least one function；Storage data area, which can be stored, uses created data etc. according to terminal.This
Outside, memory 402 may include high-speed random access memory, can also include nonvolatile memory, for example, at least one
Disk memory, flush memory device or other non-volatile solid state memory parts.In some instances, memory 402 can be into one
Step includes the memory remotely located relative to processor 401, these remote memories can pass through network connection to equipment.On
The example for stating network includes but is not limited to internet, intranet, local area network, mobile radio communication and combinations thereof.
Input unit 403 can be used for receiving the number or character information of input, and generate with the user setting of equipment with
And the related key signals input of function control.
Output device 404 may include that display screen etc. shows equipment, for example, the display screen of user terminal.
Embodiment five
The embodiment of the present invention five also provides a kind of storage medium comprising computer executable instructions, and the computer can be held
Row is instructed when being executed by computer processor for executing a kind of method for reconstructing of magnetic resonance image, this method comprises:
Magnetic resonance equipment reconstruction data collected are obtained by the data processing mould of CPU, and by the reconstruction data
It is divided into multiple groups and rebuilds data group, and every group of son is rebuild into data component and is assigned to corresponding GPU, wherein every group of son is rebuild
Data group includes the reconstruction data of the default body number of plies；
The son that the data processing module is distributed is received by the reconstruction module of GPU and rebuilds data group, and to described
Son rebuilds data group and carries out image reconstruction to generate the magnetic resonance image of corresponding body layer；
The magnetic resonance image that the reconstruction module generates is obtained by the post processing of image module of CPU, and total to the magnetic
Vibration image is post-processed, and the post-processing, which includes at least, exports the magnetic resonance image to magnetic resonance image display end.
Certainly, a kind of storage medium comprising computer executable instructions, computer provided by the embodiment of the present invention
The method operation that executable instruction is not limited to the described above, can also be performed magnetic resonance figure provided by any embodiment of the invention
Relevant operation in the method for reconstructing of picture.
By the description above with respect to embodiment, it is apparent to those skilled in the art that, the present invention
It can be realized by software and required common hardware, naturally it is also possible to which by hardware realization, but in many cases, the former is more
Good embodiment.Based on this understanding, technical solution of the present invention substantially in other words contributes to the prior art
Part can be embodied in the form of software products, which can store in computer readable storage medium
In, floppy disk, read-only memory (Read-Only Memory, abbreviation ROM), random access memory (Random such as computer
Access Memory, abbreviation RAM), flash memory (FLASH), hard disk or CD etc., including some instructions are used so that a calculating
It is total that machine equipment (can be personal computer, server or the network equipment etc.) executes magnetic described in each embodiment of the present invention
The method for reconstructing of vibration image.
It is worth noting that, in the embodiment of the reconstructing device of above-mentioned magnetic resonance image, included each unit and mould
Block is only divided according to the functional logic, but is not limited to the above division, and is as long as corresponding functions can be realized
It can；In addition, the specific name of each functional unit is also only for convenience of distinguishing each other, the protection model being not intended to restrict the invention
It encloses.
Note that the above is only a better embodiment of the present invention and the applied technical principle.It will be appreciated by those skilled in the art that
The invention is not limited to the specific embodiments described herein, be able to carry out for a person skilled in the art it is various it is apparent variation,
It readjusts and substitutes without departing from protection scope of the present invention.Therefore, although being carried out by above embodiments to the present invention
It is described in further detail, but the present invention is not limited to the above embodiments only, without departing from the inventive concept, also
It may include more other equivalent embodiments, and the scope of the invention is determined by the scope of the appended claims.
Claims (12)
1. a kind of reconstructing device of magnetic resonance image, which is characterized in that including CPU, and it is multiple with CPU communication connection
GPU, the CPU include data processing module and post processing of image module, and the GPU includes rebuilding module；
The data processing module is divided for obtaining magnetic resonance equipment reconstruction data collected, and by the reconstruction data
Data group is rebuild at multiple groups, and every group of son is rebuild into data component and is assigned to corresponding GPU, wherein the sub- reconstruction data
Group includes the reconstruction data of the default body number of plies；
The module of rebuilding is used to receive the son reconstruction data group that the data processing module is distributed, and to the sub- reconstruction
Data group carries out image reconstruction to generate the magnetic resonance image of corresponding body layer；
Described image post-processing module is used to obtain the magnetic resonance image that the reconstruction module generates, and to the magnetic resonance image
It is post-processed, the post-processing, which includes at least, exports the magnetic resonance image to magnetic resonance image display end.
2. reconstructing device according to claim 1, which is characterized in that when current reconstruction data to be allocated come from single mesh
When marking object, the reconstruction data are divided into the son of multiple reconstruction data for including the default body number of plies by the data processing module
Data group is rebuild, and every height reconstruction data component is assigned to corresponding GPU；
When current reconstruction data to be allocated come from multiple target objects, the data processing module is by the reconstruction data root
The reconstruction data are divided into multiple sub- reconstruction data groups according to target object, and every height reconstruction data component is assigned to correspondence
GPU, the sub- reconstruction data group includes all reconstruction data of single target object or the part body of single target object
The reconstruction data of layer.
3. reconstructing device according to claim 1, which is characterized in that the CPU further includes monitoring module, the monitoring mould
Block is for monitoring each GPU state, so that the data processing module divides the reconstruction data group according to each GPU state
For multiple sub- reconstruction data groups, and the sub- reconstruction data group is allocated；Or
The monitoring module is used to obtain the priority of the reconstruction data of each target object, and each GPU state of monitoring, with
Make the data processing module that the reconstruction data group are divided into multiple sub- reconstructions according to the priority and each GPU state
Data group, and the sub- reconstruction data group is allocated.
4. reconstructing device according to claim 1, which is characterized in that the GPU further includes memory application module, described interior
Application module is deposited to be used for the CPU application global memory right of possession corporeal right, so that the reconstruction module is occupied in the global memory
Image reconstruction is carried out to the sub- reconstruction data group during power, to generate the magnetic resonance image of corresponding body layer.
5. reconstructing device according to claim 1, which is characterized in that the GPU further includes Bandwidth control module, the band
Wide control module is used to be arranged the data transfer bandwidth and data transfer locations of per thread, and to avoid adjacent thread, there are numbers
According to access span.
6. reconstructing device according to claim 1, which is characterized in that the reconstruction data are passed through by the magnetic resonance equipment
3D Wave-CAIPI parallel imaging technique obtains；
The reconstruction module rebuilds to generate magnetic the received sub- reconstruction data group of institute by conjugate gradient least square method
Resonance image.
7. reconstructing device according to claim 6, which is characterized in that the reconstruction data of every individual layer include z direction gradient
Data, y direction gradient data, coil sensitivities data and k-space data；
The reconstruction module is specifically used for:
The yz plane gradient data of the body layer are sought according to the z direction gradient data of every individual layer and y direction gradient data；
Based on conjugate gradient least square method, according to the yz plane gradient data of every individual layer, the coil sensitivities data with
And k-space data seeks the magnetic resonance image of every individual layer.
8. reconstructing device according to claim 7, which is characterized in that the reconstructing device is based on CUDA platform, described heavy
Storage and calculating that block carries out data by the sparse matrix storage mode of CUDA are modeled, and by the CUDA platform
Kernel function at least complete initialization vector assignment in conjugate gradient least square method, fftshift, matrix dot multiplied by and conjugation
Dot product；
The reconstruction module, which also passes through, calls the library cufft to calculate Fourier transformation.
9. a kind of method for reconstructing of magnetic resonance image characterized by comprising
Magnetic resonance equipment reconstruction data collected are obtained by the data processing mould of CPU, and the reconstruction data are divided
Data group is rebuild at multiple groups, and every group of son is rebuild into data component and is assigned to corresponding GPU, wherein every group of son rebuilds data
Group includes the reconstruction data of the default body number of plies；
The son that the data processing module is distributed is received by the reconstruction module of GPU and rebuilds data group, and to the son weight
It builds data group and carries out image reconstruction to generate the magnetic resonance image of corresponding body layer；
The magnetic resonance image that the reconstruction module generates is obtained by the post processing of image module of CPU, and to the magnetic resonance figure
As being post-processed, the post-processing, which includes at least, exports the magnetic resonance image to magnetic resonance image display end.
10. a kind of magnetic resonance system characterized by comprising
Magnetic resonance data acquisition end, the magnetic resonance data acquisition end are used to acquire the reconstruction data of target object；
Reconstructing device, the reconstructing device are used to obtain magnetic resonance equipment reconstruction collected by the data processing module of CPU
Data, and the reconstruction data are divided into multiple groups and rebuild data group, and every group of son is rebuild into data component and is assigned to pair
The GPU answered, wherein every group of son rebuilds the reconstruction data that data group includes the default body number of plies；And it is connect by the reconstruction module of GPU
It receives the son that the data processing module is distributed and rebuilds data group, and image reconstruction is carried out to generate to the sub- reconstruction data group
The magnetic resonance image of corresponding body layer；And the magnetic resonance image that the reconstruction module generates is obtained by post processing of image module,
And the magnetic resonance image is exported to magnetic resonance image display end；
Magnetic resonance image display end, the magnetic resonance image display end is for receiving and showing that described image post-processing module exports
The magnetic resonance image.
11. a kind of computer equipment, which is characterized in that the computer equipment includes:
One or more processors；
Storage device, for storing one or more programs；
When one or more of programs are executed by one or more of processors, so that one or more of processors are real
The now method for reconstructing of magnetic resonance image as claimed in claim 9.
12. a kind of storage medium comprising computer executable instructions, which is characterized in that the computer executable instructions by
For executing the method for reconstructing of magnetic resonance image as claimed in claim 9 when computer processor executes.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810689647.8A CN108872899B (en) | 2018-06-28 | 2018-06-28 | Reconstructing device, method, magnetic resonance system, equipment and the medium of magnetic resonance image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810689647.8A CN108872899B (en) | 2018-06-28 | 2018-06-28 | Reconstructing device, method, magnetic resonance system, equipment and the medium of magnetic resonance image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108872899A CN108872899A (en) | 2018-11-23 |
| CN108872899B true CN108872899B (en) | 2019-10-18 |
Family
ID=64296345
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810689647.8A Active CN108872899B (en) | 2018-06-28 | 2018-06-28 | Reconstructing device, method, magnetic resonance system, equipment and the medium of magnetic resonance image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108872899B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111243048B (en) * | 2018-11-28 | 2024-01-30 | 西门子(深圳)磁共振有限公司 | Image reconstruction method, device and system in magnetic resonance imaging and storage medium |
| CN110833429B (en) * | 2019-12-03 | 2024-03-26 | 上海联影医疗科技股份有限公司 | Computed tomography imaging method, device, equipment and storage medium |
| CN112967167B (en) * | 2019-12-12 | 2023-04-28 | 中国科学院深圳先进技术研究院 | GPU-based image quick reconstruction method, computer-readable medium and computing device |
| WO2023108484A1 (en) * | 2021-12-15 | 2023-06-22 | 中国科学院深圳先进技术研究院 | Magnetic resonance imaging method based on wave gradient coding field and deep learning model |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7916144B2 (en) * | 2005-07-13 | 2011-03-29 | Siemens Medical Solutions Usa, Inc. | High speed image reconstruction for k-space trajectory data using graphic processing unit (GPU) |
| CN102008306B (en) * | 2010-12-22 | 2012-11-14 | 中国科学院深圳先进技术研究院 | Magnetic resonance parallel imaging method |
| CN104181486B (en) * | 2013-07-05 | 2015-07-22 | 上海联影医疗科技有限公司 | Magnetic resonance image reconstruction method and apparatus |
| CN103700123B (en) * | 2013-12-19 | 2018-05-08 | 北京唯迈医疗设备有限公司 | GPU based on CUDA frameworks accelerates x-ray image method for reconstructing and device |
| JP6809870B2 (en) * | 2016-11-04 | 2021-01-06 | 株式会社日立製作所 | Magnetic resonance imaging device and signal processing method |
-
2018
- 2018-06-28 CN CN201810689647.8A patent/CN108872899B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN108872899A (en) | 2018-11-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108872899B (en) | Reconstructing device, method, magnetic resonance system, equipment and the medium of magnetic resonance image | |
| CN104781685B (en) | For the image reconstruction using incoherent sampling and the dynamic magnetic resonance imaging of redundancy HAAR wavelets | |
| CN109325985A (en) | MR image reconstruction method, apparatus and computer readable storage medium | |
| CN100578515C (en) | Method and device for optimizing measurement data collection and image reproduction | |
| CN106680748B (en) | Magnetic resonance image acquisition methods and magnetic resonance system | |
| WO2023045482A1 (en) | Multi-layer magnetic resonance imaging method and apparatus based on long-distance attention model reconstruction | |
| CN104833931A (en) | Method and device for controlling magnetic resonance imaging system | |
| CN103767705A (en) | Magnetic resonance imaging system and magnetic resonance imaging method | |
| CN108776318A (en) | The more contrast image reconstructions of magnetic resonance | |
| Du et al. | Accelerated super-resolution MR image reconstruction via a 3D densely connected deep convolutional neural network | |
| CN102164533B (en) | diagnostic imaging system and method | |
| CN104635189A (en) | Method and system for reconstructing magnetic resonance images with multiple contrast ratios | |
| CN109959887A (en) | A kind of three-dimensional MRI method for reconstructing, device, application and readable medium | |
| CN107249454A (en) | For MRI or CT scan-geometries planing method | |
| CN110246200A (en) | Mr cardiac film imaging method, device and magnetic resonance scanner | |
| CN107121654B (en) | Mining method and device are owed in a kind of magnetic resonance | |
| CN103529414B (en) | MR imaging method and device, the method for reconstructing in K space and device | |
| EP3825711A1 (en) | Hybrid compressed sensing image reconstruction | |
| CN109975343A (en) | MR imaging method, system and storage medium | |
| US6694165B2 (en) | Method for ultra-fast MR fluoroscopy | |
| CN103514042B (en) | A kind of Dual-adjustment merge-sorting tuning method and device | |
| JP5595872B2 (en) | Magnetic resonance imaging apparatus and program | |
| US7642776B2 (en) | Method to determine an acquisition sequence in an imaging method for generation of 2D slice images | |
| US20160282435A1 (en) | Magnetic resonance imaging apparatus | |
| JP4721857B2 (en) | Magnetic resonance imaging apparatus and image reconstruction method using the same |
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 |