CN108010100A - A kind of single sweep magnetic resonance based on residual error network quantifies T2Imaging reconstruction method - Google Patents
A kind of single sweep magnetic resonance based on residual error network quantifies T2Imaging reconstruction method Download PDFInfo
- Publication number
- CN108010100A CN108010100A CN201711287890.9A CN201711287890A CN108010100A CN 108010100 A CN108010100 A CN 108010100A CN 201711287890 A CN201711287890 A CN 201711287890A CN 108010100 A CN108010100 A CN 108010100A
- Authority
- CN
- China
- Prior art keywords
- mrow
- msub
- msup
- residual error
- magnetic resonance
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T11/00—2D [Two Dimensional] image generation
- G06T11/003—Reconstruction from projections, e.g. tomography
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N24/00—Investigating or analyzing materials by the use of nuclear magnetic resonance, electron paramagnetic resonance or other spin effects
- G01N24/08—Investigating or analyzing materials by the use of nuclear magnetic resonance, electron paramagnetic resonance or other spin effects by using nuclear magnetic resonance
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10088—Magnetic resonance imaging [MRI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20048—Transform domain processing
- G06T2207/20056—Discrete and fast Fourier transform, [DFT, FFT]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2211/00—Image generation
- G06T2211/40—Computed tomography
- G06T2211/416—Exact reconstruction
Landscapes
- Physics & Mathematics (AREA)
- High Energy & Nuclear Physics (AREA)
- General Physics & Mathematics (AREA)
- Biochemistry (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Magnetic Resonance Imaging Apparatus (AREA)
Abstract
A kind of single sweep magnetic resonance based on residual error network quantifies T2Imaging reconstruction method, is related to MR imaging method.Using 4 low-angle excitation pulses with equal deflection angle, there is one evolution time after each excitation pulse so that the lateral relaxation time T of each echo-signal2It is different.Add the displacement gradient of a frequency coding peacekeeping phase code dimension after each excitation pulse so that the signal that different excitation pulses produces is different in the position of k-space.Multiple echo-signals with different lateral relaxation times are just obtained in once sampling.Then sampled signal is obtained into quantitative T by normalizing, zeroizing and be input to after Fast Fourier Transform (FFT) in trained residual error network to rebuild2Image.The training data of residual error network derives from analogue data.First random generation template, then simulated experimental environments sample to obtain the input picture of network, template passes through trained the mapping relations between input picture and output image as label.
Description
Technical field
The present invention relates to MR imaging method, is quantified more particularly, to a kind of single sweep magnetic resonance based on residual error network
T2It is imaged (T2Mapping) method for reconstructing.
Background technology
Magnetic resonance parameters are imaged (for example, T1Imaging, T2Imaging and T2 *Imaging) it can easily be used for providing for characterizing
The quantitative information of particular organization's characteristic[1].Quantitative imaging has many prominent features, and most obvious one advantage is can to eliminate solely
The influence of tissue property is stood on, such as to operator's dependence, sweep parameter difference, magnetic field space change and image scaling
Deng[2].Quantitative T2Imaging has extensive clinical practice, is related to the early diagnosis of nerve degenerative diseases[3], liver iron doping survey
Amount[4], myocardial infarction assessment[5], mark cell quantifies[6], diagnosing multiple sclerosis and epilepsy[7]Deng.Quantitative T2It is imaged on and faces
More and more concerns are obtained in bed magnetic resonance imaging (magnetic resonance imaging, MRI).Although transverse relaxation
Time T2It can be obtained by being fitted the spin echo MRI data of multiple and different echo times (echo time, TE), but from
The longer sweep time of cycle ripple MRI to obtain quantitative T in clinic2Image is challenging[8].In addition, longer sweeps
Retouching the time also causes the quantitative T based on spin echo MR I2Image is easily influenced be subject to motion artifacts[9].Base has been proposed
T is quantified in the single sweep of more echo echo-planar imagings (echo-planar imaging, EPI)2Imaging method, in function magnetic
Big advantage is shown in resonance image-forming[10,11].However, compared with single echo samples, gained image is easily subject to more
Signal attenuation and ghost artifacts influence[12,13].Other fast quantifications T2Imaging method, such as gradient spin echo
(gradient spin echo, GraSE) scheme[14]Still using more echo spin echoes (multi-echo spin-echo,
MSE it is) tactful, it usually needs a few minutes[15,16]。
Itd is proposed by overlapping echo free (overlapping-echo detachment, OLED) planar imaging
Method[17]The quantitative T of high quality can be obtained in single sweep operation2Image, its spatial and temporal resolution are schemed with traditional single sweep EPI
As quite.In addition, OLED planar imagings are also shown to motion artifacts and non-ideal B1The stronger resistance of field.However, first
Preceding sequence only includes two excitation pulses, therefore measurable T2Scope is extremely limited, this causes with larger T2Scope
Effect is poor in region (such as cerebrospinal fluid (cerebrospinal fluid, CSF)).Therefore, we are changed with four driving pulses
Kind OLED sequences, to reach the T of bigger2Measurement range.Believe however, being difficult to the overlapping echo of separation four by conventional method
Number.
Deep learning, forms more abstract high-rise expression attribute classification or feature, to find by combining low-level feature
The technology of the distributed nature of data reveals volatile popularization with the availability table of GPU powerful in recent years[18].Especially
Ground, convolutional neural networks (convolution neural network, CNN) have caused image super-resolution (super-
Resolution, SR) rebuild a series of breakthroughs[19].Different network models, such as convolutional neural networks[20], residual error network
(residual network, ResNet)[21], depth recursive convolution network (deeply-recursive convolutional
Network, DRCN)[22], efficient sub-pix convolutional neural networks (efficientsub-pixel convolutional
Neural network, ESPCN)[23]With generation confrontation network (generative adversarialnetwork, GAN)[24]
It is applied to obtain high-definition picture.Convolutional neural networks are becoming increasingly popular in the medical imaging analysis of various problems.
Bibliography：
[1]B.Zhao,F.Lam,and Z.P.Liang,"Model-based MR parameter mapping with
sparsity constraints:parameter estimation and performance bounds,"IEEE
Transactions on Medical Imaging,vol.33,no.9,pp.1832-1844,2014.
[2]H.L.Margaret Cheng,N.Stikov,N.R.Ghugre,and G.A.Wright,"Practical
medical applications of quantitative MR relaxometry,"Journal of Magnetic
Resonance Imaging,vol.36,no.4,pp.805-824,2012.
[3]S.Baudrexel,L.Nürnberger,U.Rüb,C.Seifried,J.C.Klein,T.Deller,
H.Steinmetz,R.Deichmann,and R.Hilker,"Quantitative mapping of T1and T2*
discloses nigral and brainstem pathology in early Parkinson's disease,"
Neuroimage,vol.51,no.2,pp.512-520,2010.
[4]J.S.Hankins,M.B.McCarville,R.B.Loeffler,M.P.Smeltzer,M.Onciu,
F.A.Hoffer,C.-S.Li,W.C.Wang,R.E.Ware,and C.M.Hillenbrand,"R2*magnetic
resonance imaging of the liver in patients with iron overload,"Blood,vol.113,
no.20,pp.4853-4855,2009.
[5]D.R.Messroghli,K.Walters,S.Plein,P.Sparrow,M.G.Friedrich,
J.P.Ridgway,and M.U.Sivananthan,"Myocardial T1mapping:application to patients
with acute and chronic myocardial infarction,"Magnetic Resonance in Medicine,
vol.58,no.1,pp.34-40,2007.
[6]W.Liu,H.Dahnke,J.Rahmer,E.K.Jordan,and J.A.Frank,"Ultrashort T2*
relaxometry for quantitation of highly concentrated superparamagnetic iron
oxide(SPIO)nanoparticle labeled cells,"Magnetic Resonance in Medicine,vol.61,
no.4,pp.761-766,2009.
[7]T.N.Townsend,N.Bernasconi,G.B.Pike,and A.Bernasconi,"Quantitative
analysis of temporal lobe white matter T2relaxation time in temporal lobe
epilepsy,"Neuroimage,vol.23,no.1,pp.318-324,2004.
[8]D.Ma,V.Gulani,N.Seiberlich,K.Liu,J.L.Sunshine,J.L.Duerk,and
M.A.Griswold,"Magnetic resonance fingerprinting,"Nature,vol.495,no.7440,
pp.187-193,2013.
[9]D.Le Bihan,C.Poupon,A.Amadon,and F.Lethimonnier,"Artifacts and
pitfalls in diffusion MRI,"Journal of Magnetic Resonance Imaging,vol.24,no.3,
pp.478-488,2006.
[10]S.Posse,S.Wiese,D.Gembris,K.Mathiak,C.Kessler,M.-L.Grosse-Ruyken,
B.Elghahwagi,T.Richards,S.R.Dager,and V.G.Kiselev,"Enhancement of BOLD-
contrast sensitivity by single-shot multi-echo functional MR imaging,"
Magnetic Resonance in Medicine,vol.42,no.1,pp.87-97,1999.
[11]G.Hagberg,I.Indovina,J.Sanes,and S.Posse,"Real‐time
quantification of T2*changes using multiecho planar imaging and numerical
methods,"Magnetic Resonance in Medicine,vol.48,no.5,pp.877-882,2002.
[12]J.W.Evans,P.Kundu,S.G.Horovitz,and P.A.Bandettini,"Separating
slow BOLD from non-BOLD baseline drifts using multi-echo fMRI,"Neuroimage,
vol.105,pp.189-197,2015.
[13]A.D.Halai,S.R.Welbourne,K.Embleton,and L.M.Parkes,"A comparison
of dual gradient‐echo and spin‐echo fMRI of the inferior temporal lobe,"Human
Brain Mapping,vol.35,no.8,pp.4118-4128,2014.
[14]A.M.Sprinkart,J.A.Luetkens,F.J.Doerner,J.Gieseke,
B.Schnackenburg,G.Schmitz,D.Thomas,R.Homsi,and W.Block,"Gradient Spin Echo
(GraSE)imaging for fast myocardial T2 mapping,"Journal of Cardiovascular
Magnetic Resonance,vol.17,no.1,pp.12-20,2015.
[15]T.J.Sumpf,A.Petrovic,M.Uecker,F.Knoll,and J.Frahm,"Fast T2
mapping with improved accuracy using undersampled spin-echo MRI and model-
based reconstructions with a generating function,"IEEE Transactions on
Medical Imaging,vol.33,no.12,pp.2213-2222,2014.
[16]N.Ben‐Eliezer,D.K.Sodickson,and K.T.Block,"Rapid and accurate T2
mapping from multi–spin‐echo data using Bloch‐simulation‐based
reconstruction,"Magnetic Resonance in Medicine,vol.73,no.2,pp.809-817,2015.
[17]C.B.Cai,Y.Q.Zeng,Y.C.Zhuang,S.H.Cai,L.Chen,X.H.Ding,L.J.Bao,
J.H.Zhong,and Z.Chen,"Single-shot T2 mapping through Overlapping-echo
Detachment(OLED)Planar Imaging,"IEEE Transactions on Biomedical Engineering,
vol.64,no.10,pp.2450-2461,2017.
[18]S.S.Wang,Z.H.Su,L.Ying,X.Peng,S.Zhu,F.Liang,D.G.Feng,and
D.Liang,"Accelerating magnetic resonance imaging via deep learning,"in 2016
IEEE 13th International Symposium on Biomedical Imaging(ISBI),pp.514-517,
2016.
[19]C.F.Baumgartner,O.Oktay,and D.Rueckert,"Fully Convolutional
Networks in Medical Imaging:Applications to Image Enhancement and
Recognition,"in Deep Learning and Convolutional Neural Networks for Medical
Image Computing,pp.159-179,2017.
[20]C.Dong,C.C.Loy,K.M.He,and X.O.Tang,"Learning a deep convolutional
network for image super-resolution,"in 13th European Conference on Computer
Vision(ECCV),pp.184-199,2014.
[21]K.M.He,X.Y.Zhang,S.Q.Ren,and J.Sun,"Deep residual learning for
image recognition,"in Proceedings of the IEEE Conference on Computer Vision
and Pattern Recognition(CVPR),pp.770-778,2016.
[22]J.Kim,J.Kwon Lee,and K.Mu Lee,"Deeply-recursive convolutional
network for image super-resolution,"in Proceedings of the IEEE Conference on
Computer Vision and Pattern Recognition(CVPR),pp.1637-1645,2016.
[23]W.Z.Shi,J.Caballero,F.Huszár,J.Totz,A.P.Aitken,R.Bishop,
D.Rueckert,and Z.H.Wang,"Real-time single image and video super-resolution
using an efficient sub-pixel convolutional neural network,"in Proceedings of
the IEEE Conference on Computer Vision and Pattern Recognition(CVPR),pp.1874-
1883,2016.
[24]C.Ledig,L.Theis,F.Huszár,J.Caballero,A.Cunningham,A.Acosta,
A.Aitken,A.Tejani,J.Totz,and Z.Wang,"Photo-realistic single image super-
resolution using a generative adversarial network,"arXiv preprint arXiv:
1609.04802,2016.
The content of the invention
It is an object of the invention to provide four echo-signals are obtained in single sweep operation, then come using deep learning method
Reconstruction obtains reliable quantification T2A kind of single sweep magnetic resonance based on residual error network of image quantifies T2Imaging reconstruction method.
The present invention comprises the following steps：
1) laboratory sample is got out, sample to be tested is placed on experimental bed and fixed, by the experimental bed equipped with sample
It is sent into the test chamber of magnetic resonance imager；
2) imager operation software is opened on the operation console of magnetic resonance imager, laboratory sample is carried out first interested
Zone location, is then tuned, shimming, frequency correction and capability correction；
3) compiled OLED imaging sequences are imported, according to specific experimental conditions, each of pulse train is set
Parameter；
4) the OLED imaging sequences that step 3) sets parameter are performed, start to sample, after the completion of data sampling, are performed next
Step；
5) signal that step 4) obtains is normalized, zeroized and Fast Fourier Transform (FFT), the signal of k-space is changed
To image area；
6) random mask is generated according to the feature of sample to be tested, carrying out analog sampling using simulation softward to random mask obtains
To k-space signal, then signal is normalized, is zeroized and Fast Fourier Transform (FFT) obtains training data；
7) residual error network model is built using TensorFlow deep learnings frame and Python, sets trained correlation
Parameter；
8) the data training network obtained using step 6), until data training network restrains and reaches stable, is instructed
The network model perfected, then rebuilds the experimental data that step 5) obtains using trained network model, obtaining can
The quantitative T leaned on2Image.
In step 3), the structure of the OLED imaging sequences is followed successively by：Flip angle be α excitation pulse, displacement gradient
G1, flip angle be α excitation pulse, displacement gradient G2, flip angle be α excitation pulse, displacement gradient G3, flip angle be α swash
Send out pulse, displacement gradient G4, flip angle be the reunion pulse of β, sampled echo chain；
4 angle of deflection excitation pulse combination frequency coding dimensions (RO) and 4 displacement gradient Gs of phase code dimension (PE)1、
G2、G3、G44 echo-signals are made to produce offset at the center of k-space, 4 angle excitation pulses are all tieed up (SS) with level selection
Layer choosing gradient is combined carry out layer choosing；
The sampled echo chain is made of the gradient chain for being respectively acting on frequency coding peacekeeping phase code dimension, frequency coding
The gradient chain of dimension is made of a series of positive negative gradients, and the gradient chain of phase code dimension is terraced by a series of phase code of area equations
Degree is formed；
Before sampled echo chain, frequency coding peacekeeping phase code dimension direction applies biasing gradient, frequency coding respectively
The biasing gradient area of dimension is the half of first frequency encoding gradient area, and direction is on the contrary, the biasing gradient of phase code dimension
Area ties up the half of all phase encoding gradient areas for phase code, and direction is opposite.
It is described that the signal that step 4) obtains is analyzed in step 5), and to the core under the effect of OLED imaging sequences
Spin evolution carries out theory deduction, and Observable signal strength is proportional to following expression after the reunion impulse action that flip angle is β：
Wherein,γ is nuclear spin gyromagnet
Than,It is nuclear spin in the position of the real space；By above formula understand sampling period have 4 by out of phase modulate echo-signals, 4
Different phase modulations is θ respectively4、(θ3+θ4)、(θ2+θ3+θ4) and (θ1+θ2+θ3+θ4)。
In step 6), the random mask is random raw using computer batch according to the feature distribution of experiment sample
Into, while ensure that the complexity of random mask is higher, all features of experiment sample can be included；During analog sampling,
In view of true experiment environmental change, destabilizing factor is added, improves robustness of the network model to undesirable experimental situation；Institute
Stating destabilizing factor includes excitation pulse angular deviation, displacement gradient deviation and noise etc..
In step 7), the residual error network model mainly includes：The agent structure of network and relevant training parameter,
The object function of network model is：
Wherein, N is the quantity with a collection of training image, and f () is training network, and W and b are network parameters, and x is input figure
Picture, y are the corresponding templates of input picture, ychangeIt is the matrix being set to the value of a certain threshold value of small Yu in y after the threshold value, ymask
It is the image edge information tried to achieve to y using Canny operators.
In step 8), the trained network model is trained due to the use of random mask, and generalization is stronger, energy
Suitable for the reconstruction of several samples.
The quantity of driving pulse is increased to 4 by the present invention from 2, to expand T in OLED sequences2Measurement range, and improve
The precision in CSF regions.OLED image is rebuild from 4 overlapping echo-signals using depth learning technology, so as to bypass tradition side
The difficulty that method is run into.
The present invention has one section of evolution using the excitation pulse at 4 identical small angle deflection angles after each excitation pulse
Time so that 4 echoes have different lateral relaxation times, displacement gradient are added after each excitation pulse so that 4 echoes
Signal is in signal space (k-space) off-centring, then the method progress using this k-space view data using deep learning
Reconstruction obtains a width and quantifies T2Image.This method can obtain a width in single sweep operation and reliably quantify T2Image, ensures at the same time
Larger T2Measurement range.
Brief description of the drawings
Fig. 1 is the single sweep OLED imaging sequence figures that this patent uses.
Fig. 2 is to rebuild quantitative T2The residual error network model that image uses.In fig. 2, residual error network model includes 3 subnets
Network：Input network, residual error learning network and reconstruction network.Network is inputted using the real and imaginary parts of OLED image as input, and
Represented with one group of characteristic pattern.Residual error learning network is to solve the chief component of image reconstruction task.Rebuild network and be used for handle
Multiple characteristic patterns are redeveloped into a width and quantify T2Image.
Fig. 3 illustrates the quantitative T of 3 layers of human brain2Imaging reconstruction result；In figure 3, first row represents the amplitude figure of SE, its
In 16 yellow circles represent the area-of-interest (region of interest, ROI) chosen, compiled respectively with numeral 1~16
Number；Secondary series represents the input of the original amplitude figure, that is, networks of OLED；3rd row represent the quantitative T that OLED is rebuild2Image, that is, net
The output of network；4th row represent the quantitative T that SE is fitted2It is imaged reference chart.
Fig. 4 shows the statistical result of 16 ROI.In Fig. 4, a OLED, b SE.
Embodiment
The present invention will be further described for the following drawings and specific embodiment.
The present invention provides the single sweep magnetic resonance based on residual error network to quantify T2Imaging reconstruction method, specific implementation process
In each step it is as follows：
(1) laboratory sample is got out, sample to be tested is placed on experimental bed and fixed, by the experimental bed equipped with sample
It is sent into the test chamber of magnetic resonance imager.
(2) imager operation software is opened on the operation console of magnetic resonance imager, laboratory sample is carried out feeling emerging first
Interesting zone location, is then tuned, shimming, frequency correction and capability correction.
(3) compiled OLED imaging sequences are imported, according to specific experimental conditions, each of pulse train is set
Parameter.
The structure of the OLED imaging sequences is followed successively by：Flip angle be α excitation pulse, displacement gradient G1, flip angle α
Excitation pulse, displacement gradient G2, flip angle be α excitation pulse, displacement gradient G3, flip angle be α excitation pulse, displacement
Gradient G4, flip angle be the reunion pulse of β, sampled echo chain.
Four low-angles (α) excitation pulse combination frequency coding dimension (RO) and four displacements of phase code dimension (PE)
Gradient G1、G2、G3、G4Make four echo-signals k-space center produce offset, four low-angle excitation pulses all with aspect
The layer choosing gradient of selection dimension (SS) is combined carry out layer choosing.
The sampled echo chain is made of the gradient chain for being respectively acting on frequency coding peacekeeping phase code dimension, frequency coding
The gradient chain of dimension is made of a series of positive negative gradients, and the gradient chain of phase code dimension is terraced by a series of phase code of area equations
Degree is formed.
Before sampled echo chain, frequency coding peacekeeping phase code dimension direction applies biasing gradient, frequency coding respectively
The biasing gradient area of dimension is the half of first frequency encoding gradient area, and direction is on the contrary, the biasing gradient of phase code dimension
Area ties up the half of all phase encoding gradient areas for phase code, and direction is opposite.
(4) the OLED imaging sequences that step (3) sets parameter are performed, start to sample, after the completion of data sampling, under execution
One step.
(5) signal obtained to step (4) is normalized, zeroizing turns the signal of k-space with Fast Fourier Transform (FFT)
Change to image area.
Obtained signal is analyzed, and develops to the nuclear spin under the effect of OLED imaging sequences and carries out theory deduction.
Observable signal strength is proportional to following expression after the reunion impulse action that flip angle is β：
γ is nuclear spin gyromagnetic ratio,It is nuclear spin in the position of the real space；Understand that sampling period there are four echo-signals modulated by out of phase by above formula, four are not
Same phase modulation is θ respectively4、(θ3+θ4)、(θ2+θ3+θ4) and (θ1+θ2+θ3+θ4)。
(6) random template is generated according to the feature of laboratory sample, uses template the SPROM softwares of our group developments
Carry out analog sampling obtain k-space signal, then signal is normalized, zeroize and Fast Fourier Transform (FFT) obtain train number
According to.
The random mask is to be generated according to the feature distribution of experiment sample using computer batch is random, while ensures mould
The complexity of plate is higher, can include all features of experiment sample.During analog sampling, it is contemplated that true experimental situation becomes
Change, add some destabilizing factors, such as excitation pulse angular deviation, displacement gradient deviation and noise etc. improve network mould
Robustness of the type to undesirable experimental situation.
(7) residual error network model is built using TensorFlow deep learnings frame and Python, sets trained phase
Related parameter.
The residual error network model mainly includes：The agent structure of network and relevant training parameter.Network model
Object function is：
Wherein N is the quantity with a collection of training image, and f () is training network, and W and b are network parameters, and x is input figure
Picture, y are the corresponding templates of input picture, ychangeIt is that the matrix after 0.06, y are set to the value for being less than 0.06 in ymaskIt is that y is made
The image edge information tried to achieve with Canny operators.
(8) the data training network obtained using step (6), until network convergence and reaching and stably obtaining trained net
Network model, the experimental data then obtained to step (5) are rebuild, and obtain reliably quantifying T2Image.
The trained network model is trained due to the use of random mask, and generalization is stronger, can be suitably used for a variety of
The reconstruction of sample.
Specific embodiment is given below：
T is quantified with the single sweep magnetic resonance based on residual error network2Imaging reconstruction method has carried out human brain experiment, for verifying
The feasibility of the present invention.Experiment carries out under human body nuclear magnetic resonance 3T imagers.On magnetic resonance imager operation console, beat
Open in imager operation software accordingly, area-of-interest positioning carried out to imaging object first, be then tuned, shimming,
Power and frequency correction.The validity of image is obtained in order to evaluate this method, SE imaging experiments work has been carried out under identical environment
For contrast.Compiled OLED imaging sequences (as shown in Figure 1) are then introduced into, according to specific experimental conditions, arteries and veins is set
The parameters of sequence are rushed, the experiment parameter setting of the present embodiment is as follows：Visual field FOV is 22cm × 22cm, and 15 ° excite arteries and veins
The firing time of punching is 3ms, and the firing times of 180 ° of reunion pulses is 3ms, first echo time 57.8ms, second echo
Time 83.9ms, the 3rd echo time 135.9ms, the 4th echo time 162.1ms, frequency coding peacekeeping phase code dimension
Sampling number be 192.After above experiment parameter is set, directly start to sample.In Fig. 1, α overturns for excitation pulse
Angle；β is reunion pulse flip angle；G1、G2、G3And G4For four displacement gradients of frequency coding peacekeeping phase code dimension；Gcr
The destruction gradient of 3 dimensions is tieed up for frequency coding dimension, phase code peacekeeping level selection；TE1、TE2、TE3And TE4Respectively 4
The time span of echo；Ecoh1, Ecoh2, Ecoh3 and Ecoh4 are respectively the center of 4 echo-signals.
After the completion of data sampling, data are rebuild according to step (5)~(8).Rebuild residual error network model used
As shown in Fig. 2, the residual error network model includes input network, residual error learning network and rebuilds network three parts.Inputting network will
The real and imaginary parts of OLED image represent that residual error learning network is to solve image reconstruction to appoint as input with one group of characteristic pattern
The chief component of business, rebuilds network and quantifies T for multiple characteristic patterns to be redeveloped into a width2Image.Reconstruct the quantitative T come2
Image is as shown in figure 3, (a) is the internal anatomy for the human brain adopted with spin-echo sequence (spin echo, SE) sequence, therein 16
A region marked with yellow circle is area-of-interest (region of interest, ROI), is compiled respectively with numeral 1~16
Number.(b) it is original amplitude figure with OLED sequence acquisitions, inclined stripe is caused by four echo-signals overlap in figure.
(c) it is quantitative T that OLED data reconstructions obtain2Image.(d) be with SE fit come quantitative T2Image.To 16 senses in figure
The T in interest region2Value carries out quantitative statistics, the i.e. T to each area-of-interest2Value is averaging, and obtains Fig. 4.Can be with from Fig. 3
Find out the quantitative T obtained with OLED sequences2Image is consistent with SE on the whole, and the statistical result in Fig. 4 also indicates that 16 ROI regions
T2Value has preferable accuracy.
The present invention has one section using 4 low-angle excitation pulses with equal deflection angle after each excitation pulse
Develop the time so that the lateral relaxation time T of each echo-signal2It is different.Meanwhile add one after each excitation pulse
The displacement gradient of frequency coding peacekeeping phase code dimension so that the signal that different excitation pulse produces in the position of k-space not
Equally.In this way, just obtain multiple echo-signals with different lateral relaxation times in once sampling.Then sampling is believed
Number by normalization, zeroize and Fast Fourier Transform (FFT) after be input to rebuild in trained residual error network and quantified
T2Image.The training data of residual error network derives from analogue data.First random to generate template, then simulated experimental environments sample
To the input picture of network, template obtains the mapping relations between input picture and output image as label by training.This
The method that invention proposes can obtain in single sweep operation reliably quantifies T2Image.
Claims (10)
1. a kind of single sweep magnetic resonance based on residual error network quantifies T2Imaging reconstruction method, it is characterised in that comprise the following steps：
1) laboratory sample is got out, sample to be tested is placed on experimental bed and fixed, the experimental bed equipped with sample is sent into
The test chamber of magnetic resonance imager；
2) imager operation software is opened on the operation console of magnetic resonance imager, area-of-interest is carried out to laboratory sample first
Positioning, is then tuned, shimming, frequency correction and capability correction；
3) compiled OLED imaging sequences are imported, according to specific experimental conditions, the parameters of pulse train are set；
4) the OLED imaging sequences that step 3) sets parameter are performed, starts to sample, after the completion of data sampling, performs next step
Suddenly；
5) signal that step 4) obtains is normalized, zeroized and Fast Fourier Transform (FFT), the signal of k-space is transformed into figure
Image field；
6) random mask is generated according to the feature of sample to be tested, carrying out analog sampling using simulation softward to random mask obtains k
Spacing wave, is then normalized signal, zeroizes and Fast Fourier Transform (FFT) obtains training data；
7) residual error network model is built using TensorFlow deep learnings frame and Python, sets trained related ginseng
Number；
8) the data training network obtained using step 6), until data training network restrains and reaches stable, is trained
Network model, then the experimental data that step 5) obtains is rebuild using trained network model, is obtained reliable
Quantitative T2Image.
2. a kind of single sweep magnetic resonance based on residual error network as claimed in claim 1 quantifies T2Imaging reconstruction method, its feature exist
In in step 3), the structure of the OLED imaging sequences is followed successively by：Flip angle be α excitation pulse, displacement gradient G1, upset
Angle be α excitation pulse, displacement gradient G2, flip angle be α excitation pulse, displacement gradient G3, flip angle be α excitation pulse,
Shift gradient G4, flip angle be the reunion pulse of β, sampled echo chain.
3. a kind of single sweep magnetic resonance based on residual error network as claimed in claim 2 quantifies T2Imaging reconstruction method, its feature exist
In 4 displacement gradient Gs of 4 angle of deflection excitation pulse combination frequency coding peacekeeping phase code dimensions1、G2、G3、G44 are made to return
Ripple signal produces offset at the center of k-space, and 4 angle excitation pulses are all combined progress with the layer choosing gradient of level selection dimension
Layer choosing.
4. a kind of single sweep magnetic resonance based on residual error network as claimed in claim 2 quantifies T2Imaging reconstruction method, its feature exist
It is made of in the sampled echo chain the gradient chain for being respectively acting on frequency coding peacekeeping phase code dimension, the ladder of frequency coding dimension
Degree chain is made of a series of positive negative gradients, the gradient chain of phase code dimension by a series of area equations phase encoding gradient structure
Into.
5. a kind of single sweep magnetic resonance based on residual error network as claimed in claim 2 quantifies T2Imaging reconstruction method, its feature exist
In before sampled echo chain, frequency coding peacekeeping phase code dimension direction applies biasing gradient respectively, and frequency coding is tieed up inclined
Half of the gradient area for first frequency encoding gradient area is put, direction is on the contrary, the biasing gradient area of phase code dimension is
Phase code ties up the half of all phase encoding gradient areas, and direction is opposite.
6. a kind of single sweep magnetic resonance based on residual error network as claimed in claim 1 quantifies T2Imaging reconstruction method, its feature exist
It is described that the signal that step 4) obtains is analyzed in step 5), and the nuclear spin under acting on OLED imaging sequences is drilled
Change and carry out theory deduction, observation signal intensity proportional is in following expression after the reunion impulse action that flip angle is β：
<mrow>
<mi>s</mi>
<mi>i</mi>
<mi>n</mi>
<mi>&alpha;</mi>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>-</mo>
<mi>c</mi>
<mi>o</mi>
<mi>s</mi>
<mi>&beta;</mi>
<mo>)</mo>
</mrow>
<mfenced open = "{" close = "}">
<mtable>
<mtr>
<mtd>
<mrow>
<msup>
<mi>cos</mi>
<mn>3</mn>
</msup>
<msup>
<mi>&alpha;e</mi>
<mrow>
<mo>-</mo>
<msub>
<mi>i&theta;</mi>
<mn>4</mn>
</msub>
</mrow>
</msup>
<msup>
<mi>e</mi>
<mrow>
<mo>-</mo>
<msub>
<mi>TE</mi>
<mn>1</mn>
</msub>
<mo>/</mo>
<mrow>
<mo>(</mo>
<mn>2</mn>
<msub>
<mi>T</mi>
<mn>2</mn>
</msub>
<mo>(</mo>
<mover>
<mi>r</mi>
<mo>&RightArrow;</mo>
</mover>
<mo>)</mo>
<mo>)</mo>
</mrow>
</mrow>
</msup>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mo>+</mo>
<mfrac>
<mn>1</mn>
<mn>2</mn>
</mfrac>
<msup>
<mi>cos</mi>
<mn>2</mn>
</msup>
<mi>&alpha;</mi>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>+</mo>
<mi>cos</mi>
<mi>&alpha;</mi>
<mo>)</mo>
</mrow>
<msup>
<mi>e</mi>
<mrow>
<mo>-</mo>
<mi>i</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>&theta;</mi>
<mn>3</mn>
</msub>
<mo>+</mo>
<msub>
<mi>&theta;</mi>
<mn>4</mn>
</msub>
<mo>)</mo>
</mrow>
</mrow>
</msup>
<msup>
<mi>e</mi>
<mrow>
<mo>-</mo>
<msub>
<mi>TE</mi>
<mn>2</mn>
</msub>
<mo>/</mo>
<mrow>
<mo>(</mo>
<mn>2</mn>
<msub>
<mi>T</mi>
<mn>2</mn>
</msub>
<mo>(</mo>
<mover>
<mi>r</mi>
<mo>&RightArrow;</mo>
</mover>
<mo>)</mo>
<mo>)</mo>
</mrow>
</mrow>
</msup>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mo>+</mo>
<mfrac>
<mn>1</mn>
<mn>4</mn>
</mfrac>
<mi>cos</mi>
<mi>&alpha;</mi>
<msup>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>+</mo>
<mi>cos</mi>
<mi>&alpha;</mi>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<msup>
<mi>e</mi>
<mrow>
<mo>-</mo>
<mi>i</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>&theta;</mi>
<mn>2</mn>
</msub>
<mo>+</mo>
<msub>
<mi>&theta;</mi>
<mn>3</mn>
</msub>
<mo>+</mo>
<msub>
<mi>&theta;</mi>
<mn>4</mn>
</msub>
<mo>)</mo>
</mrow>
</mrow>
</msup>
<msup>
<mi>e</mi>
<mrow>
<mo>-</mo>
<msub>
<mi>TE</mi>
<mn>3</mn>
</msub>
<mo>/</mo>
<mrow>
<mo>(</mo>
<mn>2</mn>
<msub>
<mi>T</mi>
<mn>2</mn>
</msub>
<mo>(</mo>
<mover>
<mi>r</mi>
<mo>&RightArrow;</mo>
</mover>
<mo>)</mo>
<mo>)</mo>
</mrow>
</mrow>
</msup>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mo>+</mo>
<mfrac>
<mn>1</mn>
<mn>8</mn>
</mfrac>
<msup>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>+</mo>
<mi>cos</mi>
<mi>&alpha;</mi>
<mo>)</mo>
</mrow>
<mn>3</mn>
</msup>
<msup>
<mi>e</mi>
<mrow>
<mo>-</mo>
<mi>i</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>&theta;</mi>
<mn>1</mn>
</msub>
<mo>+</mo>
<msub>
<mi>&theta;</mi>
<mn>2</mn>
</msub>
<mo>+</mo>
<msub>
<mi>&theta;</mi>
<mn>3</mn>
</msub>
<mo>+</mo>
<msub>
<mi>&theta;</mi>
<mn>4</mn>
</msub>
<mo>)</mo>
</mrow>
</mrow>
</msup>
<msup>
<mi>e</mi>
<mrow>
<mo>-</mo>
<msub>
<mi>TE</mi>
<mn>4</mn>
</msub>
<mo>/</mo>
<mrow>
<mo>(</mo>
<mn>2</mn>
<msub>
<mi>T</mi>
<mn>2</mn>
</msub>
<mo>(</mo>
<mover>
<mi>r</mi>
<mo>&RightArrow;</mo>
</mover>
<mo>)</mo>
<mo>)</mo>
</mrow>
</mrow>
</msup>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
</mrow>
Wherein, γ is nuclear spin gyromagnetic ratio,For core certainly
It is spun on the position of the real space；By above formula obtain sampling period have 4 by out of phase modulate echo-signals, 4 different modulation phases
Position is θ respectively4、(θ3+θ4)、(θ2+θ3+θ4) and (θ1+θ2+θ3+θ4)。
7. a kind of single sweep magnetic resonance based on residual error network as claimed in claim 1 quantifies T2Imaging reconstruction method, its feature exist
In in step 6), the random mask uses the random generation of computer batch, random mould according to the feature distribution of experiment sample
Plate includes all features of experiment sample, adds destabilizing factor, improves robustness of the network model to undesirable experimental situation.
8. a kind of single sweep magnetic resonance based on residual error network as claimed in claim 7 quantifies T2Imaging reconstruction method, its feature exist
Include excitation pulse angular deviation, displacement gradient deviation and noise in the destabilizing factor.
9. a kind of single sweep magnetic resonance based on residual error network as claimed in claim 1 quantifies T2Imaging reconstruction method, its feature exist
In in step 7), the residual error network model includes：The agent structure of network and relevant training parameter, network model
Object function is：
<mrow>
<mi>I</mi>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<mi>N</mi>
</mfrac>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>N</mi>
</munderover>
<mo>|</mo>
<mo>|</mo>
<mfrac>
<mrow>
<mi>f</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<mi>i</mi>
</msub>
<mo>,</mo>
<mi>W</mi>
<mo>,</mo>
<mi>b</mi>
<mo>)</mo>
</mrow>
<mo>-</mo>
<msub>
<mi>y</mi>
<mi>i</mi>
</msub>
</mrow>
<msub>
<mi>y</mi>
<mrow>
<mi>c</mi>
<mi>h</mi>
<mi>a</mi>
<mi>n</mi>
<mi>g</mi>
<mi>e</mi>
</mrow>
</msub>
</mfrac>
<mo>|</mo>
<msubsup>
<mo>|</mo>
<mn>2</mn>
<mn>2</mn>
</msubsup>
<mo>+</mo>
<mi>&lambda;</mi>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>N</mi>
</munderover>
<mo>|</mo>
<mo>|</mo>
<mfrac>
<mrow>
<mo>&dtri;</mo>
<mrow>
<mo>(</mo>
<mi>f</mi>
<mo>(</mo>
<mrow>
<msub>
<mi>x</mi>
<mi>i</mi>
</msub>
<mo>,</mo>
<mi>W</mi>
<mo>,</mo>
<mi>b</mi>
</mrow>
<mo>)</mo>
<mo>-</mo>
<msub>
<mi>y</mi>
<mi>i</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>&times;</mo>
<msub>
<mi>y</mi>
<mrow>
<mi>m</mi>
<mi>a</mi>
<mi>s</mi>
<mi>k</mi>
</mrow>
</msub>
</mrow>
<msub>
<mi>y</mi>
<mrow>
<mi>c</mi>
<mi>h</mi>
<mi>a</mi>
<mi>n</mi>
<mi>g</mi>
<mi>e</mi>
</mrow>
</msub>
</mfrac>
<mo>|</mo>
<msub>
<mo>|</mo>
<mn>1</mn>
</msub>
</mrow>
Wherein, N is the quantity with a collection of training image, and f () is training network, and W and b are network parameters, and x is input picture, y
It is the corresponding template of input picture, ychangeIt is the matrix being set to the value of a certain threshold value of small Yu in y after the threshold value, ymaskIt is to y
The image edge information tried to achieve using Canny operators.
10. a kind of single sweep magnetic resonance based on residual error network as claimed in claim 1 quantifies T2Imaging reconstruction method, its feature
It is in step 8), the trained network model is suitable for the reconstruction of several samples.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711287890.9A CN108010100B (en) | 2017-12-07 | 2017-12-07 | Single-scanning magnetic resonance quantitative T based on residual error network2Imaging reconstruction method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711287890.9A CN108010100B (en) | 2017-12-07 | 2017-12-07 | Single-scanning magnetic resonance quantitative T based on residual error network2Imaging reconstruction method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108010100A true CN108010100A (en) | 2018-05-08 |
| CN108010100B CN108010100B (en) | 2020-02-18 |
Family
ID=62057318
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711287890.9A Active CN108010100B (en) | 2017-12-07 | 2017-12-07 | Single-scanning magnetic resonance quantitative T based on residual error network2Imaging reconstruction method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108010100B (en) |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108663644A (en) * | 2018-03-14 | 2018-10-16 | 厦门大学 | Single sweep Quantitative MRI Measurement T based on residual error network reconnection2* imaging method |
| CN109100669A (en) * | 2018-07-12 | 2018-12-28 | 厦门大学 | The synchronous magnetic resonance diffusion of single sweep and T based on overlapping echo2Imaging method |
| CN109557487A (en) * | 2018-11-28 | 2019-04-02 | 厦门大学 | The method of the time-frequency conversion and phasing of magnetic resonance signal based on residual error network |
| CN109597012A (en) * | 2018-12-24 | 2019-04-09 | 厦门大学 | A kind of single sweep space-time code imaging reconstruction method based on residual error network |
| CN110151181A (en) * | 2019-04-16 | 2019-08-23 | 杭州电子科技大学 | Rapid magnetic resonance imaging method based on the U-shaped network of recurrence residual error |
| CN110782452A (en) * | 2019-11-05 | 2020-02-11 | 厦门大学 | T2 quantitative image imaging method and system |
| CN110942496A (en) * | 2019-12-13 | 2020-03-31 | 厦门大学 | Propeller sampling and neural network-based magnetic resonance image reconstruction method and system |
| WO2020118616A1 (en) * | 2018-12-13 | 2020-06-18 | 深圳先进技术研究院 | Head and neck imaging method and device based on deep prior learning |
| CN111612866A (en) * | 2020-05-20 | 2020-09-01 | 上海交通大学 | Quantitative magnetic susceptibility imaging reconstruction method and system, storage medium and terminal |
| CN112164122A (en) * | 2020-10-30 | 2021-01-01 | 哈尔滨理工大学 | Rapid CS-MRI reconstruction method for generating countermeasure network based on depth residual error |
| WO2021012972A1 (en) * | 2019-07-22 | 2021-01-28 | 深圳市安测健康信息技术有限公司 | Magnetic resonance imaging control method and apparatus, and computer readable storage medium |
| CN113743682A (en) * | 2021-11-03 | 2021-12-03 | 中国科学院精密测量科学与技术创新研究院 | NMR (nuclear magnetic resonance) relaxation time inversion method based on supervised deep neural network |
| CN114299174A (en) * | 2021-11-26 | 2022-04-08 | 武汉理工大学 | Multi-echo undersampling reconstruction-water-fat separation method based on deep unsupervised learning |
| CN116973823A (en) * | 2023-09-25 | 2023-10-31 | 之江实验室 | Line scanning magnetic resonance imaging method and system based on full steady-state balance precession |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6603989B1 (en) * | 2000-03-21 | 2003-08-05 | Dmitriy A. Yablonskiy | T2 contrast in magnetic resonance imaging with gradient echoes |
| CN105232046A (en) * | 2015-11-26 | 2016-01-13 | 厦门大学 | Method for single sweep quantitative magnetic resonance T2 imaging based on overlapping echoes |
| CN105232045A (en) * | 2015-11-26 | 2016-01-13 | 厦门大学 | Single-scanning quantitative magnetic resonance diffusion imaging method based on dual echoes |
| CN106841273A (en) * | 2017-04-27 | 2017-06-13 | 厦门大学 | A kind of reconstructing water fat separated method based on single sweep space-time code magnetic resonance imaging |
| CN107274358A (en) * | 2017-05-23 | 2017-10-20 | 广东工业大学 | Image Super-resolution recovery technology based on cGAN algorithms |
-
2017
- 2017-12-07 CN CN201711287890.9A patent/CN108010100B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6603989B1 (en) * | 2000-03-21 | 2003-08-05 | Dmitriy A. Yablonskiy | T2 contrast in magnetic resonance imaging with gradient echoes |
| CN105232046A (en) * | 2015-11-26 | 2016-01-13 | 厦门大学 | Method for single sweep quantitative magnetic resonance T2 imaging based on overlapping echoes |
| CN105232045A (en) * | 2015-11-26 | 2016-01-13 | 厦门大学 | Single-scanning quantitative magnetic resonance diffusion imaging method based on dual echoes |
| CN106841273A (en) * | 2017-04-27 | 2017-06-13 | 厦门大学 | A kind of reconstructing water fat separated method based on single sweep space-time code magnetic resonance imaging |
| CN107274358A (en) * | 2017-05-23 | 2017-10-20 | 广东工业大学 | Image Super-resolution recovery technology based on cGAN algorithms |
Non-Patent Citations (2)
| Title |
|---|
| CONGBO CAI ET.AL: "Single-Shot Mapping Through OverLapping-Echo Detachment (OLED) Planar Imaging", 《IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING》 * |
| 李晓兵: "稀疏表示下图像超分辨分析算法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108663644A (en) * | 2018-03-14 | 2018-10-16 | 厦门大学 | Single sweep Quantitative MRI Measurement T based on residual error network reconnection2* imaging method |
| CN109100669A (en) * | 2018-07-12 | 2018-12-28 | 厦门大学 | The synchronous magnetic resonance diffusion of single sweep and T based on overlapping echo2Imaging method |
| CN109557487A (en) * | 2018-11-28 | 2019-04-02 | 厦门大学 | The method of the time-frequency conversion and phasing of magnetic resonance signal based on residual error network |
| WO2020118616A1 (en) * | 2018-12-13 | 2020-06-18 | 深圳先进技术研究院 | Head and neck imaging method and device based on deep prior learning |
| CN109597012A (en) * | 2018-12-24 | 2019-04-09 | 厦门大学 | A kind of single sweep space-time code imaging reconstruction method based on residual error network |
| CN110151181A (en) * | 2019-04-16 | 2019-08-23 | 杭州电子科技大学 | Rapid magnetic resonance imaging method based on the U-shaped network of recurrence residual error |
| CN110151181B (en) * | 2019-04-16 | 2022-07-19 | 杭州电子科技大学 | Rapid magnetic resonance imaging method based on recursive residual U-shaped network |
| WO2021012972A1 (en) * | 2019-07-22 | 2021-01-28 | 深圳市安测健康信息技术有限公司 | Magnetic resonance imaging control method and apparatus, and computer readable storage medium |
| CN110782452A (en) * | 2019-11-05 | 2020-02-11 | 厦门大学 | T2 quantitative image imaging method and system |
| CN110942496B (en) * | 2019-12-13 | 2022-02-11 | 厦门大学 | Propeller sampling and neural network-based magnetic resonance image reconstruction method and system |
| CN110942496A (en) * | 2019-12-13 | 2020-03-31 | 厦门大学 | Propeller sampling and neural network-based magnetic resonance image reconstruction method and system |
| CN111612866A (en) * | 2020-05-20 | 2020-09-01 | 上海交通大学 | Quantitative magnetic susceptibility imaging reconstruction method and system, storage medium and terminal |
| CN112164122A (en) * | 2020-10-30 | 2021-01-01 | 哈尔滨理工大学 | Rapid CS-MRI reconstruction method for generating countermeasure network based on depth residual error |
| CN112164122B (en) * | 2020-10-30 | 2022-08-23 | 哈尔滨理工大学 | Rapid CS-MRI reconstruction method for generating countermeasure network based on depth residual error |
| CN113743682A (en) * | 2021-11-03 | 2021-12-03 | 中国科学院精密测量科学与技术创新研究院 | NMR (nuclear magnetic resonance) relaxation time inversion method based on supervised deep neural network |
| CN114299174A (en) * | 2021-11-26 | 2022-04-08 | 武汉理工大学 | Multi-echo undersampling reconstruction-water-fat separation method based on deep unsupervised learning |
| CN116973823A (en) * | 2023-09-25 | 2023-10-31 | 之江实验室 | Line scanning magnetic resonance imaging method and system based on full steady-state balance precession |
| CN116973823B (en) * | 2023-09-25 | 2024-01-09 | 之江实验室 | Line scanning magnetic resonance imaging method and system based on full steady-state balance precession |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108010100B (en) | 2020-02-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108010100A (en) | A kind of single sweep magnetic resonance based on residual error network quantifies T2Imaging reconstruction method | |
| CN108663644B (en) | Single-scan quantitative magnetic resonance T based on residual error network reconstruction2Imaging method | |
| CN108132274B (en) | Echo-planar imaging is without reference scan pattern distortion antidote under non-uniform magnetic field | |
| CN109100669B (en) | Single-scan synchronous magnetic resonance diffusion and T based on overlapped echoes2Image forming method | |
| Gholipour et al. | Robust super-resolution volume reconstruction from slice acquisitions: application to fetal brain MRI | |
| US7902825B2 (en) | Motion corrected tensor magnetic resonance imaging | |
| US10338178B2 (en) | System and method for high-resolution spectroscopic imaging | |
| Yang et al. | Physics-driven synthetic data learning for biomedical magnetic resonance: The imaging physics-based data synthesis paradigm for artificial intelligence | |
| Zhang et al. | Robust single-shot T 2 mapping via multiple overlapping-echo acquisition and deep neural network | |
| CN110889897B (en) | Method and system for reconstructing incoherent motion magnetic resonance imaging parameters in voxel | |
| Hu et al. | RUN‐UP: Accelerated multishot diffusion‐weighted MRI reconstruction using an unrolled network with U‐Net as priors | |
| Wang et al. | Magnetic resonance fingerprinting with compressed sensing and distance metric learning | |
| CN109597012B (en) | Single-scanning space-time coding imaging reconstruction method based on residual error network | |
| Gao et al. | PCLR: phase‐constrained low‐rank model for compressive diffusion‐weighted MRI | |
| Yang et al. | Model-based synthetic data-driven learning (MOST-DL): Application in single-shot T 2 mapping with severe head motion using overlapping-echo acquisition | |
| Cai et al. | Single-Shot ${\text {T}} _ {{2}} $ Mapping Through OverLapping-Echo Detachment (OLED) Planar Imaging | |
| Mani et al. | Model-based deep learning for reconstruction of joint kq under-sampled high resolution diffusion MRI | |
| Li et al. | A simultaneous multi‐slice T2 mapping framework based on overlapping‐echo detachment planar imaging and deep learning reconstruction | |
| Wang et al. | High‐fidelity direct contrast synthesis from magnetic resonance fingerprinting | |
| Fang et al. | Deepsti: towards tensor reconstruction using fewer orientations in susceptibility tensor imaging | |
| Lam et al. | High-dimensional mr spatiospectral imaging by integrating physics-based modeling and data-driven machine learning: Current progress and future directions | |
| US20220071490A1 (en) | System, method and computer-accessible medium for accelerated, simultaneous quantitative and non-synthetic multi-contrast imaging | |
| Guo et al. | Manifold Model for High-Resolution fMRI Joint Reconstruction and Dynamic Quantification | |
| CN114518555B (en) | Nuclear Overhauser enhanced imaging method and system | |
| Bilgic et al. | Highly Accelerated Multishot EPI through Synergistic Machine Learning and Joint 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 |