CN110322528A - Nuclear magnetic resonance brain image reconstructing blood vessel method based on 3T, 7T - Google Patents
Nuclear magnetic resonance brain image reconstructing blood vessel method based on 3T, 7T Download PDFInfo
- Publication number
- CN110322528A CN110322528A CN201910566295.1A CN201910566295A CN110322528A CN 110322528 A CN110322528 A CN 110322528A CN 201910566295 A CN201910566295 A CN 201910566295A CN 110322528 A CN110322528 A CN 110322528A
- Authority
- CN
- China
- Prior art keywords
- picture
- network
- output
- net
- net network
- 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
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformation in the plane of the image
- G06T3/40—Scaling the whole image or part thereof
- G06T3/4053—Super resolution, i.e. output image resolution higher than sensor resolution
Abstract
The present invention provides a kind of nuclear magnetic resonance brain image reconstructing blood vessel method based on 3T, 7T: the following steps are included: step 1: obtaining 3T picture and 7T picture, carry out image preprocessing, 3T picture and 7T picture after being pre-processed to 3T picture and 7T picture；Step 2: being based on U-net network, the 3T picture after pre-processing obtains the output picture by U-net as the input of U-net network；Step 3: the 7T picture after the output picture of U-net and pretreatment being inputted into VGG-16 network respectively, obtains the output that VGG-16 network is passed through in output and 7T picture of the 3T picture after VGG-16 network；Step 4: costing bio disturbance is all carried out by the output of VGG-16 network to output after VGG-16 network of the output, 3T picture of U-net network and 7T picture, parameter is obtained using stochastic gradient descent method based on loss function, U-net network is updated according to parameter, 3T picture is inputted into updated U-net network and obtains reconstructed results.The present invention can rebuild blood vessel on 3T picture.
Description
Technical field
The present invention relates to mr imaging technique fields, more particularly to a kind of nuclear magnetic resonance brain figure based on 3T, 7T
As reconstructing blood vessel method.
Background technique
The magnetic resonance image of 7T can clearly see blood vessel not available for 3T magnetic resonance image, but due to 7T equipment
It is expensive and rare, the 3T image that most of hospital still uses.Therefore, it for this problem, proposes a kind of based on 3T, 7T
Nuclear magnetic resonance brain image reconstructing blood vessel method is to overcome the problems, such as that this is very necessary.
The existing domestic and international research for image reconstruction is largely focused on super-resolution image reconstruction.2014,
Dong et al. proposes the CNN model SRCNN for being used for general nature image super-resolution rebuilding.Kim etc. is borrowed on the basis of SRCNN
Mirror is used for the VGG network structure of image classification, proposes VDSR.Bee Lim etc. proposes enhanced depth residual error network EDSR.
And it is directed to magnetic resonance image, Wang etc. is in (Accelerating magnetic resonance imagingvia
Deep learning) in CNN has been used for the reconstruction of magnetic resonance image, Chang etc. is in (Deep learning for
Undersampled MRI reconstruction) the MR image reconstruction network based on U-net is proposed in paper.But
It is all the reconstruction for carrying out high-definition picture progress k-space lack sampling as low-resolution image.
Although there are many super resolution ratio reconstruction methods, some problems are still remained in the reconstruction of 3T, 7T:
(1) most Super-resolution reconstruction establishing network is mainly used for natural image.
(2) different from the low resolution picture and high-resolution pictures of lack sampling, 3T, 7T image are not to be completely registrated
, directly it is difficult to reconstruct picture using the difference between picture.
Therefore, it is necessary to improve to the prior art.
Summary of the invention
The technical problem to be solved in the present invention is to provide a kind of efficiently nuclear magnetic resonance brain image blood vessels based on 3T, 7T
Method for reconstructing.
In order to solve the above technical problems, the present invention provides a kind of nuclear magnetic resonance brain image reconstructing blood vessel based on 3T, 7T
Method: the following steps are included:
Step 1: obtaining 3T picture and 7T picture, image preprocessing is carried out to 3T picture and 7T picture, obtains pre-processing it
3T picture and 7T picture afterwards；
Step 2: be based on U-net network, using pre-process after 3T picture as U-net network input, obtain by
The output picture of U-net；
Step 3: the 7T picture after the output picture of U-net and pretreatment being inputted into VGG-16 network respectively, is obtained
Pass through the output of VGG-16 network with 7T picture to output of the 3T picture after VGG-16 network；
Step 4: VGG- is passed through to output after VGG-16 network of the output, 3T picture of U-net network and 7T picture
The output of 16 networks all carries out costing bio disturbance, obtains parameter using stochastic gradient descent method based on loss function, more according to parameter
3T picture is inputted updated U-net network and obtains reconstructed results by new U-net network.
As to the present invention is based on the improvement of the nuclear magnetic resonance brain image reconstructing blood vessel method of 3T, 7T: returning in step 1
One changes processing method are as follows:
Wherein siIndicate each of every figure pixel value, max (si) indicate pixel maximum in this picture, in this way
Allow the value of each pixel between [- 1,1].
As to the present invention is based on the nuclear magnetic resonance brain image reconstructing blood vessel further improvements in methods of 3T, 7T: step 2
The following steps are included:
Step 2-1 constructs U-net network structure；
Tanh function is added in the output position of step 2-2, U-net network, as shown in formula 2:
Cout=tanh (xi) (formula 2)
Wherein CoutIndicate the output of U-net network；Using pre-process after 3T picture as U-net network input Xi,
Obtain the output picture by U-net；Using the 7T picture after pre-processing accordingly as label.
As to the present invention is based on the nuclear magnetic resonance brain image reconstructing blood vessel further improvements in methods of 3T, 7T: step 4
The following steps are included:
Step 4-1 carries out costing bio disturbance to the output of U-net network, uses L1loss function, objective function such as formula
Shown in 3:
WhereinIt indicates and the 7T picture after processing, yjIndicate the output of U-net network, ElIndicate the damage of U-net network
Lose function；
Step 4-2 carries out the costing bio disturbance of characteristic pattern level to the output of VGG-16 network, using mean square error function,
Its objective function is as shown in formula 4:
WhereinIndicate that 7T picture passes through the output of VGG-16 network, yvIndicate that 3T picture is defeated after VGG-16 network
Out, v indicates each pixel, and N indicates all number of pixels；EkIndicate the loss function of feature level；
Step 4-3, final objective function Equation E is as shown in formula 5:
WhereinIt is the weight for balancing two losses；
Step 4-4 carries out gradient updating using stochastic gradient descent optimization method, so that optimization object function E, carries out
Training obtains parameter, updates U-net network according to parameter, and 3T picture is inputted updated U-net network and obtains reconstructed results.
As to the present invention is based on the nuclear magnetic resonance brain image reconstructing blood vessel further improvements in methods of 3T, 7T: step
4-4 includes:
Gradient updating is carried out using stochastic gradient descent optimization method, to optimize overall objective function E, will be owned
Pretreatment after 3T picture in 90% be used as training set picture, 10% be used as test set picture, training set picture is made
It for the input of U-net network, is trained using corresponding 7T picture as label, saves and update weight and ginseng in training process
Number；
The Model Weight and parameter that the load training of U-net network obtains, obtain using test set picture as input U-net network
To reconstructed results.
The present invention is based on the technical advantages of the nuclear magnetic resonance brain image reconstructing blood vessel method of 3T, 7T are as follows:
Pre-training VGG-16 net is added by devising the error in a feature level in the present invention after U-net network
Network structure, the error for calculating the output of this feature network are difficult to be registrated to solve the problems, such as picture itself.Due on 3T picture
There is no blood vessel, and the inventive point is to present invention offers the technical advantages that can rebuild blood vessel on 3T picture.
Detailed description of the invention
Specific embodiments of the present invention will be described in further detail with reference to the accompanying drawing.
Fig. 1 is that the present invention is based on the flow diagrams of the nuclear magnetic resonance brain image reconstructing blood vessel method of 3T, 7T.
Specific embodiment
The present invention is described further combined with specific embodiments below, but protection scope of the present invention is not limited in
This.
Embodiment 1, the nuclear magnetic resonance brain image reconstructing blood vessel method based on 3T, 7T, as shown in Figure 1, including following step
It is rapid:
S1: obtaining multiple 3T, 7T nuclear magnetic resonance brain images of the same person, and it is pre- to carry out image to 3T picture and 7T picture
Processing, obtains the image after normalized；
Pretreatment is that relevant operation is carried out on picture, it is intended to improve its quality to increase the place in next stage
The precision and accuracy rate of adjustment method.(3T and the 7T figure of script is registrated to 3T, 7T nuclear magnetic resonance brain image of the same person
The slice position of piece, the differences such as angle are unified by 3T the and 7T picture after registration in angles and positions, that is,
The corresponding width 7T picture of each width 3T picture).Later to 3T picture x1,x2,…xj,…,xnWith 7T picture y1,y2,…yj,…,yn
Every picture be normalized, normalized mode is as shown in formula 1:
Wherein siIndicate each of every figure pixel value, max (si) indicate pixel maximum in this picture, in this way
Allow the value of each pixel between [- 1,1].
After normalization, 3T and 7T picture is cropped to 272*272 size using the resize function in the library OpenCV, thus
3T, 7T picture after being pre-processed.
S2: network is built
S201 constructs U-net neural network, stacks network according to U-shaped structure, and constitute down-sampling and upper sampling process.
Down-sampling can extract feature, and up-sampling can be completed to position.
S202 has carried out the normalization between [- 1,1] due to us to input and output picture, then in U-net network
Output position need to add a tanh function to guarantee the output of the network all between [- 1,1], as shown in formula 2:
Cout=tanh (xi) (formula 2)
Wherein CoutIndicate the output of U-net network；
3T picture after pre-processing obtains the output picture by U-net as the input of U-net network；To
Learn the Nonlinear Mapping relationship between 3T picture and 7T picture；Using the 7T picture after pre-processing accordingly as label.
S203 accesses the pre-training VGG-16 network of ImageNet after U-net network, forms U-net-VGG network, should
VGG-16 network is first 16 layers of VGG-19 network；By the 7T picture input after the output picture of U-net and pretreatment
Pre-training VGG-16 network obtains output and 7T picture of the 3T picture after VGG-16 network by the defeated of VGG-16 network
Out.
Due to being difficult accurately to be registrated our 3T and 7T picture completely on registration, so in the error of space pixel level
It is difficult to solve the problems, such as us.So devising the error in a feature level here, pre-training is added after U-net network
VGG-16 network structure will effectively promote our modelling effect.It is finally to use characteristic pattern as output without connecting entirely
And softmax layers.
7T picture using the 3T picture after pre-processing as the input of U-net network, after the pretreatment being accordingly registrated
As label.The 7T picture after the output picture of U-net and pretreatment is inputted into pre-training VGG-16 network respectively,
Obtain output of the 3T picture after VGG-16 network, 7T picture passes through the output of VGG-16 network.
S3: model training and test
The design of S301 loss function: costing bio disturbance is carried out to the output of U-net network, uses L1loss function, mesh
Scalar functions are as shown in formula 3:
WhereinIndicate 7T picture, yjIndicate the output of U-net network, ElIndicate the loss function of U-net network.
The costing bio disturbance that characteristic pattern level is carried out to the output of VGG-16 network, uses mean square error function, target letter
Number is as shown in formula 4:
WhereinIndicate that the 7T picture after pre-processing passes through the output of VGG-16 network, yvIndicate that 3T picture passes through VGG-
Output after 16 networks, v indicate each pixel, and N indicates all number of pixels.EkIndicate the loss function of feature level.
Final objective function Equation enters shown in formula 5:
WhereinIt is the weight for balancing two losses, E is overall objective function；
S302 carries out gradient updating using stochastic gradient descent optimization method, thus optimize overall objective function E, it will
90% in 3T picture after all pretreatments is used as training set picture, and 10% is used as test set picture.By training set figure
Input of the piece as U-net network is trained using corresponding 7T picture as label.Not to VGG-16 when paying attention to training here
Network parameter is updated.Such purpose is due to only using U-net network in test, so needing when training
Allow U-net that there is whole performance.It saves and updates weight and parameter in training process.It network training 200,000 times, is trained
.pt model file.
U-net network is used only in test phase in S303, the Model Weight and parameter that load training obtains, with test set figure
Piece obtains reconstructed results as input, obtains the blood vessel not having on 3T picture on the picture of success after reconstruction.
Optimize overall objective function E using stochastic gradient descent optimization method.To network training 200,000 times, trained
Good .pt model file.U-net network is used only in test phase, the Model Weight obtained using training is schemed with 10% 3T
Piece obtains reconstructed results as input, obtains the blood vessel not having on 3T picture on the picture of success after reconstruction.
The above list is only a few specific embodiments of the present invention for finally, it should also be noted that.Obviously, this hair
Bright to be not limited to above embodiments, acceptable there are many deformations.Those skilled in the art can be from present disclosure
All deformations for directly exporting or associating, are considered as protection scope of the present invention.
Claims (5)
1. the nuclear magnetic resonance brain image reconstructing blood vessel method based on 3T, 7T, it is characterised in that: the following steps are included:
Step 1: obtaining 3T picture and 7T picture, image preprocessing is carried out to 3T picture and 7T picture, after being pre-processed
3T picture and 7T picture；
Step 2: being based on U-net network, the 3T picture after pre-processing is obtained as the input of U-net network by U-net
Output picture；
Step 3: the 7T picture after the output picture of U-net and pretreatment being inputted into VGG-16 network respectively, obtains 3T
The output of VGG-16 network is passed through in output and 7T picture of the picture after VGG-16 network；
Step 4: VGG-16 net is passed through to output after VGG-16 network of the output, 3T picture of U-net network and 7T picture
The output of network all carries out costing bio disturbance, obtains parameter using stochastic gradient descent method based on loss function, updates U- according to parameter
3T picture is inputted updated U-net network and obtains reconstructed results by net network.
2. the nuclear magnetic resonance brain image reconstructing blood vessel method according to claim 1 based on 3T, 7T, it is characterised in that:
Normalization processing method in step 1 are as follows:
Wherein siIndicate each of every figure pixel value, max (si) indicate pixel maximum in this picture, it allows so every
The value of a pixel is between [- 1,1].
3. the nuclear magnetic resonance brain image reconstructing blood vessel method according to claim 2 based on 3T, 7T, it is characterised in that:
Step 2 the following steps are included:
Step 2-1 constructs U-net network structure；
Tanh function is added in the output position of step 2-2, U-net network, as shown in formula 2:
Cout=tanh (xi) (formula 2)
Wherein CoutIndicate the output of U-net network；3T picture after pre-processing is obtained as the input Xi of U-net network
By the output picture of U-net；Using the 7T picture after pre-processing accordingly as label.
4. the nuclear magnetic resonance brain image reconstructing blood vessel method according to claim 3 based on 3T, 7T, it is characterised in that:
Step 4 the following steps are included:
Step 4-1 carries out costing bio disturbance to the output of U-net network, uses L1loss function, objective function such as 3 institute of formula
Show:
WhereinIt indicates and the 7T picture after processing, yjIndicate the output of U-net network, ElIndicate the loss letter of U-net network
Number；
Step 4-2 carries out the costing bio disturbance of characteristic pattern level to the output of VGG-16 network, uses mean square error function, mesh
Scalar functions are as shown in formula 4:
WhereinIndicate that 7T picture passes through the output of VGG-16 network, yvIndicate output of the 3T picture after VGG-16 network, v
Indicate each pixel, N indicates all number of pixels；EkIndicate the loss function of feature level；
Step 4-3, final objective function Equation E is as shown in formula 5:
WhereinIt is the weight for balancing two losses；
Step 4-4 carries out gradient updating using stochastic gradient descent optimization method, so that optimization object function E, is trained
Parameter is obtained, U-net network is updated according to parameter, 3T picture is inputted into updated U-net network and obtains reconstructed results.
5. the nuclear magnetic resonance brain image reconstructing blood vessel method according to claim 4 based on 3T, 7T, it is characterised in that:
Step 4-4 includes:
Gradient updating is carried out using stochastic gradient descent optimization method, it, will be all pre- to optimize overall objective function E
90% in 3T picture after processing is used as training set picture, and 10% is used as test set picture, using training set picture as U-
The input of net network is trained using corresponding 7T picture as label, is saved and is updated weight and parameter in training process；
The Model Weight and parameter that the load training of U-net network obtains obtain weight using test set picture as input U-net network
Build result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910566295.1A CN110322528B (en) | 2019-06-26 | 2019-06-26 | Nuclear magnetic resonance brain image vessel reconstruction method based on 3T and 7T |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910566295.1A CN110322528B (en) | 2019-06-26 | 2019-06-26 | Nuclear magnetic resonance brain image vessel reconstruction method based on 3T and 7T |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110322528A true CN110322528A (en) | 2019-10-11 |
| CN110322528B CN110322528B (en) | 2021-05-14 |
Family
ID=68120456
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910566295.1A Active CN110322528B (en) | 2019-06-26 | 2019-06-26 | Nuclear magnetic resonance brain image vessel reconstruction method based on 3T and 7T |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110322528B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2024017111A1 (en) * | 2022-07-20 | 2024-01-25 | 中国人民解放军总医院第一医学中心 | Deep-learning-based method for generating 7t magnetic resonance image from 3t magnetic resonance image |
Citations (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20070019887A1 (en) * | 2004-06-30 | 2007-01-25 | Oscar Nestares | Computing a higher resolution image from multiple lower resolution images using model-base, robust bayesian estimation |
| WO2016019484A1 (en) * | 2014-08-08 | 2016-02-11 | Xiaoou Tang | An apparatus and a method for providing super-resolution of a low-resolution image |
| CN105637536A (en) * | 2013-07-02 | 2016-06-01 | 外科信息科学股份有限公司 | Method and system for a brain image pipeline and brain image region location and shape prediction |
| CN107977930A (en) * | 2017-12-09 | 2018-05-01 | 北京花开影视制作有限公司 | A kind of image super-resolution method and its system |
| CN108416276A (en) * | 2018-02-12 | 2018-08-17 | 浙江大学 | The abnormal gait detection method of side gait video based on people |
| CN108596065A (en) * | 2018-04-13 | 2018-09-28 | 深圳职业技术学院 | One kind is based on deep semantic segmentation marine oil spill detecting system and method |
| CN108629816A (en) * | 2018-05-09 | 2018-10-09 | 复旦大学 | The method for carrying out thin layer MR image reconstruction based on deep learning |
| CN108921851A (en) * | 2018-06-06 | 2018-11-30 | 深圳市未来媒体技术研究院 | A kind of medicine CT image dividing method based on 3D confrontation network |
| CN109215013A (en) * | 2018-06-04 | 2019-01-15 | 平安科技（深圳）有限公司 | Automatic stone age prediction technique, system, computer equipment and storage medium |
| CN109215014A (en) * | 2018-07-02 | 2019-01-15 | 中国科学院深圳先进技术研究院 | Training method, device, equipment and the storage medium of CT image prediction model |
| CN109447897A (en) * | 2018-10-24 | 2019-03-08 | 文创智慧科技（武汉）有限公司 | A kind of real scene image composition method and system |
| US10242292B2 (en) * | 2017-06-13 | 2019-03-26 | Digital Surgery Limited | Surgical simulation for training detection and classification neural networks |
| US20190098724A1 (en) * | 2017-09-24 | 2019-03-28 | Massachusetts Institute Of Technology | Methods and Apparatus for Image Analysis for Lighting Control |
| CN109635882A (en) * | 2019-01-23 | 2019-04-16 | 福州大学 | Salient object detection method based on multi-scale convolution feature extraction and fusion |
| CN109903223A (en) * | 2019-01-14 | 2019-06-18 | 北京工商大学 | A kind of image super-resolution method based on dense connection network and production confrontation network |
-
2019
- 2019-06-26 CN CN201910566295.1A patent/CN110322528B/en active Active
Patent Citations (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20070019887A1 (en) * | 2004-06-30 | 2007-01-25 | Oscar Nestares | Computing a higher resolution image from multiple lower resolution images using model-base, robust bayesian estimation |
| CN105637536A (en) * | 2013-07-02 | 2016-06-01 | 外科信息科学股份有限公司 | Method and system for a brain image pipeline and brain image region location and shape prediction |
| WO2016019484A1 (en) * | 2014-08-08 | 2016-02-11 | Xiaoou Tang | An apparatus and a method for providing super-resolution of a low-resolution image |
| US10242292B2 (en) * | 2017-06-13 | 2019-03-26 | Digital Surgery Limited | Surgical simulation for training detection and classification neural networks |
| US20190098724A1 (en) * | 2017-09-24 | 2019-03-28 | Massachusetts Institute Of Technology | Methods and Apparatus for Image Analysis for Lighting Control |
| CN107977930A (en) * | 2017-12-09 | 2018-05-01 | 北京花开影视制作有限公司 | A kind of image super-resolution method and its system |
| CN108416276A (en) * | 2018-02-12 | 2018-08-17 | 浙江大学 | The abnormal gait detection method of side gait video based on people |
| CN108596065A (en) * | 2018-04-13 | 2018-09-28 | 深圳职业技术学院 | One kind is based on deep semantic segmentation marine oil spill detecting system and method |
| CN108629816A (en) * | 2018-05-09 | 2018-10-09 | 复旦大学 | The method for carrying out thin layer MR image reconstruction based on deep learning |
| CN109215013A (en) * | 2018-06-04 | 2019-01-15 | 平安科技（深圳）有限公司 | Automatic stone age prediction technique, system, computer equipment and storage medium |
| CN108921851A (en) * | 2018-06-06 | 2018-11-30 | 深圳市未来媒体技术研究院 | A kind of medicine CT image dividing method based on 3D confrontation network |
| CN109215014A (en) * | 2018-07-02 | 2019-01-15 | 中国科学院深圳先进技术研究院 | Training method, device, equipment and the storage medium of CT image prediction model |
| CN109447897A (en) * | 2018-10-24 | 2019-03-08 | 文创智慧科技（武汉）有限公司 | A kind of real scene image composition method and system |
| CN109903223A (en) * | 2019-01-14 | 2019-06-18 | 北京工商大学 | A kind of image super-resolution method based on dense connection network and production confrontation network |
| CN109635882A (en) * | 2019-01-23 | 2019-04-16 | 福州大学 | Salient object detection method based on multi-scale convolution feature extraction and fusion |
Non-Patent Citations (8)
| Title |
|---|
| CHAO DONG .ETAL: "Learning a Deep Convolutional Network for Image Super-Resolution", 《ECCV 2014: COMPUTER VISION – ECCV 2014》 * |
| DONG NIE .ETAL: "Medical Image Synthesis with Deep Convolutional Adversarial Networks", 《IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING》 * |
| JIWON KIM .ETAL: "Accurate Image Super-Resolution Using Very Deep Convolutional Networks", 《2016 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION》 * |
| KHOSRO BAHRAMI .ETAL: "Convolutional Neural Network for Reconstruction of 7T-like Images from 3T MRI Using Appearance and Anatomical Features", 《DEEP LEARNING AND DATA LABELING FOR MEDICAL APPLICATIONS》 * |
| KHOSRO BAHRAMI .ETAL: "Reconstruction of 7T-Like Images From 3T MRI", 《 IEEE TRANSACTIONS ON MEDICAL IMAGING》 * |
| 冯兴东 著: "《分布式统计计算》", 30 April 2018 * |
| 柳杨 著: "《数字图像物体识别理论详解与实战》", 31 March 2018, 北京邮电大学出版社 * |
| 陈锟 等: "生成对抗网络在医学图像处理中的应用", 《生命科学仪器》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2024017111A1 (en) * | 2022-07-20 | 2024-01-25 | 中国人民解放军总医院第一医学中心 | Deep-learning-based method for generating 7t magnetic resonance image from 3t magnetic resonance image |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110322528B (en) | 2021-05-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Chen et al. | Brain MRI super resolution using 3D deep densely connected neural networks | |
| Fang et al. | Soft-edge assisted network for single image super-resolution | |
| CN107610194B (en) | Magnetic resonance image super-resolution reconstruction method based on multi-scale fusion CNN | |
| CN108460726B (en) | Magnetic resonance image super-resolution reconstruction method based on enhanced recursive residual network | |
| CN108734659B (en) | Sub-pixel convolution image super-resolution reconstruction method based on multi-scale label | |
| CN109214989B (en) | Single image super resolution ratio reconstruction method based on Orientation Features prediction priori | |
| CN105046672B (en) | A kind of image super-resolution rebuilding method | |
| Ma et al. | PathSRGAN: multi-supervised super-resolution for cytopathological images using generative adversarial network | |
| CN108090871A (en) | A kind of more contrast MR image reconstruction methods based on convolutional neural networks | |
| Du et al. | Accelerated super-resolution MR image reconstruction via a 3D densely connected deep convolutional neural network | |
| Fu et al. | Residual scale attention network for arbitrary scale image super-resolution | |
| CN111487573B (en) | Enhanced residual error cascade network model for magnetic resonance undersampling imaging | |
| CN109472743A (en) | The super resolution ratio reconstruction method of remote sensing images | |
| CN112991483B (en) | Non-local low-rank constraint self-calibration parallel magnetic resonance imaging reconstruction method | |
| CN111161146A (en) | Coarse-to-fine single-image super-resolution reconstruction method | |
| CN105931242B (en) | Dynamic nuclear magnetic resonance (DNMR) image rebuilding method based on dictionary learning and time gradient | |
| CN113506222A (en) | Multi-mode image super-resolution method based on convolutional neural network | |
| Yan et al. | SMIR: A Transformer-Based Model for MRI super-resolution reconstruction | |
| CN109544488B (en) | Image synthesis method based on convolutional neural network | |
| Wei et al. | Channel rearrangement multi-branch network for image super-resolution | |
| Wang et al. | Residual deep attention mechanism and adaptive reconstruction network for single image super-resolution | |
| CN110322528A (en) | Nuclear magnetic resonance brain image reconstructing blood vessel method based on 3T, 7T | |
| CN113835059A (en) | Data processing method and system for diffusion weighted magnetic resonance imaging | |
| CN107945114A (en) | Magnetic resonance image super-resolution method based on cluster dictionary and iterative backprojection | |
| Yang et al. | RSAMSR: A deep neural network based on residual self-encoding and attention mechanism for image super-resolution |
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 |