CN110060314A - A kind of CT iterative approximation accelerated method and system based on artificial intelligence - Google Patents
A kind of CT iterative approximation accelerated method and system based on artificial intelligence Download PDFInfo
- Publication number
- CN110060314A CN110060314A CN201910324763.4A CN201910324763A CN110060314A CN 110060314 A CN110060314 A CN 110060314A CN 201910324763 A CN201910324763 A CN 201910324763A CN 110060314 A CN110060314 A CN 110060314A
- Authority
- CN
- China
- Prior art keywords
- image
- neural network
- artificial intelligence
- iterative
- iterative approximation
- 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
- G06T11/008—Specific post-processing after tomographic reconstruction, e.g. voxelisation, metal artifact correction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/50—Image enhancement or restoration by the use of more than one image, e.g. averaging, subtraction
-
- 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
- G06T2207/10081—Computed x-ray tomography [CT]
Abstract
The invention discloses a kind of CT iterative approximation accelerated method and system based on artificial intelligence, the method includes the steps: utilize the image sequence of iterative process to construct image data base;Neural network is constructed according to image data base, and carries out neural metwork training;Trained neural network exports new image to iterating to calculate every time after obtained present image is handled, and according to new image update present image.Due to being handled using neural network obtained present image is iterated to calculate every time, the number of iterations can be reduced, difficulty in computation is reduced, to solve the problems, such as the computationally intensive of existing CT iterative approximation.
Description
Technical field
The present invention relates to medical imaging technology fields more particularly to a kind of CT iterative approximation based on artificial intelligence to add
Fast method and system.
Background technique
CT (computer dislocation scanning and imaging system) utilizes X-ray scanning object, obtains data for projection, and pass through tomography weight
Algorithm is built to handle these data for projection, the tomography and three-dimensional density information of object is obtained, achievees the purpose that non-destructive testing.With
The pursuit to radiological dose and picture quality in recent years, the Image Reconstruction Technology to become increasingly complex be applied in CT product,
Iterative reconstruction technique is exactly that numerous concerns are had received under this main trend.In the prior art, although iterative reconstruction technique energy
The enough higher information using ray, obtains better image quality, but there is a problem of computationally intensive.
Therefore, the existing technology needs to be improved and developed.
Summary of the invention
The technical problem to be solved in the present invention is that in view of the above drawbacks of the prior art, providing a kind of based on artificial intelligence
The CT iterative approximation accelerated method and system of energy, it is intended to solve the problems, such as that iterative reconstruction technique is computationally intensive in the prior art.
The technical proposal for solving the technical problem of the invention is as follows:
A kind of CT iterative approximation accelerated method based on artificial intelligence, wherein comprising steps of
Image data base is constructed using the image sequence of iterative process;
Neural network is constructed according to image data base, and carries out neural metwork training;
Trained neural network exports new image to being iterated to calculate after obtained present image is handled every time, and
According to new image update present image.
The CT iterative approximation accelerated method based on artificial intelligence, wherein the image sequence using iterative process
Column building image database steps specifically include:
By the input picture and target image formation image pair in the image sequence of iterative process, wherein input picture X times
The image of history iteration result, comprising: iv-th iteration image, the N-1 times iterative image, and so on to the N-X+1 times scheme
Picture, target image are the N+M times iterative image, and N is positive integer, and X and M are the positive integer more than or equal to 1;
According to multiple images to formation image data base.
The CT iterative approximation accelerated method based on artificial intelligence, wherein described that mind is constructed according to image data base
Through network, and carries out neural metwork training step and specifically includes:
The neural network based on convolution is constructed according to image data base;
Using gradient descent method, according to the parameter of objective function optimization neural network, and complete to train.
The CT iterative approximation accelerated method based on artificial intelligence, wherein the objective function are as follows:
Wherein, Img is the image crossed by Processing with Neural Network, ImgtargFor target image, k is positive integer, ImgkIt is
The k images crossed by Processing with Neural Network, Imgtarg,kFor kth target image, Σ is sum operation, and Loss () is mesh
Scalar functions.
The CT iterative approximation accelerated method based on artificial intelligence, wherein the trained neural network is to every
Secondary iterate to calculate after obtained present image is handled exports new image, and according to new image update present image step
It specifically includes:
Trained neural network is defeated after obtained present image is handled to iterating to calculate every time according to acceleration ratio
New image out;
According to new image update present image.
A kind of CT iterative approximation acceleration system based on artificial intelligence, wherein include: processor, and with the processing
The memory of device connection,
The memory is stored with the CT iterative approximation based on artificial intelligence and accelerates program, the CT based on artificial intelligence
Iterative approximation accelerates program to perform the steps of when being executed by the processor
Image data base is constructed using the image sequence of iterative process;
Neural network is constructed according to image data base, and carries out neural metwork training;
Trained neural network exports new image to being iterated to calculate after obtained present image is handled every time, and
According to new image update present image.
The CT iterative approximation acceleration system based on artificial intelligence, wherein the CT iteration based on artificial intelligence
When reconstruction accelerates program to be executed by the processor, also perform the steps of
By the input picture and target image formation image pair in the image sequence of iterative process, wherein input picture X
The image of secondary history iteration result, comprising: iv-th iteration image, the N-1 times iterative image, and so on to the N-X+1 times scheme
Picture, target image are the N+M times iterative image, and N is positive integer, and X and M are the positive integer more than or equal to 1;
According to multiple images to formation image data base.
The CT iterative approximation acceleration system based on artificial intelligence, wherein the CT iteration based on artificial intelligence
When reconstruction accelerates program to be executed by the processor, also perform the steps of
The neural network based on convolution is constructed according to image data base;
Using gradient descent method, according to the parameter of objective function optimization neural network, and complete to train.
The CT iterative approximation acceleration system based on artificial intelligence, wherein the objective function are as follows:
Wherein, Img is the image crossed by Processing with Neural Network, ImgtargFor target image, k is positive integer, ImgkIt is
The k images crossed by Processing with Neural Network, Imgtarg,kFor kth target image, Σ is sum operation, and Loss () is mesh
Scalar functions.
The CT iterative approximation acceleration system based on artificial intelligence, wherein the CT iteration based on artificial intelligence
When reconstruction accelerates program to be executed by the processor, also perform the steps of
Trained neural network is defeated after obtained present image is handled to iterating to calculate every time according to acceleration ratio
New image out;
According to new image update present image.
The utility model has the advantages that reducing the number of iterations due to estimating complicated PC using neural network, difficulty in computation is reduced, thus
Solve the problems, such as the computationally intensive of existing CT iterative approximation.
Detailed description of the invention
Fig. 1 is the flow chart of the CT iterative approximation accelerated method in the present invention based on artificial intelligence.
Fig. 2 is the flow chart of iterative approximation in the prior art.
Fig. 3 is the schematic diagram for the image sequence that iterative approximation generates in the prior art.
Fig. 4 is the schematic diagram of neural network training process in the present invention.
Fig. 5 is the schematic diagram of Processing with Neural Network process in the present invention.
Fig. 6 is the functional schematic block diagram of the CT iterative approximation acceleration system in the present invention based on artificial intelligence.
Specific embodiment
To make the objectives, technical solutions, and advantages of the present invention clearer and more explicit, right as follows in conjunction with drawings and embodiments
The present invention is further described.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, and do not have to
It is of the invention in limiting.
Please refer to Fig. 1-Fig. 5, the present invention provides a kind of CT iterative approximation accelerated method based on artificial intelligence
Some embodiments.
Technology proposed by the invention is to be used as guidance by the theoretical of traditional pre-conditioner, and pushed away
Extensively.The concrete condition of gradient descent method and pre-conditioned gradient descent method is as follows: for CT system, it is assumed that y generation
The data for projection that table scan obtains, x represent the image for the patient to be rebuild, and ideally, we can obtain formula (1).
Y=Ax (1)
But the introducing of practical situation one various errors of federation or noise, the process of solution is generally by minimum
Loss function is come the formula 2 that reaches.
Cost (x, y)=| | y-Ax | |2 (2)
For solving the loss function of simple form in this, optimal solution can be obtained with direct solution.
x*=(ATK-1A)-1ATK-1y (3)
x*=(PC) ATK-1y (4)
PC is exactly pre-conditioner.If the convergence rate that can properly dramatically speed up iteration of this PC selection.
Method proposed by the present invention replaces the middle pre-conditioner matrix of classical pre-conditioner method
A neural network is changed into, the input of this neural network is the image of preceding iteration result several times, and output is a pre-
conditioner.That is, estimating complicated PC using deep neural network, the number of iterations is reduced, reduces difficulty in computation,
To solve the problems, such as the computationally intensive of existing CT iterative approximation.
As shown in Figure 1, a kind of CT iterative approximation accelerated method based on artificial intelligence of the invention, comprising the following steps:
Step S100, image data base is constructed using the image sequence of iterative process.
Specifically, the step S100 is specifically included:
Step S110, by the input picture and target image formation image pair in the image sequence of iterative process, wherein defeated
Enter the image that image is X history iteration result, comprising: iv-th iteration image, the N-1 times iterative image, and so on to the
N-X+1 image, target image are the N+M times iterative image, and N is positive integer, and X and M are the positive integer more than or equal to 1.
Step S120, according to multiple images to formation image data base.
Iterative algorithm is exactly constantly to calculate error by each iterative process, gradually to approach final accurately knot
Fruit.Each iteration can all obtain an intermediate result image.For example it have passed through 100 iteration and just have 100 image (such as Fig. 2
With shown in Fig. 3).Each image is different, and the number of iterations is more, and image is closer to final result.Here with iteration mistake
The image sequence of journey generates training set.For example, working as X=2, input is the image of n-th and the N-1 times, and output is N+M knot
Fruit (M > 1, such as M=20), thus can be by the later result of this neural network prediction successive ignition (such as Fig. 4 institute
Show).
Here it is multiple that M value can according to need setting, for example, obtaining a training set with M=20 for reference, then with M
=5 is, referring to another training set is obtained, each training set form image data base, naturally it is also possible to the same training set is used,
When different M values, the image of corresponding the number of iterations is extracted in different ways.In image data base, training set is carried out by M value
Classification, different M values correspond to different acceleration ratios, and M value is bigger, accelerate ratio higher, and M value is smaller, accelerate ratio lower.
Therefore different M values can be selected according to the needs for accelerating ratio, rebuilds speed to achieve the purpose that adjust.
Step S200, neural network is constructed according to image data base, and carries out neural metwork training.
Specifically, the step S200 is specifically included:
Step S210, the neural network based on convolution is constructed according to image data base.
Specifically, the selected model of training is mainly based upon the neural network of convolution, such as improved ResNet,
Unet。
Step S220, it using gradient descent method, according to the parameter of objective function optimization neural network, and completes to train.
Specifically, it in order to allow the neural network image with true value closer to the image prediction currently updated, needs
Image is generated to being trained.As shown in figure 4, each iteration can basis by taking the iterative approximation using gradient descent method as an example
Last image ImgN-1 updates to obtain present image ImgN, while also having ImgN+M later using M iteration.Here
It is to utilize (ImgN, ImgN-1) for input, ImgN+M is output.Such image is to being trained.
The objective function are as follows:
Wherein, Img is the image crossed by Processing with Neural Network, ImgtargFor target image, k is positive integer, ImgkIt is
The k images crossed by Processing with Neural Network, Imgtarg,kFor kth target image, Σ is sum operation, and Loss () is mesh
Scalar functions.
After setting these parameters, it is directed to objective function using gradient descent method, the parameter of neural network is optimized.
After the completion of training, the parameter of entire model is retained, for using later.Based in training process smart network at
Ripe degree, by a certain number of test images occurred in pairs checked except image data base, to determine network training result
Quality.When the input width image in these test images is successfully converted into new images, and the new figure by smart network
As this can be showed by the standard set in training process to when the quality of target image, which can be tied in image
Beam.
Step S300, trained neural network is new to output after obtained present image is handled is iterated to calculate every time
Image, and according to new image update present image.
Specifically, the step S200 is specifically included:
Step S310, trained neural network is carried out according to acceleration ratio to obtained present image is iterated to calculate every time
New image is exported after processing.
Specifically, by setting the size of M value, adjustable acceleration ratio, to obtain different iteration speeds.
Step S320, according to new image update present image.
Present image, which can also continue to be iterated to calculate and be input in trained neural network, further to be located
Reason, until up to standard.
It is worth noting that classical PC can only accurately calculate the PC matrix rebuild based on two dimensional image, for common
Three-dimensional cone beam method, be difficult the method having had to instruct approximation.Method proposed by the present invention, as long as according to the scene of application
It is trained the matrix that can be obtained by suitable for that occasion, more general stabilization.
The preferred embodiment for the CT iterative approximation acceleration system based on artificial intelligence that the present invention also provides a kind of:
As shown in fig. 6, a kind of CT iterative approximation acceleration system based on artificial intelligence described in the embodiment of the present invention, packet
It includes: processor 10, and the memory 20 being connect with the processor 10,
The memory is stored with the CT iterative approximation based on artificial intelligence and accelerates program, the CT based on artificial intelligence
Iterative approximation accelerates program to perform the steps of when being executed by the processor 10
Image data base is constructed using the image sequence of iterative process;
Neural network is constructed according to image data base, and carries out neural metwork training;
Trained neural network exports new image to being iterated to calculate after obtained present image is handled every time, and
According to new image update present image, as detailed above.
When the CT iterative approximation based on artificial intelligence accelerates program to be executed by the processor 10, following step is also realized
It is rapid:
By the input picture and target image formation image pair in the image sequence of iterative process, wherein input picture X times
The image of history iteration result, comprising: iv-th iteration image, the N-1 times iterative image, and so on to the N-X+1 times scheme
Picture, target image are the N+M times iterative image, and N is positive integer, and X and M are the positive integer more than or equal to 1;
According to multiple images to image data base is formed, as detailed above.
When the CT iterative approximation based on artificial intelligence accelerates program to be executed by the processor 10, following step is also realized
It is rapid:
The neural network based on convolution is constructed according to image data base;
It using gradient descent method, according to the parameter of objective function optimization neural network, and completes to train, specific institute as above
It states.
The objective function are as follows:
Wherein, Img is the image crossed by Processing with Neural Network, ImgtargFor target image, k is positive integer, ImgkIt is
The k images crossed by Processing with Neural Network, Imgtarg,kFor kth target image, Σ is sum operation, and Loss () is mesh
Scalar functions, as detailed above.
When the CT iterative approximation based on artificial intelligence accelerates program to be executed by the processor 10, following step is also realized
It is rapid:
Trained neural network is defeated after obtained present image is handled to iterating to calculate every time according to acceleration ratio
New image out;
According to new image update present image, as detailed above.
In conclusion a kind of CT iterative approximation accelerated method and system based on artificial intelligence provided by the present invention, institute
Method is stated comprising steps of the image sequence using iterative process constructs image data base;Nerve net is constructed according to image data base
Network, and carry out neural metwork training;Trained neural network to iterating to calculate after obtained present image handles every time
New image is exported, and according to new image update present image.Due to being carried out using present image of the neural network to iteration
Processing can reduce the number of iterations, difficulty in computation be reduced, to solve the problems, such as the computationally intensive of existing CT iterative approximation.
It should be understood that the application of the present invention is not limited to the above for those of ordinary skills can
With improvement or transformation based on the above description, all these modifications and variations all should belong to the guarantor of appended claims of the present invention
Protect range.
Claims (10)
1. a kind of CT iterative approximation accelerated method based on artificial intelligence, which is characterized in that comprising steps of
Image data base is constructed using the image sequence of iterative process;
Neural network is constructed according to image data base, and carries out neural metwork training;
Trained neural network exports new image to iterating to calculate every time after obtained present image is handled, and according to
New image update present image.
2. the CT iterative approximation accelerated method according to claim 1 based on artificial intelligence, which is characterized in that the utilization
The image sequence building image database steps of iterative process specifically include:
By the input picture and target image formation image pair in the image sequence of iterative process, wherein input picture is gone through for X times
The image of history iteration result, comprising: iv-th iteration image, the N-1 times iterative image, and so on to the N-X+1 times image,
Target image is the N+M times iterative image, and N is positive integer, and X and M are the positive integer more than or equal to 1;
According to multiple images to formation image data base.
3. the CT iterative approximation accelerated method according to claim 1 based on artificial intelligence, which is characterized in that the basis
Image data base constructs neural network, and carries out neural metwork training step and specifically include:
The neural network based on convolution is constructed according to image data base;
Using gradient descent method, according to the parameter of objective function optimization neural network, and complete to train.
4. the CT iterative approximation accelerated method according to claim 3 based on artificial intelligence, which is characterized in that the target
Function are as follows:
Wherein, Img is the image crossed by Processing with Neural Network, ImgtargFor target image, k is positive integer, ImgkFor kth
The image crossed by Processing with Neural Network, Imgtarg,kFor kth target image, Σ is sum operation, and Loss () is target letter
Number.
5. the CT iterative approximation accelerated method according to claim 1 based on artificial intelligence, which is characterized in that the training
Good neural network exports new image to iterating to calculate every time after obtained present image is handled, and according to new image
Present image step is updated to specifically include:
Trained neural network is new to output after obtained present image is handled is iterated to calculate every time according to acceleration ratio
Image;
According to new image update present image.
6. a kind of CT iterative approximation acceleration system based on artificial intelligence characterized by comprising processor, and with it is described
The memory of processor connection,
The memory is stored with the CT iterative approximation based on artificial intelligence and accelerates program, the CT iteration based on artificial intelligence
It rebuilds and program is accelerated to perform the steps of when being executed by the processor
Image data base is constructed using the image sequence of iterative process;
Neural network is constructed according to image data base, and carries out neural metwork training;
Trained neural network exports new image to iterating to calculate every time after obtained present image is handled, and according to
New image update present image.
7. the CT iterative approximation acceleration system according to claim 6 based on artificial intelligence, which is characterized in that described to be based on
When the CT iterative approximation of artificial intelligence accelerates program to be executed by the processor, also perform the steps of
By the input picture and target image formation image pair in the image sequence of iterative process, wherein input picture is gone through for X times
The image of history iteration result, comprising: iv-th iteration image, the N-1 times iterative image, and so on to the N-X+1 times image,
Target image is the N+M times iterative image, and N is positive integer, and X and M are the positive integer more than or equal to 1;
According to multiple images to formation image data base.
8. the CT iterative approximation acceleration system according to claim 6 based on artificial intelligence, which is characterized in that described to be based on
When the CT iterative approximation of artificial intelligence accelerates program to be executed by the processor, also perform the steps of
The neural network based on convolution is constructed according to image data base;
Using gradient descent method, according to the parameter of objective function optimization neural network, and complete to train.
9. the CT iterative approximation acceleration system according to claim 8 based on artificial intelligence, which is characterized in that the target
Function are as follows:
Wherein, Img is the image crossed by Processing with Neural Network, ImgtargFor target image, k is positive integer, ImgkFor kth
The image crossed by Processing with Neural Network, Imgtarg,kFor kth target image, Σ is sum operation, and Loss () is target letter
Number.
10. the CT iterative approximation acceleration system according to claim 6 based on artificial intelligence, which is characterized in that the base
When the CT iterative approximation of artificial intelligence accelerates program to be executed by the processor, also perform the steps of
Trained neural network is new to output after obtained present image is handled is iterated to calculate every time according to acceleration ratio
Image;
According to new image update present image.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910324763.4A CN110060314B (en) | 2019-04-22 | 2019-04-22 | CT iterative reconstruction acceleration method and system based on artificial intelligence |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910324763.4A CN110060314B (en) | 2019-04-22 | 2019-04-22 | CT iterative reconstruction acceleration method and system based on artificial intelligence |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110060314A true CN110060314A (en) | 2019-07-26 |
CN110060314B CN110060314B (en) | 2023-05-16 |
Family
ID=67320260
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910324763.4A Active CN110060314B (en) | 2019-04-22 | 2019-04-22 | CT iterative reconstruction acceleration method and system based on artificial intelligence |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110060314B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2021068975A1 (en) * | 2019-10-12 | 2021-04-15 | Shanghai United Imaging Healthcare Co., Ltd. | Systems and methods for image reconstruction |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107358575A (en) * | 2017-06-08 | 2017-11-17 | 清华大学 | A kind of single image super resolution ratio reconstruction method based on depth residual error network |
CN108256555A (en) * | 2017-12-21 | 2018-07-06 | 北京达佳互联信息技术有限公司 | Picture material recognition methods, device and terminal |
US20180197317A1 (en) * | 2017-01-06 | 2018-07-12 | General Electric Company | Deep learning based acceleration for iterative tomographic reconstruction |
CN108765514A (en) * | 2018-06-06 | 2018-11-06 | 上海交通大学 | A kind of accelerating method and device of CT image reconstructions |
GB201901024D0 (en) * | 2019-01-25 | 2019-03-13 | Siemens Healthcare Ltd | Method of reconstructing magnetic resonance image data |
WO2019060843A1 (en) * | 2017-09-22 | 2019-03-28 | Nview Medical Inc. | Image reconstruction using machine learning regularizers |
US20190104940A1 (en) * | 2017-10-06 | 2019-04-11 | Toshiba Medical Systems Corporation | Apparatus and method for medical image reconstruction using deep learning for computed tomography (ct) image noise and artifacts reduction |
-
2019
- 2019-04-22 CN CN201910324763.4A patent/CN110060314B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20180197317A1 (en) * | 2017-01-06 | 2018-07-12 | General Electric Company | Deep learning based acceleration for iterative tomographic reconstruction |
CN107358575A (en) * | 2017-06-08 | 2017-11-17 | 清华大学 | A kind of single image super resolution ratio reconstruction method based on depth residual error network |
WO2019060843A1 (en) * | 2017-09-22 | 2019-03-28 | Nview Medical Inc. | Image reconstruction using machine learning regularizers |
US20190104940A1 (en) * | 2017-10-06 | 2019-04-11 | Toshiba Medical Systems Corporation | Apparatus and method for medical image reconstruction using deep learning for computed tomography (ct) image noise and artifacts reduction |
CN108256555A (en) * | 2017-12-21 | 2018-07-06 | 北京达佳互联信息技术有限公司 | Picture material recognition methods, device and terminal |
CN108765514A (en) * | 2018-06-06 | 2018-11-06 | 上海交通大学 | A kind of accelerating method and device of CT image reconstructions |
GB201901024D0 (en) * | 2019-01-25 | 2019-03-13 | Siemens Healthcare Ltd | Method of reconstructing magnetic resonance image data |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2021068975A1 (en) * | 2019-10-12 | 2021-04-15 | Shanghai United Imaging Healthcare Co., Ltd. | Systems and methods for image reconstruction |
Also Published As
Publication number | Publication date |
---|---|
CN110060314B (en) | 2023-05-16 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106796716B (en) | For providing the device and method of super-resolution for low-resolution image | |
CN109949215B (en) | Low-dose CT image simulation method | |
CN103247061B (en) | Augmented lagrangian iterative reconstruction method of X-ray image and CI image | |
CN109584290A (en) | A kind of three-dimensional image matching method based on convolutional neural networks | |
CN110598843B (en) | Training method for generating countermeasure network organization structure based on discriminant sharing | |
CN110458829B (en) | Image quality control method, device, equipment and storage medium based on artificial intelligence | |
CN107341776A (en) | Single frames super resolution ratio reconstruction method based on sparse coding and combinatorial mapping | |
CN109711401A (en) | A kind of Method for text detection in natural scene image based on Faster Rcnn | |
CN113902921A (en) | Image processing method, device, equipment and storage medium | |
WO2024032098A1 (en) | Training method and apparatus for artifact removal model, and device, medium and program product | |
CN111462020A (en) | Method, system, storage medium and device for correcting motion artifact of heart image | |
CN111695624B (en) | Updating method, device, equipment and storage medium of data enhancement strategy | |
CN110060315A (en) | A kind of image motion artifact eliminating method and system based on artificial intelligence | |
CN112598649B (en) | 2D/3D spine CT non-rigid registration method based on generation of countermeasure network | |
CN112508808A (en) | CT (computed tomography) dual-domain joint metal artifact correction method based on generation countermeasure network | |
CN107595312A (en) | Model generating method, image processing method and medical imaging devices | |
CN110490968A (en) | Based on the light field axial direction refocusing image super-resolution method for generating confrontation network | |
CN112614169A (en) | 2D/3D spine CT (computed tomography) level registration method based on deep learning network | |
CN109117860A (en) | A kind of image classification method based on subspace projection and dictionary learning | |
CN115860091A (en) | Depth feature descriptor learning method based on orthogonal constraint | |
CN110060314A (en) | A kind of CT iterative approximation accelerated method and system based on artificial intelligence | |
CN110533610A (en) | The generation method and device of image enhancement model, application method and device | |
CN113837191B (en) | Cross-star remote sensing image semantic segmentation method based on bidirectional unsupervised domain adaptive fusion | |
CN109993701A (en) | A method of the depth map super-resolution rebuilding based on pyramid structure | |
CN105608719B (en) | A kind of rapid CT image rebuilding method based on two benches projection adjustment |
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 |