CN109816747A - A kind of metal artifacts reduction method of Cranial Computed Tomography image - Google Patents
A kind of metal artifacts reduction method of Cranial Computed Tomography image Download PDFInfo
- Publication number
- CN109816747A CN109816747A CN201910136016.8A CN201910136016A CN109816747A CN 109816747 A CN109816747 A CN 109816747A CN 201910136016 A CN201910136016 A CN 201910136016A CN 109816747 A CN109816747 A CN 109816747A
- Authority
- CN
- China
- Prior art keywords
- image
- artifact
- cnn
- network
- training
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 52
- 238000002591 computed tomography Methods 0.000 title claims abstract description 45
- 239000002184 metal Substances 0.000 title claims abstract description 38
- 230000009467 reduction Effects 0.000 title claims abstract description 15
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 40
- 238000012549 training Methods 0.000 claims abstract description 26
- 238000013507 mapping Methods 0.000 claims description 9
- 230000008569 process Effects 0.000 claims description 8
- 238000003709 image segmentation Methods 0.000 claims description 5
- 238000004458 analytical method Methods 0.000 claims description 3
- 238000000605 extraction Methods 0.000 claims description 3
- 238000005259 measurement Methods 0.000 claims description 3
- 239000000203 mixture Substances 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 3
- 230000006872 improvement Effects 0.000 abstract description 3
- 238000012937 correction Methods 0.000 description 9
- 230000000694 effects Effects 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 238000012216 screening Methods 0.000 description 3
- 238000007796 conventional method Methods 0.000 description 2
- 238000006073 displacement reaction Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000002708 enhancing effect Effects 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 238000007689 inspection Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 238000011387 Li's method Methods 0.000 description 1
- 210000003484 anatomy Anatomy 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 239000003814 drug Substances 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 229910052752 metalloid Inorganic materials 0.000 description 1
- 150000002738 metalloids Chemical class 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Landscapes
- Apparatus For Radiation Diagnosis (AREA)
- Image Analysis (AREA)
Abstract
The present invention discloses a kind of metal artifacts reduction method of Cranial Computed Tomography image, including two steps: initially setting up one includes and the database not comprising artifact Cranial Computed Tomography image, using the method for deformable image registration before convolutional neural networks (CNN) training preprocessed data；Secondly, constructing simple 17 layers of CNN framework in training to learn metal artifacts, and accelerate the speed of training data using GPU, improves the learning efficiency of network；Meanwhile the experimental results showed that this method has the ability of good metal artifacts reduction, PSNR and SSIM numerical value also shows that apparent improvement.
Description
Technical field
The invention belongs to metal artifacts correcting technologies, and in particular to a kind of metal artifacts reduction method of Cranial Computed Tomography image.
Background technique
Computer tomography (computed tomography, CT) be medicine and industrial circle commonly used equipment it
One, it obtains sectioning image using specific CT algorithm for reconstructing, plays an important role to medical diagnosis on disease and defect inspection.?
In CT scan, metal is much higher than tissue as a kind of attenuation coefficient of highdensity object to X-ray, if scanning
In the process, there are the highly attenuating substance of metalloid, when X-ray passes through these objects will sharp-decay, lead to corresponding throwing
Shadow data distortion, the image reconstructed just will appear artifact.In by imaging object only single metal when, metal artifacts performance
For there are radial artifacts around it；When metal object quantity is more and metallic region area is larger, metal artifacts not only include
Radial artifact around metal object, and there is also banding artifacts on metal connecting line direction.
Since CT technology is widely used in medical treatment and industrial circle, the quality of image of reconstruction image is for checking, detecting
Effect is most important.It is suggested at present there are many correcting algorithm, the method for metal artifacts reduction can be divided into three in short
Class: physical correction, the interpolation and iterative reconstruction of projection domain.Physical correction method between X radiographic source and detector by adding
One filter absorbs the lower energy photon of light beam, to reduce artifact；But as the photon of part is absorbed, signal-to-noise ratio
It can reduce.Linear interpolation (LI) is a kind of conventional method for reducing metal artifacts, replaces Raw projection data using interpolation technique
The middle part influenced by metal, thus achieve the purpose that remove metal artifacts, but this method is easy to generate new artifact,
Cause the serious portion structure deformation of metal artifacts.Iterative approximation (LR) is the conventional method of Clinical CT, and this method is due to needing
A large amount of to calculate, the performance by experimental facilities is limited.
In conclusion existing hardening artifact bearing calibration is although various, but there is respective limitation in a particular application
Property.
Summary of the invention
Goal of the invention: it is an object of the invention to solve the deficiencies in the prior art, a kind of Cranial Computed Tomography image is provided
Metal artifacts reduction method, the present invention using based on depth convolutional neural networks residual error study, remove Cranial Computed Tomography image in
Metal artifacts, improve calculating speed.
A kind of technical solution: the metal artifacts reduction method of Cranial Computed Tomography image of the invention, comprising the following steps:
(1) database for CNN training is established；Original image is pre-processed using deformable image registration method,
And then the Cranial Computed Tomography data comprising artifact and without artifact are obtained to the database of composition；
(2) 17 layers of CNN framework are constructed in CNN training to learn metal artifacts.
Further, the detailed content of the step (1) are as follows:
It (1.1) will include that the CT picture of artifact is set as target image, according to SSIM from the CT image for not including artifact
It is worth preliminary screening and goes out the image most like with target image；
(1.2) using image obtained by deformable image registration method processing step (1.1), and then final pairing instruction is obtained
Practice data, method particularly includes:
(a) according to the clearly image of multiple groups obtained by step (1.1), feature extraction is carried out；By the input of original image 100*100
Image segmentation extracts the label of characteristic information at the image block of 30*30 from image block；
(b) according to the feature tag extracted, by the image segmentation containing artifact at the progress of an equal amount of image block
Match；
(c) image block matched is merged into complete image, carried out image procossing (i.e. picture enhancing), final
Clearly image after to registration.
In the above process, input picture is registrated according to the target image containing artifact, selects registration effect
Fruit is best, that is, takes out after registration closest to the data of target image, forms one group with the original image containing artifact
Data pair.In short, eliminating the slight displacement in CT image in mini piece between pixel using deformable image registration method.?
After completing registration, image data gets out be sent to training in network.
Further, the detailed process of the step (2) are as follows:
(2.1) construct convolutional neural networks: the depth of the convolutional neural networks is D layers, including three kinds of network layers: (A) volume
Product network+amendment linear unit (ReLU)；(B) convolutional network+crowd standardization (BN)+ReLU；(C) convolutional network；
In Type (A), the filter that 64 sizes are 3 × 3 × 1 is for generating 64 Feature Mappings；This layer is added
ReLU is pre-processed, i.e., using Element-Level operation will all pixels value in sketch map less than 0 be set as 0；
Type (B) be from the second layer to D-1 layer, using 64 having a size of 3 × 3 × 64 filter；In convolutional network and
Normalization unit BN is criticized in introducing on the basis of correcting linear unit, and BN is added between convolution sum ReLU；Number will be trained using BN
According to thoroughly upsetting, and training speed is greatly improved, allows bigger learning rate；
Type (C) is convolutional layer, the filter that convolutional layer is 3 × 3 × 64 using size, for rebuilding output；
(2.2) residual error learn, i.e., using CNN network residual error study analysis CT image in artifact, it is intended to by original image with
From the middle school's acquistion of CT image to artifact subtract each other, finally obtain and clearly correct image；
The model of CT image is f=x+n, and wherein the artifact in n representative image, a mapping function is trained using CNN
Then F (f)=n obtains the image without artifact that can be estimatedIt is instructed using the difference and network of reconstruction image
The Averaged Square Error of Multivariate between artifact after white silk needs to restore the loss function of image as measurement；Formula is as follows:
WhereinIt indicates N group data pair, is made of true picture and the image containing artifact；μ (f) represents CNN
Mapping.In order to ensure the quality of the image recovered, it is necessary to the parameter in training network, so that parameter value appropriate is obtained,
Mean square error is minimized to obtain.
The artifacts obtained by CNN network, according to formulaObtain final clearly image.
The utility model has the advantages that the present invention initially sets up one includes and the database not comprising artifact Cranial Computed Tomography image.In the step
In rapid, using the method for deformable image registration before CNN training preprocessed data.Secondly, in CNN training, building one
A simple 17 layers of CNN framework learns metal artifacts.The experimental results showed that this method has good metal artifacts reduction
Ability, PSNR and SSIM also show that apparent improvement.
Detailed description of the invention
Fig. 1 is overall flow figure of the invention.
Fig. 2 is the process schematic of deformable registration method in the present invention.
Fig. 3 is the structure chart of depth convolutional neural networks in embodiment.
Fig. 4 is in embodiment using the process of CNN network removal artifact.
Fig. 5 is that artifact process schematic in CT image is corrected in embodiment.
Fig. 6 is four kinds of comparative result schematic diagrams in embodiment after different artifact correction methods.
Specific embodiment
Technical solution of the present invention is described in detail below, but protection scope of the present invention is not limited to the reality
Apply example.
Due to the presence of metal charge, metal artifacts always affect the effect of computed tomography (CT) inspection.Cause
This, removal metal artifacts are still one of the main problem in clinical Cranial Computed Tomography image.In order to reduce the tooth regions of CT image
In metal artifacts, as shown in Figures 1 to 4, the metal artifacts reduction method of a kind of Cranial Computed Tomography image of invention, including following
Step:
(1) database for CNN training is established；Original image is pre-processed using deformable image registration method,
And then the Cranial Computed Tomography data comprising artifact and without artifact are obtained to the database of composition；
It (1.1) will include that the CT picture of artifact is set as target image, according to SSIM from the CT image for not including artifact
It is worth preliminary screening and goes out the image most like with target image (by comparing the size of the SSIM value of image, from the CT for being free of artifact
In image, preliminary screening goes out most like with target image)；
(1.2) using image obtained by deformable image registration method processing step (1.1), and then final pairing instruction is obtained
Practice data；Method particularly includes:
(a) according to the clearly image of multiple groups obtained by step (1.1), feature extraction is carried out；By the input of original image 100*100
Image segmentation extracts the label of characteristic information at the image block of 30*30 from image block；
(b) according to the feature tag extracted, by the image segmentation containing artifact at the progress of an equal amount of image block
Match；
(c) image block matched is merged into complete image, carried out image procossing (i.e. picture enhancing), final
Clearly image after to registration；
(2) 17 layers of CNN framework are constructed in CNN training to learn metal artifacts；
(2.1) construct convolutional neural networks: the depth of the convolutional neural networks is D layers, including three kinds of network layers: (A) volume
Product network+amendment linear unit (ReLU)；(B) convolutional network+crowd standardization (BN)+ReLU；(C) convolutional network；
In Type (A), the filter that 64 sizes are 3 × 3 × 1 is for generating 64 Feature Mappings；This layer is added
ReLU is pre-processed, i.e., using Element-Level operation will all pixels value in sketch map less than 0 be set as 0；
Type (B) be from the second layer to D-1 layer, using 64 having a size of 3 × 3 × 64 filter；In convolutional network and
Normalization unit BN is criticized in introducing on the basis of correcting linear unit, and BN is added between convolution sum ReLU；Number will be trained using BN
According to thoroughly upsetting, and training speed is greatly improved, allows bigger learning rate；
Type (C) is convolutional layer, the filter that convolutional layer is 3 × 3 × 64 using size, for rebuilding output；
(2.2) residual error learn, i.e., using CNN network residual error study analysis CT image in artifact, it is intended to by original image with
From the middle school's acquistion of CT image to artifact subtract each other, finally obtain and clearly correct image；
The model of CT image is f=x+n, and wherein the artifact in n representative image, a mapping function is trained using CNN
Then F (f)=n obtains the image without artifact that can be estimatedIt is instructed using the difference and network of reconstruction image
The Averaged Square Error of Multivariate between artifact after white silk needs to restore the loss function of image as measurement；Formula is as follows:
WhereinIt indicates N group data pair, is made of true picture and the image containing artifact；μ (f) represents CNN
Mapping.In order to ensure the quality of the image recovered, it is necessary to the parameter in training network, so that parameter value appropriate is obtained,
Mean square error is minimized to obtain.
The artifacts obtained by CNN network, according to formulaObtain final clearly image.
The training parameter setting of above-described embodiment is as follows:
It is trained using Adam optimizer, learning rate is set as 1e-4 to 1e-5.Epoch in network is set as
190, the size of input picture is set as 100 × 100.MatConvNet work in this Web vector graphic MATLAB 2017a environment
Have case, is realized using GTX 1080Ti graphics processor and i7-6850K CPU (3.60HZ).
The speed for being accelerated training data using GPU is enabled the training time of CNN shorten within one day, improves study
Efficiency.Simultaneously by test, proof of algorithm is carried out using CPU, every picture is detected and needs one minute or so time, and is made
In the case where GPU, detection every only needs one second, and speed is obviously improved.
In the above process, learn the Cranial Computed Tomography figure containing artifact and without artifact in the present invention using residual error learning method
Mapping as between, the method for residual error study focus on residual image (train difference) between only comprising artifact and
Noise, not comprising other significant anatomical structures.So (containing and without belonging to metal artifacts by two CT images
) it is used as training to, needing to carry out registration work before, because registration can be minimized the structural difference of training pair.
For the tooth regions in Cranial Computed Tomography image, image is further registrated using the method for deformable image registration.
As shown in Fig. 2, input picture is registrated according to the target image containing artifact, it is best to select registration effect,
It is exactly the data taking-up after being registrated closest to target image, forms one group of data pair with the original image containing artifact.Always
It, deformable image registration method attempts to eliminate the slight displacement in CT image in mini piece between pixel.Complete registration
Afterwards, image data gets out be sent to training in network.
Embodiment 1:
For experimental result as shown in figure 5, the left side is the input picture with artifact, the right is by metal artifacts reduction
Export image, it can be seen that the artifact in CT image has obtained good correction.
Meanwhile the present embodiment uses beam hardening correcting (BHC) method and linear interpolation (LI) method as comparing.Fig. 6 exhibition
Shown using four kinds of comparison results after different artifact correction methods (display window is [- 1,000 1000]).Fig. 6 (a) be using
The reference picture obtained after the method for deformable image registration, Fig. 6 (b) are the original images containing artifact；Fig. 6 (c) be using
Correction image Fig. 6 (d) after BHC method is the correction image using LI method；Fig. 6 (e) is corrected using our method
Image.
Table 1 lists the SSIM value in Fig. 6 (c)-(e) about reference Fig. 5.Table 2 lists correction image relative to reference
The PSNR of image.In contrast, there is highest accuracy using the CNN image obtained after the present invention and realizes maximum
PSNR value.
Table 1
Table 2
In short, the present invention proposes a kind of artifact correction algorithm based on depth convolutional neural networks, the DIR method of use
The use of the CNN network architecture of 17 layers of cooperation can effectively remove artifact, and accelerate the speed of training data using GPU
Degree, improves the learning efficiency of network；Meanwhile the experimental results showed that this method have good metal artifacts reduction ability,
PSNR and SSIM also shows that apparent improvement.
Claims (3)
1. a kind of metal artifacts reduction method of Cranial Computed Tomography image, it is characterised in that: the following steps are included:
(1) database for CNN training is established；Original image is pre-processed using deformable image registration method, in turn
The Cranial Computed Tomography data comprising artifact and without artifact are obtained to the database of composition；
(2) 17 layers of CNN framework are constructed in CNN training to learn metal artifacts.
2. the metal artifacts reduction method of Cranial Computed Tomography image according to claim 1, it is characterised in that: the step (1)
Detailed content are as follows:
It (1.1) will include that the CT picture of artifact is set as target image, according at the beginning of SSIM value from the CT image for not including artifact
Step filters out the image most like with target image；
(1.2) using image obtained by deformable image registration method processing step (1.1), and then final pairing training number is obtained
According to, method particularly includes:
(a) according to the clearly image of multiple groups obtained by step (1.1), feature extraction is carried out；By the input picture of original image 100*100 point
It is cut into the image block of 30*30, the label of characteristic information is extracted from image block；
(b) according to the feature tag extracted, the image segmentation containing artifact is matched at an equal amount of image block；
(c) image block matched is merged into complete image, carries out image procossing, the clearly figure after finally obtaining registration
Picture.
3. the metal artifacts reduction method of Cranial Computed Tomography image according to claim 1, it is characterised in that: the step (2)
Detailed process are as follows:
(2.1) construct convolutional neural networks: the depth of the convolutional neural networks is D layers, including three kinds of network layers: (A) convolution net
Network+amendment linear unit (ReLU)；(B) convolutional network+crowd standardization (BN)+ReLU；(C) convolutional network；
In Type (A), the filter that 64 sizes are 3 × 3 × 1 is for generating 64 Feature Mappings；This layer be added ReLU into
Row pretreatment, i.e., using Element-Level operation will all pixels value in sketch map less than 0 be set as 0；
Type (B) be from the second layer to D-1 layer, using 64 having a size of 3 × 3 × 64 filter；In convolutional network and amendment
Crowd normalization unit BN is introduced on the basis of linear unit, and BN is added between convolution sum ReLU；It is using BN that training data is thorough
Upset at bottom；
Type (C) is convolutional layer, the filter that convolutional layer is 3 × 3 × 64 using size, for rebuilding output；
(2.2) residual error learns, i.e., using the artifact in the residual error study analysis CT image of CNN network；
The model of CT image is f=x+n, and wherein the artifact in n representative image, a mapping function F (f) is trained using CNN
=n, the image without artifact then estimatedUtilize the artifact after the difference and network training of reconstruction image
Between Averaged Square Error of Multivariate, need to restore the loss function of image as measurement；Formula is as follows:
WhereinIt indicates N group data pair, is made of true picture and the image containing artifact；μ (f) represents reflecting for CNN
It penetrates；
The artifacts obtained by CNN network, according to formulaObtain final clearly image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910136016.8A CN109816747A (en) | 2019-02-25 | 2019-02-25 | A kind of metal artifacts reduction method of Cranial Computed Tomography image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910136016.8A CN109816747A (en) | 2019-02-25 | 2019-02-25 | A kind of metal artifacts reduction method of Cranial Computed Tomography image |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109816747A true CN109816747A (en) | 2019-05-28 |
Family
ID=66607331
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910136016.8A Pending CN109816747A (en) | 2019-02-25 | 2019-02-25 | A kind of metal artifacts reduction method of Cranial Computed Tomography image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109816747A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111754436A (en) * | 2020-06-24 | 2020-10-09 | 上海联影医疗科技有限公司 | Acceleration method for medical image artifact correction, computer device and storage medium |
| CN113706643A (en) * | 2020-09-09 | 2021-11-26 | 南京邮电大学 | Homomorphic adaptation learning-based head CT metal artifact correction method |
| CN113744155A (en) * | 2021-09-04 | 2021-12-03 | 重庆大学 | Rock and ore sample CT image metal artifact correction method based on triple convolution network |
-
2019
- 2019-02-25 CN CN201910136016.8A patent/CN109816747A/en active Pending
Non-Patent Citations (1)
| Title |
|---|
| XIA HUANG等: "Metal artifact reduction on cervical CT images by deep residual learning", 《BIOMEDICAL ENGINEERING ONLINE》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111754436A (en) * | 2020-06-24 | 2020-10-09 | 上海联影医疗科技有限公司 | Acceleration method for medical image artifact correction, computer device and storage medium |
| CN111754436B (en) * | 2020-06-24 | 2024-05-03 | 上海联影医疗科技股份有限公司 | Acceleration method for medical image artifact correction, computer device and storage medium |
| CN113706643A (en) * | 2020-09-09 | 2021-11-26 | 南京邮电大学 | Homomorphic adaptation learning-based head CT metal artifact correction method |
| CN113706643B (en) * | 2020-09-09 | 2023-06-30 | 南京邮电大学 | Head CT metal artifact correction method based on homomorphic adaptation learning |
| CN113744155A (en) * | 2021-09-04 | 2021-12-03 | 重庆大学 | Rock and ore sample CT image metal artifact correction method based on triple convolution network |
| CN113744155B (en) * | 2021-09-04 | 2023-09-26 | 重庆大学 | Triple convolution network-based rock-mineral sample CT image metal artifact correction method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20240037732A1 (en) | Method for enhancing quality and resolution of ct images based on deep learning | |
| CN112396672B (en) | Sparse angle cone-beam CT image reconstruction method based on deep learning | |
| Wang et al. | FBP-Net for direct reconstruction of dynamic PET images | |
| CN103150744B (en) | A kind of X ray multi-power spectrum CT data for projection process and image rebuilding method | |
| CN105118066B (en) | A kind of low-dose CT picture breakdown method based on three-dimensional distinctiveness character representation | |
| CN111292386B (en) | CT projection metal trace complement metal artifact correction method based on U-net | |
| US10489940B2 (en) | System and computer-implemented method for improving image quality | |
| CN104751429B (en) | A kind of low dosage power spectrum CT image processing methods based on dictionary learning | |
| CN112348936A (en) | Low-dose cone-beam CT image reconstruction method based on deep learning | |
| CN109816747A (en) | A kind of metal artifacts reduction method of Cranial Computed Tomography image | |
| CN104408758A (en) | Low-dose processing method of energy spectrum CT image | |
| CN117036162B (en) | Residual feature attention fusion method for super-resolution of lightweight chest CT image | |
| CN111798535B (en) | CT image enhancement display method and computer readable storage medium | |
| CN109903356A (en) | Missing CT data for projection estimation method based on the multiple parsing network of depth | |
| CN111161182B (en) | MR structure information constrained non-local mean guided PET image partial volume correction method | |
| Feng et al. | Dual residual convolutional neural network (DRCNN) for low-dose CT imaging | |
| EP4071720A1 (en) | System and method for utilizing a deep learning network to correct for a bad pixel in a computed tomography detector | |
| CN116091636A (en) | Incomplete data reconstruction method for X-ray differential phase contrast imaging based on dual-domain enhancement | |
| CN113436118B (en) | Low-dose CT image restoration method based on multi-scale convolutional coding network | |
| WO2023000244A1 (en) | Image processing method and system, and application of image processing method | |
| Mangalagiri et al. | Toward generating synthetic CT volumes using a 3D-conditional generative adversarial network | |
| Lin et al. | DdeNet: A dual-domain end-to-end network combining Pale-Transformer and Laplacian convolution for sparse view CT reconstruction | |
| CN116245969A (en) | Low-dose PET image reconstruction method based on deep neural network | |
| Li et al. | A cascade-based dual-domain data correction network for sparse view CT image reconstruction | |
| CN113744356B (en) | Low-dose SPECT chord graph recovery and scattering correction method |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190528 |