CN109886909A - A method of PET-CT image is synthesized based on CT image - Google Patents
A method of PET-CT image is synthesized based on CT image Download PDFInfo
- Publication number
- CN109886909A CN109886909A CN201910119941.XA CN201910119941A CN109886909A CN 109886909 A CN109886909 A CN 109886909A CN 201910119941 A CN201910119941 A CN 201910119941A CN 109886909 A CN109886909 A CN 109886909A
- Authority
- CN
- China
- Prior art keywords
- image
- network
- pet
- synthesis
- true
- 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
Landscapes
- Apparatus For Radiation Diagnosis (AREA)
- Nuclear Medicine (AREA)
Abstract
The present invention relates to a kind of methods based on CT image synthesis PET-CT image, the following steps are included: 1) building circulation generates confrontation network, forward circulation channel A and recycled back channel B composition including constituting a circulation, each circulation canal is a GAN network, and forward circulation channel A includes generating network G1With differentiation network D1, recycled back channel B includes generating network G2With differentiation network D2, the forward circulation channel A is to realize that CT image synthesizes PET-CT image, feedback of the recycled back channel B to realize PET-CT image synthesis CT image;2) confrontation network implementations CT image is generated according to the circulation of building and synthesizes PET-CT image.Compared with prior art, the present invention has many advantages, such as that training system is more stable, it is higher to generate the quality of image.
Description
Technical field
The present invention relates to medical images to synthesize field, more particularly, to a kind of side based on CT image synthesis PET-CT image
Method.
Background technique
With the development of medicine, computer technology and biotechnology, medical image provides a variety of for clinical diagnosis
The medical image of mode, such as CT (CR scanning), MRI (Magnetic resonance imaging), ultrasound image, PET (positive electron
Emission computed tomography), PET-CT, PET-MRI etc..Different medical images provides the different information of related internal organs,
But people are more concerned about the imaging effect (such as sensitivity, clarity, accuracy) of medical image.
CT is higher to the density resolution of tissue, can observe directly the tumour inside parenchymal viscera, provide
The accurate Anatomical orientation of lesion.PET then provides the molecular informations such as the detailed function of lesion and metabolism.And PET-CT is by CT and PET
It combines together, audit function is more powerful, and performance is more preferable.For CT, the sensitivity of PET-CT detection is higher, adds
The judgement of metabolic cost, the organ disease treatment and the good pernicious judgement of its lesion region of more early stage that facilitates patient, is convenient for doctor
Formulate targeted treatment schemes.
However the popularity rate of PET-CT Medical Devices is lower both at home and abroad, somewhat expensive, therefore, CT accurately synthesize PET-CT
With far reaching significance.
Summary of the invention
It is an object of the present invention to overcome the above-mentioned drawbacks of the prior art and provide one kind to be closed based on CT image
At the method for PET-CT image.
The purpose of the present invention can be achieved through the following technical solutions:
A method of PET-CT image is synthesized based on CT image, comprising the following steps:
1) least square circulation generation confrontation network is constructed, including one forward circulation channel A recycled of composition and reversely
Circulation canal B composition, each circulation canal are a GAN network, and forward circulation channel A includes generating network G1With differentiation network
D1, recycled back channel B includes generating network G2With differentiation network D2, the forward circulation channel A is to realize that CT image closes
At PET-CT image, feedback of the recycled back channel B to realize PET-CT image synthesis CT image;
2) it is recycled according to the least square of building and generates confrontation network implementations CT image synthesis PET-CT image.
The step 2) specifically includes the following steps:
21) in forward circulation channel, network G is generated1True CT image in the mode S of source is converted into target as far as possible
Synthesis PET-CT image in mode T, while differentiating network D1Being differentiated to synthesis PET-CT image (judges it from conjunction
At image, or come from true PET-CT image), judge that image is composograph as far as possible.Generate network as much as possible
Synthesis differentiates that network judges that it is synthesis PET-CT image as much as possible close to true PET-CT image, rather than true
PET-CT image.The mutual game of the two, final realize generate network synthesis close to true PET-CT image, differentiate that network is sentenced
Can not breaking, it is synthesis PET-CT image or true PET-CT image.And definition generates network G respectively1With differentiation network D1's
Loss function LGAN1(G1) and LGAN1(D1);
22) the PET-CT image of synthesis is sent into and generates network G2, it is converted into the synthesis CT image of source mode as far as possible, together
When differentiate network D2Synthesis CT image is differentiated and (judges that it comes from composograph, or true CT image), to the greatest extent
It may judge that image is composograph.It generates network to be synthesized as much as possible close to true CT image, and differentiates network and to the greatest extent may be used
Energy ground judges that it is synthesis CT image, rather than true CT image.The mutual game of the two, it is final to realize that generating network synthesis connects
Nearly true CT image differentiates that network can not judge that it is synthesis CT image or true CT image.And definition generates net respectively
Network G2With differentiation network D2Loss function LGAN2(G2) and LGAN2(D2);
23) it in order to minimize true CT image and the difference for recycling composograph in forward circulation channel, defines forward direction and follows
Ring loss function LcycleA(S,T)。
24) in recycled back channel, network G is generated2True PET-CT image in target modalities T is converted as far as possible
At the synthesis CT image in the mode S of source, while differentiating network D2Synthesis CT image is differentiated and (judges that it comes from synthesis
Image, or true CT image), judge that image is composograph as far as possible.Network is generated to be synthesized as much as possible close to true
Real CT image, and differentiate that network judges that it is synthesis CT image as much as possible, rather than true CT image.The two is mutually rich
It plays chess, it is final to realize that generating network synthesizes close to true CT image, differentiate that network judge not Chu that it is to synthesize CT image or very
Real CT image.And definition generates network G respectively2With differentiation network D2Loss function LGAN2(G2) and LGAN2(D2);
25) synthesis CT image is sent into and generates network G1, it is converted into the synthesis PET-CT image of target modalities, is differentiated simultaneously
Network D1Synthesis PET-CT image is differentiated and (judges that it schemes from composograph, or from true PET-CT
Picture), judge that image is composograph as far as possible.It generates network to be synthesized as much as possible close to true PET-CT image, and sentences
Other network judges that it is synthesis PET-CT image as much as possible, rather than true PET-CT image.The mutual game of the two, finally
It realizes that generating network synthesizes close to true PET-CT image, differentiates that network judge not Chu that it is to synthesize PET-CT image or very
Real PET-CT image.And definition generates network G respectively1With differentiation network D1Loss function LGAN1(G1) and LGAN1(D1);
26) anti-in order to minimize true PET-CT image and the difference for recycling composograph, definition in recycled back channel
To circulation loss function LcycleB(T,S)。
27) total losses function L is definedtotal, and circuit training is carried out, it is exported finally when total losses function minimization
PET-CT image.
The step 21) and 25) in, loss function LGAN1(G1) and LGAN1(D1) expression formula be respectively as follows:
Wherein, x is the true picture in the mode of source, and y is the true picture in target modalities, and p (x) is the probability that x is obeyed
Distribution, p (y) be y obey probability distribution, G () make a living networking complexing image, D () be differentiation network differentiation it is general
Rate, E [] are mathematical expectation.
The step 22) and 24) in, loss function LGAN2(G2) and LGAN2(D2) expression formula be respectively as follows:
The step 23) and 26) in, loss function LcycleA(S, T) and LcycleBThe expression formula of (T, S) is respectively as follows:
LcycleA(S, T)=EX~A[||G2(G1(x))-x||1]
LcycleB(T, S)=Ey~B[||G1(G2(y))-y||1]
Wherein, | | | |1For single order norm.
The total losses function LtotalExpression formula are as follows:
Ltotal=LGAN1(G1)+LGAN1(D1)+LGAN2(G2)+
LGAN2(D2)+λ1LcycleA(S,T)+λ2LcycleB(T,S)
Wherein, λ1And λ2For the weight coefficient of corresponding loss function.
Compared with prior art, the invention has the following advantages that
Least square loss function and circulation pattern are combined by the present invention, in existing technology, are only unilaterally examined
Consider a certain generate and fight latticed form, be not combined the two advantage, least square loss function is better than generating now
Cross entropy loss function general in network is fought, least square loss function is essentially consisted in and only reaches saturation on one point, thus
So that entire training system is more stable, along with the feedback element in circulation pattern, increases input and generate confrontation network
Feedback information so that generate image quality it is higher, the two advantage is combined by the present invention, is followed using least square
Ring generates the purpose that confrontation network finally realizes CT image synthesis PET-CT image.
Detailed description of the invention
Fig. 1 is medical image synthesis flow schematic diagram of the invention.
Fig. 2 is the comparison of true CT image and synthesis PET-CT image of the invention.
Specific embodiment
The present invention is described in detail with specific embodiment below in conjunction with the accompanying drawings.
The problems such as in view of being directed to its popularity rate where the advantage of PET-CT, it is an object of the invention to pass through minimum
Two, which multiply circulation, generates confrontation network to realize that CT image synthesizes PET-CT image.
Technical scheme is as follows:
Least square circulation generates confrontation network and is mainly made of two GAN networks.It is logical that it corresponds respectively to forward circulation
Road A and recycled back channel B, two channels constitute a circulation.Wherein containing there are two generate network G1、G2With two differentiation nets
Network D1、D2.Forward channel is mainly responsible for CT synthesis PET-CT, and backward channel is mainly responsible for the feedback of PET-CT synthesis CT, so that
Forward channel is more accurate true from CT synthesis PET-CT image.
(1) in forward circulation channel, network G is generated1True CT in the mode S of source is converted into target modalities T
PET-CT is synthesized, while differentiating network D1Synthesis PET-CT image is differentiated, judges that it comes from composograph, still
True PET-CT image.To reach optimum efficiency, realizes and generate network synthesis close to true PET-CT image, differentiate network
It can not judge that it is synthesis PET-CT image or true PET-CT image, and then minimize and generate network G1With differentiation network D1
Loss function, it is as follows:
Wherein, x is the true picture in the mode of source, and y is the true picture in target modalities.
Then synthesis PET-CT is re-fed into generation network G2, it is converted into the synthesis CT of source mode, while differentiating network D2It is right
Synthesis CT image is differentiated, judges that it comes from composograph, or true CT image.To keep composograph closer
True picture, and then minimize and generate network G2With differentiation network D2Loss function, it is as follows:
At the same time, the difference for minimizing true CT image and circulation composograph in forward circulation channel, increases one
A forward circulation loss function, as follows:
LcycleA(S, T)=EX~A[||G2(G1(x))-x||1]
(2) in recycled back channel, network G is generated2True PET-CT in target modalities T is converted into source mode S
In synthesis CT, while differentiating network D2Synthesis CT image is differentiated, judges that it comes from composograph, or true
CT image.To reach optimum efficiency, make composograph closer to true picture, and then minimize and generate network G2With differentiation net
Network D2Loss function LGAN2(G2) and LGAN2(D2)。
Then synthesis CT is re-fed into generation network G1, it is converted into the synthesis PET-CT of target modalities, while differentiating network D1
Synthesis PET-CT image is differentiated, judges that it comes from composograph, or true PET-CT image.To make to synthesize
Image minimizes closer to true picture and generates network G1With differentiation network D1Loss function LGAN1(G1) and LGAN1
(D1)。
At the same time, the difference for minimizing true PET-CT image and circulation composograph in recycled back channel, increases
One recycled back loss function, as follows:
LcycleB(T, S)=EY~B[||G1(G2(y))-y||1]
Total losses function is constituted finally, four loss functions of definition are added, and distributes different weight coefficients, it is minimum
Change total loss function, as follows:
Ltotal=LGAN1(G1)+LGAN1(D1)+LGAN2(G2)+
LGAN2(D2)+λ1LcycleA(S,T)+λ2LcycleB(T,S)
Wherein, λ1And λ2For the weight coefficient of corresponding loss function, can rule of thumb set.
Forward circulation channel and recycled back channel form entire circulation, by constantly circuit training, eventually by following
Ring generates confrontation network implementations CT image and synthesizes accurate PET-CT image.
Embodiment:
As shown in FIG. 1, FIG. 1 is preferred embodiment medical image synthesis flow schematic diagram of the invention, dot dotted line is forward direction
Circulation canal, short-term dotted line are recycled back channel, and the two constitutes entire circuit training system.In addition, solid line correspond to it is each
Partial loss function.This experiment is established on Tensor Flow workbench, carries out weight to the true CT and PET-CT of acquisition
256 × 256 pixels newly are sampled, then carry out circuit training according to attached drawing 1.
In forward circulation channel, network G is generated1True CT in the mode of source is converted into the synthesis in target modalities
PET-CT, while differentiating network D1Synthesis PET-CT image is differentiated, judges that it comes from composograph, or true
PET-CT image.To reach optimum efficiency, make composograph closer to true picture, and then minimize and generate network G1With sentence
Other network D1Loss function LGAN1(G1) and LGAN1(D1)。
Then synthesis PET-CT is re-fed into generation network G2, it is converted into the synthesis CT of source mode, while differentiating network D2It is right
Synthesis CT image is differentiated, judges that it comes from composograph, or true CT image.To keep composograph closer
True picture, and then minimize and generate network G2With differentiation network D2Loss function LGAN2(G2) and LGAN2(D2)。
At the same time, the difference for minimizing true CT image and circulation composograph in forward circulation channel, increases one
A forward circulation loss function LcycleA(S,T)。
In recycled back channel, network G is generated2The conjunction true PET-CT in target modalities being converted into the mode of source
At CT, while differentiating network D2Synthesis CT image is differentiated, judges that it comes from composograph, or true CT figure
Picture.To reach optimum efficiency, make composograph closer to true picture, and then minimize and generate network G2With differentiation network D2's
Loss function LGAN2(G2) and LGAN2(D2)。
Then synthesis CT is re-fed into generation network G1, it is converted into the synthesis PET-CT of target modalities, while differentiating network D1
Synthesis PET-CT image is differentiated, judges that it comes from composograph, or true PET-CT image.To make to synthesize
Image minimizes closer to true picture and generates network G1With differentiation network D1Loss function LGAN1(G1) and LGAN1
(D1)。
At the same time, the difference for minimizing true PET-CT image and circulation composograph in recycled back channel, increases
One recycled back loss function LcycleB(T,S)。
Total losses function is constituted finally, six loss functions of definition are added, and distributes different weight coefficients, it is minimum
Change total loss function Ltotal。
Forward circulation channel and recycled back channel form entire circulation, by constantly circuit training, eventually by following
Ring generates confrontation network implementations CT image and synthesizes accurate PET-CT image.
As shown in Fig. 2, figure (2a) is true CT image, figure (2b) is that least square circulation generates confrontation network synthesis
PET-CT effect picture.As seen from the figure, common CT image is lower to the tumor tissues sensitivity that more early stage is small, does not see
Lesion region can not judge whether it is tumor tissues, be unfavorable for the organ disease treatment of patient.And it is imitated in the PET-CT of synthesis
On fruit figure, tumor region is more obvious, facilitates the diagnosis of doctor, the treatment for more early stage that is more advantageous to patient.
Since in actual PET-CT image, tumor tissues metabolism is vigorous, a large amount of metabolite PET-CT is taken in
Tracer, the aggregation for being rendered as tracer is dense poly-, so the lesion for having accumulated a large amount of tracers is easier to be found.And it is of the invention
Least square circulation generate confrontation network, exactly learnt mapping relations therein, to realize true PET-CT image
Effect.
Claims (6)
1. a kind of method based on CT image synthesis PET-CT image, which comprises the following steps:
1) building circulation generates confrontation network, forward circulation channel A and recycled back channel B group including constituting a circulation
At each circulation canal is a GAN network, and forward circulation channel A includes generating network G1With differentiation network D1, recycled back
Channel B includes generating network G2With differentiation network D2, the forward circulation channel A is to realize CT image synthesis PET-CT figure
Picture, feedback of the recycled back channel B to realize PET-CT image synthesis CT image;
2) confrontation network implementations CT image is generated according to the circulation of building and synthesizes PET-CT image.
2. a kind of method based on CT image synthesis PET-CT image according to claim 1, which is characterized in that described
Step 2) specifically includes the following steps:
21) in forward circulation channel, network G is generated1True CT image in the mode S of source is converted into the conjunction in target modalities T
At PET-CT image, while differentiating network D1To synthesis PET-CT image differentiate, judge its from composograph, still
From true PET-CT image, and respectively, definition generates network G1With differentiation network D1Loss function LGAN1(G1) and LGAN1
(D1);
22) the PET-CT image of synthesis is sent into and generates network G2, it is converted into the synthesis CT image of source mode, while differentiating network
D2Synthesis CT image is differentiated, judges that it comes from composograph, or true CT image, and definition generates respectively
Network G2With differentiation network D2Loss function LGAN2(G2) and LGAN2(D2);
23) it is the difference for minimizing true CT image and circulation composograph in forward circulation channel, defines forward circulation loss
Function LcycleA(S,T);
24) in recycled back channel, network G is generated2True PET-CT image in target modalities T is converted into the mode S of source
Synthesis CT image, while differentiating network D2Synthesis CT image is differentiated, judges that it, from composograph, still comes from
True CT image, and definition generates network G respectively2With differentiation network D2Loss function LGAN2(G2) and LGAN2(D2);
25) synthesis CT image is sent into and generates network G1, it is converted into the synthesis PET-CT image of target modalities, while differentiating network
D1Synthesis PET-CT image is differentiated, judges that it comes from composograph, or true PET-CT image, and respectively
Definition generates network G1With differentiation network D1Loss function LGAN1(G1) and LGAN1(D1);
26) it is reversely followed to minimize true PET-CT image and the difference for recycling composograph, definition in recycled back channel
Ring loss function LcycleB(T,S)。
27) total losses function L is definedtotal, and circuit training is carried out, final PET-CT is exported when total losses function minimization
Image.
3. a kind of method based on CT image synthesis PET-CT image according to claim 2, which is characterized in that described
Step 21) and 25) in, loss function LGAN1(G1) and LGAN1(D1) expression formula be respectively as follows:
Wherein, x is the true picture in the mode of source, and y is the true picture in target modalities, and p (x) is the probability distribution that x is obeyed,
P (y) be y obey probability distribution, G () make a living networking complexing image, D () be differentiation network differentiation probability, E
[] is mathematical expectation.
4. a kind of method based on CT image synthesis PET-CT image according to claim 3, which is characterized in that described
Step 22) and 24) in, loss function LGAN2(G2) and LGAN2(D2) expression formula be respectively as follows:
5. a kind of method based on CT image synthesis PET-CT image according to claim 2, which is characterized in that described
Step 23) and 26) in, loss function LcycleA(S, T) and LcycleBThe expression formula of (T, S) is respectively as follows:
LcycleA (S, T)=EX~A[||G2(G1(x))-x||1]
LcycleB (T, S)=EY~B[||G1(G2(y))-y||1]
Wherein, | | | |1For single order norm.
6. a kind of method based on CT image synthesis PET-CT image according to claim 5, which is characterized in that described
Total losses function LtotalExpression formula are as follows:
Ltotal=LGAN1(G1)+LGAN1(D1)+LGAN2(G2)+
LGAN2(D2)+λ1LcycleA(S,T)+λ2LcycleB(T,S)
Wherein, λ1And λ2For the weight coefficient of corresponding loss function.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910119941.XA CN109886909A (en) | 2019-02-18 | 2019-02-18 | A method of PET-CT image is synthesized based on CT image |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910119941.XA CN109886909A (en) | 2019-02-18 | 2019-02-18 | A method of PET-CT image is synthesized based on CT image |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109886909A true CN109886909A (en) | 2019-06-14 |
Family
ID=66928255
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910119941.XA Pending CN109886909A (en) | 2019-02-18 | 2019-02-18 | A method of PET-CT image is synthesized based on CT image |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109886909A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111178502A (en) * | 2019-12-12 | 2020-05-19 | 中国科学院深圳先进技术研究院 | Training method and system for circularly generating confrontation network model and electronic equipment |
CN116502701A (en) * | 2023-06-29 | 2023-07-28 | 合肥锐世数字科技有限公司 | Attenuation correction method and device, training method and device, imaging method and system |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107220929A (en) * | 2017-06-23 | 2017-09-29 | 深圳市唯特视科技有限公司 | A kind of non-paired image method for transformation using the consistent confrontation network of circulation |
CN108229349A (en) * | 2017-12-21 | 2018-06-29 | 中国科学院自动化研究所 | Reticulate pattern facial image identification device |
CN108416752A (en) * | 2018-03-12 | 2018-08-17 | 中山大学 | A method of image is carried out based on production confrontation network and removes motion blur |
CN108446667A (en) * | 2018-04-04 | 2018-08-24 | 北京航空航天大学 | Based on the facial expression recognizing method and device for generating confrontation network data enhancing |
CN108986067A (en) * | 2018-05-25 | 2018-12-11 | 上海交通大学 | Pulmonary nodule detection method based on cross-module state |
CN109325951A (en) * | 2018-08-13 | 2019-02-12 | 深圳市唯特视科技有限公司 | A method of based on the conversion and segmenting medical volume for generating confrontation network |
-
2019
- 2019-02-18 CN CN201910119941.XA patent/CN109886909A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107220929A (en) * | 2017-06-23 | 2017-09-29 | 深圳市唯特视科技有限公司 | A kind of non-paired image method for transformation using the consistent confrontation network of circulation |
CN108229349A (en) * | 2017-12-21 | 2018-06-29 | 中国科学院自动化研究所 | Reticulate pattern facial image identification device |
CN108416752A (en) * | 2018-03-12 | 2018-08-17 | 中山大学 | A method of image is carried out based on production confrontation network and removes motion blur |
CN108446667A (en) * | 2018-04-04 | 2018-08-24 | 北京航空航天大学 | Based on the facial expression recognizing method and device for generating confrontation network data enhancing |
CN108986067A (en) * | 2018-05-25 | 2018-12-11 | 上海交通大学 | Pulmonary nodule detection method based on cross-module state |
CN109325951A (en) * | 2018-08-13 | 2019-02-12 | 深圳市唯特视科技有限公司 | A method of based on the conversion and segmenting medical volume for generating confrontation network |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111178502A (en) * | 2019-12-12 | 2020-05-19 | 中国科学院深圳先进技术研究院 | Training method and system for circularly generating confrontation network model and electronic equipment |
CN116502701A (en) * | 2023-06-29 | 2023-07-28 | 合肥锐世数字科技有限公司 | Attenuation correction method and device, training method and device, imaging method and system |
CN116502701B (en) * | 2023-06-29 | 2023-10-20 | 合肥锐世数字科技有限公司 | Attenuation correction method and device, training method and device, imaging method and system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Spadea et al. | Deep learning based synthetic‐CT generation in radiotherapy and PET: a review | |
US8787648B2 (en) | CT surrogate by auto-segmentation of magnetic resonance images | |
Robb | Virtual endoscopy: development and evaluation using the Visible Human datasets | |
CN107072595A (en) | Adaptive restatement based on multi-modality imaging is drawn | |
CN109961834A (en) | The generation method and equipment of diagnostic imaging report | |
CN105451663A (en) | Ultrasound acquisition feedback guidance to a target view | |
CN110020710B (en) | Beam direction and weight multi-objective optimization method based on artificial bee colony algorithm | |
CN110503626A (en) | Based on space-semantic significance constraint CT image modalities alignment schemes | |
CN108778418A (en) | Composite calulation machine tomographic imaging | |
CN109886909A (en) | A method of PET-CT image is synthesized based on CT image | |
Rammohan et al. | History of technological advancements towards MR-Linac: the future of image-guided radiotherapy | |
CN114881848A (en) | Method for converting multi-sequence MR into CT | |
Wu et al. | UC-GAN for MR to CT image synthesis | |
CN113378991A (en) | Medical data generation method and device, electronic equipment and storage medium | |
CN116993848B (en) | CBCT image reconstruction method, CBCT image reconstruction system, CBCT image reconstruction computer device and CBCT image storage medium | |
Vajpayee et al. | Structurally-constrained optical-flow-guided adversarial generation of synthetic CT for MR-only radiotherapy treatment planning | |
Robb | Virtual endoscopy: evaluation using the visible human datasets and comparison with real endoscopy in patients | |
Yang et al. | The birth of the first CT scanner | |
Qiu et al. | Visualization of the temporal bone of the Chinese Visible Human | |
Iddrisu et al. | 3D reconstructions of brain from MRI scans using neural radiance fields | |
Qiu et al. | Chest CBCT-based synthetic CT using cycle-consistent adversarial network with histogram matching | |
Zheng et al. | Unsupervised Cross-Modality Domain Adaptation Network for X-Ray to CT Registration | |
CN113421323B (en) | Bimodal volume data region-of-interest fusion method based on ray projection | |
CN110111395B (en) | Method for synthesizing PET-MRI image based on MRI image | |
CN114141354A (en) | Standardized inspection project name making method based on image equipment characteristics |
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: 20190614 |