CN110074813A - A kind of ultrasonic image reconstruction method and system - Google Patents
A kind of ultrasonic image reconstruction method and system Download PDFInfo
- Publication number
- CN110074813A CN110074813A CN201910342064.2A CN201910342064A CN110074813A CN 110074813 A CN110074813 A CN 110074813A CN 201910342064 A CN201910342064 A CN 201910342064A CN 110074813 A CN110074813 A CN 110074813A
- Authority
- CN
- China
- Prior art keywords
- image
- layer
- network model
- ultrasonic
- radio frequency
- 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
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B8/00—Diagnosis using ultrasonic, sonic or infrasonic waves
- A61B8/52—Devices using data or image processing specially adapted for diagnosis using ultrasonic, sonic or infrasonic waves
- A61B8/5215—Devices using data or image processing specially adapted for diagnosis using ultrasonic, sonic or infrasonic waves involving processing of medical diagnostic data
- A61B8/5238—Devices using data or image processing specially adapted for diagnosis using ultrasonic, sonic or infrasonic waves involving processing of medical diagnostic data for combining image data of patient, e.g. merging several images from different acquisition modes into one image
- A61B8/5261—Devices using data or image processing specially adapted for diagnosis using ultrasonic, sonic or infrasonic waves involving processing of medical diagnostic data for combining image data of patient, e.g. merging several images from different acquisition modes into one image combining images from different diagnostic modalities, e.g. ultrasound and X-ray
Abstract
The present invention discloses a kind of ultrasonic image reconstruction method and system.The method for building up of image reconstruction network model includes: acquisition training sample set;Training sample set includes multiple samples pair, and each sample is to including the one group of ultrasonic radio frequency training data obtained according to natural image and/or medical image and corresponding width training gray scale image.Building generates confrontation network model, and generating confrontation network model includes generator and arbiter.Confrontation network model is generated using training sample set training, so that generation confrontation network model is reached nash banlance, using the generator under nash banlance state as image reconstruction network model, is used for ultrasonic image reconstruction.Natural image and medical image of the present invention using resolution ratio much higher than ultrasound image generate the data basis of confrontation network model as training, then learn Optimal Parameters automatically by generating confrontation network model, the mapping relations from initial data to ultrasound image are obtained, image quality can be effectively improved.
Description
Technical field
The present invention relates to imaging fields, more particularly to a kind of ultrasonic image reconstruction method and system.
Background technique
Ultrasonic imaging is a kind of convenient, real-time, lossless imaging method.This method is taken using in ultrasound echo signal
The acoustic impedance different information of different parts is imaged inside the object to be detected of band.Conventional ultrasound imaging process is divided into two
Step, the first step are that ultrasonic radio frequency (Radio Frequency, RF) signal is obtained from ultrasonic transducer, and second step is believed using RF
Number generate ultrasound image, that is, ultrasonic image reconstruction.Ultrasonic radio frequency signal is the original signal in ultrasound imaging procedure, easily
It is interfered, influences image quality.Moreover, ultrasonic image reconstruction process is related to multiple steps, each step requires setting and is permitted
Multi-parameter, current method are mostly by rule of thumb come the parameter that each step is arranged, and every step parameter adjusts generated effect phase
It mutually influences, is difficult to find a kind of effective method and integrated planning is carried out to these parameters, cause final image quality poor.
Summary of the invention
The object of the present invention is to provide a kind of ultrasonic image reconstruction method and system, can effectively improve image quality.
To achieve the above object, the present invention provides following schemes:
A kind of ultrasonic image reconstruction method, which comprises
Obtain the ultrasonic radio frequency signal data of imageable target;
The ultrasonic radio frequency signal data input picture is rebuild into network model, obtains the ultrasound figure of the imageable target
Picture;Wherein, it is ultrasonic radio frequency signal data that described image, which rebuilds the input of network model, and described image rebuilds the defeated of network model
It is out ultrasound image;Described image, which rebuilds network model, to be established based on the deep neural network algorithm for generating confrontation network;
Described image rebuild network model method for building up include:
Obtain training sample set;The training sample set includes multiple samples pair, and each sample is to super including one group
Sound radio frequency training data and a width trained gray scale image corresponding with the ultrasonic radio frequency training data;Wherein, the ultrasound is penetrated
Frequency training data is the ultrasonic radio frequency data obtained according to the natural image and/or medical image of sample;The medical image packet
Include at least one of CT image and magnetic resonance image;
Building generates confrontation network model, and the generation confrontation network model includes generator and arbiter;
Network model is fought using the training sample set training generation, reaches the generation confrontation network model
Nash banlance, using the generator under nash banlance state as image reconstruction network model.
Optionally, the method for obtaining the ultrasonic radio frequency training data includes:
Obtain the natural image and/or medical image of sample;
The natural image and/or the medical image are inputted into ultrasonic sound field simulation software, the ultrasound for obtaining sample is penetrated
Frequency training data.
Optionally, the generator includes seven-layer structure;Wherein, first layer is full articulamentum;The second layer is batch normalizing
Change layer;Third layer is full articulamentum;4th layer is batch normalization layer;Layer 5 is two-dimentional warp lamination, and convolution kernel size is 4
× 4, step-length 2;Layer 6 is batch normalization layer;Layer 7 is warp lamination, and convolution kernel size is 4 × 4, and step-length 2 swashs
Function living is sigmoid function.
Optionally, the arbiter includes 5 layers of structure;Wherein, first layer and the second layer are two-dimensional convolution layer, and convolution kernel is big
Small is 4 × 4, step-length 2, and activation primitive is that linear unit function is corrected in band leakage;Third layer, the 4th layer and layer 5 are Quan Lian
Connect layer.
A kind of ultrasonic image reconstruction system, the system comprises:
Rf data obtains module, for obtaining the ultrasonic radio frequency signal data of imageable target;
Image reconstruction module, for the ultrasonic radio frequency signal data input picture to be rebuild network model, described in acquisition
The ultrasound image of imageable target;Wherein, it is ultrasonic radio frequency signal data, described image that described image, which rebuilds the input of network model,
The output for rebuilding network model is ultrasound image;It is based on the depth nerve for generating confrontation network that described image, which rebuilds network model,
What network algorithm was established;Described image rebuild network model subsystem of establishing include:
Sample set obtains module, for obtaining training sample set;The training sample set includes multiple samples pair, Mei Yisuo
Sample is stated to including one group of ultrasonic radio frequency training data and a width trained gray-scale figure corresponding with the ultrasonic radio frequency training data
Picture;Wherein, the ultrasonic radio frequency training data is the ultrasonic radio frequency number obtained according to the natural image and/or medical image of sample
According to;The medical image includes at least one of CT image and magnetic resonance image;
Network struction module is fought, generates confrontation network model for constructing, the generation confrontation network model includes life
It grows up to be a useful person and arbiter;
Training module makes the generation pair for fighting network model using the training sample set training generation
Anti- network model reaches nash banlance, using the generator under nash banlance state as image reconstruction network model.
Optionally, the sample set acquisition module includes:
Sample image acquiring unit, for obtaining the natural image and/or medical image of sample;
Training data determination unit, for emulating the natural image and/or medical image input ultrasonic sound field
Software obtains the ultrasonic radio frequency training data of sample.
Optionally, the generator includes seven-layer structure;Wherein, first layer is full articulamentum;The second layer is batch normalizing
Change layer;Third layer is full articulamentum;4th layer is batch normalization layer;Layer 5 is two-dimentional warp lamination, and convolution kernel size is 4
× 4, step-length 2;Layer 6 is batch normalization layer;Layer 7 is warp lamination, and convolution kernel size is 4 × 4, and step-length 2 swashs
Function living is sigmoid function.
Optionally, the arbiter includes 5 layers of structure;Wherein, first layer and the second layer are two-dimensional convolution layer, and convolution kernel is big
Small is 4 × 4, step-length 2, and activation primitive is that linear unit function is corrected in band leakage;Third layer, the 4th layer and layer 5 are Quan Lian
Connect layer.
The specific embodiment provided according to the present invention, the invention discloses following technical effects:
In ultrasonic image reconstruction method and system provided by the invention, the method for building up of image reconstruction network model includes:
Obtain training sample set;Training sample set includes multiple samples pair, each sample to include one group of ultrasonic radio frequency training data and
One corresponding trained gray scale image;Wherein, ultrasonic radio frequency training data is the natural image and/or medicine according to sample
The ultrasonic radio frequency data that image obtains;Medical image includes at least one of CT image and magnetic resonance image.Building generates confrontation
Network model, generating confrontation network model includes generator and arbiter.Confrontation network mould is generated using training sample set training
Type makes generation confrontation network model reach nash banlance, using the generator under nash banlance state as image reconstruction network mould
Type is used for ultrasonic image reconstruction.As it can be seen that natural image and medical image of the present invention using resolution ratio much higher than ultrasound image are made
The data basis of confrontation network model is generated for training, is then learnt Optimal Parameters automatically by generating confrontation network model, is obtained
The mapping relations from initial data to ultrasound image are obtained, image quality can be effectively improved.
Detailed description of the invention
It in order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, below will be to institute in embodiment
Attached drawing to be used is needed to be briefly described, it should be apparent that, the accompanying drawings in the following description is only some implementations of the invention
Example, for those of ordinary skill in the art, without any creative labor, can also be according to these attached drawings
Obtain other attached drawings.
Fig. 1 is a kind of flow chart of ultrasonic image reconstruction method provided in an embodiment of the present invention;
Fig. 2 is a kind of flow chart of the method for building up of image reconstruction network model provided in an embodiment of the present invention;
Fig. 3 is a kind of structural block diagram of ultrasonic image reconstruction system provided in an embodiment of the present invention;
Fig. 4 is a kind of structural block diagram for establishing subsystem of image reconstruction network model provided in an embodiment of the present invention;
Fig. 5 is a kind of flow chart for obtaining ultrasonic radio frequency training data provided in an embodiment of the present invention;
Fig. 6 is the network structure of generator provided in an embodiment of the present invention;
Fig. 7 is the network structure of arbiter provided in an embodiment of the present invention;
Fig. 8 is the flow chart that training provided in an embodiment of the present invention generates confrontation network model.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
The object of the present invention is to provide a kind of ultrasonic image reconstruction method and system, can effectively improve image quality.
In order to make the foregoing objectives, features and advantages of the present invention clearer and more comprehensible, with reference to the accompanying drawing and specific real
Applying mode, the present invention is described in further detail.
Fig. 1 is a kind of flow chart of ultrasonic image reconstruction method provided in an embodiment of the present invention.As shown in Figure 1, a kind of super
Acoustic image method for reconstructing, which comprises
Step 101: obtaining the ultrasonic radio frequency signal data of imageable target.
Step 102: the ultrasonic radio frequency signal data input picture being rebuild into network model, obtains the imageable target
Ultrasound image;Wherein, it is ultrasonic radio frequency signal data that described image, which rebuilds the input of network model, and described image rebuilds network mould
The output of type is ultrasound image;Described image, which rebuilds network model, is built based on the deep neural network algorithm for generating confrontation network
Vertical.
Fig. 2 is a kind of flow chart of the method for building up of image reconstruction network model provided in an embodiment of the present invention.Such as Fig. 2 institute
Show, the method for building up that described image rebuilds network model includes:
Step 201: obtaining training sample set;The training sample set includes multiple samples pair, and each sample is to packet
Include one group of ultrasonic radio frequency training data and a width trained gray scale image corresponding with the ultrasonic radio frequency training data;Wherein, institute
Stating ultrasonic radio frequency training data is the ultrasonic radio frequency data obtained according to the natural image and/or medical image of sample;The doctor
Learning image includes at least one of CT image and magnetic resonance image.
Step 202: building generates confrontation network model, and the generation confrontation network model includes generator and arbiter.
The generator includes seven-layer structure;Wherein, first layer is full articulamentum;The second layer is batch normalization layer;Third layer is complete
Articulamentum;4th layer is batch normalization layer;Layer 5 is two-dimentional warp lamination, and convolution kernel size is 4 × 4, step-length 2, no
Use activation primitive;Layer 6 is batch normalization layer;Layer 7 is warp lamination, and convolution kernel size is 4 × 4, and step-length 2 swashs
Function living is sigmoid function.The arbiter includes 5 layers of structure;Wherein, first layer and the second layer are two-dimensional convolution layer, volume
Product core size is 4 × 4, step-length 2, and activation primitive is band leakage amendment linear unit (Leaky ReLU) function;Third layer,
Four layers are full articulamentum with layer 5.
Step 203: fighting network model using the training sample set training generation, make the generation confrontation network
Model reaches nash banlance, using the generator under nash banlance state as image reconstruction network model.
Specifically, the method for obtaining the ultrasonic radio frequency training data includes:
Obtain the natural image and/or medical image of sample;
The natural image and/or the medical image are inputted into ultrasonic sound field simulation software, the ultrasound for obtaining sample is penetrated
Frequency training data.
Fig. 3 is a kind of structural block diagram of ultrasonic image reconstruction system provided in an embodiment of the present invention.As shown in figure 3, a kind of
Ultrasonic image reconstruction system, the system comprises:
Rf data obtains module 301, for obtaining the ultrasonic radio frequency signal data of imageable target.
Image reconstruction module 302 obtains institute for the ultrasonic radio frequency signal data input picture to be rebuild network model
State the ultrasound image of imageable target;Wherein, it is ultrasonic radio frequency signal data, the figure that described image, which rebuilds the input of network model,
Output as rebuilding network model is ultrasound image;It is based on the depth mind for generating confrontation network that described image, which rebuilds network model,
It is established through network algorithm.
Fig. 4 is a kind of structural block diagram for establishing subsystem of image reconstruction network model provided in an embodiment of the present invention.Such as
Shown in Fig. 4, the subsystem of establishing that described image rebuilds network model includes:
Sample set obtains module 401, for obtaining training sample set;The training sample set includes multiple samples pair, often
One sample is to including one group of ultrasonic radio frequency training data and a width training ash corresponding with the ultrasonic radio frequency training data
Rank image;Wherein, the ultrasonic radio frequency training data is to be penetrated according to the ultrasound that the natural image and/or medical image of sample obtain
Frequency evidence;The medical image includes at least one of CT image and magnetic resonance image.
Network struction module 402 is fought, generates confrontation network model for constructing, the generation confrontation network model includes
Generator and arbiter.The generator includes seven-layer structure;Wherein, first layer is full articulamentum;The second layer is batch normalizing
Change layer;Third layer is full articulamentum;4th layer is batch normalization layer;Layer 5 is two-dimentional warp lamination, and convolution kernel size is 4
× 4, step-length 2 does not have to activation primitive;Layer 6 is batch normalization layer;Layer 7 is warp lamination, and convolution kernel size is 4
× 4, step-length 2, activation primitive is sigmoid function.The arbiter includes 5 layers of structure;Wherein, first layer and the second layer are
Two-dimensional convolution layer, convolution kernel size are 4 × 4, and step-length 2, activation primitive is Leaky ReLU function;Third layer, the 4th layer and
Layer 5 is full articulamentum.
Training module 403 makes the generation for fighting network model using the training sample set training generation
Confrontation network model reaches nash banlance, using the generator under nash banlance state as image reconstruction network model.
Specifically, the sample set acquisition module 401 includes:
Sample image acquiring unit, for obtaining the natural image and/or medical image of sample.
Training data determination unit, for emulating the natural image and/or medical image input ultrasonic sound field
Software obtains the ultrasonic radio frequency training data of sample.
Specific implementation process of the invention is as follows:
(1) ultrasonic radio frequency training data is obtained using ultrasonic sound field simulation software.
The network training stage of the invention needs a large amount of sample data to be trained, and requires sample data while having
RF data high-resolution gray scale image corresponding with its, input of the RF data as network, corresponding high-resolution gray scale image
As the corresponding label figure of RF data.In order to improve the image quality of ultrasonic image reconstruction, present invention uses two class images to come
Generate sample data, one kind is high-resolution natural image, it is another kind of for CT image and/or magnetic resonance image (MRI,
Magnetic Resonance Imaging) etc. high-resolution medical image.Using above-mentioned two classes image data, pass through ultrasound
Sound field simulation software can produce corresponding RF data.The simulation software that the present embodiment is selected is sound field simulation software Field II.
Fig. 5 is a kind of flow chart for obtaining ultrasonic radio frequency training data provided in an embodiment of the present invention.As shown in figure 5, sharp
With FieldII emulation tool packet, with natural image, CT or MRI image are as the template that imitative body parameter is arranged, by image
The position of scattering point and size in imitative body is arranged in the position of pixel and gray value, then by setting related probes parameter,
To obtain imitative body and ultrasonic wave the ultrasonic RF signal data received of popping one's head in after reciprocation occurs for scanning mode.To RF signal number
According to being rebuild, and the validity for being able to demonstrate that above-mentioned RF data creating method is compared with original image.
(2) building generates confrontation network.
Generating confrontation network is composed of generator and arbiter, and generator generates analogue data, and arbiter is sentenced
Medium well at data be it is true or simulation.The generator for generating analogue data, which will be continued to optimize, oneself allows arbiter to judge
It does not come out, arbiter will also optimize that oneself to allow oneself interpretation to obtain more acurrate, and both sides relation forms confrontation.
The flattening of RF 2-D data is become the one of 1*16 after flattening at one-dimensional vector z, such as the two-dimensional matrix of 4*4
Matrix is tieed up, generator restores the analogue data of higher-dimension from the RF data of low-dimensional.Due to first to restore high from low-dimensional data
Dimension data, so to use warp Product function, in the network architecture to simulate the generation of high dimensional data.Fig. 6 is that the present invention is implemented
The network structure for the generator that example provides.As shown in fig. 6, generator has altogether by 7 layers of structure composition, wherein first 4 layers corresponding
Structure composition is two full articulamentums, connects a batch normalization (Batch Normalization) behind each full articulamentum
Layer.5th layer is two-dimentional warp lamination, and convolution kernel size is 4 × 4, and step-length 2 does not have to activation primitive.6th layer is Batch
Normalization layers.7th layer is warp lamination, and convolution kernel size is 4 × 4, and step-length 2, activation primitive is sigmoid letter
Number.
Arbiter is used to judge that the data of input are the analogue datas that true sample or generator generate, i.e. estimation sample
Originally belong to the conditional probability distribution of certain class.Wherein, true sample is that training sample concentrates the corresponding gray-scale figure of RF data.In net
Will be by convolution twice in network structure, then full connection twice is connect, the differentiation result that last output layer generates is between 0 to 1.Wherein
1 representative is judged as authentic specimen, and 0 representative is judged as analogue data.Numerical value indicates that input data is authentic specimen closer to 1
Probability is higher, indicates that the probability that input data is analog sample is higher closer to 0.Fig. 7 sentences to be provided in an embodiment of the present invention
The network structure of other device.As shown in fig. 7, arbiter, altogether by 5 layers of structure composition, first two layers is two-dimensional convolution layer, convolution kernel
Size is 4 × 4, step-length 2, and activation primitive is LeakyReLU function, and latter three layers are full articulamentum, and last full articulamentum is defeated
Result is differentiated out.
(3) network training study is carried out using sample data, obtains image reconstruction network model.
Fig. 8 is the flow chart that training provided in an embodiment of the present invention generates confrontation network model.As shown in figure 8, by flat
RF one-dimensional vector afterwards is put into generator, generates analog sample, then by analog sample and authentic specimen (gray scale image) point
It is not input in arbiter, generates and differentiate result.The purpose of generator is to make analog sample close to authentic specimen, so simulation
The differentiation result of sample numerically otherwise disconnecting nearly 1, and the purpose of arbiter be can accurate judgement analog sample or true sample
This, thus the differentiation result of analog sample numerically however disconnecting nearly 0, the differentiation result of authentic specimen are numerically continuous
Close to 1.The present invention optimizes above-mentioned training process using Adam optimization algorithm, and the learning rate of arbiter is 0.0001, generates
The learning rate of device is 0.001.
Adam optimization algorithm is one kind of gradient descent method, and minimum deflection mould is approached for recursiveness in deep learning
Type.In the training process, each forward-propagating can all obtain the penalty values of output valve and true value, this penalty values is smaller, generation
Table model is better.
Generator and arbiter are trained simultaneously, simulate generator as far as possible by the repetitive exercise of a large amount of numbers
Out close to the sample of raw ultrasound image, and arbiter has the ability for data of more accurately discerning the false from the genuine, final entire raw
Nash banlance can be reached at confrontation network, i.e., arbiter is to the differentiation result of analogue data and truthful data all close to 0.5.By
Above-mentioned training process can be obtained the generator with reconstruction ability, using this generator as image reconstruction network model into
The subsequent image reconstruction of row.
(4) image reconstruction is carried out using image reconstruction network model.
The ultrasonic radio frequency signal data input picture of imageable target is rebuild into network model, image reconstruction network model
Export the ultrasound image of imageable target.
The present invention generates the data base of confrontation network model using the natural image of high resolution and medical image as training
Then plinth learns Optimal Parameters by generating confrontation network model automatically, obtain the mapping from initial data to ultrasound image and close
System, can effectively improve image quality.
Each embodiment in this specification is described in a progressive manner, the highlights of each of the examples are with other
The difference of embodiment, the same or similar parts in each embodiment may refer to each other.
Used herein a specific example illustrates the principle and implementation of the invention, and above embodiments are said
It is bright to be merely used to help understand method and its core concept of the invention;At the same time, for those skilled in the art, foundation
Thought of the invention, there will be changes in the specific implementation manner and application range.In conclusion the content of the present specification is not
It is interpreted as limitation of the present invention.
Claims (8)
1. a kind of ultrasonic image reconstruction method, which is characterized in that the described method includes:
Obtain the ultrasonic radio frequency signal data of imageable target;
The ultrasonic radio frequency signal data input picture is rebuild into network model, obtains the ultrasound image of the imageable target;Its
In, the input that described image rebuilds network model is ultrasonic radio frequency signal data, and the output that described image rebuilds network model is
Ultrasound image;Described image, which rebuilds network model, to be established based on the deep neural network algorithm for generating confrontation network;It is described
The method for building up of image reconstruction network model includes:
Obtain training sample set;The training sample set includes multiple samples pair, and each sample is penetrated to including one group of ultrasound
Frequency training data and a width trained gray scale image corresponding with the ultrasonic radio frequency training data;Wherein, the ultrasonic radio frequency instruction
Practicing data is the ultrasonic radio frequency data obtained according to the natural image and/or medical image of sample;The medical image includes CT
At least one of image and magnetic resonance image;
Building generates confrontation network model, and the generation confrontation network model includes generator and arbiter;
Fight network model using training sample set training generations, make the generation fight network model reach receive it is assorted
Balance, using the generator under nash banlance state as image reconstruction network model.
2. ultrasonic image reconstruction method according to claim 1, which is characterized in that obtain the ultrasonic radio frequency training data
Method include:
Obtain the natural image and/or medical image of sample;
The natural image and/or the medical image are inputted into ultrasonic sound field simulation software, obtain the ultrasonic radio frequency instruction of sample
Practice data.
3. ultrasonic image reconstruction method according to claim 1, which is characterized in that the generator includes seven-layer structure;
Wherein, first layer is full articulamentum;The second layer is batch normalization layer;Third layer is full articulamentum;4th layer is batch normalizing
Change layer;Layer 5 is two-dimentional warp lamination, and convolution kernel size is 4 × 4, step-length 2;Layer 6 is batch normalization layer;7th
Layer is warp lamination, and convolution kernel size is 4 × 4, and step-length 2, activation primitive is sigmoid function.
4. ultrasonic image reconstruction method according to claim 1, which is characterized in that the arbiter includes five-layer structure;
Wherein, first layer and the second layer are two-dimensional convolution layer, and convolution kernel size is 4 × 4, step-length 2, and activation primitive is band leakage amendment
Linear unit function;Third layer, the 4th layer and layer 5 are full articulamentum.
5. a kind of ultrasonic image reconstruction system, which is characterized in that the system comprises:
Rf data obtains module, for obtaining the ultrasonic radio frequency signal data of imageable target;
Image reconstruction module obtains the imaging for the ultrasonic radio frequency signal data input picture to be rebuild network model
The ultrasound image of target;Wherein, it is ultrasonic radio frequency signal data that described image, which rebuilds the input of network model, and described image is rebuild
The output of network model is ultrasound image;It is based on the deep neural network for generating confrontation network that described image, which rebuilds network model,
What algorithm was established;Described image rebuild network model subsystem of establishing include:
Sample set obtains module, for obtaining training sample set;The training sample set includes multiple samples pair, each sample
This is to including one group of ultrasonic radio frequency training data and a width trained gray scale image corresponding with the ultrasonic radio frequency training data;Its
In, the ultrasonic radio frequency training data is the ultrasonic radio frequency data obtained according to the natural image and/or medical image of sample;Institute
Stating medical image includes at least one of CT image and magnetic resonance image;
Network struction module is fought, generates confrontation network model for constructing, the generation confrontation network model includes generator
And arbiter;
Training module makes the generation confrontation net for fighting network model using the training sample set training generation
Network model reaches nash banlance, using the generator under nash banlance state as image reconstruction network model.
6. ultrasonic image reconstruction system according to claim 5, which is characterized in that the sample set obtains module and includes:
Sample image acquiring unit, for obtaining the natural image and/or medical image of sample;
Training data determination unit, for the natural image and/or the medical image to be inputted ultrasonic sound field simulation software,
Obtain the ultrasonic radio frequency training data of sample.
7. ultrasonic image reconstruction system according to claim 5, which is characterized in that the generator includes seven-layer structure;
Wherein, first layer is full articulamentum;The second layer is batch normalization layer;Third layer is full articulamentum;4th layer is batch normalizing
Change layer;Layer 5 is two-dimentional warp lamination, and convolution kernel size is 4 × 4, step-length 2;Layer 6 is batch normalization layer;7th
Layer is warp lamination, and convolution kernel size is 4 × 4, and step-length 2, activation primitive is sigmoid function.
8. ultrasonic image reconstruction system according to claim 5, which is characterized in that the arbiter includes 5 layers of structure;Its
In, first layer and the second layer are two-dimensional convolution layer, and convolution kernel size is 4 × 4, step-length 2, and activation primitive is band leakage modified line
Property unit function;Third layer, the 4th layer and layer 5 are full articulamentum.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910342064.2A CN110074813B (en) | 2019-04-26 | 2019-04-26 | Ultrasonic image reconstruction method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910342064.2A CN110074813B (en) | 2019-04-26 | 2019-04-26 | Ultrasonic image reconstruction method and system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110074813A true CN110074813A (en) | 2019-08-02 |
CN110074813B CN110074813B (en) | 2022-03-04 |
Family
ID=67416866
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910342064.2A Active CN110074813B (en) | 2019-04-26 | 2019-04-26 | Ultrasonic image reconstruction method and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110074813B (en) |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110517249A (en) * | 2019-08-27 | 2019-11-29 | 中山大学 | Imaging method, device, equipment and the medium of ultrasonic elastic image |
CN111091603A (en) * | 2019-11-04 | 2020-05-01 | 深圳先进技术研究院 | Ultrasonic imaging method and device, readable storage medium and terminal equipment |
CN111402266A (en) * | 2020-03-13 | 2020-07-10 | 中国石油大学(华东) | Method and system for constructing digital core |
CN111444830A (en) * | 2020-03-25 | 2020-07-24 | 腾讯科技(深圳)有限公司 | Imaging method and device based on ultrasonic echo signal, storage medium and electronic device |
CN111489404A (en) * | 2020-03-20 | 2020-08-04 | 深圳先进技术研究院 | Image reconstruction method, image processing device and device with storage function |
CN111798400A (en) * | 2020-07-20 | 2020-10-20 | 福州大学 | Non-reference low-illumination image enhancement method and system based on generation countermeasure network |
WO2021120071A1 (en) * | 2019-12-18 | 2021-06-24 | 深圳迈瑞生物医疗电子股份有限公司 | Ultrasonic imaging method, ultrasonic imaging system, and computer-readable storage medium |
CN113112559A (en) * | 2021-04-07 | 2021-07-13 | 中国科学院深圳先进技术研究院 | Ultrasonic image segmentation method and device, terminal equipment and storage medium |
CN113239978A (en) * | 2021-04-22 | 2021-08-10 | 科大讯飞股份有限公司 | Method and device for correlating medical image preprocessing model and analysis model |
CN113436109A (en) * | 2021-07-08 | 2021-09-24 | 清华大学 | Ultrafast high-quality plane wave ultrasonic imaging method based on deep learning |
CN113470139A (en) * | 2020-04-29 | 2021-10-01 | 浙江大学 | CT image reconstruction method based on MRI |
CN114739673A (en) * | 2022-03-18 | 2022-07-12 | 河北工业大学 | Bearing vibration signal characteristic interpretable dimension reduction and fault diagnosis method |
CN116681790A (en) * | 2023-07-18 | 2023-09-01 | 脉得智能科技(无锡)有限公司 | Training method of ultrasound contrast image generation model and image generation method |
Citations (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101099681A (en) * | 2007-06-07 | 2008-01-09 | 复旦大学 | Small scale thermo-acoustic imaging de-convolution reconstruction method |
US20120220875A1 (en) * | 2010-04-20 | 2012-08-30 | Suri Jasjit S | Mobile Architecture Using Cloud for Hashimoto's Thyroiditis Disease Classification |
CN104055536A (en) * | 2014-04-25 | 2014-09-24 | 深圳大学 | Ultrasound and magnetic resonance image fusion and registration method |
US20170337682A1 (en) * | 2016-05-18 | 2017-11-23 | Siemens Healthcare Gmbh | Method and System for Image Registration Using an Intelligent Artificial Agent |
CN107610193A (en) * | 2016-06-23 | 2018-01-19 | 西门子保健有限责任公司 | Use the image rectification of depth production machine learning model |
CN108268870A (en) * | 2018-01-29 | 2018-07-10 | 重庆理工大学 | Multi-scale feature fusion ultrasonoscopy semantic segmentation method based on confrontation study |
CN108447049A (en) * | 2018-02-27 | 2018-08-24 | 中国海洋大学 | A kind of digitlization physiology organism dividing method fighting network based on production |
CN108550118A (en) * | 2018-03-22 | 2018-09-18 | 深圳大学 | Fuzzy processing method, device, equipment and the storage medium of motion blur image |
CN109087327A (en) * | 2018-07-13 | 2018-12-25 | 天津大学 | A kind of thyroid nodule ultrasonic image division method cascading full convolutional neural networks |
CN109242865A (en) * | 2018-09-26 | 2019-01-18 | 上海联影智能医疗科技有限公司 | Medical image auto-partition system, method, apparatus and storage medium based on multichannel chromatogram |
CN109377520A (en) * | 2018-08-27 | 2019-02-22 | 西安电子科技大学 | Cardiac image registration arrangement and method based on semi-supervised circulation GAN |
CN109493308A (en) * | 2018-11-14 | 2019-03-19 | 吉林大学 | The medical image synthesis and classification method for generating confrontation network are differentiated based on condition more |
CN109544517A (en) * | 2018-11-06 | 2019-03-29 | 中山大学附属第医院 | Method and system are analysed in multi-modal ultrasound group credit based on deep learning |
CN109637634A (en) * | 2018-12-11 | 2019-04-16 | 厦门大学 | A kind of medical image synthetic method based on generation confrontation network |
US20190122120A1 (en) * | 2017-10-20 | 2019-04-25 | Dalei Wu | Self-training method and system for semi-supervised learning with generative adversarial networks |
-
2019
- 2019-04-26 CN CN201910342064.2A patent/CN110074813B/en active Active
Patent Citations (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101099681A (en) * | 2007-06-07 | 2008-01-09 | 复旦大学 | Small scale thermo-acoustic imaging de-convolution reconstruction method |
US20120220875A1 (en) * | 2010-04-20 | 2012-08-30 | Suri Jasjit S | Mobile Architecture Using Cloud for Hashimoto's Thyroiditis Disease Classification |
CN104055536A (en) * | 2014-04-25 | 2014-09-24 | 深圳大学 | Ultrasound and magnetic resonance image fusion and registration method |
US20170337682A1 (en) * | 2016-05-18 | 2017-11-23 | Siemens Healthcare Gmbh | Method and System for Image Registration Using an Intelligent Artificial Agent |
CN107610193A (en) * | 2016-06-23 | 2018-01-19 | 西门子保健有限责任公司 | Use the image rectification of depth production machine learning model |
US20190122120A1 (en) * | 2017-10-20 | 2019-04-25 | Dalei Wu | Self-training method and system for semi-supervised learning with generative adversarial networks |
CN108268870A (en) * | 2018-01-29 | 2018-07-10 | 重庆理工大学 | Multi-scale feature fusion ultrasonoscopy semantic segmentation method based on confrontation study |
CN108447049A (en) * | 2018-02-27 | 2018-08-24 | 中国海洋大学 | A kind of digitlization physiology organism dividing method fighting network based on production |
CN108550118A (en) * | 2018-03-22 | 2018-09-18 | 深圳大学 | Fuzzy processing method, device, equipment and the storage medium of motion blur image |
CN109087327A (en) * | 2018-07-13 | 2018-12-25 | 天津大学 | A kind of thyroid nodule ultrasonic image division method cascading full convolutional neural networks |
CN109377520A (en) * | 2018-08-27 | 2019-02-22 | 西安电子科技大学 | Cardiac image registration arrangement and method based on semi-supervised circulation GAN |
CN109242865A (en) * | 2018-09-26 | 2019-01-18 | 上海联影智能医疗科技有限公司 | Medical image auto-partition system, method, apparatus and storage medium based on multichannel chromatogram |
CN109544517A (en) * | 2018-11-06 | 2019-03-29 | 中山大学附属第医院 | Method and system are analysed in multi-modal ultrasound group credit based on deep learning |
CN109493308A (en) * | 2018-11-14 | 2019-03-19 | 吉林大学 | The medical image synthesis and classification method for generating confrontation network are differentiated based on condition more |
CN109637634A (en) * | 2018-12-11 | 2019-04-16 | 厦门大学 | A kind of medical image synthetic method based on generation confrontation network |
Non-Patent Citations (5)
Title |
---|
GEEKZW: "超声图像模拟", 《CSDN》 * |
GIBSON, E; LI, WQ;: "NiftyNet: a deep-learning platform for medical imaging", 《COMPUTER METHODS AND PROGRAMS IN BIOMEDICINE》 * |
VEDULA S , ET AL.: "Towards CT-quality Ultrasound Imaging using Deep Learning", 《ARXIV PREPRINT》 * |
吴洋洋 等: "生成对抗网络的血管内超声图像超分辨率重建", 《南方医科大学学报》 * |
孙博: "基于生成对抗网络的文本自动生成方法研究", 《中国优秀硕士学位论文全文数据库 (信息科技辑)》 * |
Cited By (19)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110517249A (en) * | 2019-08-27 | 2019-11-29 | 中山大学 | Imaging method, device, equipment and the medium of ultrasonic elastic image |
CN111091603A (en) * | 2019-11-04 | 2020-05-01 | 深圳先进技术研究院 | Ultrasonic imaging method and device, readable storage medium and terminal equipment |
CN114401676A (en) * | 2019-12-18 | 2022-04-26 | 深圳迈瑞生物医疗电子股份有限公司 | Ultrasonic imaging method, ultrasonic imaging system, and computer-readable storage medium |
WO2021120071A1 (en) * | 2019-12-18 | 2021-06-24 | 深圳迈瑞生物医疗电子股份有限公司 | Ultrasonic imaging method, ultrasonic imaging system, and computer-readable storage medium |
CN111402266A (en) * | 2020-03-13 | 2020-07-10 | 中国石油大学(华东) | Method and system for constructing digital core |
CN111489404B (en) * | 2020-03-20 | 2023-09-05 | 深圳先进技术研究院 | Image reconstruction method, image processing device and device with storage function |
CN111489404A (en) * | 2020-03-20 | 2020-08-04 | 深圳先进技术研究院 | Image reconstruction method, image processing device and device with storage function |
CN111444830A (en) * | 2020-03-25 | 2020-07-24 | 腾讯科技(深圳)有限公司 | Imaging method and device based on ultrasonic echo signal, storage medium and electronic device |
CN111444830B (en) * | 2020-03-25 | 2023-10-31 | 腾讯科技(深圳)有限公司 | Method and device for imaging based on ultrasonic echo signals, storage medium and electronic device |
CN113470139A (en) * | 2020-04-29 | 2021-10-01 | 浙江大学 | CT image reconstruction method based on MRI |
CN111798400A (en) * | 2020-07-20 | 2020-10-20 | 福州大学 | Non-reference low-illumination image enhancement method and system based on generation countermeasure network |
CN111798400B (en) * | 2020-07-20 | 2022-10-11 | 福州大学 | Non-reference low-illumination image enhancement method and system based on generation countermeasure network |
CN113112559A (en) * | 2021-04-07 | 2021-07-13 | 中国科学院深圳先进技术研究院 | Ultrasonic image segmentation method and device, terminal equipment and storage medium |
CN113239978A (en) * | 2021-04-22 | 2021-08-10 | 科大讯飞股份有限公司 | Method and device for correlating medical image preprocessing model and analysis model |
CN113436109B (en) * | 2021-07-08 | 2022-10-14 | 清华大学 | Ultrafast high-quality plane wave ultrasonic imaging method based on deep learning |
CN113436109A (en) * | 2021-07-08 | 2021-09-24 | 清华大学 | Ultrafast high-quality plane wave ultrasonic imaging method based on deep learning |
CN114739673A (en) * | 2022-03-18 | 2022-07-12 | 河北工业大学 | Bearing vibration signal characteristic interpretable dimension reduction and fault diagnosis method |
CN116681790A (en) * | 2023-07-18 | 2023-09-01 | 脉得智能科技(无锡)有限公司 | Training method of ultrasound contrast image generation model and image generation method |
CN116681790B (en) * | 2023-07-18 | 2024-03-22 | 脉得智能科技(无锡)有限公司 | Training method of ultrasound contrast image generation model and image generation method |
Also Published As
Publication number | Publication date |
---|---|
CN110074813B (en) | 2022-03-04 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110074813A (en) | A kind of ultrasonic image reconstruction method and system | |
Tom et al. | Simulating patho-realistic ultrasound images using deep generative networks with adversarial learning | |
US11354791B2 (en) | Methods and system for transforming medical images into different styled images with deep neural networks | |
CN103908300B (en) | Pin enhancing in diagnostic ultrasound imaging | |
CN109978778A (en) | Convolutional neural networks medicine CT image denoising method based on residual error study | |
EP3510564B1 (en) | Ray-tracing methods for realistic interactive ultrasound simulation | |
CN108829639B (en) | Magnetic resonance imaging method and equipment | |
Zhou et al. | High spatial–temporal resolution reconstruction of plane-wave ultrasound images with a multichannel multiscale convolutional neural network | |
Magee et al. | An augmented reality simulator for ultrasound guided needle placement training | |
CN109191476A (en) | The automatic segmentation of Biomedical Image based on U-net network structure | |
US20210134028A1 (en) | Method and system for simultaneous quantitative multiparametric magnetic resonance imaging (mri) | |
JP2012503501A (en) | Simulation of medical image diagnosis | |
Tang et al. | Plane-wave image reconstruction via generative adversarial network and attention mechanism | |
KR20040102038A (en) | A method for encoding image pixels, a method for processing images and a method for processing images aimed at qualitative recognition of the object reproduced by one or more image pixels | |
CN107847217A (en) | The method and apparatus shown for generating ultrasonic scattering body surface | |
CN114983468A (en) | Imaging system and method using a real-time inspection completeness monitor | |
CN108985366A (en) | B ultrasound image recognition algorithm and system based on convolution depth network | |
WO2007100263A1 (en) | Method for simulation of ultrasound images | |
Brickson et al. | Reverberation noise suppression in the aperture domain using 3D fully convolutional neural networks | |
CN105303537B (en) | A kind of medical image three-dimensional blood vessel display Enhancement Method | |
US20210204904A1 (en) | Ultrasound diagnostic system | |
CN106023277A (en) | Magnetic induction magnetoacoustic endoscopic image modeling and simulation method | |
CN107817492A (en) | The imaging method and device of wide angle synthetic aperture radar | |
CN108596900A (en) | Thyroid-related Ophthalmopathy medical image data processing unit, method, computer readable storage medium and terminal device | |
Starkov et al. | Ultrasound simulation with animated anatomical models and on-the-fly fusion with real images via path-tracing |
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 |