CN110322399A - A kind of ultrasound image method of adjustment, system, equipment and computer storage medium - Google Patents
A kind of ultrasound image method of adjustment, system, equipment and computer storage medium Download PDFInfo
- Publication number
- CN110322399A CN110322399A CN201910604343.1A CN201910604343A CN110322399A CN 110322399 A CN110322399 A CN 110322399A CN 201910604343 A CN201910604343 A CN 201910604343A CN 110322399 A CN110322399 A CN 110322399A
- Authority
- CN
- China
- Prior art keywords
- image
- ultrasound image
- initial
- neural network
- network model
- 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
- 238000002604 ultrasonography Methods 0.000 title claims abstract description 262
- 238000000034 method Methods 0.000 title claims abstract description 60
- 238000003860 storage Methods 0.000 title claims abstract description 16
- 238000003062 neural network model Methods 0.000 claims abstract description 100
- 238000012800 visualization Methods 0.000 claims abstract description 34
- 238000012549 training Methods 0.000 claims description 69
- 230000005540 biological transmission Effects 0.000 claims description 18
- 238000004422 calculation algorithm Methods 0.000 claims description 10
- 238000004590 computer program Methods 0.000 claims description 10
- 238000002372 labelling Methods 0.000 claims description 10
- 239000011159 matrix material Substances 0.000 claims description 9
- 238000013519 translation Methods 0.000 claims description 9
- 230000004927 fusion Effects 0.000 claims description 6
- 238000000605 extraction Methods 0.000 claims description 5
- 238000013459 approach Methods 0.000 claims description 3
- 210000005036 nerve Anatomy 0.000 claims description 2
- 230000004069 differentiation Effects 0.000 claims 1
- 230000001105 regulatory effect Effects 0.000 abstract description 3
- 238000010586 diagram Methods 0.000 description 16
- 238000013528 artificial neural network Methods 0.000 description 7
- 238000004891 communication Methods 0.000 description 7
- 230000002708 enhancing effect Effects 0.000 description 7
- 238000005516 engineering process Methods 0.000 description 6
- 230000000694 effects Effects 0.000 description 5
- 235000013399 edible fruits Nutrition 0.000 description 4
- 238000004519 manufacturing process Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 210000004218 nerve net Anatomy 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 208000037170 Delayed Emergence from Anesthesia Diseases 0.000 description 1
- 210000000436 anus Anatomy 0.000 description 1
- 239000011248 coating agent Substances 0.000 description 1
- 238000000576 coating method Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 210000004061 pubic symphysis Anatomy 0.000 description 1
- 210000000664 rectum Anatomy 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/20—Linear translation of whole images or parts thereof, e.g. panning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/90—Dynamic range modification of images or parts thereof
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10132—Ultrasound image
- G06T2207/10136—3D ultrasound image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20212—Image combination
- G06T2207/20221—Image fusion; Image merging
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
This application discloses a kind of ultrasound image method of adjustment, system, equipment and computer storage mediums, obtain the initial ultrasound image of ultrasonic device shooting；Initial ultrasound image is transmitted to preparatory trained image recognition neural network model；Receive the feature point image that image recognition neural network model identifies in initial ultrasound image；Three-dimensional visualization adjustment is carried out to initial ultrasound image based on feature point image.Ultrasound image method of adjustment provided by the present application, it is identified by the initial ultrasound image that trained image recognition neural network model automatic butt is received in advance, obtain characteristic pattern point picture, three-dimensional visualization adjustment is carried out to initial ultrasound image based on feature point image again, so that the regulated efficiency for carrying out three-dimensional visualization to ultrasound image can be improved without manually three-dimensional visualization adjustment can be carried out to ultrasound image.Ultrasound image adjustment system, equipment and computer readable storage medium provided by the present application also solve the problems, such as relevant art.
Description
Technical field
This application involves ultrasound image processing technology fields, more specifically to a kind of ultrasound image method of adjustment, are
System, equipment and computer media.
Background technique
When observing using ultrasonic device ultrasound images such as basin bottoms, the ultrasound shot to ultrasonic device is needed
Image carries out three-dimensional visualization and needs to adjust ultrasound image to suitable position in the process, for example adjusts to horizontal position
It sets, consequently facilitating observing ultrasound image.
A kind of existing ultrasound image method of adjustment is: the characteristic area in manual identified ultrasound image, according to characteristic area
Domain is adjusted ultrasound image.
However, need manually to identify the characteristic area in ultrasound image in existing ultrasound image method of adjustment,
Recognition efficiency is slow, also needs to manually adjust after identification, so that the efficiency for carrying out three-dimensional visualization to ultrasound image is lower.
In conclusion how to improve to ultrasound image carry out three-dimensional visualization efficiency be current those skilled in the art urgently
Problem to be solved.
Summary of the invention
The purpose of the application is to provide a kind of ultrasound image method of adjustment, can solve how to improve to a certain extent pair
Ultrasound image carries out the technical issues of three-dimensional visualization.Present invention also provides a kind of ultrasound image adjustment system, equipment and meters
Calculation machine readable storage medium storing program for executing.
To achieve the goals above, the application provides the following technical solutions:
A kind of ultrasound image method of adjustment, comprising:
Obtain the initial ultrasound image of ultrasonic device shooting；
The initial ultrasound image is transmitted to preparatory trained image recognition neural network model；
Receive the feature point image that described image identification neural network model identifies in the initial ultrasound image；
Three-dimensional visualization adjustment is carried out to the initial ultrasound image based on the feature point image.
Preferably, described that three-dimensional visualization adjustment, packet are carried out to the initial ultrasound image based on the feature point image
It includes:
Obtain the view-finder parameter of the ultrasonic device；
The translation rotation matrix of the initial ultrasound image is calculated based on the feature point image and the view-finder parameter；
The translation rotation matrix is multiplied with the initial ultrasound image, it is super to obtain three-dimensional visualization target adjusted
Acoustic image.
Preferably, the transmission initial ultrasound image to preparatory trained image recognition neural network model it
Before, further includes:
Obtain initial ultrasound image training sample；
Obtain the corresponding sample label of the initial ultrasound image training sample；
Based on the initial ultrasound image training sample and the sample label, the initial described image built is identified
Neural network model is trained, and obtains trained described image identification neural network model in advance.
It is preferably, described to obtain the corresponding sample label of the initial ultrasound image training sample, comprising:
It obtains to the initial ultrasound image labeling sample after the feature object mark in the initial ultrasound image training sample
This；
Gauss operation is carried out to the initial ultrasound image labeling sample after the mark, obtains the sample label.
Preferably, described to obtain to the initial ultrasound after the feature object mark in the initial ultrasound image training sample
Image labeling sample, comprising:
It obtains to the initial ultrasound image obtained after the characteristic area mark in the initial ultrasound image training sample
Characteristic area marks sample；
Obtain the spy to the initial ultrasound image obtained after the characteristic point mark in the initial ultrasound image training sample
Sign point mark sample；Wherein, on the characteristic area, the characteristic area and characteristic point respectively include at least the characteristic point
2.
Preferably, described image identification neural network model includes outline identification network and Feature point recognition network；It is described
Outline identification network is used to carry out feature extraction to the initial ultrasound image, obtains feature regional images；The characteristic point is known
Other network is used to characteristic area mark sample and the initial ultrasound image carrying out Fusion Features, obtains the characteristic point
Image；
It is described to be based on the initial ultrasound image training sample and the sample label, to the initial described image built
Identification neural network model is trained, and obtains trained described image identification neural network model in advance, comprising:
Sample is marked based on the initial ultrasound image training sample and the characteristic area, to the outline identification network
It is trained, obtains the trained outline identification network；
Sample is marked based on the initial ultrasound image training sample, characteristic area mark sample and the characteristic point
This, is trained the Feature point recognition network, obtains the trained Feature point recognition network.
Preferably, described image identification neural network model further includes image enhancement network, for according to other characteristic points
The target feature point is enhanced with the positional relationship of target feature point, obtains characteristic point enhancing image；
Described image enhances network and is used to carry out convolution algorithm to a feature point image, obtains convolution algorithm result；By institute
It states convolution algorithm result to be cascaded with another feature point image, obtains cascade result；Residual error fortune is carried out to the cascade result
It calculates, obtains the characteristic point enhancing image of another feature point image.
Preferably, described to be based on the initial ultrasound image training sample and the sample label, it is initial to what is built
Described image identification neural network model is trained, and obtains trained described image identification neural network model in advance, packet
It includes:
The initial ultrasound image training sample and the sample label are transmitted to described image identification neural network mould
Type；
Obtain the label result of described image identification neural network model output；
The sample label and the label result are transmitted to the arbiter built in advance, the arbiter is for judging
Whether the label result and the sample label meet default approach condition；
The judging result for receiving the arbiter output transmits the judging result to described image and identifies neural network mould
Type, so that described image identification neural network model is based on the judging result and adjusts inherent parameters；
Repeat it is described acquisition described image identification neural network model output label result and later the step of, directly
Sentence to trained described image identification neural network model, described image identification neural network model in advance is obtained including described
Other device.
A kind of ultrasound image adjustment system, comprising:
First obtains module, for obtaining the initial ultrasound image of ultrasonic device shooting；
First transmission module is used for transmission the initial ultrasound image to preparatory trained image recognition neural network mould
Type；
First receiving module identifies in the initial ultrasound image for receiving described image identification neural network model
Obtained feature point image；
The first adjustment module, for carrying out three-dimensional visualization tune to the initial ultrasound image based on the feature point image
It is whole.
A kind of ultrasound image adjustment equipment, comprising:
Memory, for storing computer program；
Processor realizes the step of as above any ultrasound image method of adjustment when for executing the computer program
Suddenly.
A kind of computer readable storage medium is stored with computer program in the computer readable storage medium, described
The step of as above any ultrasound image method of adjustment is realized when computer program is executed by processor.
A kind of ultrasound image method of adjustment provided by the present application obtains the initial ultrasound image of ultrasonic device shooting；Transmission
Initial ultrasound image is to preparatory trained image recognition neural network model；Image recognition neural network model is received initial
The feature point image identified in ultrasound image；Three-dimensional visualization tune is carried out to initial ultrasound image based on feature point image
It is whole.A kind of ultrasound image method of adjustment provided by the present application, it is automatic by preparatory trained image recognition neural network model
Received initial ultrasound image is identified, obtains feature point image, then based on feature point image to initial ultrasound image into
Row three-dimensional visualization adjustment, so that without manually three-dimensional visualization adjustment can be carried out to ultrasound image, with existing skill
Art is compared, and the regulated efficiency that three-dimensional visualization is carried out to ultrasound image can be improved.A kind of ultrasound image tune provided by the present application
Whole system, equipment and computer readable storage medium also solve the problems, such as relevant art.
Detailed description of the invention
In order to illustrate the technical solutions in the embodiments of the present application or in the prior art more clearly, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
The embodiment of application for those of ordinary skill in the art without creative efforts, can also basis
The attached drawing of offer obtains other attached drawings.
Fig. 1 is a kind of first pass figure of ultrasound image method of adjustment provided by the embodiments of the present application；
Fig. 2 is a kind of second flow chart of Ultrasound Image Recognition Method provided by the embodiments of the present application；
Fig. 3 is the flow chart that image recognition neural network model identifies initial ultrasound image；
Fig. 4 is the acquisition schematic diagram of characteristic point candidate's figure in basin baselap acoustic image；
Fig. 5 is characterized a candidate figure and obtains the schematic diagram of feature point diagram；
Fig. 6 is the effect diagram of ultrasonic device method of adjustment provided by the present application；
Fig. 7 is the structural schematic diagram that a kind of ultrasound image provided by the embodiments of the present application adjusts system；
Fig. 8 is the structural schematic diagram that a kind of ultrasound image provided by the embodiments of the present application adjusts equipment；
Fig. 9 is another structural schematic diagram that a kind of ultrasound image provided by the embodiments of the present application adjusts equipment.
Specific embodiment
Below in conjunction with the attached drawing in the embodiment of the present application, technical solutions in the embodiments of the present application carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of embodiments of the present application, instead of all the embodiments.It is based on
Embodiment in the application, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall in the protection scope of this application.
Referring to Fig. 1, Fig. 1 is a kind of first pass figure of ultrasound image method of adjustment provided by the embodiments of the present application.
A kind of ultrasound image method of adjustment provided by the embodiments of the present application, may comprise steps of:
Step S101: the initial ultrasound image of ultrasonic device shooting is obtained.
In practical application, the initial ultrasound image of ultrasonic device shooting, the ultrasound figure of ultrasonic device shooting can be first obtained
The type of picture can determine according to actual needs, and the application is not specifically limited herein.
Step S102: transmission initial ultrasound image to preparatory trained image recognition neural network model.
It, can be by initial ultrasound image transmitting to training in advance after obtaining initial ultrasound image in practical application
Image recognition neural network model in, initial ultrasound image is identified by image recognition neural network model, is obtained
Characteristic image in initial ultrasound image.In concrete application scene, the structure of image recognition neural network model can be according to figure
Type as accuracy of identification, neural network model etc. determines.
Step S103: the feature point diagram that image recognition neural network model identifies in initial ultrasound image is received
Picture.
Step S104: three-dimensional visualization adjustment is carried out to initial ultrasound image based on feature point image.
In practical application, after image recognition neural network model obtains the feature point image in initial ultrasound image, just
It can receive the feature point image that image recognition neural network model obtains, to be based on feature point image to initial ultrasound image
Only three-dimensional visualization adjusts, for example can carry out three-dimensional visualization to initial ultrasound image according to the location information of feature point image
Adjustment etc., for example initial ultrasound image translated, rotated etc. and the information of target object in initial ultrasound image to the greatest extent may be used
Observer being presented to energy, so that observer such as observes at the target object more.
A kind of ultrasound image method of adjustment provided by the present application obtains the initial ultrasound image of ultrasonic device shooting；Transmission
Initial ultrasound image is to preparatory trained image recognition neural network model；Image recognition neural network model is received initial
The feature point image identified in ultrasound image；Three-dimensional visualization tune is carried out to initial ultrasound image based on feature point image
It is whole.A kind of ultrasound image method of adjustment provided by the present application, it is automatic by preparatory trained image recognition neural network model
Received initial ultrasound image is identified, obtains feature point image, then based on feature point image to initial ultrasound image into
Row three-dimensional visualization adjustment, so that without manually three-dimensional visualization adjustment can be carried out to ultrasound image, with existing skill
Art is compared, and the regulated efficiency that three-dimensional visualization is carried out to ultrasound image can be improved.
Referring to Fig. 2, Fig. 2 is a kind of second flow chart of Ultrasound Image Recognition Method provided by the embodiments of the present application.
In practical application, a kind of Ultrasound Image Recognition Method provided by the embodiments of the present application be may comprise steps of:
Step S201: the initial ultrasound image of ultrasonic device shooting is obtained.
Step S202: transmission initial ultrasound image to preparatory trained image recognition neural network model.
Step S203: the feature point diagram that image recognition neural network model identifies in initial ultrasound image is received
Picture.
Step S204: the view-finder parameter of ultrasonic device is obtained.
Step S205: the translation rotation matrix of initial ultrasound image is calculated based on feature point image and view-finder parameter.
Step S206: translation rotation matrix is multiplied with initial ultrasound image, obtains three-dimensional visualization target adjusted
Ultrasound image.
In practical application, during the three-dimensional visualization adjustment to initial ultrasound image, due to the ultrasound image of acquisition
Information influenced by view-finder, in simple terms, when view-finder from different perspectives finds a view to target object, shooting is obtained
Ultrasound image be different, and observer be initial ultrasound image is observed by view-finder, then, for the ease of see
The person of examining observes initial ultrasound image by view-finder, when carrying out three-dimensional visualization adjustment to initial ultrasound image, can combine
The relevant parameter for shooting obtained initial ultrasound image and view-finder is adjusted initial ultrasound image, such as in order to ajust just
Position of the beginning ultrasound image in view-finder can calculate the flat of initial ultrasound image based on feature point image and view-finder parameter
Spin matrix is moved, translation rotation matrix is multiplied with initial ultrasound image, obtains three-dimensional visualization target ultrasound figure adjusted
Picture.The parameter of view-finder can be apex location parameter, the location parameter of view-finder line of symmetry etc. of view-finder.It should be pointed out that
In concrete application scene, can also according to initial ultrasound image and view-finder parameter to view-finder photographic subjects object when position
It is adjusted, so that the ultrasound image of shooting view-finder next time meets observation requirement of observer etc..
In the application Fig. 1 and embodiment shown in Fig. 2, the training process of image neural network model determines that image is known
Other neural network model is to the identification method of ultrasound image and accuracy of identification etc., for this purpose, ultrasound image provided by the present application adjusts
In the process, it before transmission initial ultrasound image to preparatory trained image recognition neural network model, can also obtain just
Beginning ultrasound image training sample；Obtain the corresponding sample label of initial ultrasound image training sample；It is instructed based on initial ultrasound image
Practice sample and sample label, initial image recognition neural network model is trained, obtains trained image in advance and know
Other neural network model.It should be pointed out that initial image recognition neural network model can be the neural network mould currently built
Type, or the neural network model etc. generated in historical process.To initial image recognition neural network model progress
In trained process, the external world such as user can set the default training parameter such as learning rate of image recognition neural network model, scheme
As identifying that neural network model after receiving initial ultrasound image training sample and sample label, can first be based on initial ultrasound
Image training sample generating process label, Kernel-based methods label and sample label determine the real-time training parameter of itself, judge reality
When training parameter it is whether consistent with default training parameter, if it is not, then adjust own net model parameter, return and execute based on initial
Ultrasound image training sample generating process label and later step, if so, obtaining trained image recognition neural network
Model.
In practical application, during being trained to image recognition neural network model, for the ease of obtaining training
Sample can obtain training sample by mathematical logic, then obtain the corresponding sample label of initial ultrasound image training sample
When, the initial ultrasound image labeling sample after the available feature object mark in initial ultrasound image training sample；It is right
Initial ultrasound image labeling sample after mark carries out Gauss operation, obtains sample label.Also it can pass through Gauss operation pair
Initial ultrasound image training sample carries out operation, obtains sample label only comprising feature object.
In practical application, due to including the diversity of information in ultrasound image, directly to the feature object in ultrasound image
If being identified, recognition speed is slow, and precision may be poor, in order to improve such situation, can only extract in ultrasound image
Characteristic point so as to ultrasound image carry out three-dimensional visualization adjustment, then obtain to the spy in initial ultrasound image training sample
Levy object marking after initial ultrasound image labeling sample when, the available characteristic area in initial ultrasound image training sample
The characteristic area mark sample of the initial ultrasound image obtained after the mark of domain；It obtains to the spy in initial ultrasound image training sample
The characteristic point mark sample of the initial ultrasound image obtained after sign point mark；Wherein, characteristic point is on characteristic area, characteristic area
At least two is respectively included with characteristic point, and characteristic point can be corresponded with characteristic area.Namely first initial ultrasound image is instructed
The characteristic area practiced in sample is labeled, and obtains initial ultrasound image characteristic region mark sample, then to initial ultrasound image
Characteristic point in training sample is labeled, and obtains initial ultrasound image characteristic point mark sample, special by initial ultrasound image
Sign area marking sample and initial ultrasound image characteristic point mark sample are trained image recognition neural network model, this
When, since characteristic point is characterized the point in region, so image recognition neural network model is patrolled what ultrasonic device was identified
Volume are as follows: image recognition neural network model first identifies the characteristic area in ultrasound image, identifies to characteristic area, obtains
To satisfactory characteristic point.
In concrete application scene, image recognition neural network model may include outline identification network and Feature point recognition net
Network；Outline identification network is used to carry out feature extraction to initial ultrasound image, obtains feature regional images；Feature point recognition network
For characteristic area mark sample and initial ultrasound image to be carried out Fusion Features, feature point image is obtained；Correspondingly, being based on
Initial ultrasound image training sample and sample label are trained the initial image recognition neural network model built, obtain
When to preparatory trained image recognition neural network model, initial ultrasound image training sample and characteristic area mark can be based on
Sample is infused, outline identification network is trained, trained outline identification network is obtained；Based on initial ultrasound image training sample
Originally, characteristic area mark sample and characteristic point mark sample, are trained to Feature point recognition network, obtain trained feature
Point identification network.
In practical application, in order to further ensure image recognition neural network model output feature point image it is accurate
Property, image recognition neural network model can also include image enhancement network, for according to other characteristic points and target feature point
Positional relationship the target feature point is enhanced, obtain characteristic point enhancing image, also can be according to other characteristic points
Positional relationship between target feature point is modified the position of target feature point；Image enhancement network is used for a feature
Point image carries out convolution algorithm, obtains convolution algorithm result；Convolution algorithm result and another feature point image are cascaded, obtained
To cascade result；Residual error operation is carried out to cascade result, obtains the characteristic point enhancing image of another feature point image.
In concrete application scene, when the quantity of characteristic area and characteristic point is 2 and corresponds, know receiving image
When the characteristic image that other neural network model identifies in initial ultrasound image, image recognition neural network mould can receive
The fisrt feature point image and second feature point image that type identifies in initial ultrasound image；Image recognition neural network mould
The process that type identifies initial ultrasound image can be refering to Fig. 3.
Fig. 3 is the flow chart that image recognition neural network model identifies initial ultrasound image.
The process that image recognition neural network model identifies initial ultrasound image can be with are as follows:
Step S301: feature extraction is carried out to initial ultrasound image, obtains fisrt feature area image and second feature area
Area image.
Step S302: fisrt feature area image and initial ultrasound image are subjected to Fusion Features, obtain fisrt feature point
Candidate image.
Step S303: second feature area image and initial ultrasound image are subjected to Fusion Features, obtain second feature point
Candidate image.
In order to make it easy to understand, being illustrated so that initial ultrasound image is basin baselap acoustic image as an example, in basin baselap acoustic image
Characteristic area can be lower edge and anus rectum angle leading edge after pubic symphysis, referring to Fig. 4, Fig. 4 is special in basin baselap acoustic image
The acquisition schematic diagram of the candidate figure of sign point.
Step S304: second feature point candidate image is subjected to convolution operation, and is carried out with fisrt feature point candidate image
Cascade, then residual error operation is carried out to the first level link fruit and obtains fisrt feature point image.
Step S305: fisrt feature point candidate image is subjected to convolution operation, and is carried out with second feature point candidate image
Cascade, then residual error operation is carried out to the second level link fruit and obtains second feature point image.
In practical application, there may be correlations between the characteristic point in ultrasound image, it is possible to by between characteristic point
Correlation handles characteristic point candidate's figure, obtains more accurate feature point diagram, please refers to Fig. 5 and Fig. 6, and Fig. 5 is spy
The candidate figure of sign point obtains the schematic diagram of feature point diagram, and wherein conv indicates convolution operation,Indicate cascade operation, residual
Block indicates residual error operation, and a indicates that fisrt feature point candidate figure, b indicate that second feature point candidate figure, A indicate fisrt feature point
Enhancing figure；Fig. 6 is the effect diagram of ultrasonic device method of adjustment provided by the present application, it will be appreciated from fig. 6 that provided by the present application super
The adjustment effect of acoustic equipment method of adjustment is preferable.
In practical application, in order to further increase the accuracy of image recognition neural network model, to image recognition mind
In training process through network model, the principle that network can be fought based on production carries out image recognition neural network model
Training, image recognition neural network model includes arbiter at this time, then is based on initial ultrasound image training sample and sample label,
Initial image recognition neural network model is trained, when obtaining preparatory trained image recognition neural network model,
Production confrontation net can be built using initial image recognition neural network model as the generator in production confrontation network
Arbiter in network is trained generator based on initial ultrasound image training sample, sample label and arbiter, until
To preparatory trained image recognition neural network model；Wherein, arbiter may include sequentially connected first operation layer,
Two operation layers, third operation layer, the first full articulamentum, relu (Rectified Linear Units) layer, the second full articulamentum,
Active coating；First operation layer, the second operation layer and third operation layer are by convolutional layer, BN (Batch Normalization) layer
With LRelu ((Leaky-ReLU) layer composition.
It, can be by scheming when being trained using arbiter to image recognition neural network model in concrete application scene
As identification neural network model exports the label obtained after itself identifies initial ultrasound image training sample as a result, label
As a result type can be determined according to the type of sample label, for example when sample label is characterized point image, label result is just
Feature point image, when sample label is characterized enhancing image, label result is just characterized enhancing image etc., transmits label knot
For fruit to arbiter, arbiter compares label result again and whether sample label meets default approach condition, for example compares label knot
Whether the similitude between fruit and sample mark meets similarity Condition, which is fed back to image recognition nerve net by arbiter
Network model, image recognition neural network model adjust the parameter of itself further according to the similitude, repeat the above process, until image
Identification neural network model does not adjust the parameter of itself further according to the similitude, just obtains trained image recognition neural network
Model；In the process, arbiter can export Boolean according to the similitude between sample label and label result, by boolean
Value indicates similitude between sample label and sample results etc..
Present invention also provides a kind of ultrasound images to adjust system, with a kind of ultrasound figure provided by the embodiments of the present application
The correspondence effect having as method of adjustment.Referring to Fig. 7, Fig. 7 is a kind of ultrasound image adjustment system provided by the embodiments of the present application
The structural schematic diagram of system.
A kind of ultrasound image provided by the embodiments of the present application adjusts system, may include:
First obtains module 101, for obtaining the initial ultrasound image of ultrasonic device shooting；
First transmission module 102 is used for transmission initial ultrasound image to preparatory trained image recognition neural network mould
Type；
First receiving module 103 is identified for receiving image recognition neural network model in initial ultrasound image and is obtained
Feature point image；
The first adjustment module 104, for carrying out three-dimensional visualization adjustment to initial ultrasound image based on feature point image.
A kind of ultrasound image provided by the embodiments of the present application adjusts system, and the first adjustment module may include:
First acquisition submodule, for obtaining the view-finder parameter of ultrasonic device；
First computational submodule, for calculating the translation rotation of initial ultrasound image based on feature point image and view-finder parameter
Torque battle array；
First operation submodule obtains three-dimensional visualization tune for translation rotation matrix to be multiplied with initial ultrasound image
Target ultrasound image after whole.
A kind of ultrasound image provided by the embodiments of the present application adjusts system, can also include:
Second obtains module, transmits initial ultrasound image to trained image recognition mind in advance for the first transmission module
Before network model, initial ultrasound image training sample is obtained；
Third obtains module, for obtaining the corresponding sample label of initial ultrasound image training sample；
First training module, for being based on initial ultrasound image training sample and sample label, to the initial figure built
As identifying that neural network model is trained, preparatory trained image recognition neural network model is obtained.
A kind of ultrasound image provided by the embodiments of the present application adjusts system, and third obtains module and may include:
Second acquisition submodule, for obtaining to initial after the feature object mark in initial ultrasound image training sample
Ultrasound image marks sample；
Second operation submodule obtains sample for carrying out Gauss operation to the initial ultrasound image labeling sample after mark
This label.
A kind of ultrasound image provided by the embodiments of the present application adjusts system, and the second acquisition submodule may include:
First acquisition unit, it is first to being obtained after the characteristic area mark in initial ultrasound image training sample for obtaining
The characteristic area of beginning ultrasound image marks sample；
Second acquisition unit, it is initial to being obtained after the characteristic point mark in initial ultrasound image training sample for obtaining
The characteristic point of ultrasound image marks sample；Wherein, on characteristic area, characteristic area and characteristic point respectively include at least characteristic point
2.
A kind of ultrasound image provided by the embodiments of the present application adjusts system, and image recognition neural network model may include wheel
Exterior feature identification network and Feature point recognition network；Outline identification network is used to carry out feature extraction to initial ultrasound image, obtains spy
Levy area image；Feature point recognition network is used to characteristic area mark sample and initial ultrasound image carrying out Fusion Features, obtains
To feature point image；
Correspondingly, the first training module may include:
First training submodule, for marking sample based on initial ultrasound image training sample and characteristic area, to profile
Identification network is trained, and obtains trained outline identification network；
Second training submodule, for based on initial ultrasound image training sample, characteristic area mark sample and characteristic point
Sample is marked, Feature point recognition network is trained, trained Feature point recognition network is obtained.
A kind of ultrasound image provided by the embodiments of the present application adjusts system, and image recognition neural network model can also include
Image enhancement network；
Image enhancement network is used to carry out convolution algorithm to a feature point image, obtains convolution algorithm result；Convolution is transported
It calculates result to be cascaded with another feature point image, obtains cascade result；Residual error operation is carried out to cascade result, obtains another spy
The characteristic point for levying point image enhances image.
A kind of ultrasound image provided by the embodiments of the present application adjusts system, and the first training module may include:
First transmission unit, for initial ultrasound image training sample and sample label to be transmitted to image recognition nerve net
Network model；
Third acquiring unit, for obtaining the label result of image recognition neural network model output；
Second transmission unit, for sample label and label result to be transmitted to the arbiter built in advance, arbiter is used
In judging label result and whether sample label meets preset condition；
First receiving unit, for receiving the judging result of arbiter output, transmission judging result to image recognition nerve
Network model, so that image recognition neural network model is based on judging result and adjusts inherent parameters；
First execution unit, for repeating the label result and later for obtaining the output of image recognition neural network model
The step of, until obtaining trained described image identification neural network model in advance.
Present invention also provides a kind of ultrasound image adjustment equipment and computer readable storage mediums, all have the application
A kind of correspondence effect that ultrasound image method of adjustment has that embodiment provides.Referring to Fig. 8, Fig. 8 mentions for the embodiment of the present application
A kind of structural schematic diagram of ultrasound image adjustment equipment supplied.
A kind of ultrasound image provided by the embodiments of the present application adjusts equipment, comprising:
Memory 201, for storing computer program；
Processor 202 realizes that ultrasound image described in any embodiment as above adjusts when for executing computer program
The step of method.
Referring to Fig. 9, can also include: in another kind ultrasound image adjustment equipment provided by the embodiments of the present application and handle
The input port 203 that device 202 connects is used for transmission the extraneous order inputted to processor 202；What is connect with processor 202 is aobvious
Show unit 204, the processing result for video-stream processor 202 is to the external world；The communication module 205 connecting with processor 202, is used for
Realize ultrasound image adjustment equipment and extraneous communication.Display unit 204 can make display for display panel, laser scanning
Deng；Communication mode used by communication module 205 includes but is not limited to that mobile high definition chained technology (HML), general serial are total
Line (USB), is wirelessly connected high-definition media interface (HDMI): adopting wireless fidelity technology (WiFi), Bluetooth Communication Technology, low-power consumption
Bluetooth Communication Technology, the communication technology based on IEEE802.11s.
A kind of computer readable storage medium provided by the embodiments of the present application is stored with meter in computer readable storage medium
Calculation machine program realizes ultrasound image method of adjustment described in any embodiment as above when computer program is executed by processor
Step.
Computer readable storage medium involved in the application includes random access memory (RAM), memory, read-only memory
(ROM), electrically programmable ROM, electrically erasable ROM, register, hard disk, moveable magnetic disc, CD-ROM or technical field
Any other form of storage medium well known to interior.
A kind of ultrasound image provided by the embodiments of the present application adjusts related in system, equipment and computer readable storage medium
Partial explanation refers to the detailed description of corresponding part in a kind of ultrasound image method of adjustment provided by the embodiments of the present application,
This is repeated no more.In addition, in above-mentioned technical proposal provided by the embodiments of the present application with correspond in the prior art technical solution realize
The consistent part of principle is simultaneously unspecified, in order to avoid excessively repeat.
It should also be noted that, herein, relational terms such as first and second and the like are used merely to one
Entity or operation are distinguished with another entity or operation, without necessarily requiring or implying between these entities or operation
There are any actual relationship or orders.Moreover, the terms "include", "comprise" or its any other variant are intended to contain
Lid non-exclusive inclusion, so that the process, method, article or equipment including a series of elements is not only wanted including those
Element, but also including other elements that are not explicitly listed, or further include for this process, method, article or equipment
Intrinsic element.In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded that
There is also other identical elements in process, method, article or equipment including the element.
The foregoing description of the disclosed embodiments makes those skilled in the art can be realized or use the application.To this
A variety of modifications of a little embodiments will be apparent for a person skilled in the art, and the general principles defined herein can
Without departing from the spirit or scope of the application, to realize in other embodiments.Therefore, the application will not be limited
It is formed on the embodiments shown herein, and is to fit to consistent with the principles and novel features disclosed in this article widest
Range.
Claims (11)
1. a kind of ultrasound image method of adjustment characterized by comprising
Obtain the initial ultrasound image of ultrasonic device shooting；
The initial ultrasound image is transmitted to preparatory trained image recognition neural network model；
Receive the feature point image that described image identification neural network model identifies in the initial ultrasound image；
Three-dimensional visualization adjustment is carried out to the initial ultrasound image based on the feature point image.
2. the method according to claim 1, wherein described be based on the feature point image to the initial ultrasound
Image carries out three-dimensional visualization adjustment, comprising:
Obtain the view-finder parameter of the ultrasonic device；
The translation rotation matrix of the initial ultrasound image is calculated based on the feature point image and the view-finder parameter；
The translation rotation matrix is multiplied with the initial ultrasound image, obtains three-dimensional visualization target ultrasound figure adjusted
Picture.
3. method according to claim 1 or 2, which is characterized in that the transmission initial ultrasound image to preparatory instruction
Before the image recognition neural network model perfected, further includes:
Obtain initial ultrasound image training sample；
Obtain the corresponding sample label of the initial ultrasound image training sample；
Based on the initial ultrasound image training sample and the sample label, nerve is identified to the initial described image built
Network model is trained, and obtains trained described image identification neural network model in advance.
4. according to the method described in claim 3, it is characterized in that, the acquisition initial ultrasound image training sample is corresponding
Sample label, comprising:
It obtains to the initial ultrasound image labeling sample after the feature object mark in the initial ultrasound image training sample；
Gauss operation is carried out to the initial ultrasound image labeling sample after the mark, obtains the sample label.
5. according to the method described in claim 4, it is characterized in that, the acquisition is in the initial ultrasound image training sample
Feature object mark after initial ultrasound image labeling sample, comprising:
Obtain the feature to the initial ultrasound image obtained after the characteristic area mark in the initial ultrasound image training sample
Area marking sample；
Obtain the characteristic point to the initial ultrasound image obtained after the characteristic point mark in the initial ultrasound image training sample
Mark sample；Wherein, for the characteristic point on the characteristic area, the characteristic area and characteristic point respectively include at least two.
6. according to the method described in claim 5, it is characterized in that, described image identification neural network model includes outline identification
Network and Feature point recognition network；The outline identification network is used to carry out feature extraction to the initial ultrasound image, obtains
Feature regional images；The Feature point recognition network be used for by the characteristic area mark sample and the initial ultrasound image into
Row Fusion Features obtain the feature point image；
It is described to be based on the initial ultrasound image training sample and the sample label, the initial described image built is identified
Neural network model is trained, and obtains trained described image identification neural network model in advance, comprising:
Sample is marked based on the initial ultrasound image training sample and the characteristic area, the outline identification network is carried out
Training, obtains the trained outline identification network；
Sample is marked based on the initial ultrasound image training sample, characteristic area mark sample and the characteristic point, it is right
The Feature point recognition network is trained, and obtains the trained Feature point recognition network.
7. according to the method described in claim 6, it is characterized in that, described image identification neural network model further includes that image increases
Strong network is obtained for being enhanced according to the positional relationship of other characteristic points and target feature point the target feature point
Characteristic point enhances image；
Described image enhances network and is used to carry out convolution algorithm to a feature point image, obtains convolution algorithm result；By the volume
Product operation result is cascaded with another feature point image, obtains cascade result；Residual error operation is carried out to the cascade result, is obtained
Characteristic point to another feature point image enhances image.
8. the method according to the description of claim 7 is characterized in that described be based on the initial ultrasound image training sample and institute
Sample label is stated, the initial described image identification neural network model built is trained, preparatory trained institute is obtained
State image recognition neural network model, comprising:
The initial ultrasound image training sample and the sample label are transmitted to described image identification neural network model；
Obtain the label result of described image identification neural network model output；
The sample label and the label result are transmitted to the arbiter built in advance, the arbiter is described for judging
Whether label result and the sample label meet default approach condition；
The judging result for receiving the arbiter output transmits the judging result to described image and identifies neural network model,
So that described image identification neural network model is based on the judging result and adjusts inherent parameters；
Repeat it is described acquisition described image identification neural network model output label result and later the step of, until
Neural network model is identified to trained described image in advance, and described image identifies that neural network model includes the differentiation
Device.
9. a kind of ultrasound image adjusts system characterized by comprising
First obtains module, for obtaining the initial ultrasound image of ultrasonic device shooting；
First transmission module is used for transmission the initial ultrasound image to preparatory trained image recognition neural network model；
First receiving module is identified in the initial ultrasound image and is obtained for receiving described image identification neural network model
Feature point image；
The first adjustment module, for carrying out three-dimensional visualization adjustment to the initial ultrasound image based on the feature point image.
10. a kind of ultrasonic device characterized by comprising
Memory, for storing computer program；
Processor realizes the ultrasound image adjustment side as described in any one of claim 1 to 8 when for executing the computer program
The step of method.
11. a kind of computer readable storage medium, which is characterized in that be stored with computer in the computer readable storage medium
Program realizes the ultrasound image method of adjustment as described in any one of claim 1 to 8 when the computer program is executed by processor
The step of.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910604343.1A CN110322399B (en) | 2019-07-05 | 2019-07-05 | Ultrasonic image adjustment method, system, equipment and computer storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910604343.1A CN110322399B (en) | 2019-07-05 | 2019-07-05 | Ultrasonic image adjustment method, system, equipment and computer storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110322399A true CN110322399A (en) | 2019-10-11 |
| CN110322399B CN110322399B (en) | 2023-05-05 |
Family
ID=68122801
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910604343.1A Active CN110322399B (en) | 2019-07-05 | 2019-07-05 | Ultrasonic image adjustment method, system, equipment and computer storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110322399B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110840482A (en) * | 2019-10-28 | 2020-02-28 | 苏州佳世达电通有限公司 | Ultrasonic imaging system and method thereof |
| CN111222560A (en) * | 2019-12-30 | 2020-06-02 | 深圳大学 | Image processing model generation method, intelligent terminal and storage medium |
| 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 |
| CN112184683A (en) * | 2020-10-09 | 2021-01-05 | 深圳度影医疗科技有限公司 | Ultrasonic image identification method, terminal equipment and storage medium |
| TWI719687B (en) * | 2019-10-25 | 2021-02-21 | 佳世達科技股份有限公司 | Ultrasound imaging system and ultrasound imaging method |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107174342A (en) * | 2017-03-21 | 2017-09-19 | 哈尔滨工程大学 | A kind of area of computer aided fracture reduction degree measure |
| CN109117773A (en) * | 2018-08-01 | 2019-01-01 | Oppo广东移动通信有限公司 | A kind of characteristics of image point detecting method, terminal device and storage medium |
-
2019
- 2019-07-05 CN CN201910604343.1A patent/CN110322399B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107174342A (en) * | 2017-03-21 | 2017-09-19 | 哈尔滨工程大学 | A kind of area of computer aided fracture reduction degree measure |
| CN109117773A (en) * | 2018-08-01 | 2019-01-01 | Oppo广东移动通信有限公司 | A kind of characteristics of image point detecting method, terminal device and storage medium |
Non-Patent Citations (1)
| Title |
|---|
| 刘晓等: "基于BP神经网络的超声图像识别方法的研究", 《中国医疗器械杂志》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TWI719687B (en) * | 2019-10-25 | 2021-02-21 | 佳世達科技股份有限公司 | Ultrasound imaging system and ultrasound imaging method |
| CN110840482A (en) * | 2019-10-28 | 2020-02-28 | 苏州佳世达电通有限公司 | Ultrasonic imaging system and method thereof |
| CN111222560A (en) * | 2019-12-30 | 2020-06-02 | 深圳大学 | Image processing model generation method, intelligent terminal and storage medium |
| CN111222560B (en) * | 2019-12-30 | 2022-05-20 | 深圳大学 | Image processing model generation method, intelligent terminal and storage medium |
| 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 |
| CN112184683A (en) * | 2020-10-09 | 2021-01-05 | 深圳度影医疗科技有限公司 | Ultrasonic image identification method, terminal equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110322399B (en) | 2023-05-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110322399A (en) | A kind of ultrasound image method of adjustment, system, equipment and computer storage medium | |
| CN110338844B (en) | Three-dimensional imaging data display processing method and three-dimensional ultrasonic imaging method and system | |
| CN104036488B (en) | Binocular vision-based human body posture and action research method | |
| CN102525381B (en) | The recording equipment of image processing apparatus, image processing method and embodied on computer readable | |
| Xu et al. | Predicting animation skeletons for 3d articulated models via volumetric nets | |
| US20180174311A1 (en) | Method and system for simultaneous scene parsing and model fusion for endoscopic and laparoscopic navigation | |
| CN110196053A (en) | A kind of real-time field robot vision navigation method and system based on FPGA | |
| CN109410168A (en) | For determining the modeling method of the convolutional neural networks model of the classification of the subgraph block in image | |
| WO2020023467A1 (en) | Unique identification of freely swimming fish in an aquaculture environment | |
| CN106355147A (en) | Acquiring method and detecting method of live face head pose detection regression apparatus | |
| CN108898634A (en) | Pinpoint method is carried out to embroidery machine target pinprick based on binocular camera parallax | |
| CN102567734B (en) | Specific value based retina thin blood vessel segmentation method | |
| CN109508787A (en) | Neural network model training method and system for ultrasound displacement estimation | |
| CN111768375A (en) | Asymmetric GM multi-mode fusion significance detection method and system based on CWAM | |
| CN111055275A (en) | Action simulation method and device, computer readable storage medium and robot | |
| CN109171817A (en) | Three-dimensional breast ultrasound scan method and ultrasonic scanning system | |
| CN105574812A (en) | Multi-angle three-dimensional data registration method and device | |
| CN109146868A (en) | 3D Lung neoplasm generation method, device and electronic equipment | |
| CN110363170A (en) | Video is changed face method and apparatus | |
| CN112348909A (en) | Target positioning method, device, equipment and storage medium | |
| CN110211069A (en) | A kind of image denoising model and training method, system, equipment, computer media | |
| JP6944492B2 (en) | Image acquisition method, related equipment and readable storage medium | |
| CN106096565B (en) | The task cooperative method of mobile robot and static sensor based on sensing network | |
| CN113723380B (en) | Face recognition method, device, equipment and storage medium based on radar technology | |
| CN111783637B (en) | Key point labeling method and device, and target object space pose determining method and device |
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 |