CN107909622A - Model generating method, the scanning planing method of medical imaging and medical image system - Google Patents
Model generating method, the scanning planing method of medical imaging and medical image system Download PDFInfo
- Publication number
- CN107909622A CN107909622A CN201711238100.8A CN201711238100A CN107909622A CN 107909622 A CN107909622 A CN 107909622A CN 201711238100 A CN201711238100 A CN 201711238100A CN 107909622 A CN107909622 A CN 107909622A
- Authority
- CN
- China
- Prior art keywords
- sample
- scan images
- model
- distance field
- smart network
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T11/00—2D [Two Dimensional] image generation
- G06T11/003—Reconstruction from projections, e.g. tomography
- G06T11/005—Specific pre-processing for tomographic reconstruction, e.g. calibration, source positioning, rebinning, scatter correction, retrospective gating
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2211/00—Image generation
- G06T2211/40—Computed tomography
- G06T2211/428—Real-time
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
An embodiment of the present invention provides model generating method, the scanning planing method of medical imaging and medical image system, it is related to medical image automatically scanning planning technology field, smart network's model is determined based on deep learning method, and then sweep parameter is obtained by smart network's model, it ensure that the precision of sweep parameter.This method includes model and generates and use process：Model generating process includes：Obtain sample pre-scan images and the corresponding scan sample parameter of sample pre-scan images；Sample distance fields are generated according to scan sample parameter；Sample pre-scan images are learnt with sample distance fields by smart network, generate smart network's model.Model is included using process：Obtain the pre-scan images of person under inspection target area；Pre-scan images are handled by smart network's model, obtain distance field；According to distance field, sweep parameter is determined.Technical solution provided in an embodiment of the present invention, suitable for scanning planning process.
Description
【Technical field】
The present invention relates to medical image automatically scanning planning technology field, more particularly to model generating method, medical imaging
Scanning planing method and medical image system.
【Background technology】
In clinical medicine domain, it is often necessary to person under inspection target area is scanned and obtains medical scanning image, with auxiliary
Help diagnosis, treatment of disease etc..
For conventional magnetic resonance scan sequences, usually (it is also referred to as by the pre-scan images of low resolution by doctor first
For locating plate) identify anatomical location, reference location line and reference center are then demarcated by hand, so that it is determined that required sweep
Retouch parameter, such as scanning direction, sweep center point, scanning center's axis and the size for scanning window.After sweep parameter is determined, root
Scanning sequence is applied to anatomical location according to sweep parameter and carries out accurate scan.
In current magnetic resonance scan sequences, by manually demarcating definite sweep parameter, this method is less efficient, extends
The whole scanning process required time, and due to labour variance, experience difference, it is likely to result in recognition result, calibration
As a result inconsistent, the precision of sweep parameter can not ensure.
【The content of the invention】
In view of this, an embodiment of the present invention provides a kind of model generating method, medical imaging scanning planing method and
Medical image system, based on deep learning, by smart network's model to the pre-scan images of person under inspection target area into
Row processing, and then determine sweep parameter, the time required for scanning process can be reduced, and ensure the precision of sweep parameter.
In a first aspect, the embodiment of the present invention provides a kind of model generating method, the described method includes：
Obtain sample pre-scan images and the corresponding scan sample parameter of the sample pre-scan images, the sample are swept
Parameter is retouched including at least the one of which in sweep center point, scanning center's axis, scanning window size；
According to the scan sample parameter, sample distance fields are generated；
The sample pre-scan images are learnt with the sample distance fields by smart network, are obtained described
The mapping relations of sample pre-scan images and the sample distance fields；
Smart network's model is generated according to the mapping relations.
Aspect as described above and any possible implementation, it is further provided a kind of implementation, passes through described
Smart network learns the sample pre-scan images with the sample distance fields, obtains the sample prescan figure
As with before the mapping relations of the sample distance fields, the method further includes：
The characteristic information of the sample pre-scan images is extracted, the characteristic information of the sample pre-scan images includes at least
The three-dimensional gradient information of the sample pre-scan images, gradient block message, image block message, structure tensor information, 3 D wavelet
One of which in coefficient, 2-d wavelet coefficient；
Then the sample pre-scan images are learnt with the sample distance fields by smart network, obtain institute
The mapping relations of sample pre-scan images and the sample distance fields are stated, including：
Pass through characteristic information of the smart network to the sample pre-scan images and the sample distance fields
Practise, obtain the characteristic information of the sample pre-scan images and the mapping relations of the sample distance fields.
Second aspect, the embodiment of the present invention provide a kind of scanning planing method of medical imaging, the described method includes：
Obtain the pre-scan images of person under inspection target area；
The pre-scan images are handled by smart network's model, obtain distance field, the artificial intelligence
Network model includes the pre-scan images and the mapping relations of the distance field；
According to the distance field, sweep parameter is determined, the sweep parameter includes at least sweep center point, scanning center
One of which in axis, scanning window size, scanning direction.
Aspect as described above and any possible implementation, it is further provided a kind of implementation, in the basis
The distance field, after determining sweep parameter, the method further includes：
Scanning is performed to the person under inspection target area based on the sweep parameter.
Aspect as described above and any possible implementation, it is further provided a kind of implementation, passes through described
Before smart network's model handles the pre-scan images, the method further includes：
Extract the characteristic information of the pre-scan images, the characteristic informations of the pre-scan images includes at least described pre- sweep
The three-dimensional gradient information of tracing picture, gradient block message, image block message, structure tensor information, 3 D wavelet coefficient, 2-d wavelet
One of which in coefficient, gradient magnitude information, gradient units vector information；
It is then described that the pre-scan images are handled by smart network's model, distance field is obtained, including：
The characteristic information of the pre-scan images is handled by smart network's model, obtains distance field, institute
State smart network's model and include the characteristic information of the pre-scan images and the mapping relations of the distance field.
Aspect as described above and any possible implementation, it is further provided a kind of implementation, it is described to pass through people
Work intelligent network model handles the pre-scan images, obtains distance field, including：
The pre-scan images are handled by smart network's model, obtain the distance based on sweep center point
；
Then according to the distance field, sweep parameter is determined, including：
According to the distance field based on sweep center point, sweep center point is determined.
Aspect as described above and any possible implementation, it is further provided a kind of implementation, it is described to pass through people
Work intelligent network model handles the pre-scan images, obtains distance field, including：
The pre-scan images are handled by smart network's model, obtain the distance based on scanning center's axis
；
Then according to the distance field, sweep parameter is determined, including：
According to the distance field based on scanning center's axis, scanning center's axis is determined.
Aspect as described above and any possible implementation, it is further provided a kind of implementation, it is described to pass through people
Work intelligent network model handles the pre-scan images, obtains distance field, including：
The pre-scan images are handled by smart network's model, obtain non-linear adding window distance field；
Then according to the distance field, sweep parameter is determined, including：
According to the non-linear adding window distance field, scanning window size is determined.
Aspect as described above and any possible implementation, it is further provided a kind of implementation, it is described according to institute
Distance field is stated, determines sweep parameter, including：
Fitting operation is weighted to the distance field, determines sweep parameter；Or,
Hough ballot is carried out to the distance field, determines sweep parameter.
The third aspect, the embodiment of the present invention provide a kind of medical image system, and the medical image system includes processor
And memory；The memory for storing instruction, when described instruction is performed by the processor, causes the medical imaging
System realizes as above either side and the method described in any possible implementation.
An embodiment of the present invention provides a kind of model generating method, the scanning planing method of medical imaging and medical imaging system
System, the method based on deep learning determines smart network's model, by smart network's model to person under inspection target area
The pre-scan images in domain are handled, and obtain corresponding distance field, and then determine sweep parameter by distance field.The present invention is implemented
The technical solution that example provides obtains distance field compared with the prior art, by smart network's model, and then determines scanning ginseng
Number, reduces the time required for whole scanning process, and machine learning is not in the error that artificial or experience is brought, therefore also
It can ensure the precision of sweep parameter.
【Brief description of the drawings】
In order to illustrate the technical solution of the embodiments of the present invention more clearly, below will be to needed in the embodiment attached
Figure is briefly described, it should be apparent that, drawings in the following description are only some embodiments of the present invention, for this area
For those of ordinary skill, without creative efforts, other attached drawings can also be obtained according to these attached drawings.
Fig. 1 is a kind of method flow diagram of model generation provided in an embodiment of the present invention；
Fig. 2 is the schematic diagram of a kind of brain pre-scan images provided in an embodiment of the present invention and anatomical location scanning window；
Fig. 3 is a kind of pixel schematic diagram provided in an embodiment of the present invention；
Fig. 4 is another pixel schematic diagram provided in an embodiment of the present invention；
Fig. 5 is the method flow diagram of another model generation provided in an embodiment of the present invention；
Fig. 6 is a kind of scanning planning method flow diagram of medical imaging provided in an embodiment of the present invention；
Fig. 7 is three-dimensional distance field provided in an embodiment of the present invention schematic diagram；
Fig. 8 is the scanning planning method flow diagram of another medical imaging provided in an embodiment of the present invention；
Fig. 9 is the scanning planning method flow diagram of another medical imaging provided in an embodiment of the present invention；
Figure 10 is the scanning planning method flow diagram of another medical imaging provided in an embodiment of the present invention；
Figure 11 is a kind of flow signal that sweep center point is determined based on three-dimensional gradient information provided in an embodiment of the present invention
Figure；
Figure 12 is the scanning planning method flow diagram of another medical imaging provided in an embodiment of the present invention；
Figure 13 is the scanning planning method flow diagram of another medical imaging provided in an embodiment of the present invention；
Figure 14 is the scanning planning method flow diagram of another medical imaging provided in an embodiment of the present invention；
Figure 15 is the scanning planning method flow diagram of another medical imaging provided in an embodiment of the present invention；
Figure 16 is a kind of schematic diagram of medical image system provided in an embodiment of the present invention；
Figure 17 is a kind of specific medical image system figure provided in an embodiment of the present invention.
【Embodiment】
For a better understanding of the technical solution of the present invention, the embodiment of the present invention is retouched in detail below in conjunction with the accompanying drawings
State.
It will be appreciated that described embodiment is only part of the embodiment of the present invention, instead of all the embodiments.Base
Embodiment in the present invention, those of ordinary skill in the art obtained without creative efforts it is all its
Its embodiment, belongs to the scope of protection of the invention.
The term used in embodiments of the present invention is only merely for the purpose of description specific embodiment, and is not intended to be limiting
The present invention.In the embodiment of the present invention and " one kind " of singulative used in the attached claims, " described " and "the"
It is also intended to including most forms, unless context clearly shows that other implications.
Depending on linguistic context, word as used in this " if " can be construed to " ... when " or " when ...
When " or " in response to determining " or " in response to detection ".Similarly, depending on linguistic context, phrase " if it is determined that " or " if detection
(condition or event of statement) " can be construed to " when definite " or " in response to determining " or " when the detection (condition of statement
Or event) when " or " in response to detecting (condition or event of statement) ".
It should be appreciated that term "and/or" used herein is only a kind of incidence relation for describing affiliated partner, represent
There may be three kinds of relations, for example, A and/or B, can represent：Individualism A, while there are A and B, individualism B these three
Situation.In addition, character "/" herein, it is a kind of relation of "or" to typically represent forward-backward correlation object.
An embodiment of the present invention provides a kind of model generating method, the generation of network model suitable for scanning planning
Journey, as shown in Figure 1, the described method includes：
101st, sample pre-scan images and the corresponding scan sample parameter of the sample pre-scan images are obtained.
Wherein, pre-scan images are the entire scan images to including a larger position for treating anatomical location, are point
The relatively low three-dimensional reconnaissance image of resolution or positioning picture.Sample pre-scan images can be preserved in database it is multiple by
Sample image known to the early period of inspection person or same person under inspection.
Wherein, the corresponding scan sample parameter of the sample pre-scan images includes at least sweep center point, scanning center
One of which in axis, scanning window size, scan sample parameter is given value or determines value.
It should be noted that in order to carry out accurate scan to anatomical location, it usually needs determine to solve by pre-scan images
The suitable scanning window that degree is put is cutd open, the suitable window that scans can determine candidate region in pre-scan images.It is of the invention real
Apply example and be based on this purpose, there is provided the process of corresponding smart network's model, Jin Erke are obtained by sample learning
Directly to obtain scanning window by pre-scan images by smart network's model.
Wherein, the corresponding scan sample parameter of sample pre-scan images (including sweep center point, scanning center's axis, scanning
Window size) it is to determine scan sample window (namely being to determine the distance field of scan sample window) required parameter.
Specifically, in order to become apparent from contrast of the pre-scan images with scanning window, specifically carried out by head scanning
Illustrate, as shown in Fig. 2, (1) is the head pre-scan images of low resolution in figure, it is pre-scan images successively from left to right
Sagittal plane, coronal-plane and cross section, correspondingly, showing that anatomical location scans window at three from left to right in (2) in figure
Schematic diagram on direction.
102nd, according to the scan sample parameter, sample distance fields are generated.
In step 102, multiple sample distance fields are automatically generated by scan sample parameter, sample distance fields can be based on sweeping
Retouch central point or scanning center's axis.Specifically, sample distance fields can be sweep center point or the scanning of foregoing scan sample window
Obtained after central shaft progress range conversion.The range conversion of sample distance fields can be Euclidean Distance Transform (eudlidean
Distance transfrom), chessboard distance conversion (chessboard distance transform) or street range conversion
(city block distance transform).Certainly, it is contemplated that scan coordinate system performs scanning in three directions, removes
Single shaft distance field, can also be made of L-shaped twin shaft distance field the scan axis of both direction.
Also, sample distance fields can be two-dimensional distance field or three-dimensional distance field.Two dimensions of two-dimensional distance field can wrap
Include：Euclidean distance and chessboard distance, either Euclidean distance and city block distance or chessboard distance and city block distance.
Three dimensions of three-dimensional distance field may include：Euclidean distance, chessboard distance and city block distance.
During a concrete implementation, sample distance fields can be three-dimensional distance field D (d1, d2, d3).Wherein d1 pairs
The first dimension is answered, can be replaced with Euclidean distance；D2 corresponds to the second dimension, can be replaced with chessboard distance；D3 corresponds to third dimension, can
Replaced with city block distance.Each dimension it is as follows apart from calculating process：
1) a scan sample window can determine that according to scan sample parameter, the scan sample window is in sample pre-scan images
It can determine that a candidate region, certain sample pre-scan images may also pass through binary conversion treatment.One is included in candidate region
Connected region S, the target pixel points in candidate region are classified, are divided into internal point, external point and isolated point.As shown in figure 3,
By taking four neighborhoods of center pixel as an example, if center pixel is object pixel (value 1) and four neighborhoods are all that (value is object pixel
1), then the point is internal point.If as shown in figure 4, the center pixel is object pixel, four neighborhoods are background pixel (value is 0),
Then the central point is isolated point.Target pixel points in candidate region in addition to internal point and isolated point are then boundary point.
2) all internal points and non-internal point in candidate region are determined, wherein, the collection of internal point composition is combined into R1；Non- inside
The collection of point composition is combined into R2.
3) for the every bit in R1, minimum range of this in R2 is calculated, these minimum ranges form set S3.
Make (x1,y1) represent R1 in any point, make (x2,y2) represent R2 in any point,
When minimum range corresponds to Euclidean distance,
When minimum range corresponds to chessboard distance, d2 ((x1,y1),(x2,y2))=| x1-x2|+|y1-y2|；
When minimum range corresponds to city block distance, d3 ((x1,y1),(x2,y2))=max (| x1-x2|,|y1-y2|)。
103rd, the sample pre-scan images are learnt with the sample distance fields by smart network, obtained
The mapping relations of the sample pre-scan images and the sample distance fields.
Alternatively, artificial intelligence (Artificial Intelligence, AI) network can select BP (Back
Propagation, backpropagation) neutral net, radial base neural net, perceptron neural network, linear neural network, from group
Knit neutral net, Feedback Neural Network, clustering network, deep learning network, feedforward neural network etc..
In a kind of specific embodiment, smart network selects convolutional neural networks (Convolutional
Neural Network, CNN), sample distance fields are three-dimensional samples distance field, and step 103 is passed through with supervised learning method
Recurrent networks learning sample pre-scan images and three-dimensional samples distance field in CNN networks, obtain the mapping relations between them,
Can be with all parameters in automated setting CNN networks, namely generation smart network's model according to the mapping relations.
104th, smart network's model is generated according to the mapping relations.
In above-mentioned model generating method, by smart network to several sample pre-scan images and sample distance fields
Deep learning/machine learning is carried out, network parameter is constantly adjusted and obtains the mapping pass of sample pre-scan images and sample distance fields
System, and then smart network's model is generated according to mapping relations, complete whole model generating process.
Furthermore, it is understood that with reference to preceding method flow, in order to reduce the difficulty of smart network's study, study is improved
Success rate, can also be learnt with sample distance fields according to the characteristic information of sample pre-scan images, establish artificial intelligence net
Network model, therefore the alternatively possible implementation of the embodiment of the present invention additionally provides following methods flow, performs in step
Before 103, as shown in figure 5, including：
105th, the characteristic information of the sample pre-scan images is extracted.
Wherein, the characteristic information of the sample pre-scan images includes at least the three-dimensional gradient of the sample pre-scan images
Wherein one in information, gradient block message, image block message, structure tensor information, 3 D wavelet coefficient, 2-d wavelet coefficient
Kind.
It is following steps that then step 103, which specifically performs,：
1031st, by smart network to the characteristic informations of the sample pre-scan images and the sample distance fields into
Row study, obtains the characteristic information of the sample pre-scan images and the mapping relations of the sample distance fields.
Wherein, the algorithm which uses includes gradient back propagation, gradient descent algorithm etc..
An embodiment of the present invention provides a kind of scanning planing method of medical imaging, by artificial suitable for scanning planning
During intelligent network model determines sweep parameter, scanning planning can be performed by medical imaging devices, as shown in fig. 6, institute
The method of stating includes：
201st, the pre-scan images of person under inspection target area are obtained.
Wherein, the pre-scan images of person under inspection target area are pending pre-scan images, pre-scan images it is specific
Step 101 is shown in explanation, and details are not described herein again.
202nd, the pre-scan images are handled by smart network's model, obtains distance field.
Wherein, smart network's model can indicate that the pre-scan images and the mapping of the distance field are closed
System.The training generation method of smart network's model is shown in above-mentioned model generating method embodiment.
Specifically, in step 202, can be by the characteristic information of pre-scan images, pre-scan images block or pre-scan images
Input smart network's model, and then the dissection degree by being included in smart network's Model Identification pre-scan images
Put, obtain the distance field of corresponding anatomical location scanning window.
203rd, according to the distance field, sweep parameter is determined.
Wherein, the sweep parameter is included at least in sweep center point, scanning center's axis, scanning window size, scanning direction
One of which.It should be noted that scanning direction refers to the application direction of scanning sequence, can after distance field is obtained
Scanning direction is judged with the shape according to distance field.
It should be noted that pre-scan images, distance field, sweep parameter in the embodiment of the present invention, scanning window with it is above-mentioned
Sample pre-scan images, sample distance fields, scan sample parameter, scan sample window in model generation embodiment, actually represent
Meaning does not have difference, and simply the effect in the method flow of place is different.For example sample pre-scan images are carried out as sample
The image of study, and pre-scan images are pending images in the embodiment of the present invention.
The scanning planing method of medical imaging provided in an embodiment of the present invention, passes through the people determined based on deep learning method
Work intelligent network model handles the pre-scan images of person under inspection target area, obtains corresponding distance field, and then pass through
Distance field determines sweep parameter.Technical solution provided in an embodiment of the present invention compared with the prior art, passes through smart network
Model obtains distance field, and then determines sweep parameter, can reduce the time required for whole scanning process, and machine learning will not
There are the error that artificial or experience is brought, therefore it can also ensure the precision of sweep parameter.And in the embodiment of the present invention, distance
Can be variously-shaped distance field, therefore the scanning planing method of medical imaging that the technical program provides, it can be applicable in
In a variety of scanning areas.
Furthermore, it is understood that in a specific example, as shown in fig. 7, with kneed pre-scan images (in figure (1), its
Middle upper left one, upper right one and one small figure of bottom right are pre-scan images of the knee joint on three-dimensional respectively, be respectively cross section,
Coronal-plane and sagittal plane view, the white frame in the every small small frame schematic diagram in the figure lower left corner have substantially marked out dissection degree
Put) exemplified by, by the three-dimensional distance field generated by technical solution provided in an embodiment of the present invention (in figure (2), relevant position in figure
Distance field figure respectively with (1) correspond, wherein the deeper black shaded area of color is three-dimensional distance field, is specifically shown in every
White frame mark part in the small small frame schematic diagram in the figure lower left corner), with the three-dimensional distance generated by known sweep parameter
(in figure (3), the distance field figure of relevant position is corresponded with (1) respectively in figure, wherein the deeper black shaded area of color for field
For three-dimensional distance field, the white frame mark part that is specifically shown in the small frame schematic diagram in the every small figure lower left corner) contrasted, can be with
See, the three-dimensional distance field of three-dimensional distance field and the sweep parameter generation of the technical program generation is about identical, this is fine
The accuracy and practicality that illustrate the technical program.It should be noted that one small figure of lower-left is equal in (1) (2) (3) in Fig. 7
The meaningless supplement figure set for the beauty of schematic diagram.
Furthermore, it is understood that with reference to preceding method flow, after sweep parameter is determined, system will also according to sweep parameter into
Row scanning, therefore the alternatively possible implementation of the embodiment of the present invention also provides following methods flow, performs in step 203
Afterwards, as shown in figure 8, including：
204th, scanning is performed to the person under inspection target area based on the sweep parameter.
Furthermore, it is understood that with reference to preceding method flow, in order to reduce the intractability of artificial nerve network model, Ke Yigen
Distance field is obtained according to the characteristic information of pre-scan images, therefore the alternatively possible implementation of the embodiment of the present invention also provides
Following methods flows, perform before step 202, as shown in figure 9, including：
205th, the characteristic information of the pre-scan images is extracted.
Wherein, the characteristic information of the pre-scan images includes at least the three-dimensional gradient information of the pre-scan images, ladder
Spend the one of which in block message, image block message, structure tensor information, 3 D wavelet coefficient, 2-d wavelet coefficient.
It is following steps that then step 202, which specifically performs,：
2021st, the characteristic information of the pre-scan images is handled by smart network's model, obtains distance
.
Wherein, smart network's model can indicate the characteristic information of the pre-scan images and the distance field
Mapping relations.
Furthermore, it is understood that with reference to preceding method flow, in order to more specifically introduce the determination process of each sweep parameter,
Based on sweep center point, scanning center's axis, scanning window size, the alternatively possible implementation of the embodiment of the present invention is for step
Rapid 202 realization, additionally provides following three kinds of possible implementation methods.
The first implementation method, based on sweep center point, as shown in Figure 10, including：
2022nd, the pre-scan images are handled by smart network's model, obtains being based on sweep center point
Distance field.
Wherein, in step 2022, smart network's model is according to above-mentioned model generating method embodiment, by sample
This pre-scan images and the network model that study generation is carried out by the sample distance fields of sweep center point generation.
It is following steps that then step 203, which specifically performs,：
2031st, according to the distance field based on sweep center point, sweep center point is determined.
In order to make the process for determining sweep center point more easily understand, the embodiment of the present invention is directed to be believed based on three-dimensional gradient
(gradient x, gradient y, gradient z) determine the process of sweep center point three-dimensional coordinate (x, y, z) to breath, there is provided stream as shown in figure 11
Journey schematic diagram.
Second of implementation method, based on scanning center's axis, as shown in figure 12, including：
2023rd, the pre-scan images are handled by smart network's model, obtains being based on scanning center's axis
Distance field.
Wherein, in step 2023, smart network's model is according to above-mentioned model generating method embodiment, by sample
This pre-scan images carries out the network model of study generation with the sample distance fields generated by scanning center's axis.
It is following steps that then step 203, which specifically performs,：
2032nd, according to the distance field based on scanning center's axis, scanning center's axis is determined.
The third implementation method, based on window size is scanned, as shown in figure 13, including：
2024th, the pre-scan images are handled by smart network's model, obtains non-linear adding window distance
.
Wherein, in step 2024, smart network's model is according to above-mentioned model generating method embodiment, by sample
This pre-scan images and the network mould that learning training generation is carried out by the sample distance fields of sweep center point, the size generation of scanning window
Type.Specifically, generating non-linear windowed function by sweep center point and scanning window size, obtained by non-linear windowed function non-
Linear adding window distance field, and then learning sample pre-scan images and non-linear adding window distance field, obtain smart network's model.
It is following steps that then step 203, which specifically performs,：
2033rd, according to the non-linear adding window distance field, scanning window size is determined.
Furthermore, it is understood that with reference to preceding method flow, for being determined in step 203 in a manner of which kind of computing by distance field
The specific implementation process of sweep parameter, the alternatively possible implementation of the embodiment of the present invention additionally provide following two possibility
Implementation method：
First method, as shown in figure 14, including：
2034th, fitting operation is weighted to the distance field, determines sweep parameter.
Specifically, such as scanning center can be determined by weighted linear fitting operation when distance field is cylinder
Axis.
Second method, as shown in figure 15, including：
2035th, Hough ballot is carried out to the distance field, determines sweep parameter.
It should be noted that based in above-described embodiment, voted by Hough (Hough) determine sweep parameter when, can be with
According to the characteristics of organizational structure of anatomical location, each locus is weighted in field of adjusting the distance, to shield representativeness
The region of difference.
An embodiment of the present invention provides a kind of medical image system, and as shown in figure 16, the medical image system includes place
Manage device 31 and memory 32；The memory 32 for storing instruction, when described instruction is performed by the processor 31, causes
The relevant method of any embodiment in the medical image system implementation model generation method, the scanning planing method of medical imaging
Flow.
, can be by smart network to multiple sample pre-scan images and sample distance in above-mentioned medical image system
Field carries out deep learning, obtains the correspondence of sample pre-scan images and sample three-dimensional distance field, and then according to mapping relations
Smart network's model is generated, completes whole model generating process.And then can by smart network's model to by
The pre-scan images of inspection person target area are handled, and obtain corresponding distance field, and sweep parameter is determined by distance field.This hair
The technical solution that bright embodiment provides obtains distance field compared with the prior art, by smart network's model, and then determines
Sweep parameter, can reduce the time required for whole scanning process, and machine learning is not in the mistake that artificial or experience is brought
Difference, therefore can also ensure the precision of sweep parameter.
In a kind of possible implementation, for the scanning planing method of medical imaging, as shown in figure 17, medical imaging
System can include scanner 41 and computer 42.
Wherein, scanner 41 can be that digital radiography (digital radiography, DR) scanner, magnetic are common
Shake (magnetic resonance, MR) scanner, computed tomography (computed tomography, CT) scanner
Deng.
Specifically, scanner can perform person under inspection target area prescan generation pre-scan images, or based on calculating
The sweep parameter that machine 42 generates performs image scanning to person under inspection target area.
Wherein, computer 42 can be used to realize the ad hoc approach and device for implementing to disclose in the embodiment of the present invention.
In some embodiments, computer 42 can realize this hair by its hardware device, software program, firmware and combinations thereof
The specific implementation of bright embodiment.In certain embodiments, computer 42 can be the computer of a general purpose, or one has
The computer of specific purpose.
As shown in figure 17, computer 42 can include internal communication bus 421, and processor (processor) 422 is read-only
Memory (ROM) 423, random access memory (RAM) 424, communication port 425, input output assembly 426, hard disk 427, with
And user interface 428.Internal communication bus 421 can realize the data communication of 42 inter-module of computer.Processor 422 can be into
Row judges and sends prompting.
In certain embodiments, processor 422 can be made of one or more processors.Communication port 425 can be real
Existing computer 42 and miscellaneous part (not shown) are for example：External equipment, image capture device, database, external storage with
And into row data communication between image processing workstations etc..
In certain embodiments, computer 42 can send and receive information and data by communication port 425 from network.
Input output assembly 426 supports the input/output data stream between computer 42 and miscellaneous part.User interface 428 can be real
Existing interaction and information exchange between computer 42 and user.Computer 42 can also include various forms of program storage units
And data storage element, such as hard disk 427, read-only storage (ROM) 423, random access memory (RAM) 424, Neng Goucun
The various data files that storage computer disposal and/or communication use, and the possible programmed instruction performed by processor 422.
In embodiments of the present invention, the instruction of processor 422 is used to perform following operation：Obtain person under inspection target area
Pre-scan images；Pre-scan images are handled by smart network's model, obtain distance field, smart network's mould
Type includes the mapping relations between pre-scan images and distance field；According to the distance field, sweep parameter is determined.Further,
Pre-scan images can be stored in advance in read-only storage 423 either in random access memory 424 or to pass through scanner
41 scannings immediately obtain.
Alternatively, processor 422, which can also carry out, is trained smart network：Obtain sample pre-scan images and
The corresponding scan sample parameter of sample pre-scan images, scan sample parameter include at least sweep center point, scanning center's axis, sweep
Retouch the one of which in window size；According to scan sample parameter, sample distance fields are generated；It is pre- to sample by smart network
Scan image is learnt with sample distance fields, and the mapping for obtaining the sample pre-scan images and the sample distance fields is closed
System；Smart network's model is generated according to the mapping relations.In this embodiment, processor 422 is performed to artificial intelligence
The process of network training further includes：Extract the characteristic information of sample pre-scan images；It is pre- by smart network's learning sample
The characteristic information and sample distance fields of scan image, obtain the characteristic information of the sample pre-scan images and the sample distance
The mapping relations of field, correspondingly, the processing for the pre-scan images to person under inspection target area may include：Extraction is swept in advance
The characteristic information of tracing picture, and this feature information is inputted in smart network's model, export the distance field of prediction.It is wherein pre-
The characteristic information of scan image may include the three-dimensional gradient information of sample pre-scan images, gradient block message, image block message, knot
Structure tensor information, 3 D wavelet coefficient, 2-d wavelet coefficient etc..
Alternatively, above-mentioned sample pre-scan images, scan sample parameter or smart network's model can be deposited in advance
Storage is in computer 42.It is apparent to those skilled in the art that for convenience and simplicity of description, foregoing description
System, apparatus, and unit specific work process, may be referred to the corresponding process in preceding method embodiment, it is no longer superfluous herein
State.
In several embodiments provided by the present invention, it should be understood that disclosed system, apparatus and method can be with
Realize by another way.For example, device embodiment described above is only schematical, for example, the unit
Division, is only a kind of division of logic function, can have other dividing mode when actually realizing, for example, multiple units or group
Part can combine or be desirably integrated into another system, or some features can be ignored, or not perform.It is another, it is shown
Or the mutual coupling, direct-coupling or communication connection discussed can be by some interfaces, device or unit it is indirect
Coupling or communication connection, can be electrical, machinery or other forms.
The unit illustrated as separating component may or may not be physically separate, be shown as unit
The component shown may or may not be physical location, you can with positioned at a place, or can also be distributed to multiple
In network unit.Some or all of unit therein can be selected to realize the mesh of this embodiment scheme according to the actual needs
's.
In addition, each functional unit in each embodiment of the present invention can be integrated in a processing unit, can also
That unit is individually physically present, can also two or more units integrate in a unit.Above-mentioned integrated list
Member can both be realized in the form of hardware, can also be realized in the form of hardware adds SFU software functional unit.
The above-mentioned integrated unit realized in the form of SFU software functional unit, can be stored in one and computer-readable deposit
In storage media.Above-mentioned SFU software functional unit is stored in a storage medium, including some instructions are used so that a computer
It is each that device (can be personal computer, server, or network equipment etc.) or processor (Processor) perform the present invention
The part steps of embodiment the method.And foregoing storage medium includes：USB flash disk, mobile hard disk, read-only storage (Read-
Only Memory, ROM), random access memory (Random Access Memory, RAM), magnetic disc or CD etc. it is various
Can be with the medium of store program codes.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all essences in the present invention
God and any modification, equivalent substitution, improvement and etc. within principle, done, should be included within the scope of protection of the invention.
Claims (10)
- A kind of 1. model generating method, it is characterised in that the described method includes：Obtain sample pre-scan images and the corresponding scan sample parameter of the sample pre-scan images, the scan sample ginseng Number is including at least the one of which in sweep center point, scanning center's axis, scanning window size；According to the scan sample parameter, sample distance fields are generated；The sample pre-scan images are learnt with the sample distance fields by smart network, obtain the sample The mapping relations of pre-scan images and the sample distance fields；Smart network's model is generated according to the mapping relations.
- 2. according to the method described in claim 1, it is characterized in that, the sample is swept in advance by smart network described As being learnt with the sample distance fields, the mapping for obtaining the sample pre-scan images and the sample distance fields is closed for tracing Before system, the method further includes：The characteristic information of the sample pre-scan images is extracted, the characteristic information of the sample pre-scan images is including at least described The three-dimensional gradient information of sample pre-scan images, gradient block message, image block message, structure tensor information, 3 D wavelet coefficient, One of which in 2-d wavelet coefficient；Then the sample pre-scan images are learnt with the sample distance fields by smart network, obtain the sample The mapping relations of this pre-scan images and the sample distance fields, including：The characteristic information of the sample pre-scan images is learnt with the sample distance fields by smart network, is obtained To the characteristic information and the mapping relations of the sample distance fields of the sample pre-scan images.
- A kind of 3. scanning planing method of medical imaging, it is characterised in that the described method includes：Obtain the pre-scan images of person under inspection target area；The pre-scan images are handled by smart network's model, obtain distance field, the smart network Model includes the pre-scan images and the mapping relations of the distance field；According to the distance field, sweep parameter is determined, the sweep parameter includes at least sweep center point, scanning center's axis, sweeps Retouch the one of which in window size, scanning direction.
- 4. according to the method described in claim 3, it is characterized in that, described according to the distance field, determine sweep parameter it Afterwards, the method further includes：Scanning is performed to the person under inspection target area based on the sweep parameter.
- 5. according to the method described in claim 3, it is characterized in that, pre- swept to described by smart network's model described Tracing is as before being handled, the method further includes：The characteristic information of the pre-scan images is extracted, the characteristic information of the pre-scan images includes at least the prescan figure The three-dimensional gradient information of picture, gradient block message, image block message, structure tensor information, 3 D wavelet coefficient, 2-d wavelet system One of which in number, gradient magnitude information, gradient units vector information；It is then described that the pre-scan images are handled by smart network's model, distance field is obtained, including：The characteristic information of the pre-scan images is handled by smart network's model, obtains distance field, the people Work intelligent network model includes the characteristic information of the pre-scan images and the mapping relations of the distance field.
- 6. according to the method described in claim 3, it is characterized in that, it is described by smart network's model to the prescan Image is handled, and obtains distance field, including：The pre-scan images are handled by smart network's model, obtain the distance field based on sweep center point；Then according to the distance field, sweep parameter is determined, including：According to the distance field based on sweep center point, sweep center point is determined.
- 7. according to the method described in claim 3, it is characterized in that, it is described by smart network's model to the prescan Image is handled, and obtains distance field, including：The pre-scan images are handled by smart network's model, obtain the distance field based on scanning center's axis；Then according to the distance field, sweep parameter is determined, including：According to the distance field based on scanning center's axis, scanning center's axis is determined.
- 8. according to the method described in claim 3, it is characterized in that, it is described by smart network's model to the prescan Image is handled, and obtains distance field, including：The pre-scan images are handled by smart network's model, obtain non-linear adding window distance field；Then according to the distance field, sweep parameter is determined, including：According to the non-linear adding window distance field, scanning window size is determined.
- 9. according to the method described in claim 3, it is characterized in that, described according to the distance field, determine sweep parameter, wrap Include：Fitting operation is weighted to the distance field, determines sweep parameter；Or,Hough ballot is carried out to the distance field, determines sweep parameter.
- 10. a kind of medical image system, it is characterised in that the medical image system includes processor and memory；It is described Memory for storing instruction, when described instruction is performed by the processor, causes the medical image system to realize such as right It is required that 1 to 9 any one of them method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711238100.8A CN107909622B (en) | 2017-11-30 | 2017-11-30 | Model generation method, medical imaging scanning planning method and medical imaging system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711238100.8A CN107909622B (en) | 2017-11-30 | 2017-11-30 | Model generation method, medical imaging scanning planning method and medical imaging system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107909622A true CN107909622A (en) | 2018-04-13 |
| CN107909622B CN107909622B (en) | 2021-05-14 |
Family
ID=61849395
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711238100.8A Active CN107909622B (en) | 2017-11-30 | 2017-11-30 | Model generation method, medical imaging scanning planning method and medical imaging system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107909622B (en) |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108597589A (en) * | 2018-04-27 | 2018-09-28 | 上海联影医疗科技有限公司 | Model generating method, object detection method and medical image system |
| CN108989684A (en) * | 2018-08-23 | 2018-12-11 | 阿里巴巴集团控股有限公司 | The method and apparatus for controlling shooting distance |
| CN109009211A (en) * | 2018-06-22 | 2018-12-18 | 联想(北京)有限公司 | Smart machine, the method and device based on ultrasound examination |
| CN109381212A (en) * | 2018-09-27 | 2019-02-26 | 上海联影医疗科技有限公司 | A kind of image formation control method and system |
| CN109948607A (en) * | 2019-02-21 | 2019-06-28 | 电子科技大学 | Candidate frame based on deep learning deconvolution network generates and object detection method |
| CN110084805A (en) * | 2019-04-30 | 2019-08-02 | 东软医疗系统股份有限公司 | FOV parameter setting method, device and image processing equipment |
| CN110728274A (en) * | 2018-06-29 | 2020-01-24 | 通用电气公司 | Medical device computer-assisted scanning method, medical device and readable storage medium |
| CN110960241A (en) * | 2019-12-09 | 2020-04-07 | 上海联影医疗科技有限公司 | Method and device for determining scanning parameters of medical image scanning and computer equipment |
| CN111161371A (en) * | 2019-10-25 | 2020-05-15 | 上海联影智能医疗科技有限公司 | Imaging system and method |
| CN111540036A (en) * | 2020-03-12 | 2020-08-14 | 李新福 | Automatic modeling method and system for objects in closed container |
| CN112749593A (en) * | 2019-10-31 | 2021-05-04 | 通用电气精准医疗有限责任公司 | Medical imaging system, method for recognizing body position of detected object, and storage medium |
| CN112820383A (en) * | 2019-11-15 | 2021-05-18 | 上海联影医疗科技股份有限公司 | Medical imaging method, system and storage medium |
| CN115919285A (en) * | 2023-02-28 | 2023-04-07 | 山东奥新医疗科技有限公司 | Nuclear magnetic resonance positioning method, device, equipment and storage medium |
| CN117500434A (en) * | 2021-04-02 | 2024-02-02 | Aixscan公司 | Progressive scanning using multiple motion pulse X-ray source tomosynthesis imaging systems |
| US11935162B2 (en) | 2019-05-14 | 2024-03-19 | Koninklijke Philips N.V. | Protocol-dependent 2-D pre-scan projection image based on 3-D pre-scan volumetric image data |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7245691B2 (en) * | 2004-01-07 | 2007-07-17 | Ge Medical Systems Global Technology Company, Llc | Radiation tomographic imaging apparatus and imaging method using it |
| US7576738B2 (en) * | 2005-05-27 | 2009-08-18 | California Institute Of Technology | Method for constructing surface parameterizations |
| CN104346821A (en) * | 2013-08-09 | 2015-02-11 | 西门子公司 | Automatic Planning For Medical Imaging |
| CN105496563A (en) * | 2015-12-04 | 2016-04-20 | 上海联影医疗科技有限公司 | Method for calibrating medical image positioning line |
| CN106910194A (en) * | 2015-12-22 | 2017-06-30 | 上海联影医疗科技有限公司 | The dividing method of the intracardiac adventitia in cardiac function MRI |
-
2017
- 2017-11-30 CN CN201711238100.8A patent/CN107909622B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7245691B2 (en) * | 2004-01-07 | 2007-07-17 | Ge Medical Systems Global Technology Company, Llc | Radiation tomographic imaging apparatus and imaging method using it |
| US7576738B2 (en) * | 2005-05-27 | 2009-08-18 | California Institute Of Technology | Method for constructing surface parameterizations |
| CN104346821A (en) * | 2013-08-09 | 2015-02-11 | 西门子公司 | Automatic Planning For Medical Imaging |
| CN105496563A (en) * | 2015-12-04 | 2016-04-20 | 上海联影医疗科技有限公司 | Method for calibrating medical image positioning line |
| CN106910194A (en) * | 2015-12-22 | 2017-06-30 | 上海联影医疗科技有限公司 | The dividing method of the intracardiac adventitia in cardiac function MRI |
Non-Patent Citations (2)
| Title |
|---|
| KSENIJA GRÄFE* 等: "2D Images Recorded With a Single-Sided Magnetic Particle Imaging Scanner", 《IEEE TRANSACTIONS ON MEDICAL IMAGING》 * |
| 朱立人 等: "一种提高N线模型探头标定精度的解算方法", 《中国生物医学工程学报》 * |
Cited By (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108597589B (en) * | 2018-04-27 | 2022-07-05 | 上海联影医疗科技股份有限公司 | Model generation method, target detection method and medical imaging system |
| CN108597589A (en) * | 2018-04-27 | 2018-09-28 | 上海联影医疗科技有限公司 | Model generating method, object detection method and medical image system |
| CN109009211A (en) * | 2018-06-22 | 2018-12-18 | 联想(北京)有限公司 | Smart machine, the method and device based on ultrasound examination |
| CN109009211B (en) * | 2018-06-22 | 2023-07-21 | 联想(北京)有限公司 | Intelligent equipment, ultrasonic detection-based method and device |
| CN110728274A (en) * | 2018-06-29 | 2020-01-24 | 通用电气公司 | Medical device computer-assisted scanning method, medical device and readable storage medium |
| CN108989684A (en) * | 2018-08-23 | 2018-12-11 | 阿里巴巴集团控股有限公司 | The method and apparatus for controlling shooting distance |
| CN109381212A (en) * | 2018-09-27 | 2019-02-26 | 上海联影医疗科技有限公司 | A kind of image formation control method and system |
| CN109948607A (en) * | 2019-02-21 | 2019-06-28 | 电子科技大学 | Candidate frame based on deep learning deconvolution network generates and object detection method |
| CN110084805B (en) * | 2019-04-30 | 2023-05-05 | 东软医疗系统股份有限公司 | FOV parameter setting method and device and image processing equipment |
| CN110084805A (en) * | 2019-04-30 | 2019-08-02 | 东软医疗系统股份有限公司 | FOV parameter setting method, device and image processing equipment |
| US11935162B2 (en) | 2019-05-14 | 2024-03-19 | Koninklijke Philips N.V. | Protocol-dependent 2-D pre-scan projection image based on 3-D pre-scan volumetric image data |
| CN111161371A (en) * | 2019-10-25 | 2020-05-15 | 上海联影智能医疗科技有限公司 | Imaging system and method |
| CN111161371B (en) * | 2019-10-25 | 2024-05-03 | 上海联影智能医疗科技有限公司 | Imaging system and method |
| CN112749593A (en) * | 2019-10-31 | 2021-05-04 | 通用电气精准医疗有限责任公司 | Medical imaging system, method for recognizing body position of detected object, and storage medium |
| CN112820383A (en) * | 2019-11-15 | 2021-05-18 | 上海联影医疗科技股份有限公司 | Medical imaging method, system and storage medium |
| CN112820383B (en) * | 2019-11-15 | 2023-04-28 | 上海联影医疗科技股份有限公司 | Medical imaging method, system and storage medium |
| CN110960241A (en) * | 2019-12-09 | 2020-04-07 | 上海联影医疗科技有限公司 | Method and device for determining scanning parameters of medical image scanning and computer equipment |
| CN111540036A (en) * | 2020-03-12 | 2020-08-14 | 李新福 | Automatic modeling method and system for objects in closed container |
| CN117500434A (en) * | 2021-04-02 | 2024-02-02 | Aixscan公司 | Progressive scanning using multiple motion pulse X-ray source tomosynthesis imaging systems |
| CN115919285A (en) * | 2023-02-28 | 2023-04-07 | 山东奥新医疗科技有限公司 | Nuclear magnetic resonance positioning method, device, equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107909622B (en) | 2021-05-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107909622A (en) | Model generating method, the scanning planing method of medical imaging and medical image system | |
| CN111402228B (en) | Image detection method, device and computer readable storage medium | |
| JP6947759B2 (en) | Systems and methods for automatically detecting, locating, and semantic segmenting anatomical objects | |
| CN110136809A (en) | A kind of medical image processing method, device, electromedical equipment and storage medium | |
| WO2020063986A1 (en) | Method and apparatus for generating three-dimensional model, device, and storage medium | |
| CN107545309A (en) | Scored using the picture quality of depth generation machine learning model | |
| CN107203989A (en) | End-to-end chest CT image dividing method based on full convolutional neural networks | |
| JP2019114262A (en) | Medical image processing apparatus, medical image processing program, learning apparatus and learning program | |
| CN111369525A (en) | Image analysis method, apparatus and storage medium | |
| CN114424246A (en) | Method, system and computer-readable storage medium for registering intraoral measurements | |
| CN110689521A (en) | Automatic identification method and system for human body part to which medical image belongs | |
| WO2021027152A1 (en) | Image synthesis method based on conditional generative adversarial network, and related device | |
| CN113570619B (en) | Computer-aided pancreas pathology image diagnosis system based on artificial intelligence | |
| CN113793301A (en) | Training method of fundus image analysis model based on dense convolution network model | |
| CN113159040A (en) | Method, device and system for generating medical image segmentation model | |
| Čular et al. | Dental age estimation from panoramic X-ray images using statistical models | |
| CN114708237A (en) | Detection algorithm for hair health condition | |
| CN108597589B (en) | Model generation method, target detection method and medical imaging system | |
| Chen et al. | Detection of various dental conditions on dental panoramic radiography using Faster R-CNN | |
| CN113610746A (en) | Image processing method and device, computer equipment and storage medium | |
| CN113222996A (en) | Heart segmentation quality evaluation method, device, equipment and storage medium | |
| CN116975779A (en) | Neural network-based oral cavity full-scene feature recognition method, system and terminal | |
| CN106530386A (en) | Volume rendering method and system for medical images | |
| CN110910409A (en) | Gray scale image processing method and device and computer readable storage medium | |
| CN109886320A (en) | A kind of human femur under loading X-ray intelligence reading method and system |
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 | ||
| CB02 | Change of applicant information |
Address after: 201807 Shanghai city Jiading District Industrial Zone Jiading Road No. 2258 Applicant after: Shanghai Lianying Medical Technology Co., Ltd Address before: 201807 Shanghai city Jiading District Industrial Zone Jiading Road No. 2258 Applicant before: SHANGHAI UNITED IMAGING HEALTHCARE Co.,Ltd. |
|
| CB02 | Change of applicant information | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |