CN108597589A - Model generating method, object detection method and medical image system - Google Patents
Model generating method, object detection method and medical image system Download PDFInfo
- Publication number
- CN108597589A CN108597589A CN201810395323.3A CN201810395323A CN108597589A CN 108597589 A CN108597589 A CN 108597589A CN 201810395323 A CN201810395323 A CN 201810395323A CN 108597589 A CN108597589 A CN 108597589A
- Authority
- CN
- China
- Prior art keywords
- distance field
- medical image
- frame
- sample
- target
- 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
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H30/00—ICT specially adapted for the handling or processing of medical images
- G16H30/20—ICT specially adapted for the handling or processing of medical images for handling medical images, e.g. DICOM, HL7 or PACS
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/50—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for simulation or modelling of medical disorders
Abstract
An embodiment of the present invention provides a kind of model generating method, object detection method and medical image systems, are related to technical field of medical image processing, determine target frame by distance field, reduce the complexity of target detection, improve the stability of solution.This method includes model generating process and target detection process：Model generating process includes：Obtain sample medical image and the corresponding sample object frame of sample medical image；According to sample object frame, distance field is generated；Sample medical image is learnt with distance field by smart network, obtains their mapping relations；Smart network's model is generated according to mapping relations.Target detection process includes：Obtain the medical image of subject target area；Medical image is handled by smart network's model, obtains distance field；According to distance field, target frame is determined in medical image.Technical solution provided in an embodiment of the present invention is suitable for determining medical image during target area.
Description
【Technical field】
The present invention relates to technical field of medical image processing more particularly to a kind of model generating method, object detection methods
And medical image system.
【Background technology】
Target detection is computer vision and a kind of particular task of image procossing, and the purpose is to be determined in medical image
Target, and one frame of label (the Bounding Box, also known as target frame) in target.In technical field of medical image processing
In, there are many important applications, such as automatic positioning organ or the specific lesion of detection for target detection.
Currently, have much object detection methods based on convolutional neural networks, current method (such as RCNN (Regions
With CNN features, convolutional neural networks characteristic area), Yolo (You Only Look Once, frame choosing identification are unified)
Etc. be all based on it is a kind of be called frame recurrence technology (bounding-box regression), i.e., target frame is defined as
Four parameters (x, y, w, h), wherein (x, y) is the center point coordinate of target frame, w and h are the width and height of target frame, and with
Central point translates and the wide high method scaled as mathematical modeling.Predictably, if by this model extension to three dimensions,
There will be six parameters, each parameter is required for individual regression equation.Another defect of technology that frame returns is, only
When being selected frame close to target frame, wide high scale transformation can be considered as just linear transformation.Due to the size and length of target area
Wide ratio is likely to difference, every time when progress frame recurrence, is required for fully sampling, most methods suggestion from nine sizes not
Start to return target frame with the different frame pattern of length-width ratio；And each parameter of frame pattern is independent one
Dimension needs an individual passage to go the value of dimension where returning each parameter in convolutional neural networks；It is more complicated,
When establishing model, balance is established between multiple parameters, each parameter learning can be uneven when otherwise training, and is easy generation error
Model.
In realizing process of the present invention, inventor has found that at least there are the following problems in the prior art：
Object detection method in the prior art, needed when determining target frame multiple parameters (nine frame patterns, and
Each target frame four or six parameters), the complexity of problem is increased, the stability of solution is reduced.
【Invention content】
In view of this, an embodiment of the present invention provides a kind of model generating method, object detection method and medical imaging systems
System, during target detection, determines target frame by distance field, reduces the complexity of target detection, improve solution
Stability.
In a first aspect, the embodiment of the present invention provides a kind of model generating method, the method includes：
Obtain sample medical image and the corresponding sample object frame of the sample medical image；
According to the sample object frame, distance field is generated；
The sample medical image is learnt with the distance field by smart network, obtains the sample doctor
Learn the mapping relations of image and the distance field；
Smart network's model is generated according to the mapping relations.
The aspect and any possible implementation manners as described above, it is further provided a kind of realization method,
The distance field includes distance field in frame；Alternatively,
The distance field includes distance field and the outer distance field of frame in frame；Alternatively,
The distance field includes the weighted array of distance field and the outer distance field of frame in frame.
The aspect and any possible implementation manners as described above, it is further provided a kind of realization method, it is described according to institute
Sample object frame is stated, distance field is generated, including：
Respectively to carrying out binary conversion treatment inside and outside the sample object frame, and obtain binaryzation result；
Range conversion is carried out to the binaryzation result, obtains distance field.
The aspect and any possible implementation manners as described above, it is further provided a kind of realization method, when the sample
When target frame is three-dimensional frame,
It is described that distance field is generated according to the sample object frame, including：It is inside and outside to the sample object frame respectively
Binary conversion treatment is carried out, binaryzation result is obtained；And three-dimensional distance transformation is carried out to the binaryzation result, obtain three-dimensional distance
；
It is described that the sample medical image is learnt with the distance field by smart network, obtain the sample
The mapping relations of this medical image and the distance field, including：By smart network to the sample medical image and institute
The each layer for stating three-dimensional distance field is learnt, and the mapping relations of the sample medical image and the three-dimensional distance field are obtained.
The aspect and any possible implementation manners as described above, it is further provided a kind of realization method, the artificial intelligence
Energy network includes at least convolutional neural networks, reverse transmittance nerve network, radial base neural net, perceptron neural network, line
In nerve network, self organizing neural network, Feedback Neural Network, clustering network, deep learning network, feedforward neural network
It is a kind of.
Second 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 executed by the processor, leads to the medical imaging
System realizes the method described in either side as above or any possible realization method.
The third aspect, the embodiment of the present invention provide a kind of object detection method, the method includes：
Obtain the medical image of subject target area；
The medical image is handled by smart network's model, obtains the corresponding distance field of target frame,
Smart network's model includes the mapping relations of medical image distance field corresponding with the target frame；
According to the distance field, the target frame is determined in the medical image.
The aspect and any possible implementation manners as described above, it is further provided a kind of realization method, the distance field
Including distance field in frame；Alternatively, the distance field includes distance field and the outer distance field of frame in frame.
The aspect and any possible implementation manners as described above, it is further provided a kind of realization method, it is described according to institute
Distance field is stated, the target frame is determined in the medical image, including：
To the distance field into row distance inverse transformation, the target frame is obtained.
The aspect and any possible implementation manners as described above, it is further provided a kind of realization method, the target side
Include at least one of bone discrete regions, Lung neoplasm region, tumor region in frame.
Fourth aspect, the embodiment of the present invention provide a kind of medical image system, which is characterized in that the medical image system
Including processor and memory；The memory for storing instruction, when described instruction is executed by the processor, leads to institute
It states medical image system and realizes method described in either side as above or any possible realization method.
An embodiment of the present invention provides a kind of model generating method, object detection method and medical image systems, in target
In detection process, medical image is handled by smart network's model, obtains the corresponding distance field of target frame, into
And target frame is determined by distance field, compared with the prior art in multiple parameters determine target frame, present invention implementation carries
The method of confession, it is only necessary to one parameter of distance field determines target frame, and calculative number of parameters is fallen below one, to
The complexity for reducing target detection improves the stability of solution.
【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 invention, for this field
For those of ordinary skill, without creative efforts, other attached drawings are can also be obtained according to these attached drawings.
Fig. 1 is the method flow diagram that a kind of model provided in an embodiment of the present invention generates；
Fig. 2 is the schematic diagram of a kind of medical image provided in an embodiment of the present invention, target frame and distance field；
Fig. 3 is the schematic diagram of distance field and the outer distance field of frame in a kind of frame provided in an embodiment of the present invention；
Fig. 4 is the method flow diagram that another model provided in an embodiment of the present invention generates；
Fig. 5 is the method flow diagram that another model provided in an embodiment of the present invention generates；
Fig. 6 is a kind of method flow diagram of target detection provided in an embodiment of the present invention；
Fig. 7 is the method flow diagram of another target detection provided in an embodiment of the present invention；
Fig. 8 is a kind of composition frame chart of model generating means provided in an embodiment of the present invention；
Fig. 9 is a kind of composition frame chart of object detecting device provided in an embodiment of the present invention；
Figure 10 is a kind of entity composition figure of medical image system provided in an embodiment of the present invention；
Figure 11 is the entity composition figure of another medical image system provided in an embodiment of the present invention；
Figure 12 is a kind of schematic diagram of specific medical image system provided in an embodiment of the present invention.
【Specific implementation mode】
For a better understanding of the technical solution of the present invention, being retouched in detail to the embodiment of the present invention below in conjunction with the accompanying drawings
It states.
It will be appreciated that described embodiments are only a part of the embodiments 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, shall fall within the protection scope of the present invention.
The term used in embodiments of the present invention is the purpose only merely for description specific embodiment, is not intended to be limiting
The present invention.In the embodiment of the present invention and "an" of singulative used in the attached claims, " described " and "the"
It is also intended to including most forms, unless context clearly shows that other meanings.
It will be appreciated that though in embodiments of the present invention processing module, but this may be described using term first, second
A little processing modules should not necessarily be limited by these terms.These terms are only used for processing module being distinguished from each other out.For example, not departing from this
In the case of inventive embodiments range, first processing module can also be referred to as Second processing module, similarly, second processing mould
Block can also be referred to as first processing module.
Depending on context, word as used in this " if " can be construed to " ... when " or " when ...
When " or " in response to determination " or " in response to detection ".Similarly, depend on context, phrase " if it is determined that " or " if detection
(condition or event of statement) " can be construed to " when determining " or " in response to determination " or " when the detection (condition of statement
Or event) when " or " in response to detection (condition or event of statement) ".
It should be appreciated that term "and/or" used herein is only a kind of incidence relation of description affiliated partner, indicate
There may be three kinds of relationships, for example, A and/or B, can indicate：Individualism A, exists simultaneously A and B, individualism B these three
Situation.In addition, character "/" herein, it is a kind of relationship of "or" to typically represent forward-backward correlation object.
An embodiment of the present invention provides a kind of model generating methods, and the network model suitable for target detection generated
Journey, as shown in Figure 1, the method includes：
101, sample medical image and the corresponding sample object frame of the sample medical image are obtained.
Wherein, sample medical image refers to the sample as smart network, which includes the doctor of target area
Image is learned, is medical image known to early period.Sample medical image can be two-dimensional medical images, can also be 3 D medical figure
Picture.Sample medical image can be MR (Magnetic Resonance, Magnetic resonance imaging), PET (Positron
Emission Computed Tomography, positron emission computerized tomography), SPECT (Single-Photon
Emission Computed Tomography, single photon emission computed tomography), CT (Computed
Tomography, computed tomography), DR (Digital Radiography, digital flat panel x-ray imaging), ultrasound
(Ultrasound) blending image between any medical supply generates image or aforementioned any appliance such as.Sample object
Frame refers to be determined by sample medical image with the frame of target area in marker samples medical image, being early period
Known target frame.Correspondingly, sample object frame can also be two-dimentional frame, or three-dimensional frame.
In an optional implementation manner, two-value can be carried out to sample medical image before determining sample object frame
Change is handled, such as：After binary conversion treatment, the point gray value within sample object frame is 0, except sample object frame
Point gray value is 1；Alternatively, after binary conversion treatment, the point gray value within sample object frame is 1, sample object frame
Except point gray value be 0.
As shown in Fig. 2, (1) is sample medical image, wherein being target area, target area in the rectangular small frame of mark director
Including at bone/framework collapse, and breaking part is located at the centre position of entire target frame；(2) binaryzation of target area is indicated
Image, the gray value that the pixel in sample object frame is belonged in figure are 1, the gray value of the pixel outside sample object frame
It is 0, is region defined by sample object frame by the black rectangles region that binary conversion treatment identifies.
102, according to the sample object frame, distance field is generated.
Specifically, sample object frame is indicated that distance field is indicated by gray-value image by binary image, what distance field referred to
It is the set of the minimum range of the boundary pixel point outside each pixel (i.e. Chosen Point) distance objective region of selection area, because
This can generate distance field by the method for range conversion by sample object frame.Optionally, distance field can be European
(Euclidean) range conversion, chamfering (Chamfer) range conversion, Minkowsky range conversions, block (city-block)
The combination of one or more of range conversion.
In the first possible embodiment, distance field is distance field in frame.Specifically, distance field refers in frame
It is that each pixel in target frame is Chosen Point, each pixel outside target frame is background dot, carries out range conversion
Obtained distance field.
In this embodiment it is assumed that the image in target frame is A, include multiple pixel p in A；Outside target frame
Image be A comprising the minimum range of multiple pixel q, every bit to boundary point in target frame are corresponding for the point
DT values are as follows with formulae express：
DT1(p)=min { d (p, q) } (formula 1)
Wherein, min expressions are minimized operation, and d indicates to determine the operation of the distance between two pixels, herein embodiment party
In formula, p is Chosen Point, and q is the boundary point of variation.
As (3) are shown at a distance from corresponding to the sample object frame with Fig. 2 (2) obtained using above-mentioned formula in Fig. 2
Field picture, due to being in the position at center in skeleton for the boundary of object, theoretically closest in object
The point of the heart should have maximum DT1Value, therefore the brighter expression DT of pixel of distance field image1Value is bigger, which corresponds to mesh
Mark center or the bone fracture position of frame；The darker expression DT of pixel of distance field image1It is worth smaller, which corresponds to bone
The distance of bone fracture position is bigger.And the part of black completely then corresponds to except target frame.Above by the distance field of target frame
Determine that the mode on boundary has preferable locating effect to target frame central area.
In second of possible embodiment, distance field includes distance field and the outer distance field of frame in frame.It is exemplary
Ground, the outer distance field of frame refer to that each pixel outside target frame is Chosen Point, each pixel in target frame is
Background dot carries out the distance field that range conversion obtains.
In this embodiment, it also assumes that the image in target frame is A, includes multiple pixel p in A；Target side
The image of outer frame isIt includes multiple pixel q, and the minimum range of the every bit in target frame to boundary point is the point pair
The DT answered2Value is as follows with formulae express：
DT2(q)=min { d (q, p) } (formula 2)
Wherein, min expressions are minimized operation, and d indicates to determine the operation of the distance between two pixels, herein embodiment party
In formula, q is Chosen Point, and p is the boundary point of variation.
As shown in figure 3, to carry out in the obtained frame of range conversion distance field outside distance field and frame by target frame
Schematic diagram.Wherein, (1) is to indicate two kinds of target frame different binary images (definition of distance field Chosen Point is different, and left figure is fixed
Each pixel arrives the distance field of target frame in justice target frame, right figure define outside target frame each pixel to target
The distance field of frame), (2) are obtained by left figure in (1), are the gray-value image of distance field in frame, (3) are obtained by right figure in (1)
It arrives, is the gray-value image of the outer distance field of frame.
In the third possible embodiment, distance field includes the set of weights of distance field and the outer distance field of frame in frame
It closes.In this embodiment, distance field can be used following formula to indicate：
DT3(p, q)=α DT1(p,q)+βDT2(q, p) (formula 3)
Wherein, DT3Indicate the DT values of the corresponding pixel p in Weighted distance field；DT1Indicate the corresponding picture of distance field in frame
The DT values of vegetarian refreshments p；DT1Indicate the DT values of the corresponding pixel p of the outer distance field of frame；α is DT1Weight, β DT2Weight, 0
≤ α≤1,0≤β≤1, and alpha+beta=1.It sets different weights or obtains different distance fields, on the one hand can enrich trained sample
This；On the other hand the precision of network training can be improved.
103, the sample medical image is learnt by smart network with the distance field, obtains the sample
The mapping relations of this medical image and the distance field.
Wherein, artificial intelligence (Artificial Intelligence, the AI) network can select convolutional neural networks
(Convolutional Neural Network, CNN), backpropagation (Back Propagation, BP) neural network, radial direction
Base neural net, perceptron neural network, linear neural network, self organizing neural network, Feedback Neural Network, clustering network,
Deep learning network or feedforward neural network.Specifically, CNN can be RCNN (Region-based Fully again
Convolutional Networks, the complete convolutional network in feature based region), fast-RCNN (fast convolution neural networks
Characteristic area), faster-RCNN (faster convolutional neural networks characteristic area), Yolo, Yolo2, Yolo9000 or SSD
(Single Shot MultiBox Detector, the choosing identification of multi-layer frame are unified).
In a kind of possible embodiment, step 103 passes through the Recurrent networks in CNN with supervised learning method
Sample medical image and distance field are practised, the mapping relations between them are obtained, it can be with automated setting CNN according to the mapping relations
In all parameters, can also generate smart network's model.
104, smart network's model is generated according to the mapping relations.
Above-mentioned model generating method carries out machine learning to several by AI networks to sample medical image and distance field,
Constantly adjustment network parameter obtains the mapping relations of sample medical image and distance field, and then generates AI networks according to mapping relations
Model completes whole network model training process.
It should be noted that when distance field includes distance field outside distance field and frame in frame, AI networks can learn
The mapping relations of sample medical image and two distance fields, the AI network models obtained from when in use, can be by medicine figures
As generating two distance fields, further target frame is determined by two distance fields, more accurately.
Furthermore, it is understood that in conjunction with preceding method flow, in step 102, sample object frame generates the tool of distance field
Body realizes that process, the alternatively possible realization method of the embodiment of the present invention additionally provide following methods flow, as shown in figure 4,
Step 102 includes：
1021, respectively to the inside and outside carry out binary conversion treatment of the sample object frame, and binaryzation result is obtained.
1022, range conversion is carried out to the binaryzation result, obtains distance field.
Specifically, sample object frame here can be two-dimentional frame or three-dimensional frame, when sample object frame is
When two-dimentional frame, two-dimensional distance transformation can be carried out, two-dimensional distance field is obtained；It, can when sample object frame is three-dimensional frame
To carry out three-dimensional distance transformation, three-dimensional distance field is obtained.
Wherein, the range conversion in step 1022 can be Euclidean Distance Transform (Euclidean Distance
Transform) either chessboard distance transformation (Chessboard Distance Transform) or street range conversion
(City Block Distance Transform)。
Furthermore, it is understood that in conjunction with preceding method flow, when the sample object frame is three-dimensional frame, the present invention is implemented
The technical solution that example provides can be based on three-dimensional frame and obtain three-dimensional distance field, then simplify god by the method for Layered Learning
Learning process through network.Therefore the alternatively possible realization method of the embodiment of the present invention, for step 102 and step
103 realization additionally provides following methods flow, as shown in figure 5,
Step 102 includes：
1023, binaryzation is obtained as a result, and to institute to the inside and outside carry out binary conversion treatment of the sample object frame respectively
It states binaryzation result and carries out three-dimensional distance transformation, obtain three-dimensional distance field.
Step 103 includes：
1031, by smart network to each layer of the sample medical image and the three-dimensional distance field
It practises, obtains the mapping relations of the sample medical image and the three-dimensional distance field.
Three dimensions is layered in step 1031, obtains at least one layer of three-dimensional distance field, and inputs smart network
Practise the mapping relations of itself and sample medical image.
Above-described embodiment can use two dimension to calculate and solve three-dimensional problem by the method for three-dimensional distance field order training method, section
Video card memory is saved.
An embodiment of the present invention provides a kind of object detection methods, the process suitable for determining target area medical image
In, as shown in fig. 6, the method includes：
201, the medical image of subject target area is obtained.
Wherein, the medical image of subject target area refers to including the pending medicine figure of subject target area
Picture can be two-dimensional medical images, can also be 3 d medical images.The explanation of medical image is shown in that step 101 is no longer superfluous herein
It states.
202, the medical image is handled by smart network's model, obtains the corresponding distance of target frame
.
Wherein, the above-mentioned model of smart network's model generates embodiment or any possible realization method is realized
What method generated, include the mapping relations of medical image distance field corresponding with target frame.
Wherein, target frame refers to mark the frame of subject target area in pending medical image.Target
Can be bone discrete regions, Lung neoplasm region or tumor region etc. in frame.
The corresponding distance field of target frame can be distance field in frame；Alternatively, may include distance field and side in frame
Outer frame distance field.The specific explanations of distance field are shown in step 102, and details are not described herein again.
203, according to the distance field, the target frame is determined in the medical image.
Specifically, target frame can be generated by distance field by the method apart from inverse transformation.
Above-mentioned object detection method embodiment is that the use process of smart network's model passes through in target detection
Smart network's model handles medical image, obtains the corresponding distance field of target frame, and then true by distance field
Set the goal frame, compared with the prior art in multiple parameters determine target frame, the present invention implements the method provided, it is only necessary to
One parameter of distance field determines target frame, calculative number of parameters is fallen below one, to reduce target detection
Complexity, improve the stability of solution.
Furthermore, it is understood that in conjunction with preceding method flow, in step 203, distance field generates the specific reality of target frame
Existing process, the alternatively possible realization method of the embodiment of the present invention additionally provides following methods flow, as shown in fig. 7, step
203 include：
2031, the target frame is obtained into row distance inverse transformation to the distance field.
Wherein, in step 2031 apart from inverse transformation can be based on Euclidean Distance Transform or
The inverse transformation of Chessboard Distance Transform or City Block Distance Transform.
An embodiment of the present invention provides a kind of model generating means, are suitable for model and generate correlation technique flow, such as Fig. 8 institutes
Show, described device includes：
Acquiring unit 31, for obtaining sample medical image and the corresponding sample object side of the sample medical image
Frame.
First generation unit 32, for according to the sample object frame, generating distance field.
Unit 33, for passing through smart network to the sample medical image and the distance field
It practises, obtains the mapping relations of the sample medical image and the distance field.
Second generation unit 34, for generating smart network's model according to the mapping relations.
Optionally, as shown in figure 8, first generation unit 32 includes：
First processing module 321 is used for respectively to the inside and outside carry out binary conversion treatment of the sample object frame, and obtains
Binaryzation result.
Second processing module 322 obtains distance field for carrying out range conversion to the binaryzation result.
Optionally, as shown in figure 8, when the sample object frame is three-dimensional frame, first generation unit 32,
Specifically for the inside and outside carry out binary conversion treatment of the sample object frame, obtaining binaryzation result respectively；And to the two-value
Change result and carry out three-dimensional distance transformation, obtains three-dimensional distance field.
The unit 33, be specifically used for by smart network to the sample medical image with it is described it is three-dimensional away from
The each layer left the theatre is learnt, and the mapping relations of the sample medical image and the three-dimensional distance field are obtained.
Above-mentioned model generating means carry out machine learning to several by AI networks to sample medical image and distance field,
Constantly adjustment network parameter obtains the mapping relations of sample medical image and distance field, and then generates AI networks according to mapping relations
Model completes whole network model training process.
An embodiment of the present invention provides a kind of object detecting devices, are suitable for target detection correlation technique flow, such as Fig. 9 institutes
Show, described device includes：
Acquiring unit 41, the medical image for obtaining subject target area.
Processing unit 42 handles the medical image for passing through smart network's model, obtains target side
The corresponding distance field of frame, smart network's model includes medical image to be reflected with the corresponding distance field of target frame
Penetrate relationship.
Determination unit 43, for according to the distance field, the target frame to be determined in the medical image.
Optionally, as shown in figure 9, the determination unit 43 includes：
Processing module 431, for, into row distance inverse transformation, obtaining the target frame to the distance field.
Optionally, include in the target frame in bone discrete regions, Lung neoplasm region, tumor region at least
It is a kind of.
Above-mentioned object detecting device is handled medical image by smart network's model in target detection,
Obtain the corresponding distance field of target frame, and then target frame determined by distance field, compared with the prior art in multiple ginsengs
Number determines that target frames, the present invention implement the method provided, it is only necessary to which one parameter of distance field determines target frame, needing to count
The number of parameters of calculation falls below one, to reduce the complexity of target detection, improves the stability of solution.
An embodiment of the present invention provides a kind of medical image systems, and as shown in Figure 10, the medical image system includes place
Manage device 51 and memory 52；The memory 52 for storing instruction, when described instruction is executed by the processor 51, causes
The medical image system implementation model generates relevant method flow.
An embodiment of the present invention provides a kind of medical image systems, and as shown in figure 11, the medical image system includes place
Manage device 61 and memory 62；The memory 62 for storing instruction, when described instruction is executed by the processor 61, causes
The medical image system realizes the relevant method flow of target detection.
It should be noted that the medical imaging system of correlation technique flow is generated for realizing model in actual application
System can be the integrated same medical image system with the medical image system for realizing target detection correlation technique flow.
In a specific embodiment, as shown in figure 12, medical image system provided in an embodiment of the present invention can be a calculating
Machine 71, computer 71 be used to realize the ad hoc approach and device for implementing to disclose in the embodiment of the present invention.
Optionally, computer 71 can be the computer or a computer for having specific purpose of a general purpose.
Computer 71 can be realized of the invention real by its hardware device, software program, firmware and combination thereof
Apply the specific implementation of example.
As shown in figure 12, computer 71 may include the internal communication bus 711, (processing of processor (processor) 712
Device 712 can be made of one or more processors), read-only memory (ROM) 713, random access memory (RAM) 714 leads to
Believe port 715, input output assembly 716, hard disk 717 and user interface 718.Internal communication bus 711 may be implemented to count
The data of 71 inter-module of calculation machine communicate, and processor 712 can be judged and be sent out prompt.Communication port 715 may be implemented to count
Calculation machine 71 and other component (not shown), such as external equipment, image capture device, database, external storage and figure
As processing workstation etc., data communication.Input output assembly 716 supports the input between computer 71 and other component/defeated
Go out data flow.The interaction between computer 71 and user and information exchange may be implemented in user interface 718.
Optionally, computer 71 can also send and receive data information by communication port 715 from high in the clouds.
It should be noted that computer 71 may include various forms of program storage units and data storage element,
Such as hard disk 717, read-only memory (ROM) 713, random access memory (RAM) 714, can store computer disposal and/or
Communicate the possible program instruction performed by the various data files used and processor 712.
In embodiments of the present invention, correlation technique flow or use of the instruction of processor 712 for executing model generation
In the correlation technique flow of performance objective detection.
It is apparent to those skilled in the art that for convenience and simplicity of description, the system of foregoing description,
The specific work process of device and unit, can refer to corresponding processes in the foregoing method embodiment, and details are not described herein.
In several embodiments provided by the present invention, it should be understood that disclosed system, device and method can be with
It realizes by another way.For example, the apparatus embodiments described above are merely exemplary, for example, the unit
It divides, only a kind of division of logic function, formula that in actual implementation, there may be another division manner, for example, multiple units or group
Part can be combined or can be integrated into another system, or some features can be ignored or not executed.Another point, 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 separated, aobvious as unit
The component shown may or may not be physical unit, you can be located at a place, or may be distributed over multiple
In network element.Some or all of unit therein can be selected according to the actual needs to realize the mesh of this embodiment scheme
's.
In addition, each functional unit in each embodiment of the present invention can be integrated in a processing unit, it can also
It is that each unit physically exists alone, it can also be during two or more units be integrated in one unit.Above-mentioned integrated list
The form that hardware had both may be used in member is realized, can also be realized in the form of hardware adds SFU software functional unit.
The above-mentioned integrated unit being 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) execute the present invention
The part steps of embodiment the method.And storage medium above-mentioned includes：USB flash disk, mobile hard disk, read-only memory (Read-
Only Memory, ROM), random access memory (Random Access Memory, RAM), magnetic disc or CD etc. it is various
The medium of program code can be stored.
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
With within principle, any modification, equivalent substitution, improvement and etc. done should be included within the scope of protection of the invention god.
Claims (10)
1. a kind of model generating method, which is characterized in that the method includes：
Obtain sample medical image and the corresponding sample object frame of the sample medical image；
According to the sample object frame, distance field is generated；
The sample medical image is learnt with the distance field by smart network, obtains the sample medicine figure
As the mapping relations with the distance field；
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 distance field includes distance field in frame；Alternatively, described
Distance field includes distance field and the outer distance field of frame in frame.
3. according to the method described in claim 1, it is characterized in that, described according to the sample object frame, distance field is generated,
Including：
Respectively to the inside and outside carry out binary conversion treatment of the sample object frame, and obtain binaryzation result；
Range conversion is carried out to the binaryzation result, obtains distance field.
4. according to the method described in claim 2, it is characterized in that, when the sample object frame is three-dimensional frame,
It is described that distance field is generated according to the sample object frame, including：Respectively to the inside and outside progress of the sample object frame
Binary conversion treatment obtains binaryzation result；And three-dimensional distance transformation is carried out to the binaryzation result, obtain three-dimensional distance field；
It is described that the sample medical image is learnt with the distance field by smart network, obtain the sample doctor
The mapping relations of image and the distance field are learned, including：By smart network to the sample medical image and described three
Each layer of dimension distance field is learnt, and the mapping relations of the sample medical image and the three-dimensional distance field are obtained.
5. a kind of object detection method, which is characterized in that the method includes：
Obtain the medical image of subject target area；
The medical image is handled by smart network's model, obtains the corresponding distance field of target frame, it is described
Smart network's model includes the mapping relations of medical image distance field corresponding with the target frame；
According to the distance field, the target frame is determined in the medical image.
6. according to the method described in claim 5, it is characterized in that,
The distance field includes distance field in frame；Alternatively,
The distance field includes distance field and the outer distance field of frame in frame；Alternatively,
The distance field includes the weighted array of distance field and the outer distance field of frame in frame.
7. according to the method described in claim 5, it is characterized in that, described according to the distance field, in the medical image
Determine the target frame, including：
To the distance field into row distance inverse transformation, the target frame is obtained.
8. the method according to the description of claim 7 is characterized in that including bone discrete regions, lung in the target frame
At least one of knuckle areas, tumor region.
9. a kind of medical image system, which is characterized in that the medical image system includes processor and memory；It is described to deposit
Reservoir for storing instruction, when described instruction is executed by the processor, causes the medical image system to be realized as right is wanted
Seek 1 to 4 any one of them method.
10. a kind of medical image system, which is characterized in that the medical image system includes processor and memory；It is described
Memory for storing instruction, when described instruction is executed by the processor, causes the medical image system to realize such as right
It is required that 5 to 8 any one of them methods.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810395323.3A CN108597589B (en) | 2018-04-27 | 2018-04-27 | Model generation method, target detection method and medical imaging system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810395323.3A CN108597589B (en) | 2018-04-27 | 2018-04-27 | Model generation method, target detection method and medical imaging system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108597589A true CN108597589A (en) | 2018-09-28 |
| CN108597589B CN108597589B (en) | 2022-07-05 |
Family
ID=63610925
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810395323.3A Active CN108597589B (en) | 2018-04-27 | 2018-04-27 | Model generation method, target detection method and medical imaging system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108597589B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109993066A (en) * | 2019-03-06 | 2019-07-09 | 开易（北京）科技有限公司 | Vehicle positioning method and system towards sideline |
| WO2021155661A1 (en) * | 2020-02-05 | 2021-08-12 | 华为技术有限公司 | Image processing method and related device |
| CN114677502A (en) * | 2022-05-30 | 2022-06-28 | 松立控股集团股份有限公司 | License plate detection method with any inclination angle |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20070116357A1 (en) * | 2005-11-23 | 2007-05-24 | Agfa-Gevaert | Method for point-of-interest attraction in digital images |
| WO2016086744A1 (en) * | 2014-12-02 | 2016-06-09 | Shanghai United Imaging Healthcare Co., Ltd. | A method and system for image processing |
| CN106485704A (en) * | 2016-09-30 | 2017-03-08 | 上海联影医疗科技有限公司 | The extracting method of vessel centerline |
| CN107909622A (en) * | 2017-11-30 | 2018-04-13 | 上海联影医疗科技有限公司 | Model generating method, the scanning planing method of medical imaging and medical image system |
-
2018
- 2018-04-27 CN CN201810395323.3A patent/CN108597589B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20070116357A1 (en) * | 2005-11-23 | 2007-05-24 | Agfa-Gevaert | Method for point-of-interest attraction in digital images |
| WO2016086744A1 (en) * | 2014-12-02 | 2016-06-09 | Shanghai United Imaging Healthcare Co., Ltd. | A method and system for image processing |
| CN106485704A (en) * | 2016-09-30 | 2017-03-08 | 上海联影医疗科技有限公司 | The extracting method of vessel centerline |
| CN107909622A (en) * | 2017-11-30 | 2018-04-13 | 上海联影医疗科技有限公司 | Model generating method, the scanning planing method of medical imaging and medical image system |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109993066A (en) * | 2019-03-06 | 2019-07-09 | 开易（北京）科技有限公司 | Vehicle positioning method and system towards sideline |
| CN109993066B (en) * | 2019-03-06 | 2021-05-14 | 开易（北京）科技有限公司 | Sideline-oriented vehicle positioning method and system |
| WO2021155661A1 (en) * | 2020-02-05 | 2021-08-12 | 华为技术有限公司 | Image processing method and related device |
| CN114677502A (en) * | 2022-05-30 | 2022-06-28 | 松立控股集团股份有限公司 | License plate detection method with any inclination angle |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108597589B (en) | 2022-07-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20210110135A1 (en) | Method and system for artificial intelligence based medical image segmentation | |
| Zanjani et al. | Deep learning approach to semantic segmentation in 3D point cloud intra-oral scans of teeth | |
| CN111369525B (en) | Image analysis method, apparatus and storage medium | |
| CN110276745B (en) | Pathological image detection algorithm based on generation countermeasure network | |
| CN110047056A (en) | With the cross-domain image analysis and synthesis of depth image to image network and confrontation network | |
| CN106462963B (en) | System and method for being sketched outline automatically in adaptive radiation therapy | |
| JP2020516427A (en) | RECIST assessment of tumor progression | |
| JP2017174039A (en) | Image classification device, method, and program | |
| CN107358600A (en) | Automatic hook Target process, device and electronic equipment in radiotherapy planning | |
| Ali et al. | Echocardiographic image segmentation using deep Res-U network | |
| US9390502B2 (en) | Positioning anatomical landmarks in volume data sets | |
| CN110246580B (en) | Cranial image analysis method and system based on neural network and random forest | |
| KR102458324B1 (en) | Data processing method using a learning model | |
| CN107909622A (en) | Model generating method, the scanning planing method of medical imaging and medical image system | |
| JP6824845B2 (en) | Image processing systems, equipment, methods and programs | |
| CN111316290B (en) | System and method for interactive representation learning migration through deep learning of feature ontologies | |
| US20220335600A1 (en) | Method, device, and storage medium for lesion segmentation and recist diameter prediction via click-driven attention and dual-path connection | |
| CN108597589A (en) | Model generating method, object detection method and medical image system | |
| US11308613B2 (en) | Synthesis of contrast enhanced medical images | |
| JP2007530088A (en) | Object-specific segmentation | |
| WO2008141293A2 (en) | Image segmentation system and method | |
| JP6643821B2 (en) | Image processing device | |
| CN111568451A (en) | Exposure dose adjusting method and system | |
| Gu et al. | Pulmonary nodules detection based on deformable convolution | |
| CN111563496A (en) | Continuous learning for automatic view planning for image acquisition |
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 | ||
| 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. |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |