CN109087357A - Scan orientation method, apparatus, computer equipment and computer readable storage medium - Google Patents
Scan orientation method, apparatus, computer equipment and computer readable storage medium Download PDFInfo
- Publication number
- CN109087357A CN109087357A CN201810835317.5A CN201810835317A CN109087357A CN 109087357 A CN109087357 A CN 109087357A CN 201810835317 A CN201810835317 A CN 201810835317A CN 109087357 A CN109087357 A CN 109087357A
- Authority
- CN
- China
- Prior art keywords
- plane
- image data
- point
- history
- distance
- 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
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
- G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10088—Magnetic resonance imaging [MRI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30048—Heart; Cardiac
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The embodiment of the invention discloses a kind of Scan orientation method, apparatus, computer equipment and computer readable storage medium, method includes: the 3 d image data for obtaining position to be positioned;3 d image data is input in preparatory trained location point characteristic model, each position point feature data of location point characteristic model output are obtained;Plane fitting point is determined according to each position point feature data, and plane equation to be positioned is determined according to plane fitting point.The positioning of the plane of scanning motion only can be realized in method provided by the embodiment of the present invention by the 3 d image data at position to be positioned, be applicable to different scanning portions are scanned by different scanning mode situation when positioning, so that the positioning of the plane of scanning motion is more quickly, accurately, the time required to reducing scanning process, improve the accuracy of scan image, and overcome the inaccuracy positioned according to anatomic landmarks point, it is not full-time in prescan range, remain to the positioning for preferably realizing the plane of scanning motion.
Description
Technical field
The present embodiments relate to medical image scanning technique field more particularly to a kind of Scan orientation method, apparatus, meter
Calculate machine equipment and computer readable storage medium.
Background technique
Medical imaging inspection has very important booster action in the clinical diagnosis of disease, is carrying out medical scanning
When, it is often necessary to the scanning of position/plane is fixed to subject target area, with the diagnosis of aided disease, treatment etc..
By taking magnetic resonance imaging as an example, for conventional magnetic resonance scan sequences, prescan figure is usually passed through by doctor first
As demarcating reference location line by hand after identifying anatomical location, scanning sequence is applied to anatomical location according to reference location line
Carry out accurate scan.
By taking cardiac magnetic resonance scans as an example, cardiac magnetic resonance positioning scanning at least needs to hold one's breath for patient 5 times, and respectively three is disconnected
Face, multilayer axle position, false two chambers, false four chambers and multilayer short axle, breath-hold scans time at least need 4~5 minutes, and centre also needs to cure
Raw six manual positioning, each positioning result can all influence subsequent scanning result, if technician's experience is insufficient, may need
Multiple scanning, resetting, positioning time is long and last positioning result may be difficult to meet the needs of clinical diagnosis.Together
When, result disunity when carrying out heart positioning scanning by manual positioning method, to the same subject, between different operation person
The positioning result of the time different with the same operator can may all have very big difference.It can be seen that existing Scan orientation
Mode efficiency is lower, extends the time required for entire scanning process, and due to labour variance, experience difference, may
Cause recognition result, calibration result inconsistent, the precision of sweep parameter not can guarantee.
Currently, the positioning of plane can also be scanned according to the anatomical location points at position to be positioned.Specifically, passing through
The 3D investigation image for treating positioning position carries out the detection of each anatomical location, is swept by the determination of anatomical fixed position point
Retouch plane.For example, investigating the bicuspid valve point that image determines heart area, cardiac apical by 3D, flow out point for left room to aorta, right
Room maximum angle point equipotential is set a little, and plane composed by specific location point is determined as the plane of scanning motion.If brchypinacoid is perpendicular to two points
The line of valve point and the apex of the heart, by left ventricle center, four Cavity surfaces pass through bicuspid valve and apex of the heart line, and by right ventricle's maximum diameter
Point.But these points are that the Position Approximate point come is summed up according to doctors experience, and when actual scanning, according to anatomical location points, institute is true
Fixed face may not be optimal face, and certain pre-scan images possible range deficiencies cause certain solution plane location points to lack
It loses, so that the position inaccurate of the plane of scanning motion.
Summary of the invention
The embodiment of the invention provides a kind of Scan orientation method, apparatus, computer equipment and computer-readable storage mediums
Matter, to realize the positioning for fast and accurately realizing the plane of scanning motion.
In a first aspect, the embodiment of the invention provides a kind of Scan orientation methods, comprising:
Obtain the 3 d image data at position to be positioned;
The 3 d image data is input in preparatory trained location point characteristic model, it is special to obtain the location point
Levy each position point feature data of model output;
Plane fitting point is extracted from each location point according to each location point characteristic, and according to described flat
Face match point determines plane equation to be positioned.
Second aspect, the embodiment of the invention also provides a kind of Scan orientation devices, comprising:
Data acquisition module, for obtaining the 3 d image data at position to be positioned;
Characteristic module, for the 3 d image data to be input to preparatory trained location point characteristic model
In, obtain each position point feature data of the location point characteristic model output;
Plane determining module, for determining plane fitting point according to each position point feature data, and according to plane fitting point
Determine plane equation to be positioned.
The third aspect, the embodiment of the invention also provides a kind of computer equipment, the equipment includes:
One or more processors;
Storage device, for storing one or more programs;
When one or more of programs are executed by one or more of processors, so that one or more of processing
Device realizes such as Scan orientation method provided by any embodiment of the invention.
Fourth aspect, the embodiment of the invention also provides a kind of computer readable storage mediums, are stored thereon with computer
Program realizes such as Scan orientation method provided by any embodiment of the invention when the program is executed by processor.
The 3 d image data that the embodiment of the present invention passes through acquisition position to be positioned;The 3 d image data is input to
In preparatory trained location point characteristic model, each position point feature data of the location point characteristic model output are obtained;Root
Plane fitting point is determined according to each position point feature data, and plane equation to be positioned is determined according to plane fitting point, the present invention
The 3 d image data that Scan orientation method only passes through position to be positioned provided by embodiment can be realized position to be positioned and sweep
The positioning for retouching plane is applicable to be scanned different scanning portions by different scanning mode situation when positioning, makes
The positioning for obtaining the plane of scanning motion is more quick, accurate, and then reduces scanning process required time, improves the accurate of scan image
Degree, and the inaccuracy positioned according to anatomic landmarks point is overcome, it is not full-time in prescan range, it remains to preferable
Realize the positioning of the plane of scanning motion.
Detailed description of the invention
Fig. 1 is the flow chart of Scan orientation method provided by the embodiment of the present invention one;
Fig. 2 a is the flow chart of Scan orientation method provided by the embodiment of the present invention two;
Fig. 2 b is to be scanned positioning using apart from computation model in Scan orientation method provided by the embodiment of the present invention
Schematic diagram;
Fig. 3 a is the flow chart of Scan orientation method provided by the embodiment of the present invention three;
Fig. 3 b is that the training process of distance field computation model in Scan orientation method provided by the embodiment of the present invention three is shown
It is intended to;
Fig. 3 c is the signal that computation model is trained of adjusting the distance in Scan orientation method provided by the embodiment of the present invention
Figure;
Fig. 4 a is the flow chart of Scan orientation method provided by the embodiment of the present invention four;
Fig. 4 b is that using face parted pattern is scanned positioning in Scan orientation method provided by the embodiment of the present invention
Schematic diagram;
Fig. 5 a is the flow chart of Scan orientation method provided by the embodiment of the present invention five;
Fig. 5 b is the schematic diagram that opposite parted pattern is trained in Scan orientation method provided by the embodiment of the present invention;
Fig. 6 is the structural schematic diagram of Scan orientation device provided by the embodiment of the present invention six;
Fig. 7 is the structural schematic diagram of computer equipment provided by the embodiment of the present invention seven.
Specific embodiment
The present invention is described in further detail with reference to the accompanying drawings and examples.It is understood that this place is retouched
The specific embodiment stated is used only for explaining the present invention rather than limiting the invention.It also should be noted that in order to just
Only the parts related to the present invention are shown in description, attached drawing rather than entire infrastructure.
Embodiment one
Fig. 1 is the flow chart of Scan orientation method provided by the embodiment of the present invention one, and the present embodiment is applicable to pass through
Various scanning mode carry out situation when positioning scanning to each medical imaging position, are particularly suitable for using cardiac magnetic resonance imaging
System is scanned situation when positioning to heart.This method can be executed by Scan orientation device, which can
It is realized in a manner of using software and/or hardware, for example, the Scan orientation device is configured in computer equipment.Such as Fig. 1 institute
Show, this method specifically includes:
S110, the 3 d image data for obtaining position to be positioned.
In the present embodiment, the positioning of plane is scanned by the 3 d image data at position to be positioned.Optionally, to
The 3 d image data at positioning position is to treat positioning position progress prescan to be formed by 3D investigation data.Illustratively, when
When position to be positioned is heart area, 3 d image data is that the 3D of heart area investigates data.
Optionally, the acquisition modes of the 3 d image data at position to be positioned include: to obtain same direction in scanning area
Multi-layer image data, multi-layer image data are subjected to the 3 d image data that image reconstruction forms scanning area, using default
Image zooming-out algorithm the 3 d image data at position to be positioned is extracted from the 3 d image data of scanning area.It is optional
, the 3 d image data of scanning area can be split using automatic threshold algorithm, by the 3-D image of scanning area
Data are divided into foreground data and background data, and the 3-D image number at position to be positioned is then extracted using morphological method
According to.
Illustratively, Multi Slice Mode can be carried out along the z-axis direction, obtained the multi-layer image data on z-axis direction, used figure
As algorithm for reconstructing on z-axis direction multi-layer image data carry out image reconstruction after obtain scanning area 3 d image data, so
The 3 d image data at position to be positioned is extracted from the 3 d image data of scanning area afterwards.When position to be positioned is heart
When region, the segmentation of foreground and background can be carried out to the 3 d image data of scanning area using automatic threshold segmentation algorithm,
Then the 3 d image data of heart area is extracted using morphological method.
S120,3 d image data is input in preparatory trained location point characteristic model, obtains position point feature
The each position point feature data of model output.
In the present embodiment, after the 3 d image data for obtaining position to be positioned, by the 3-D image number at position to be positioned
According to being input in preparatory trained location point characteristic model, the spy of each position point in position 3 d image data to be positioned is obtained
Levy data.
Optionally, the characteristic of each position point includes between each position point coordinate and each position point and plane to be positioning
Distance parameter.Wherein, the distance parameter of each position point and plane to be positioning can be between each position point and plane to be positioning
Distance value, or the distance between each position point and plane to be positioning correlation.
Optionally, before 3 d image data to be input to preparatory trained location point characteristic model, further includes:
It is normalized by the 3 d image data that pre-set image data normalization algorithm treats positioning position,
Obtain image normalization data.
In the present embodiment, before the 3 d image data input position point feature model by position to be positioned, to three
Dimensional data image is normalized.It is to advise 3 d image data by setting that 3 d image data, which is normalized,
Scaling is then carried out, falls in it in small specific sections, facilitates the data handling procedure in follow-up data treatment process.Its
In, the algorithm that 3 d image data is normalized is herein with no restrictions.For example, normalization algorithm can be min-max
Standardized algorithm, zero-mean value standardized algorithm or decimal calibrate standardized algorithm.Wherein, min-max standardized algorithm is logical
Data are normalized in the maxima and minima for crossing data, and zero-mean value standardized algorithm is being averaged by data
Data are normalized in value and standard deviation, and decimal calibration standardized algorithm is the scaling position pair by mobile data
Data are normalized.
Optionally, image data normalization algorithm is min-max standardized algorithm, it can is advised by min-max
The 3 d image data that generalized algorithm treats positioning position is normalized.If the pixel value of each position point in position to be positioned
Maximum value is Imax, minimum value Imin, location point pixel value in position to be positioned is Ic, it is after location point pixel value normalization
I, then each position point pixel value in position to be positioned normalizes calculation formula are as follows: I=(Ic-Imin)/(Imax-Imin)。
Optionally, image data normalization algorithm is zero-mean value standardized algorithm, it can is standardized by zero-mean value
The 3 d image data that algorithm treats positioning position is normalized.If the pixel value mean value of each position point in position to be positioned
ForStandard deviation is σ, and location point pixel value in position to be positioned is Ic, it is I after location point pixel value normalization, then it is to be positioned
Position each position point pixel value normalizes calculation formula are as follows:
S130, plane fitting point is determined according to each position point feature data, and determined according to plane fitting point to be positioned
Plane equation.
In the present embodiment, portion point can be filtered out from each position point feature data as plane fitting point,
Plane equation to be positioned is determined according to the plane fitting point filtered out;Each position point can not also be screened, direct root
According to whole location points or portion point is randomly selected as plane fitting point, and plane to be positioned is determined according to plane fitting point
Equation.
Optionally, according to plane fitting point determine plane equation to be positioned include: according to each position point feature parameter from
Plane fitting point is filtered out in each position point, and each plane fitting point coordinate value is fitted using preset fitting algorithm, is obtained
To plane equation to be positioned.
Optionally, it is put down according to pre-determined distance parameter area and position 3 d image data each position point to be positioned with to be positioned
The distance between face parameter screens each position point, will meet corresponding to the distance parameter of pre-determined distance parameter area condition
Location point as plane fitting point, fit plane equation to be positioned further according to the coordinate value of each plane fitting point.Example
Property, the coordinate value of each plane fitting point can be fitted using least square fitting algorithm, calculate to be positioned put down
Face equation.
Optionally, determine that plane equation to be positioned includes: by optimization algorithm according to each flat according to plane fitting point
The coordinate value of face match point determines plane equation to be positioned.Optionally, plane side to be positioned is determined by optimization algorithm
Cheng Shi, plane fitting point can be all location points in position 3 d image data to be positioned, or 3-D image number
Portion point in.
Illustratively, the parameter of plane to be positioning equation can be calculated using optimization algorithm.In the present embodiment, right
Optimization algorithm is with no restrictions.For example, optimization algorithm can be gradient descent method, Newton method, conjugate direction method or conjugation ladder
Degree method scheduling algorithm.
Illustratively, pass throughDetermine plane to be positioning side
The parameter of journey.Wherein, n is the number of all plane fitting points in position to be positioned, xiFor i-th of plane fitting of plane to be positioning
The x-axis coordinate value of point, yiFor the y-axis coordinate value of i-th of plane fitting point of plane to be positioning, ziFor i-th of plane of plane to be positioning
The z-axis coordinate value of match point, DiFor when plane to be positioning equation is ax+by+cz+d=0, i-th of plane in position to be positioned is quasi-
The distance between chalaza and plane to be positioning value, will meet all plane fitting points and ax+by+cz+d=0 in position to be positioned
It is difference between the distance between plane value and the distance value obtained by distance field computation model and minimum, and meet a2+b2
+c2Parameter of=1 a, b, c, d value as plane to be positioning.
Optionally, determine that plane equation to be positioned includes: by ballot mode according to each plane according to plane fitting point
The coordinate value of match point determines plane equation to be positioned.Optionally, when determining plane equation to be positioned by ballot mode,
Plane fitting point can be all location points in position 3 d image data to be positioned, or in 3 d image data
Portion point.All or part of plane fitting point can be selected to vote, the parameter combination of maximum probability is determined as
The parameter of plane to be positioning equation.
The 3 d image data that the embodiment of the present invention passes through acquisition position to be positioned;The 3 d image data is input to
In preparatory trained location point characteristic model, each position point feature data of the location point characteristic model output are obtained;Root
Plane fitting point is determined according to each position point feature data, and plane equation to be positioned is determined according to plane fitting point, the present invention
Determining for the plane of scanning motion can be realized in the 3 d image data that Scan orientation method only passes through position to be positioned provided by embodiment
Position, is applicable to be scanned different scanning portions by different scanning mode situation when positioning, so that the plane of scanning motion
Positioning more quickly, it is accurate, and then the time required to reducing scanning process, improve the accuracy of scan image, and gram
The inaccuracy positioned according to anatomic landmarks point has been taken, it is not full-time in prescan range, it remains to preferably realize scanning
The positioning of plane.
Embodiment two
Fig. 2 a is the flow chart of Scan orientation method provided by the embodiment of the present invention two, and the present embodiment is in above-described embodiment
On the basis of further optimized.As shown in Figure 2 a, which comprises
S210, the 3 d image data for obtaining position to be positioned.
S220,3 d image data is input in preparatory trained distance field computation model, is obtained apart from field computation
The distance matrix of model output.
In the present embodiment, preparatory trained location point characteristic model is embodied as distance field computation model, it will be each
Location point characteristic is embodied as the distance between each position point and plane to be positioning value, is calculated by distance field computation model
The distance between each position point and plane to be positioning are worth in position 3 d image data to be positioned, each position point and plane to be positioning
The distance between value composition distance matrix.
It should be noted that if before the 3 d image data transmission range field computation model by position to be positioned, it is right
Data have carried out normalized, then the distance value for including in the distance matrix of distance field computation model output can be normalization
Distance value after processing, or the actual distance value between each position point and plane to be positioning.Preferably, it will normalize
The distance value in distance matrix that distance value after processing is exported as distance field computation model, makes the positioning of plane to be positioning
It is more accurate.
S230, each position point is screened according to pre-determined distance range and each distance value, the position of fitting condition will be met
A conduct plane fitting point is set, and plane equation to be positioned is determined according to plane fitting point.
Optionally, using location point corresponding to the distance value within the scope of pre-determined distance as plane fitting point, using pre-
If fitting algorithm each plane fitting is fitted, plane equation to be positioned is calculated.Optionally, pre-determined distance range
It can be determined according to the concrete condition at position to be positioned.Illustratively, when position to be positioned is heart, pre-determined distance range can
Think (0,0.1).Optionally, preset fitting algorithm can be least square fitting algorithm.
In another embodiment of the invention, each position point can not also be screened, directly according to whole positions
Portion point is set a little or randomly selected as plane fitting point, according to plane fitting point by way of optimization algorithm or ballot
Determine plane equation to be positioned.
Optionally, by way of preset fitting algorithm, optimization algorithm or ballot according to each plane fitting point determine to
The more detailed content of the plane equation of positioning can be found in above-described embodiment, and details are not described herein.
Fig. 2 b is to be scanned positioning using apart from computation model in Scan orientation method provided by the embodiment of the present invention
Schematic diagram.The positioning that plane is scanned by preparatory trained distance field computation model is schematically illustrated in figure
Process.As shown in Figure 2 b, the 3 d image data at the position to be positioned extracted from prescan figure is input to and is trained in advance
Distance field computation model in, obtain distance field computation model output distance field, plane to be positioning is determined according to distance field
Plane parameter.
It should be noted that being carried out using Scan orientation method provided by the embodiment of the present invention to 57 subjects
The test of cardiac magnetic resonance imaging, test result are as shown in table 1.It is shown in table 1 in the different planes of scanning motion, uses this hair
The normal vector mean error of Scan orientation plane determined by Scan orientation method provided by bright embodiment and putting down for distance field
Equal error.Normal vector mean error corresponding to each Scan orientation plane and distance field average error value can be obtained from table 1.
And the heart is carried out using each heart scanning plane of the Scan orientation method provided by the embodiment of the present invention to different subjects
Dirty magnetic resonance imaging, imaging results pass through the assessment of doctor, and only an example needs doctor to finely tune, other are clinical acceptable.
Table 1
Brchypinacoid | Two Cavity surfaces | Three Cavity surfaces | Four Cavity surfaces | |
Normal vector mean error (°) | 5.7 | 5.4 | 7.2 | 5.4 |
Distance field mean error (mm) | 5.4 | 3.7 | 4.7 | 6.1 |
Wherein, normal vector angular error calculation are as follows:
Wherein,The normal vector in face is marked for craft,Localization method institute is determined for scanning provided by the embodiment of the present invention
The normal vector of the determining plane of scanning motion.In the present embodiment, each test data is corresponding with a brchypinacoid, two chambers
Face, three Cavity surfaces and four Cavity surfaces, the normal vector mean error of each plane of scanning motion refers to all test data institutes in table 1
The average value of the angular error of the normal vector of the corresponding Scan orientation plane.
Distance field error calculation mode are as follows:Wherein, D0For cardia each position
The distance between point and manual mark face, D1For Scan orientation side provided by cardia each position point and the embodiment of the present invention
Actual range between the plane of scanning motion determined by method.Optionally, the distance matrix that can be exported according to distance field computation model
Determine the actual range between each position point and the plane of scanning motion.
Optionally, if the distance value in the distance matrix of distance field computation model output is the distance after not normalizing
Value, then the distance value in distance matrix directly exported distance field computation model is as corresponding each position point and the plane of scanning motion
Between actual range.If the distance value in the distance matrix of distance field computation model output is the distance value after normalization,
The distance between each position point and the plane of scanning motion of the field computation model that can then adjust the distance output carry out retrospectively calculate, obtain heart
Actual range between position each position point and the plane of scanning motion.Specifically, each distance value that distance field computation model is exported with
Product of the matrix of adjusting the distance when being normalized between used preset threshold is as between each position point and the plane of scanning motion
Actual distance.In table 1 distance field mean error refer to each position point corresponding to each distance value within preset threshold away from
Average value from value error.
The technical solution of the embodiment of the present invention is input to by the 3 d image data on the basis of the above embodiments
In preparatory trained location point characteristic model, each position point feature data of the location point characteristic model output, root are obtained
It determines that plane fitting point is embodied according to each position point feature data, passes through preparatory trained distance field computation model meter
The distance between position each position point and plane to be positioning to be positioned value is calculated, and plane fitting point is determined according to each distance value, is made
The determination for obtaining plane fitting point is more accurate, and then keeps the positioning of plane to be positioning more accurate.
Embodiment three
Fig. 3 a is the flow chart of Scan orientation method provided by the embodiment of the present invention three, and the present embodiment is in above-described embodiment
On the basis of further optimized.As shown in Figure 3a, which comprises
S310, the plane ginseng for obtaining history 3 d image data and the corresponding plane to be positioning of history 3 d image data
Number.
In the present embodiment, based on history 3 d image data and the corresponding plane to be positioning of history 3 d image data
Plane parameter is trained the distance field computation model pre-established.Wherein, history 3 d image data and history are three-dimensional
The plane parameter of the corresponding plane to be positioning of image data is for trained data.Optionally, history 3 d image data
The 3 d image data that can be extracted by the history scan data at position to be positioned, history 3 d image data it is corresponding to
The plane parameter for positioning plane can be the normal vector of plane to be positioning or the plane equation of plane to be positioning.
Optionally, the acquisition modes of the plane parameter of the corresponding plane to be positioning of history 3 d image data include:
By image processing software, each plane to be positioning is positioned using position line from history 3 d image data, is obtained
The plane parameter of each plane to be positioning.Optionally, position line manually can be used to plane to be positioning by image processing software
Positioning, obtains the plane parameter of corresponding plane to be positioning.
Optionally, the acquisition modes of the plane parameter of the corresponding plane to be positioning of history 3 d image data include:
According to the figure of each sub-portion bit image in position to be positioned in scan image information corresponding to history 3 d image data
As information calculates the plane parameter of plane to be positioning.
Optionally, scan image information is to treat the high resolution scanning image obtained when positioning position is formally scanned
Information.Optionally, obtained from true high resolution scanning image information when scanning history 3 d image data it is corresponding to
Position the plane parameter of plane.In general, treating the more general prescan of scan data formal when positioning position is scanned
The information content of data is more abundant, and includes doctor in formal scan data in the data formally scanned to scanning area
The markup information (for example, the localization process such as the various sections at position to be positioned, chamber) of progress, by formal scan data and doctor
Markup information as " goldstandard " during training pattern, the distance field computation model pre-established is trained.
Optionally, pre-scan images are the image of DICOM format, include the figure of current layer in every width DICOM image
Picture information, for example, the top left co-ordinate position of current layer and current layer x-axis unit vector vx and y-axis unit vector vy, it will be with vx
Unit normal vector of the unit vector vertical with vy as plane to be positioning, by determining unit normal vector (a, b, c) as to
Position the plane parameter of plane.In addition, in conjunction with current layer top left co-ordinate information (x0,y0,z0), determine parameter d=-ax0-by0-
cz0, the plane parameter for finally obtaining plane to be positioning is (a, b, c, d), and the plane equation of plane to be positioning is ax+by+cz+d
=0.
Preferably, by way of being positioned manually, plane to be positioning is positioned using position line using image processing software,
Obtain the plane parameter of corresponding plane to be positioning.The acquisition of the plane parameter of plane to be positioning is carried out using image processing software
It is more convenient, accurate.
S320, calculated according to history 3 d image data and plane parameter the corresponding history of history 3 d image data away from
From matrix.
Optionally, by it is preset apart from computational algorithm calculate in history 3 d image data each position point with it is to be positioned
History distance value between plane, the history distance value between each position point and plane to be positioning form history 3 d image data
Corresponding history distance matrix.It optionally, can be Euclidean distance algorithm apart from computational algorithm.Illustratively, if it is to be positioned flat
Face equation is ax+by+cz+d=0, then point (x in any position in history 3 d image datai,yi,zi) and plane to be positioning between
Distance are as follows:
S330, training sample set is generated based on history 3 d image data and history distance matrix, uses training sample set
The distance field computation model pre-established is trained, trained distance field computation model is obtained.
In the present embodiment, history corresponding to usage history 3 d image data and history 3 d image data is apart from square
Battle array is trained as sample to the distance field computation model pre-established.Optionally, the distance field meter pre-established
Calculation model is convolutional neural networks model.
Fig. 3 b is that the training process of distance field computation model in Scan orientation method provided by the embodiment of the present invention three is shown
It is intended to.Illustratively, convolutional neural networks used by the embodiment of the present invention are as shown in Figure 3b.Wherein, solid line (short) straight arrow
Channel before indicating arrow obtains the channel after arrow, dotted line through the operation of convolutional layer, block homogenization layer and relu active coating
Curved arrow and plus sige represent superposition and relu active coating, and solid line (length) curved arrow represents concat layers, i.e., by the channel of front and
Subsequent channel is side by side.
Illustratively, the calculating process of 16 channel A is obtained by 1 channel A are as follows: 1 channel A is uniformed into layer through convolutional layer, block
And relu active coating operation obtains 16 channel A.The calculating process of 32 channel A is obtained by 16 channel A and 16 channel B are as follows: is led to 16
Road A be superimposed with 16 channel B after through relu active coating operation, the operation result obtained through relu active coating is again through convolutional layer, block
Homogenization layer and relu active coating operation obtain 32 channel A.The calculating process of 32 channel Cs is obtained by 16 channel B and 64 channel Cs
Are as follows: 64 channel Cs are obtained into 16 channel D after convolutional layer, block homogenization layer and relu active coating operation, by 16 channel B and are drained into
After 16 channel D, 32 channel Cs are obtained.The calculating process of 64 channel Cs is obtained by 64 channel A, 64 channel B and 32 channel B are as follows: will
64 channel A be superimposed with 64 channel B after through relu active coating operation, the operation result obtained through relu active coating is again through convolution
Layer, block homogenization layer and relu active coating operation obtain 32 channel E, and by 32 channel B and after draining into 32 channel E, it is logical to obtain 64
Road C.The calculating process of 1 channel B is obtained by 32 channel Cs and 32 channel D operations are as follows: is passed through after being superimposed 32 channel Cs with 32 channel D
Relu active coating operation, the operation result obtained through relu active coating are transported through convolutional layer, block homogenization layer and relu active coating again
Calculation obtains 1 channel B.
Optionally, the method for field computation of adjusting the distance model training is herein with no restrictions.Illustratively, field computation of adjusting the distance mould
The training method of type can be back-propagation algorithm, stochastic gradient descent method or randomized optimization process.It in the present embodiment, can be with
Training method using randomized optimization process (adam method) as distance field computation model.
Optionally, distance field computation model is exported into IoutWith goldstandard IlabelBetween 1 norm be used as apart from field computation mould
Cost function in type training process.Wherein, goldstandard IlabelFor history distance matrix corresponding to history 3 d image data,
When history 3 d image data being input in distance field computation model, the gauged distance matrix that should export.Illustratively, generation
Valence function loss=| Iout-Ilabel|.Optionally, the cost function in distance field computation model training process can also be distance
Field computation model exports IoutWith goldstandard IlabelBetween 2 norms, weighting 1 norm or weighting 2 norms, herein with no restrictions.
Optionally, in order to enable trained distance field computation model is suitable for a plurality of types of 3 d image datas.It can
To be expanded to training data, history 3 d image data and corresponding history distance matrix are deformed, after deformation
Data be also used as the training data of distance field computation model.It illustratively, can be to history 3 d image data and corresponding
History distance matrix translated (such as along tri- directions x, y or z random translation, range of translation are ± 50mm), rotation (such as around
Random-Rotation axis rotates random angles, and angular range is ± 20 °), scaling (such as random scaling 0.7-1.3 times) handle, will locate
As training sample, field computation of adjusting the distance model is trained data after reason.As training after being deformed to training data
Data can expand limited training data, increase the training sample of distance field computation model, make the distance field trained
Computation model still is able to export accurate distance matrix in 3 d image data inaccuracy, the generation offset of input.
Optionally, the history distance matrix corresponding to usage history 3 d image data and history 3 d image data is raw
Before training sample set, further includes:
Take slip gauge then to history distance corresponding to history 3 d image data and history 3 d image data according to default
Matrix carries out taking block, using taking history distance matrix corresponding to the history three-dimensional data after block and history 3 d image data raw
At training sample set.
Optionally, the image block of same size, the object of image block are extracted using sliding window or by the way of randomly selecting
Reason may range from 50mm*50mm*50mm-120mm*120mm*120mm, and the spatial resolution of image block may range from 2mm*
2mm*2mm-5mm*5mm*5mm.Illustratively, the size of image block is 100mm*100mm*100mm, spatial resolution 3mm*
3mm*3mm.Using taking history distance matrix corresponding to the history three-dimensional data after block and history 3 d image data to generate instruction
Practice sample set, the size of each training sample pair is reduced, so that the training process speed based on training sample set is faster.
Optionally, before usage history 3 d image data and history distance matrix generate training sample, further includes:
History 3 d image data and history distance matrix data are normalized respectively, history of forming image is returned
One changes data and history range normalization matrix.
Optionally, above-described embodiment can be found in the more detailed content that history 3 d image data is normalized,
Details are not described herein.
Optionally, normalization algorithm identical with history 3 d image data can be used to return history distance matrix
One change processing, also can be used the normalization algorithm different from history 3 d image data and history distance matrix is normalized
Processing.In the present embodiment, the data distribution rule of history 3 d image data and history distance matrix is different, using being different from
History distance matrix is normalized in the normalization algorithm of history 3 d image data.
Illustratively, distance threshold T can be preset1, according to distance matrix to history distance matrix carry out truncation and
Normalization.Specifically, if any distance value is D in history distance matrixi, then by Di>T1Distance value be set as T1, then will
All distance values are divided by T in history distance matrix1, obtain history range normalization matrix.It can be seen that history is apart from normalizing
Change the history range normalization value in matrix between 0-1.Optionally, distance threshold T1Value range is (30mm, 200mm).
History range normalization matrix size after normalization is identical as history 3 d image data size, history range normalization matrix
In it is each value be corresponding position point to plane to be positioning normalized distance.
Fig. 3 c is the signal that computation model is trained of adjusting the distance in Scan orientation method provided by the embodiment of the present invention
Figure.The process that computation model of adjusting the distance is trained is schematically illustrated in figure.As shown in Figure 3c, according to plane to be positioning
Plane parameter calculates distance field corresponding to 3 d image data, by distance field and the position to be positioned extracted from prescan figure
3 d image data computation model of adjusting the distance is trained as training sample, obtain it is trained apart from computation model.
S340, the 3 d image data for obtaining position to be positioned.
S350,3 d image data is input in preparatory trained distance field computation model, is obtained apart from field computation
The distance matrix of model output.
S360, each position point is screened according to pre-determined distance range and each distance value, the position of fitting condition will be met
It sets a conduct and determines plane equation to be positioned apart from plane fitting point, and according to apart from plane fitting point.
It should be noted that the training method of distance field computation model provided by the embodiment of the present invention can be individually performed.
That is, the operating procedure that can be used alone in S310-S330 provided by the embodiment of the present invention completes field meter of adjusting the distance
The training of model is calculated, the 3 d image data based on position to be positioned passes through distance in no longer execution subsequent step S340-S360
Computation model determines the operation of plane equation to be positioned.
The technical solution of the embodiment of the present invention increases field computation model progress of adjusting the distance on the basis of the above embodiments
Trained operation, the plane by obtaining history 3 d image data and the corresponding plane to be positioning of history 3 d image data are joined
Number;The corresponding history distance matrix of history 3 d image data is calculated according to history 3 d image data and plane parameter;Base
In history 3 d image data and history distance matrix generate training sample set, using training sample set to pre-establish away from
Computation model of leaving the theatre is trained, and obtains trained distance field computation model, so that training is obtained apart from computation model
It is more accurate.
Example IV
Fig. 4 a is the flow chart of Scan orientation method provided by the embodiment of the present invention four, and the present embodiment is in above-described embodiment
On the basis of further optimized.As shown in fig. 4 a, which comprises
S410, the 3 d image data for obtaining position to be positioned.
S420,3 d image data is input in preparatory trained face parted pattern, obtains the output of face parted pattern
Apart from subdivision matrix.
In the present embodiment, preparatory trained location point characteristic model is embodied as face parted pattern, by each position
Point feature data materialization is the distance between each position point and plane to be positioning partition value, is calculated by face parted pattern undetermined
The distance between each position point and plane to be positioning partition value, each position point and plane to be positioning in the 3 d image data of position position
The distance between partition value composition apart from subdivision matrix.
Optionally, face parted pattern is disaggregated model, and output includes prospect output channel and background output channel, is obtained
Prospect output channel output data, as corresponding to position 3 d image data to be positioned apart from subdivision matrix.
S430, basis are preset the first segmentation threshold and are respectively screened apart from partition value to each position point, divide meeting
The location point of condition is as plane fitting point.
Optionally, obtain face parted pattern output after subdivision matrix, according to default first segmentation threshold and distance
Each position point is treated each in positioning position 3 d image data with the distance between plane to be positioning partition value in subdivision matrix
Location point is screened, and will be greater than the location point corresponding to the partition value of default first segmentation threshold as plane fitting
Point.Optionally, presetting the first segmentation threshold can be adjusted according to specific position to be positioned.Schematically, when to be positioned
When position is heart area, presetting the first segmentation threshold can be 0.5.
S440, each plane fitting point is fitted by preset fitting algorithm, obtains plane equation to be positioned.
In the present embodiment, each plane fitting point is fitted using fitting algorithm, obtains plane equation to be positioned.
Optionally, be fitted according to each plane fitting point to form plane to be positioning equation with it is true according to plane fitting point in above-described embodiment
The mode of fixed plane equation to be positioned is similar, and more detailed content can be found in above-described embodiment, and details are not described herein.
Fig. 4 b is that using face parted pattern is scanned positioning in Scan orientation method provided by the embodiment of the present invention
Schematic diagram.The process that the positioning of plane is scanned by preparatory trained face parted pattern is schematically illustrated in figure.
As shown in Figure 4 b, the 3 d image data at the position to be positioned extracted from prescan figure is input to preparatory trained face point
Cut in model, obtain the output of face parted pattern apart from subdivision matrix, determine the flat of plane to be positioning according to apart from subdivision matrix
Face parameter.
The technical solution of the embodiment of the present invention is input to by the 3 d image data on the basis of the above embodiments
In preparatory trained location point characteristic model, each position point feature data of the location point characteristic model output, root are obtained
Determine that plane fitting point is embodied according to each position point feature data, by trained face parted pattern in advance calculate to
The distance between position each position point and plane to be positioning partition value are positioned, and will be met by default first segmentation threshold and be preset
The location point corresponding to the partition value of segmentation condition is as segmentation plane match point, so that the determination of segmentation plane match point
It is more accurate, and then keep the positioning of plane to be positioning more accurate, and filter out portion point conduct from each position point
Segmentation plane match point makes the fitting speed of plane to be positioning faster, the time required to reducing scanning process.
Embodiment five
Fig. 5 a is the flow chart of Scan orientation method provided by the embodiment of the present invention five, and the present embodiment is in above-described embodiment
On the basis of further optimized.As shown in Figure 5 a, which comprises
S510, the plane ginseng for obtaining history 3 d image data and the corresponding plane to be positioning of history 3 d image data
Number.
S520, calculated according to history 3 d image data and plane parameter the corresponding history of history 3 d image data away from
From matrix.
In the present embodiment, history 3 d image data and the corresponding plane to be positioning of history 3 d image data are obtained
The mode of plane parameter, and the mode of calculating history distance matrix are similar with above-described embodiment, and specific detailed content can be found in
Above-described embodiment, details are not described herein.
S530, history distance matrix is split according to preset second segmentation threshold, obtains history distance segmentation square
Battle array.
In the present embodiment, after being split according to preset second segmentation threshold to history distance matrix, segmentation is used
Obtained history carries out the training of face parted pattern apart from subdivision matrix.Optionally, history is apart from subdivision matrix by history three-dimensional
The segmentation distance value of each position point and plane to be positioning forms in image data.
Optionally, history distance matrix is split according to preset second segmentation threshold, obtains history distance segmentation
Matrix, comprising:
History distance value each in history distance matrix is adjusted according to preset second segmentation threshold, it will be adjusted
Matrix composed by history distance value is as history apart from subdivision matrix.
Illustratively, if the second segmentation threshold is T, according to the second segmentation threshold T to each history in history distance matrix
Distance value is adjusted.Optionally, if any history distance value is D in history distance matrixi, then by DiThe history distance value of < T
It is adjusted to 1, other history distance values are adjusted to 0.Optionally, the value of the second segmentation threshold can according to position to be positioned into
Row adjustment.Illustratively, when position to be positioned is heart area, the second segmentation threshold can be 5mm.
S540, training sample set is generated based on history 3 d image data and history segmentation distance matrix, uses training sample
This collection is trained the face parted pattern pre-established, obtains trained face parted pattern.
Optionally, the cost function in face parted pattern training process can be focal loss function focalloss, dice
Deng the cost function for classification.In the present embodiment, it adjusts the distance in the training method of opposite parted pattern and above-described embodiment
The training method of computation model is similar, and more detailed content can be found in above-described embodiment, and details are not described herein.
Fig. 5 b is the schematic diagram that opposite parted pattern is trained in Scan orientation method provided by the embodiment of the present invention.
The process that opposite parted pattern is trained is schematically illustrated in figure.As shown in Figure 5 b, according to the plane of plane to be positioning
Parameter calculates corresponding to 3 d image data apart from subdivision matrix, will be extracted apart from subdivision matrix and from prescan figure to
The 3 d image data at positioning position is trained opposite parted pattern as training sample, obtains trained face segmentation mould
Type.
S550, the 3 d image data for obtaining position to be positioned.
S560,3 d image data is input in preparatory trained face parted pattern, obtains the output of face parted pattern
Apart from subdivision matrix.
S570, basis are preset the first segmentation threshold and are respectively screened apart from partition value to each position point, divide meeting
The location point of condition is as plane fitting point.
S580, each plane fitting point is fitted by preset fitting algorithm, obtains plane equation to be positioned.
It should be noted that the training method of face parted pattern provided by the embodiment of the present invention can be individually performed.Also
It is to say, the operating procedure that can be used alone in S510-S540 provided by the embodiment of the present invention completes opposite parted pattern
It is true by face parted pattern no longer to execute the 3 d image data based on position to be positioned in subsequent step S550-S580 for training
The operation of fixed plane equation to be positioned.
The technical solution of the embodiment of the present invention increases opposite parted pattern on the basis of the above embodiments and is trained
Operation, pass through obtain history 3 d image data and the corresponding plane to be positioning of history 3 d image data plane parameter;
The corresponding history distance matrix of history 3 d image data is calculated according to history 3 d image data and plane parameter;According to pre-
If the second segmentation threshold history distance matrix is split, obtain history apart from subdivision matrix;Based on history 3-D image
Data and history segmentation distance matrix generate training sample set, using training sample set to the face parted pattern pre-established into
Row training, obtains trained face parted pattern, so that the obtained face parted pattern of training is more accurate.
It, can also be directly according to history 3 d image data and history three-dimensional figure in another embodiment of the invention
The plane parameter as corresponding to data determines that model is trained to the plane pre-established;When needs are scanned positioning
When, it is directly that the coordinates matrix of position 3 d image data to be positioned and the x, y, z axis of position 3 d image data to be positioned is defeated
Enter to trained plane and determine in model, obtains the plane parameter that plane determines the plane to be positioning of model output.Optionally,
Above-described embodiment can be found in the training method and data processing method of areal model, details are not described herein.
Embodiment six
Fig. 6 is the structural schematic diagram of Scan orientation device provided by the embodiment of the present invention six.The Scan orientation device can
It is realized in a manner of using software and/or hardware, such as the Scan orientation device can be configured in computer equipment, such as Fig. 6
Shown, described device includes: data acquisition module 610, characteristic module 620 and plane determining module 630, in which:
Data acquisition module 610, for obtaining the 3 d image data at position to be positioned;
Characteristic module 620, for the 3 d image data to be input to preparatory trained location point character modules
In type, each position point feature data of the location point characteristic model output are obtained;
Plane determining module 630, for determining plane fitting point according to each position point feature data, and according to plane fitting
Point determines plane equation to be positioned.
The embodiment of the present invention obtains the 3 d image data at position to be positioned by data acquisition module;Characteristic module
The 3 d image data is input in preparatory trained location point characteristic model, it is defeated to obtain the location point characteristic model
Each position point feature data out;Plane determining module determines plane fitting point according to each position point feature data, and according to flat
Face match point determines plane equation to be positioned, and Scan orientation method provided by the embodiment of the present invention only passes through position to be positioned
3 d image data the positioning of the plane of scanning motion can be realized, be applicable to through different scanning mode to different scanning portions
It is scanned situation when positioning, so that the positioning of the plane of scanning motion is more quickly, accurately, and then scanning process is reduced and is taken
Between, the accuracy of scan image is improved, and overcome the inaccuracy positioned according to anatomic landmarks point, swept in advance
It retouches that range is not full-time, remains to the positioning for preferably realizing the plane of scanning motion.
On the basis of above scheme, the characteristic module 620 is specifically used for:
The 3 d image data is input in preparatory trained distance field computation model, the distance field meter is obtained
The distance matrix of model output is calculated, the distance matrix is made of the distance value of each location point and the plane to be positioning.
On the basis of above scheme, the plane determining module 630 is specifically used for:
Each location point is screened according to pre-determined distance range and each distance value, fitting condition will be met
Location point determines plane equation to be positioned as plane fitting point, and according to the plane fitting point.
On the basis of above scheme, described device further include:
Historical data acquiring unit, for it is three-dimensional to obtain history before the 3 d image data for obtaining position to be positioned
The plane parameter of image data and the corresponding plane to be positioning of the history 3 d image data;
Distance matrix determination unit, it is described for being calculated according to the history 3 d image data and the plane parameter
The corresponding history distance matrix of history 3 d image data, the distance matrix by each position point in history 3 d image data with
The distance value of the plane to be positioning forms;
Distance field model training unit, for being generated based on the history 3 d image data and the history distance matrix
Training sample set is trained the distance field computation model pre-established using the training sample set, is trained
Distance field computation model.
On the basis of above scheme, the characteristic module 620 is specifically used for:
The 3 d image data is input in preparatory trained face parted pattern, it is defeated to obtain the face parted pattern
Out apart from subdivision matrix, it is described to be made of apart from subdivision matrix each location point partition value at a distance from plane to be positioning.
On the basis of above scheme, the plane determining module 630 includes:
Match point determination unit is preset the first segmentation threshold for basis and is respectively sieved apart from partition value to each position point
Choosing, will meet the location point of segmentation condition as plane fitting point;
Plane fitting unit obtains to be positioned put down for being fitted each plane fitting point by preset fitting algorithm
Face equation.
On the basis of above scheme, described device further include:
Historical data acquiring unit, for it is three-dimensional to obtain history before the 3 d image data for obtaining position to be positioned
The plane parameter of image data and the corresponding plane to be positioning of the history 3 d image data;
Distance matrix determination unit, it is described for being calculated according to the history 3 d image data and the plane parameter
The corresponding history distance matrix of history 3 d image data, the history distance matrix is by each in the history 3 d image data
The distance value of location point and the plane to be positioning forms;
Subdivision matrix determination unit, for being divided according to preset second segmentation threshold the history distance matrix
It cuts, obtains history apart from subdivision matrix, the history is apart from subdivision matrix by each position point in the history 3 d image data
It is formed with the segmentation distance value of the plane to be positioning;
Face parted pattern training unit, for dividing distance matrix based on the history 3 d image data and the history
Training sample set is generated, the face parted pattern pre-established is trained using the training sample set, is trained
Face parted pattern.
Scan orientation method provided by any embodiment can be performed in Scan orientation device provided by the embodiment of the present invention,
Have the corresponding functional module of execution method and beneficial effect.
Embodiment seven
Fig. 7 is the structural schematic diagram of computer equipment provided by the embodiment of the present invention seven.Fig. 7, which is shown, to be suitable for being used in fact
The block diagram of the exemplary computer device 712 of existing embodiment of the present invention.The computer equipment 712 that Fig. 7 is shown is only one
Example, should not function to the embodiment of the present invention and use scope bring any restrictions.
As shown in fig. 7, computer equipment 712 is showed in the form of universal computing device.The component of computer equipment 712 can
To include but is not limited to: one or more processor 716, system storage 728 connect different system components (including system
Memory 728 and processor 716) bus 718.
Bus 718 indicates one of a few class bus structures or a variety of, including memory bus or Memory Controller,
Peripheral bus, graphics acceleration port, processor 716 or total using the local of any bus structures in a variety of bus structures
Line.For example, these architectures include but is not limited to industry standard architecture (ISA) bus, microchannel architecture
(MAC) bus, enhanced isa bus, Video Electronics Standards Association (VESA) local bus and peripheral component interconnection (PCI) are total
Line.
Computer equipment 712 typically comprises a variety of computer system readable media.These media can be it is any can
The usable medium accessed by computer equipment 712, including volatile and non-volatile media, moveable and immovable Jie
Matter.
System storage 728 may include the computer system readable media of form of volatile memory, such as deposit at random
Access to memory (RAM) 730 and/or cache memory 732.Computer equipment 712 may further include it is other it is removable/
Immovable, volatile/non-volatile computer system storage medium.Only as an example, storage device 734 can be used for reading
Write immovable, non-volatile magnetic media (Fig. 7 do not show, commonly referred to as " hard disk drive ").Although being not shown in Fig. 7,
The disc driver for reading and writing to removable non-volatile magnetic disk (such as " floppy disk ") can be provided, and non-easy to moving
The CD drive that the property lost CD (such as CD-ROM, DVD-ROM or other optical mediums) is read and write.In these cases, each
Driver can be connected by one or more data media interfaces with bus 718.Memory 728 may include at least one
Program product, the program product have one group of (for example, at least one) program module, these program modules are configured to perform this
Invent the function of each embodiment.
Program/utility 740 with one group of (at least one) program module 742, can store in such as memory
In 728, such program module 742 includes but is not limited to operating system, one or more application program, other program modules
And program data, it may include the realization of network environment in each of these examples or certain combination.Program module 742
Usually execute the function and/or method in embodiment described in the invention.
Computer equipment 712 can also be with one or more external equipments 714 (such as keyboard, sensing equipment, display
724 etc.) it communicates, the equipment interacted with the computer equipment 712 communication can be also enabled a user to one or more, and/or
(such as network interface card is adjusted with any equipment for enabling the computer equipment 712 to be communicated with one or more of the other calculating equipment
Modulator-demodulator etc.) communication.This communication can be carried out by input/output (I/O) interface 722.Also, computer equipment
712 can also by network adapter 720 and one or more network (such as local area network (LAN), wide area network (WAN) and/or
Public network, such as internet) communication.As shown, network adapter 720 passes through its of bus 718 and computer equipment 712
The communication of its module.It should be understood that although not shown in the drawings, other hardware and/or software can be used in conjunction with computer equipment 712
Module, including but not limited to: microcode, device driver, redundant processing unit, external disk drive array, RAID system, magnetic
Tape drive and data backup storage system etc..
Processor 716 by the program that is stored in system storage 728 of operation, thereby executing various function application and
Data processing, such as realize Scan orientation method provided by the embodiment of the present invention, this method comprises:
Obtain the 3 d image data at position to be positioned;
The 3 d image data is input in preparatory trained location point characteristic model, it is special to obtain the location point
Levy each position point feature data of model output;
Plane fitting point is determined according to each position point feature data, and plane side to be positioned is determined according to plane fitting point
Journey.
Certainly, it will be understood by those skilled in the art that processor can also realize it is provided by any embodiment of the invention
The technical solution of Scan orientation method.
Embodiment eight
The embodiment of the present invention eight additionally provides a kind of computer readable storage medium, is stored thereon with computer program, should
The Scan orientation method as provided by the embodiment of the present invention is realized when program is executed by processor, this method comprises:
Obtain the 3 d image data at position to be positioned;
The 3 d image data is input in preparatory trained location point characteristic model, it is special to obtain the location point
Levy each position point feature data of model output;
Plane fitting point is determined according to each position point feature data, and plane side to be positioned is determined according to plane fitting point
Journey.
Certainly, a kind of computer readable storage medium provided by the embodiment of the present invention, the computer program stored thereon
The method operation being not limited to the described above, can also be performed the phase in Scan orientation method provided by any embodiment of the invention
Close operation.
The computer storage medium of the embodiment of the present invention, can be using any of one or more computer-readable media
Combination.Computer-readable medium can be computer-readable signal media or computer readable storage medium.It is computer-readable
Storage medium for example may be-but not limited to-the system of electricity, magnetic, optical, electromagnetic, infrared ray or semiconductor, device or
Device, or any above combination.The more specific example (non exhaustive list) of computer readable storage medium includes: tool
There are electrical connection, the portable computer diskette, hard disk, random access memory (RAM), read-only memory of one or more conducting wires
(ROM), erasable programmable read only memory (EPROM or flash memory), optical fiber, portable compact disc read-only memory (CD-
ROM), light storage device, magnetic memory device or above-mentioned any appropriate combination.In this document, computer-readable storage
Medium can be any tangible medium for including or store program, which can be commanded execution system, device or device
Using or it is in connection.
Computer-readable signal media may include in a base band or as carrier wave a part propagate data-signal,
Wherein carry computer-readable program code.The data-signal of this propagation can take various forms, including but unlimited
In electromagnetic signal, optical signal or above-mentioned any appropriate combination.Computer-readable signal media can also be that computer can
Any computer-readable medium other than storage medium is read, which can send, propagates or transmit and be used for
By the use of instruction execution system, device or device or program in connection.
The program code for including on computer-readable medium can transmit with any suitable medium, including --- but it is unlimited
In wireless, electric wire, optical cable, RF etc. or above-mentioned any appropriate combination.
The computer for executing operation of the present invention can be write with one or more programming languages or combinations thereof
Program code, described program design language include object oriented program language-such as Java, Smalltalk, C++,
It further include conventional procedural programming language-such as " C " language or similar programming language.Program code can be with
It fully executes, partly execute on the user computer on the user computer, being executed as an independent software package, portion
Divide and partially executes or executed on a remote computer or server completely on the remote computer on the user computer.?
Be related in the situation of remote computer, remote computer can pass through the network of any kind --- including local area network (LAN) or
Wide area network (WAN)-be connected to subscriber computer, or, it may be connected to outer computer (such as mentioned using Internet service
It is connected for quotient by internet).
Note that the above is only a better embodiment of the present invention and the applied technical principle.It will be appreciated by those skilled in the art that
The invention is not limited to the specific embodiments described herein, be able to carry out for a person skilled in the art it is various it is apparent variation,
It readjusts and substitutes without departing from protection scope of the present invention.Therefore, although being carried out by above embodiments to the present invention
It is described in further detail, but the present invention is not limited to the above embodiments only, without departing from the inventive concept, also
It may include more other equivalent embodiments, and the scope of the invention is determined by the scope of the appended claims.
Claims (10)
1. a kind of Scan orientation method characterized by comprising
Obtain the 3 d image data at position to be positioned;
The 3 d image data is input in preparatory trained location point characteristic model, the location point character modules are obtained
The each position point feature data of type output;
Plane fitting point is determined according to each position point feature data, and plane equation to be positioned is determined according to plane fitting point.
2. the method according to claim 1, wherein described be input to preparatory training for the 3 d image data
In good location point characteristic model, each position point feature data of the location point characteristic model output are obtained, comprising:
The 3 d image data is input in preparatory trained distance field computation model, is obtained described apart from field computation mould
The distance matrix of type output, the distance matrix are made of the distance value of each location point and the plane to be positioning.
3. according to the method described in claim 2, it is characterized in that, described determine plane fitting according to each position point feature data
Point, and plane equation to be positioned is determined according to plane fitting point, comprising:
Each location point is screened according to pre-determined distance range and each distance value, the position of fitting condition will be met
Point is used as plane fitting point, and determines plane equation to be positioned according to the plane fitting point.
4. according to the method described in claim 2, it is characterized in that, before the 3 d image data for obtaining position to be positioned,
Further include:
Obtain the plane parameter of history 3 d image data and the corresponding plane to be positioning of the history 3 d image data;
Calculate that the history 3 d image data is corresponding to be gone through according to the history 3 d image data and the plane parameter
History distance matrix, the distance matrix by each position point and the plane to be positioning in history 3 d image data distance value group
At;
Training sample set is generated based on the history 3 d image data and the history distance matrix, uses the training sample
Collection is trained the distance field computation model pre-established, obtains trained distance field computation model.
5. the method according to claim 1, wherein described be input to preparatory training for the 3 d image data
In good location point characteristic model, each position point feature data of the location point characteristic model output are obtained, comprising:
The 3 d image data is input in preparatory trained face parted pattern, the face parted pattern output is obtained
It is described to be made of apart from subdivision matrix each location point partition value at a distance from plane to be positioning apart from subdivision matrix.
6. according to the method described in claim 5, it is characterized in that, described determine plane fitting according to each position point feature data
Point, and plane equation to be positioned is determined according to plane fitting point, comprising:
According to default first segmentation threshold and respectively each position point is screened apart from partition value, the position of segmentation condition will be met
Point is used as plane fitting point;
Each plane fitting point is fitted by preset fitting algorithm, obtains plane equation to be positioned.
7. according to the method described in claim 5, it is characterized in that, before the 3 d image data for obtaining position to be positioned,
Further include:
Obtain the plane parameter of history 3 d image data and the corresponding plane to be positioning of the history 3 d image data;
Calculate that the history 3 d image data is corresponding to be gone through according to the history 3 d image data and the plane parameter
History distance matrix, the history distance matrix is by each position point in the history 3 d image data and the plane to be positioning
Distance value composition;
The history distance matrix is split according to preset second segmentation threshold, obtains history apart from subdivision matrix, institute
State segmentation distance of the history apart from subdivision matrix by each position point and the plane to be positioning in the history 3 d image data
Value composition;
Training sample set is generated based on the history 3 d image data and history segmentation distance matrix, uses the training
Sample set is trained the face parted pattern pre-established, obtains trained face parted pattern.
8. a kind of Scan orientation device characterized by comprising
Data acquisition module, for obtaining the 3 d image data at position to be positioned;
Characteristic module is obtained for the 3 d image data to be input in preparatory trained location point characteristic model
Obtain each position point feature data of the location point characteristic model output;
Plane determining module for determining plane fitting point according to each position point feature data, and is determined according to plane fitting point
Plane equation to be positioned.
9. a kind of computer equipment, which is characterized in that the equipment includes:
One or more processors;
Storage device, for storing one or more programs;
When one or more of programs are executed by one or more of processors, so that one or more of processors are real
The now Scan orientation method as described in any in claim 1-7.
10. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the program is by processor
The Scan orientation method as described in any in claim 1-7 is realized when execution.
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810835317.5A CN109087357B (en) | 2018-07-26 | 2018-07-26 | Scanning positioning method and device, computer equipment and computer readable storage medium |
CN202110943320.0A CN113610923A (en) | 2018-07-26 | 2018-07-26 | Scanning positioning method and device, computer equipment and computer readable storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810835317.5A CN109087357B (en) | 2018-07-26 | 2018-07-26 | Scanning positioning method and device, computer equipment and computer readable storage medium |
Related Child Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110943320.0A Division CN113610923A (en) | 2018-07-26 | 2018-07-26 | Scanning positioning method and device, computer equipment and computer readable storage medium |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109087357A true CN109087357A (en) | 2018-12-25 |
CN109087357B CN109087357B (en) | 2021-06-29 |
Family
ID=64830863
Family Applications (2)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810835317.5A Active CN109087357B (en) | 2018-07-26 | 2018-07-26 | Scanning positioning method and device, computer equipment and computer readable storage medium |
CN202110943320.0A Pending CN113610923A (en) | 2018-07-26 | 2018-07-26 | Scanning positioning method and device, computer equipment and computer readable storage medium |
Family Applications After (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110943320.0A Pending CN113610923A (en) | 2018-07-26 | 2018-07-26 | Scanning positioning method and device, computer equipment and computer readable storage medium |
Country Status (1)
Country | Link |
---|---|
CN (2) | CN109087357B (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110163857A (en) * | 2019-05-24 | 2019-08-23 | 上海联影医疗科技有限公司 | Detection method, device, storage medium and the x-ray system of image background regions |
CN110223352A (en) * | 2019-06-14 | 2019-09-10 | 浙江明峰智能医疗科技有限公司 | A kind of medical image scanning automatic positioning method based on deep learning |
CN111443318A (en) * | 2019-01-16 | 2020-07-24 | 上海联影医疗科技有限公司 | Magnetic resonance image processing method, magnetic resonance image processing device, storage medium and magnetic resonance imaging system |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101034480A (en) * | 2007-04-05 | 2007-09-12 | 上海交通大学 | Method for computer-assisted rebuilding heart mitral annulus |
CN101315661A (en) * | 2008-07-18 | 2008-12-03 | 东南大学 | Fast three-dimensional face recognition method for reducing expression influence |
CN101344373A (en) * | 2008-08-14 | 2009-01-14 | 中国人民解放军总后勤部军需装备研究所 | Standardization processing method based on three-dimensional head and face curved surface modeling |
CN101836235A (en) * | 2007-08-16 | 2010-09-15 | 皇家飞利浦电子股份有限公司 | Imaging method for sampling a cross-section plane in a three-dimensional (3d) image data volume |
JP2014127788A (en) * | 2012-12-26 | 2014-07-07 | Nippon Hoso Kyokai <Nhk> | Device of correcting stereoscopic image, and program of the same |
CN104166978A (en) * | 2013-12-27 | 2014-11-26 | 上海联影医疗科技有限公司 | Blood vessel extracting method |
CN105389813A (en) * | 2015-10-30 | 2016-03-09 | 上海联影医疗科技有限公司 | Medical image organ recognition method and segmentation method |
CN105900140A (en) * | 2013-11-05 | 2016-08-24 | 皇家飞利浦有限公司 | Automated segmentation of tri-plane images for real time ultrasonic imaging |
CN106204514A (en) * | 2015-04-30 | 2016-12-07 | 中国科学院深圳先进技术研究院 | A kind of liver localization method based on three-dimensional CT image and device |
-
2018
- 2018-07-26 CN CN201810835317.5A patent/CN109087357B/en active Active
- 2018-07-26 CN CN202110943320.0A patent/CN113610923A/en active Pending
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101034480A (en) * | 2007-04-05 | 2007-09-12 | 上海交通大学 | Method for computer-assisted rebuilding heart mitral annulus |
CN101836235A (en) * | 2007-08-16 | 2010-09-15 | 皇家飞利浦电子股份有限公司 | Imaging method for sampling a cross-section plane in a three-dimensional (3d) image data volume |
CN101315661A (en) * | 2008-07-18 | 2008-12-03 | 东南大学 | Fast three-dimensional face recognition method for reducing expression influence |
CN101344373A (en) * | 2008-08-14 | 2009-01-14 | 中国人民解放军总后勤部军需装备研究所 | Standardization processing method based on three-dimensional head and face curved surface modeling |
JP2014127788A (en) * | 2012-12-26 | 2014-07-07 | Nippon Hoso Kyokai <Nhk> | Device of correcting stereoscopic image, and program of the same |
CN105900140A (en) * | 2013-11-05 | 2016-08-24 | 皇家飞利浦有限公司 | Automated segmentation of tri-plane images for real time ultrasonic imaging |
CN104166978A (en) * | 2013-12-27 | 2014-11-26 | 上海联影医疗科技有限公司 | Blood vessel extracting method |
CN106204514A (en) * | 2015-04-30 | 2016-12-07 | 中国科学院深圳先进技术研究院 | A kind of liver localization method based on three-dimensional CT image and device |
CN105389813A (en) * | 2015-10-30 | 2016-03-09 | 上海联影医疗科技有限公司 | Medical image organ recognition method and segmentation method |
Non-Patent Citations (1)
Title |
---|
张波: "基于三维人体骨架模型的动作识别", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111443318A (en) * | 2019-01-16 | 2020-07-24 | 上海联影医疗科技有限公司 | Magnetic resonance image processing method, magnetic resonance image processing device, storage medium and magnetic resonance imaging system |
CN111443318B (en) * | 2019-01-16 | 2022-08-02 | 上海联影智能医疗科技有限公司 | Magnetic resonance image processing method, magnetic resonance image processing device, storage medium and magnetic resonance imaging system |
CN110163857A (en) * | 2019-05-24 | 2019-08-23 | 上海联影医疗科技有限公司 | Detection method, device, storage medium and the x-ray system of image background regions |
CN110223352A (en) * | 2019-06-14 | 2019-09-10 | 浙江明峰智能医疗科技有限公司 | A kind of medical image scanning automatic positioning method based on deep learning |
Also Published As
Publication number | Publication date |
---|---|
CN113610923A (en) | 2021-11-05 |
CN109087357B (en) | 2021-06-29 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110956635B (en) | Lung segment segmentation method, device, equipment and storage medium | |
US11576645B2 (en) | Systems and methods for scanning a patient in an imaging system | |
CN107480677B (en) | Method and device for identifying interest region in three-dimensional CT image | |
CN107909622B (en) | Model generation method, medical imaging scanning planning method and medical imaging system | |
US11576578B2 (en) | Systems and methods for scanning a patient in an imaging system | |
CN111553892B (en) | Lung nodule segmentation calculation method, device and system based on deep learning | |
Jones | Facial Reconstruction Using Volumetric Data. | |
US10275895B2 (en) | Mechanism for advanced structure generation and editing | |
CN109741312A (en) | A kind of Lung neoplasm discrimination method, device, equipment and medium | |
CN110246580B (en) | Cranial image analysis method and system based on neural network and random forest | |
Li et al. | Automatic lumbar vertebrae detection based on feature fusion deep learning for partial occluded C-arm X-ray images | |
CN109087357A (en) | Scan orientation method, apparatus, computer equipment and computer readable storage medium | |
US20210271914A1 (en) | Image processing apparatus, image processing method, and program | |
CN111738989A (en) | Organ delineation method and device | |
EP3047455B1 (en) | Method and system for spine position detection | |
CN112950648A (en) | Method and apparatus for determining the midsagittal plane in a magnetic resonance image | |
EP2584970A2 (en) | Mechanism for advanced structure generation and editing | |
WO2015040547A1 (en) | Method and system for spine position detection | |
Yang et al. | A multi-stage progressive learning strategy for COVID-19 diagnosis using chest computed tomography with imbalanced data | |
TWI697010B (en) | Method of obtaining medical sagittal image, method of training neural network and computing device | |
CN113706514A (en) | Focus positioning method, device and equipment based on template image and storage medium | |
EP4167184A1 (en) | Systems and methods for plaque identification, plaque composition analysis, and plaque stability detection | |
US9401051B2 (en) | Mechanism for dynamically propagating real-time alterations of medical images | |
Cheng et al. | Ground truth delineation for medical image segmentation based on Local Consistency and Distribution Map analysis | |
CN113657420A (en) | Body representation |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |