CN107590510A - A kind of image position method, device, computer and storage medium - Google Patents
A kind of image position method, device, computer and storage medium Download PDFInfo
- Publication number
- CN107590510A CN107590510A CN201710757889.1A CN201710757889A CN107590510A CN 107590510 A CN107590510 A CN 107590510A CN 201710757889 A CN201710757889 A CN 201710757889A CN 107590510 A CN107590510 A CN 107590510A
- Authority
- CN
- China
- Prior art keywords
- probability
- pixel
- forest
- lifting
- random chance
- 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.)
- Pending
Links
Abstract
The embodiment of the invention discloses a kind of image position method, device, computer and storage medium, wherein method includes:Medical image is obtained, and determines the characteristic information of each pixel of the medical image;Each pixel is classified according to the characteristic information and random chance lifting forest classified device, and calculates the probability that each pixel belongs to target location;The positional information for the pixel for meeting predetermined probabilities condition is defined as to the positioning result of the target location.The embodiment of the present invention solves the problems, such as that framing accuracy is low in the prior art, poor robustness, improves the precision and robustness of framing.
Description
Technical field
The present embodiments relate to framing technology, more particularly to a kind of image position method, device, computer and deposit
Storage media.
Background technology
With the continuous development of image processing techniques, image processing techniques be widely used in medical image, robot vision,
The field such as Aero-Space and communication.
The state of an illness of doctor acquisition patient, such as medical image is aided in include CT and scheme by various medical images in medical domain
Picture, MRI and ultrasonoscopy etc..In the post processing of medical image, to single in medical image or multiple targets
Positioning be image segmentation and image registration etc. ought to prerequisite steps, there is important clinical value, but by
Easily influenceed in the image quality of medical image by various factors, for example, noise in imaging process, the pathology of patient, position,
Contrast agent that posture is acted or taken etc. can influence the image quality of medical image, the difficulty of increase image object positioning
Degree.
In recent years, the artificial intelligence technology using machine learning as representative is constantly developed, and is gradually applied to medical science
In the target positioning of image.At present, classification learning, learning model poor robustness, positioning accurate are carried out based on single tree-model
Spend low.
The content of the invention
The embodiment of the present invention provides a kind of image position method, device, computer and storage medium, to realize raising image
The robustness of positioning.
In a first aspect, the embodiments of the invention provide a kind of image position method, this method includes:
Medical image is obtained, and determines the characteristic information of each pixel of the medical image;
Each pixel is classified according to the characteristic information and random chance lifting forest classified device, and calculated
Each pixel belongs to the probability of target location;
The positional information for the pixel for meeting predetermined probabilities condition is defined as to the positioning result of the target location.
Further, before being classified according to random chance boosted tree forest classified device to each pixel, also
Including:
It is determined that meet the pixel of intensity value ranges corresponding to target location.
Further, the probability that each pixel belongs to target location is calculated, including:
The probability of each pixel each probability lifting root vertex in random chance lifts forest is calculated, is defined as institute
The probability that each pixel belongs to target location in the probability lifts root vertex is stated, wherein, each pixel is each general
The probability of rate lifting root vertex is equal to the probability of the subtree root node of the probability boosted tree and the weight of its corresponding root node
The sum of products;
The average value of the probability of each probability lifting root vertex in random chance lifting forest is defined as described each
The probability of pixel.
Further, before medical image is obtained, in addition to:
At least two training sample sets are obtained, random chance lifting forest classified device is trained according to the training sample set.
Further, random chance lifting forest classified device is trained according to the training sample set, including:
By the training sample according to the random chance lifted forest classified device classified, by classification results with it is described
The goldstandard of training sample set is compared, and determines the training error of the random chance lifting forest;
If the training error of the random chance lifting forest is more than default error threshold, where misclassification sample
The band of position increases sample collection quantity, updates the training sample set;
The random chance lifting forest classified device is trained according to the training sample set after renewal.
After random chance lifting forest classified device is trained according to the training sample set, in addition to:
The training sample of other probability boosted trees in forest in addition to current probability boosted tree is lifted according to random chance
Collection tests the current probability lifting Tree Classifier, obtains test result;
The training result of the current probability boosted tree is updated according to the test result.
Further, the medical image includes:Angiographic image or heart contrastographic picture;
Accordingly, the target location includes vessel seed point or aorta petal.
Second aspect, the embodiment of the present invention additionally provide a kind of image positioning device, and the device includes:
Image collection module, for obtaining medical image, and determine the characteristic information of each pixel of the medical image;
Probability determination module, forest classified device is lifted according to the characteristic information and random chance each pixel is clicked through
Row classification, and calculate the probability that each pixel belongs to target location;
Positioning result determining module, for the positional information for meeting the pixel of predetermined probabilities condition to be defined as into the mesh
The positioning result of cursor position.
Further, described device also includes:
Pixel determining module, for classifying according to random chance lifting forest classified device to each pixel
Before, it is determined that meeting the pixel of intensity value ranges corresponding to target location.
Further, test module is specifically used for:
The probability of each pixel each probability lifting root vertex in random chance lifts forest is calculated, is defined as institute
The probability that each pixel belongs to target location in probability lifts root vertex is stated, wherein, each pixel carries in each probability
The probability for rising root vertex is equal to the probability of subtree root node and the weight product of its corresponding root node of the probability boosted tree
Sum;
The average value of the probability of each predetermined probabilities lifting root vertex of random chance lifting forest kind is defined as institute
State the probability of each pixel.
Further, described device also includes:
Probability boosted tree training module, for before medical image is obtained, obtaining at least two training sample sets, according to
The training sample set training random chance lifting forest classified device.
Further, described device also includes:
Training error determining module, for according to the training sample set train random chance lifting forest classified device it
Afterwards, the training sample is lifted into forest classified device according to the random chance to be classified, by classification results and the training
Sample is compared, and determines training error of each training sample in probability boosted tree;
Sample set update module, if the training error for the training sample is more than default error threshold, increase institute
The sample collection quantity in sample position region is stated, updates the training sample set;
Grader circuit training module, according to the training sample set training random chance lifting forest classified device after renewal.
Further, described device also includes:
Random chance lifts forest test module, for training random chance lifting forest according to the training sample set
After grader, the training sample of other probability boosted trees in forest in addition to current probability boosted tree is lifted according to random chance
This collection tests the grader of the current probability boosted tree, obtains test result;
Training result update module, for updating the training knot of the current probability boosted tree according to the test result
Fruit.
Further, the medical image includes:Angiographic image or heart contrastographic picture;
Accordingly, the target location includes vessel seed point or aorta petal.
The third aspect, the embodiment of the present invention additionally provide a kind of computer, it is characterised in that including memory, processor
And store the computer program that can be run on a memory and on a processor, it is characterised in that described in the computing device
During program any described methods of 1-7 are required available for perform claim.
Fourth aspect, the embodiment of the present invention additionally provide a kind of computer-readable recording medium, are stored thereon with computer
Program, it is characterised in that require any described methods of 1-7 available for perform claim when the program is executed by processor.
The embodiment of the present invention lifts forest by the way that the characteristic information of bit image undetermined is inputted into random chance, determines each pixel
Point belongs to the probability of target location, according to predetermined probabilities condition, is defined as the positioning result of target location, is carried by random chance
Rise forest and instead of traditional classification tree, while avoid the situation that single classification tree there may be error, solve prior art
The problem of middle framing accuracy is low, poor robustness, improve the precision and robustness of framing.
Brief description of the drawings
Fig. 1 is a kind of flow chart for image position method that the embodiment of the present invention one provides;
Fig. 2 is a kind of flow chart for image position method that the embodiment of the present invention two provides;
Fig. 3 is a kind of structural representation for image positioning device that the embodiment of the present invention three provides;
Fig. 4 is a kind of structural representation for computer that the embodiment of the present invention four provides.
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 limitation of the invention.It also should be noted that in order to just
Part related to the present invention rather than entire infrastructure are illustrate only in description, accompanying drawing.
Embodiment one
Fig. 1 is a kind of flow chart for image position method that the embodiment of the present invention one provides, and the present embodiment is applicable to pair
The situation that image is accurately positioned, this method can be performed by image positioning device provided in an embodiment of the present invention, the dress
Putting can be realized by the way of software and/or hardware.Referring to Fig. 1, this method specifically includes:
S110, medical image is obtained, and determine the characteristic information of each pixel of medical image.
Wherein, medical image can be the medical image gathered by medical imaging devices.Exemplary, medical image can be with
It is MRI (Magnetic Resonance Imaging, mr techniques) image, CT (Computed Tomography, electronics
Computer tomography technology) image or ultrasonoscopy etc..In the present embodiment, medical image is gray level image, if
The image of collection is coloured image, then the coloured image is converted into gray level image.
Wherein, the characteristic information of medical image refer to the grey scale pixel value by each pixel, pixel grey scale gradient or
Relation between each grey scale pixel value of person characterizes the feature of image content information.Optionally, the characteristic information of medical image is each
The Haar features of pixel, it is preferred that the characteristic information of medical image is that 3D (Three Dimensions, three-dimensional) class Haar is special
Sign.Wherein, Haar features are capable of the grey scale change of phenogram picture, Haar feature templates can be edge feature, linear character, in
The template of heart feature and diagonal feature composition, white and two kinds of matrixes of black are included in feature templates, and Haar characteristic values are
White rectangle pixel and with black rectangle pixel and difference.3D Like-Fenton Oxidations are calculated between the multiple image of sequential
Haar characteristic values, the movable information of scene can be preferably described, catch the motor pattern of objects in images, in the present embodiment, if
Bit image undetermined is medical image, then 3D Like-Fenton Oxidations can preferably describe each organ of patient's body beating situation or
Body involuntary movement caused by the physiological phenomenons such as the breathing of patient, improves the degree of accuracy of the extraction of characteristics of image to be positioned,
Improve image position accuracy.
In the present embodiment, the characteristic information of bit image undetermined is gathered in the following way:Using current pixel point as sample point,
In rectangular image block centered on the sample point, the pixel ash of the rectangular image block of two optional positions and arbitrary size is calculated
Angle value sum, then both differences are calculated, obtain the 3D Like-Fenton Oxidations of current sample point.
S120, according to characteristic information and random chance lifting forest classified device each pixel is classified, and calculated each
Pixel belongs to the probability of target location.
Wherein, tree graph refers to a kind of connected graph of loop free, including the feature such as root, branch, node and tip, such as wraps
Include binary tree or decision tree etc..Probability tree refers to being labeled with the tree graph of the chance phenomenon of the probability of each branch in each branch
Shape, wherein, the probability of all child nodes of a root node and for 1.In the present embodiment, probability boosted tree is the excellent of probability tree
Change, be different from common probabilistic tree, binary tree or decision tree etc., each root node is provided with probability interval, probability boosted tree
Each root node can include multiple child nodes, optionally, each root node of probability boosted tree includes left and right two
Individual child node.Exemplary, a root node includes two child nodes, for common probabilistic tree, if the probability of sampled point is more than
0.5, then the sampled point is divided to left child node, if the probability of sampled point is less than middle probable value 0.5, by the sampling dot-dash
Divide to right child node.Be set to 0.1 for probability boosted tree, such as probability, if the probability of sampled point be less than 0.4, i.e., in
Between the difference of probable value and sampled point probability be more than probability interval, then the sampled point is divided to left child node;If sampled point
Probability be more than 0.6, i.e. the difference of sampled point probable value and middle probable value is more than probability interval, then is divided to the sampled point
Right child node;If the probability of sampled point is less than or equal to 0.6, and more than or equal to 0.4, i.e. sampled point probable value and middle probable value
Difference is less than probability interval, then the sampled point is divided into right child node and left child node simultaneously, and two child nodes are carried out
The probabilistic classification of next stage.Wherein, the setting at probability interval needs to be determined according to specifically location tasks, such as positions sustainer
In the task of valve, the probability interval could be arranged to 0.1.
In the present embodiment, at least two probability boosted trees, each probability lifting are included in random chance lifting forest
Tree used training sample set in training carries out sampling with replacement from population sample collection and obtained, each probability boosted tree
In training, used feature set obtains from the feature set progress random sampling of totality.In test phase, random chance carries
Its positioning probable value can be obtained to test sample by rising each probability boosted tree in forest, and each probability boosted tree is asked
It is average to can obtain final positioning probable value of the random chance lifting forest to test sample.
In the present embodiment, the tree constructions such as common probabilistic tree, binary tree or decision tree instead of by probability boosted tree, lead to
Cross setting probability interval sampled point is divided in left and right child node, avoid due to the limited caused over-fitting of training sample
Phenomenon, and because systematic error causes the situation of sampled point misclassification, improve the robustness and sampled point of disaggregated model
Nicety of grading.
Wherein, target location refers to the area to be targeted in bit image undetermined, and in the present embodiment, the feature of image is believed
In each probability boosted tree of breath input random chance lifting forest, the output information of random chance lifting forest is to be positioned
Each pixel is the probable value of location information in image.
Optionally, the probability that each pixel belongs to target location is calculated, including:
The probability of each pixel each probability lifting root vertex in random chance lifts forest is calculated, is defined as institute
The probability that each pixel belongs to target location in each probability lifting root vertex during random chance lifts forest is stated, wherein, institute
Each pixel is stated in the probability of each probability lifting root vertex equal to the probability of the subtree root node of the probability boosted tree and its
The weight sum of products of root node is corresponded to, thus leaf node of the recursive calculation to probability boosted tree;The random chance is carried
The average value for rising the probability of each probability lifting root vertex of forest kind is defined as the probability of each pixel.
In the present embodiment, the child node in random chance lifting forest in each probability boosted tree is divided into node of divergence and leaf
Node.Include grader in node of divergence, exemplary, the grader can be AdaBoost graders.Pass through institute
The destination probability of sample point can be obtained and be divided to according to destination probability value in child node by stating Adaboost graders.
In leaf node, not comprising grader, priori destination probability during this node is only reached comprising sample point.Exemplary,
In two classification tasks, if a probability boosted tree in random chance lifting forest only includes a root node and two
Leaf node, an AdaBoost grader, positive label of the grader in the leaf node of left and right are included in root node
Probability is respectively 0.2 and 0.8.Assuming that what the test sample of a Unknown Label was obtained by the AdaBoost graders of root node
Positive label probability value is 0.55, then according to the algorithm of probability boosted tree, it is 0.55 that it, which is divided into the probability of right subtree, is divided into a left side
The probability of subtree is 0.45, and the two probability sum is 1.Its final positive label probability value is 0.45*0.2+0.55*0.8=
0.53.The positive label is the pixel for belonging to target location, and the negative label is the pixel for being not belonging to target location.At this
In example, the probability of the root node of probability boosted tree determines according to the probability step-by-step calculation of all leaf nodes in tree, improves
The accuracy of classification.
In the present embodiment, by same bit image input random chance lifting forest undetermined, random chance is lifted each in forest
Each pixel belongs to the probability of target location in the exportable positioning image of probability boosted tree, by the probability of each corresponding pixel
Value average is defined as final output result.The testing result of more probability boosted trees, keeps away in comprehensive random chance lifting forest
Exempt from single probability boosted tree and error in classification or calculation error be present, improve the robust of probability lifting tree-model
Property, improve image position accuracy.
S130, the positioning result that the positional information for the pixel for meeting predetermined probabilities condition is defined as to target location.
Wherein, predetermined probabilities condition refers to the probable value condition for screening target location positioning result.Optionally, in advance
If Probability Condition is most probable value or predetermined probabilities value scope.Exemplary, forest generation is lifted by random chance and treated
Each pixel belongs to the probable value of target location in positioning image, and the pixel of most probable value is determined to belong into target location
Pixel, record the coordinate position of the pixel.Exemplary, will if pixel quantity corresponding to target location is more than 1
Meet that multiple pixels are determined to belong to target location corresponding to the probable value of predetermined probabilities value scope, record the seat of each pixel
Cursor position.
Optionally, medical image includes:Angiographic image or heart contrastographic picture;
Accordingly, target location includes vessel seed point or aorta petal.
Wherein, angiographic image and heart contrastographic picture can be by that will contain organic compound under roentgen radiation x thoroughly
Bright contrast agent is rapidly injected blood flow, Heart and great blood vessel chamber is developed under roentgen radiation x, while the image of quick obtaining.Blood
Pipe seed point or aorta petal are typically the detection target of angiographic image or heart contrastographic picture.With heart in the present embodiment
It is described exemplified by contrastographic picture, it is exemplary, in the clinical detection of heart or clinical operation, it is thus necessary to determine that sustainer
The position of valve and scope, it is necessary first to aorta petal is positioned, and according to positioning result in heart contrastographic picture
The operation such as image segmentation and registration is carried out to aorta petal.High-precision framing is the follow-up premise precisely operated,
By improving the precision and robustness of framing in the present embodiment, the degree of accuracy of clinical medicine operation is improved, reduces medical science
Error.
The technical scheme of the present embodiment, forest is lifted by the way that the characteristic information of bit image undetermined is inputted into random chance, really
Fixed each pixel belongs to the probability of target location, according to predetermined probabilities condition, is defined as the positioning result of target location, by with
Machine probability lifting forest instead of traditional classification tree, while avoid the situation that single classification tree there may be error, solve
The problem of framing accuracy is low in the prior art, poor robustness, improve the precision and robustness of framing.
On the basis of above-mentioned technical proposal, optionally, also include before step S120:
It is determined that meet the pixel of intensity value ranges corresponding to target location.
In the present embodiment, before by bit image undetermined input random chance lifting forest, according to the pixel of each pixel ash
Angle value excludes the pixel region in the range of target location.Exemplary, the ash in the region of aorta petal in heart contrastographic picture
Angle value is distributed in 50-800, then screening is in the pixel region of the tonal range, as localization region.
In the present embodiment, positioning units are screened by the grey scale pixel value scope of target location, reduced with the picture calculated
Vegetarian refreshments quantity, the interference of inessential pixel is avoided, accelerate framing speed, improve framing efficiency.
Embodiment two
Fig. 2 is a kind of flow chart for image position method that the embodiment of the present invention two provides, on the basis of above-described embodiment
On, further image position method is optimized,
S210, at least two training sample sets are obtained, random chance lifting forest classified device is trained according to training sample set.
In the present embodiment, before being positioned to image, training random chance lifting forest.In the present embodiment, at random
The training process of each probability boosted tree is separate in probability lifting forest, and in random chance lifting forest each it is general
Rate boosted tree corresponds to independent training sample set, is advantageous to improve the independence of each probability boosted tree, improves random chance
Lift the degree of accuracy of the synthesis positioning result of forest.The training sample set for training each probability boosted tree is carried out from total sample
Sampling with replacement obtains, and may have part sample between the training sample set of each probability boosted tree in random chance lifting forest
Point is overlapping, and the repetitive rate of the training sample set of each probability boosted tree is accounted for the ratio totally sampled by sampling samples and determined.
S220, by training sample according to random chance lifted forest classified device classified, by classification results with training sample
The training error for determining random chance lifting forest is compared in the goldstandard of this collection.
Wherein, the goldstandard of training sample is the true tag value of each pixel in training sample, in two classification problems,
The corresponding width bianry image of goldstandard, i.e., foreground and background, prospect are that 1 background is 0.
In the present embodiment, the target location in classification results is compared with known target position in training sample, really
Fixed correct and classification error pixel quantity of classifying, the ratio of the pixel quantity of classification error and pixel total quantity is true
It is set to the error in classification of the training sample.
Optionally, the sampling put back to by having concentrates selection training sample in training sample.
If the error in classification of S230, training sample is more than default error threshold, in the position area where misclassification sample
Domain increases sample collection quantity, more new training sample set.
Wherein, default error threshold can be determined according to history classification results, if the error in classification of training sample is more than
Default error threshold, it is determined that the nicety of grading of current random chance lifting forest classified device is low, need to further optimize training.This
In embodiment, training is optimized by increasing corresponding training sample.Optionally, default error threshold is more than to the error in classification
The training sample of value carries out region division, determines the larger region of error in classification, and increase includes the training of the region corresponding content
Sample, more new training sample set.
In the present embodiment, according to the purposive increase training sample of error in classification, the specific aim of optimization training is improved, is added
The training process of speed random chance lifting forest classified device, improves training effectiveness.
S240, random chance lifting forest classified device is trained according to the training sample set after renewal.
Repetitive exercise is optimized to random chance lifting forest classified device according to the training sample of renewal, improved random general
Rate lifts the nicety of grading of forest classified device, improves the stationkeeping ability of random chance lifting forest.
S250, medical image is obtained, and determine the characteristic information of each pixel of medical image.
S260, according to characteristic information and random chance lifting forest classified device each pixel is classified, and calculated each
Pixel belongs to the probability of target location.
S270, the positioning result that the positional information for the pixel for meeting predetermined probabilities condition is defined as to target location.
On the basis of above-described embodiment, after step S240, and include before step S250:
The training sample of other probability boosted trees in forest in addition to current probability boosted tree is lifted according to random chance
Collection test current probability lifting Tree Classifier, obtains test result, the training of current probability boosted tree is updated according to test result
As a result, strengthening the robustness of model.
If such as the grader of current probability boosted tree to the test result of the training sample set of other probability boosted trees with
The test result of current probability boosted tree hair training sample set differs by more than predetermined threshold value, using the training of other probability boosted trees
The test result that sample set obtains.
In the present embodiment, after random chance lifting forest classified device is trained successfully, lifted according to random chance in forest
The grader of the training sample set pair current probability boosted tree of other probability boosted trees in addition to current probability boosted tree is carried out
Test.In test process, test result is updated to the training result of current probability boosted tree, avoids in training process
Existing over-fitting, causes error in classification very little in training process, or even error in classification is zero, but actual location precision is low
Situation, improve the positioning precision of probability boosted tree.
It should be noted that perform image position method by S210-S270 in the present embodiment, simply one it is preferable real
Example is applied, performs image position method in other embodiments or by S210, S250-S270.
The technical scheme of the present embodiment, Tree Classifier is individually lifted to different probability by different training sample sets and entered
Row training, and the misclassification situation in training process is detected, corresponding training sample is increased according to misclassification, after renewal
Training sample set pair probability lifting Tree Classifier carries out further optimization training, improves the positioning of random chance lifting forest
Precision.
Embodiment three
Fig. 3 is a kind of structural representation for image positioning device that the embodiment of the present invention three provides, and the device specifically includes:
Image collection module 310, for obtaining medical image, and determine the characteristic information of each pixel of medical image;
Probability determination module 320, for being clicked through according to characteristic information and random chance lifting forest classified device to each pixel
Row classification, and calculate the probability that each pixel belongs to target location;
Positioning result determining module 330, for the positional information for meeting the pixel of predetermined probabilities condition to be defined as into mesh
The positioning result of cursor position.
Further, device also includes:
Pixel determining module, for carrying out classifying it to each pixel according to random chance lifting forest classified device
Before, it is determined that meeting the pixel of intensity value ranges corresponding to target location.
Further, probability determination module 320 is specifically used for:
The probability of each pixel each probability lifting root vertex in random chance lifts forest is calculated, is defined as institute
The probability that each pixel belongs to target location in each probability lifting root vertex during random chance lifts forest is stated, wherein, institute
State the son that each pixel probability of each probability lifting root vertex in random chance lifts forest is equal to the probability boosted tree
The weight sum of products of the probability of root vertex and its corresponding root node;
The average value that random chance is lifted to the probability of each probability lifting root vertex in forest is defined as each pixel
Probability.
Further, device also includes:
Classifier training module, for before medical image is obtained, at least two training sample sets being obtained, according to training
Sample set training random chance lifting forest classified device.
Further, device also includes:
Error in classification determining module, carried out for the training sample to be lifted into forest classified device according to the random chance
Classification, classification results are compared with the goldstandard of the training sample set, determine the training error of the probability boosted tree;
Sample set update module, if the training error for random chance lifting forest is more than default error threshold,
Then increase sample collection quantity in the band of position where misclassification sample, update the training sample set;
Grader circuit training module, according to the training sample set training random chance lifting forest classified device after renewal.
Further, device also includes:
Grader test module, for according to training sample set train random chance lifting forest classified device after, root
The training sample set test that other probability boosted trees in forest in addition to current probability boosted tree are lifted according to random chance is current
The grader of probability boosted tree;
Training result update module, for test result to be updated to the training result of current probability boosted tree.
Further, medical image includes:Angiographic image or heart contrastographic picture;
Accordingly, target location includes vessel seed point or aorta petal.
Image positioning device provided in an embodiment of the present invention can perform the framing that any embodiment of the present invention is provided
Method, possess and perform the corresponding functional module of image position method and beneficial effect.
Example IV
Fig. 4 is a kind of structural representation for computer that the embodiment of the present invention four provides, and Fig. 4 is shown suitable for being used for realizing
The block diagram of the exemplary medical imaging device of embodiment of the present invention, the computer that Fig. 4 is shown are only an example, should not be right
The function and use range of the embodiment of the present invention bring any restrictions.
Computer 400 can be used to realize the ad hoc approach and device for implementing to disclose in some embodiments of the invention.This
Specific device in embodiment illustrates a hardware platform for including display module using functional block diagram.In some embodiments
In, computer 400 can realize some realities of the invention by its hardware device, software program, firmware and combinations thereof
Apply the specific implementation of example.In certain embodiments, computer 400 can be the computer of a general purpose, or one has spy
Determine the computer of purpose.
As shown in figure 4, computer 400 can include internal communication bus 401, processor (processor) 402 is read-only
Memory (ROM) 403, random access memory (RAM) 404, COM1 405, input output assembly 406, hard disk 407, with
And user interface 408.Internal communication bus 401 can realize the data communication of the inter-module of computer 400.Processor 402 can be with
Judged and send prompting.In certain embodiments, processor 402 can be made up of one or more processors.Communication ends
Mouth 405 can realize computer 400 with miscellaneous part (not shown) for example:External equipment, image capture device, data
Enter row data communication between storehouse, external storage and image processing workstations etc..In certain embodiments, computer 400 can be with
Sent by COM1 405 from network and receive information and data.Input output assembly 406 supports computer 400 and other
Input/output data stream between part.User interface 408 can realize the interaction and information between computer 400 and user
Exchange.Computer 400 can also include various forms of program storage units and data storage element, such as hard disk 407, only
Read memory (ROM) 403, random access memory (RAM) 404, can store computer disposal and/or communication use it is various
Data file, and the possible programmed instruction performed by processor 402.
It can be used for performing a kind of image position method during the computing device program, methods described includes:
Medical image is obtained, and determines the characteristic information of each pixel in the medical image;
Each pixel is classified according to the characteristic information and random chance lifting forest classified device, and calculated
Each pixel belongs to target location probability;
The positional information for the pixel for meeting predetermined probabilities condition is defined as to the positioning result of the target location.
Although the present invention is disclosed as above with preferred embodiment, it is not for limiting the present invention, any this area
Technical staff without departing from the spirit and scope of the present invention, may be by the methods and technical content of the disclosure above to this hair
Bright technical scheme makes possible variation and modification, therefore, every content without departing from technical solution of the present invention, according to the present invention
Any simple modifications, equivalents, and modifications made to above example of technical spirit, belong to technical solution of the present invention
Protection domain.
Meanwhile the application has used particular words to describe embodiments herein.Such as " one embodiment ", " one implements
Example ", and/or " some embodiments " mean a certain feature, structure or the feature related at least one embodiment of the application.Cause
This, it should be highlighted that and it is noted that " embodiment " or " implementation that are referred to twice or repeatedly in diverse location in this specification
Example " or " alternate embodiment " are not necessarily meant to refer to the same embodiment.In addition, in one or more embodiments of the application
Some features, structure or feature can carry out appropriate combination.
In addition, it will be understood by those skilled in the art that each side of the application can be by some with patentability
Species or situation are illustrated and described, including any new and useful process, the combination of machine, product or material, or right
Their any new and useful improvement.Correspondingly, the various aspects of the application can be performed completely by hardware, can be complete
Performed, can also be performed by combination of hardware by software (including firmware, resident software, microcode etc.).Hardware above is soft
Part is referred to alternatively as " data block ", " module ", " submodule ", " engine ", " unit ", " subelement ", " component " or " system ".This
Outside, each side of the application may show as the computer product being located in one or more computer-readable mediums, the product
Encoded including computer-readable program.
Embodiment five
The embodiment of the present invention five provides a kind of computer-readable recording medium, is stored thereon with computer program, the journey
The image position method provided such as all inventive embodiments of the application is realized when sequence is executed by processor, this method includes:
Medical image is obtained, and determines the characteristic information of each pixel in the medical image;
Each pixel is classified according to the characteristic information and random chance lifting forest classified device, and calculated
Each pixel belongs to target location probability;
The positional information for the pixel for meeting predetermined probabilities condition is defined as to the positioning result of the target location.
Computer-readable signal media may include the propagation data signal containing computer program code in one, such as
A part in base band or as carrier wave.The transmitting signal may have many forms, including electromagnetic form, light form etc.
Deng or suitable combining form.Computer-readable signal media can be any meter in addition to computer-readable recording medium
Calculation machine computer-readable recording medium, the medium can by be connected to an instruction execution system, device or equipment with realize communication, propagate or
Transmit the program for using.Program coding in computer-readable signal media can be carried out by any suitable medium
Propagate, include the combination of radio, cable, fiber optic cables, radiofrequency signal or similar mediums or any of above medium.
Computer program code needed for the operation of the application each several part can use any one or more programming language,
Including Object-Oriented Programming Language such as Java, Scala, Smalltalk, Eiffel, JADE, Emerald, C++, C#, VB.NET,
Python etc., conventional procedural programming language for example C language, Visual Basic, Fortran 2003, Perl, COBOL 2002,
PHP, ABAP, dynamic programming language such as Python, Ruby and Groovy, or other programming languages etc..The program coding can be with complete
Entirely on the user computer run run on the user computer as independent software kit or partly in subscriber computer
Run in remote computer or run completely on remote computer or server in upper operation part.In the latter cases, remotely
Computer can be connected by any latticed form with subscriber computer, such as LAN (LAN) or wide area network (WAN), or even
Outer computer (such as passing through internet) is connected to, or is serviced using such as software in cloud computing environment, or as service
(SaaS)。
In addition, except clearly stating in non-claimed, the order of herein described processing element and sequence, digital alphabet
Using or other titles use, be not intended to limit the order of the application flow and method.Although by each in above-mentioned disclosure
Kind of example discusses some it is now recognized that useful inventive embodiments, but it is to be understood that, such details only plays explanation
Purpose, appended claims are not limited in the embodiment disclosed, on the contrary, claim is intended to cover and all meets the application
The amendment of embodiment spirit and scope and equivalent combinations.For example, although system component described above can be set by hardware
It is standby to realize, but only can also be achieved by the solution of software, such as pacify on existing server or mobile device
The described system of dress.
Similarly, it is noted that real to one or more invention so as to help in order to simplify herein disclosed statement
Apply the understanding of example, above in the description of the embodiment of the present application, sometimes by various features merger to one embodiment, accompanying drawing or
In descriptions thereof.But this disclosure method is not meant to carry in the aspect ratio claim required for the application object
And feature it is more.In fact, the feature of embodiment will be less than whole features of the single embodiment of above-mentioned disclosure.
Description composition, the numeral of number of attributes are used in some embodiments, it should be appreciated that such to be used for embodiment
The numeral of description, qualifier " about ", " approximation " or " generally " has been used to modify in some instances.Unless say in addition
Bright, " about ", " approximation " or the numeral that " generally " shows allow the change for having ± 20%.Correspondingly, in some embodiments
In, the numerical parameter used in description and claims is approximation, approximation feature according to needed for separate embodiment
It can change.In certain embodiments, numerical parameter is considered as defined significant digit and retained using general digit
Method.Although it is used to confirm that the Numerical Range of its scope range and parameter are approximation in some embodiments of the application, specific real
Apply in example, being set in for such numerical value is reported as precisely as possible in feasible region.
Each patent, patent application, patent application publication thing and the other materials quoted for the application, such as article, book
Entire contents, are incorporated herein as reference by nationality, specification, publication, document etc. hereby.It is inconsistent with teachings herein
Or except producing the application history file of conflict, to the conditional file of the application claim widest scope (currently or afterwards
Be additional in the application) also except.If it should be noted that description, definition, and/or art in the application attaching material
The use of language with it is herein described it is interior have place that is inconsistent or conflicting, with making for the description of the present application, definition and/or term
With being defined.
Finally, it will be understood that embodiment described herein is only illustrating the principle of the embodiment of the present application.Other
Deformation may also belong to scope of the present application.Therefore, unrestricted as example, the alternative configuration of the embodiment of the present application is visual
To be consistent with teachings of the present application.Correspondingly, embodiments herein is not limited only to the implementation that the application is clearly introduced and described
Example.
Pay attention to, above are only presently preferred embodiments of the present invention and institute's application technology principle.It will be appreciated by those skilled in the art that
The invention is not restricted to specific embodiment described here, can carry out for a person skilled in the art various obvious changes,
Readjust and substitute without departing from protection scope of the present invention.Therefore, although being carried out by above example to the present invention
It is described in further detail, but the present invention is not limited only to above example, without departing from the inventive concept, also
Other more equivalent embodiments can be included, and the scope of the present invention is determined by scope of the appended claims.
Claims (10)
- A kind of 1. image position method, it is characterised in that including:Medical image is obtained, and determines the characteristic information of each pixel in the medical image;Each pixel is classified according to the characteristic information and random chance lifting forest classified device, and described in calculating Each pixel belongs to the probability of target location;The positional information for the pixel for meeting predetermined probabilities condition is defined as to the positioning result of the target location.
- 2. according to the method for claim 1, it is characterised in that forest classified device is being lifted to described each according to random chance Before pixel is classified, in addition to:It is determined that meet the pixel of intensity value ranges corresponding to target location.
- 3. according to the method for claim 1, it is characterised in that the probability that each pixel belongs to target location is calculated, Including:The probability of the root node of each probability boosted tree of each pixel in random chance lifts forest is calculated, it is determined that Belong to the probability of target location in the probability lifts root vertex for each pixel, wherein, each pixel exists The probability of each probability lifting root vertex is equal to the probability and its corresponding root node of the subtree root node of the probability boosted tree The weight sum of products;The average value that the random chance is lifted to the probability of each probability lifting root vertex in forest is defined as each pixel Point belongs to the probability of target location.
- 4. according to the method for claim 1, it is characterised in that before medical image is obtained, in addition to:At least two training sample sets are obtained, random chance lifting forest classified device is trained according to the training sample set.
- 5. according to the method for claim 4, it is characterised in that train random chance lifting gloomy according to the training sample set Woods grader, including:The training sample is lifted into forest classified device according to the random chance to be classified, by classification results and the training The goldstandard of sample set is compared, and determines the training error of the random chance lifting forest;If the training error of the random chance lifting forest is more than default error threshold, the position where misclassification sample Region increases sample collection quantity, updates the training sample set;The random chance lifting forest classified device is trained according to the training sample set after renewal.
- 6. according to the method for claim 4, it is characterised in that random chance lifting is being trained according to the training sample set After forest classified device, in addition to:The training sample set that other probability boosted trees in forest in addition to current probability boosted tree are lifted according to random chance is surveyed The current probability lifting Tree Classifier is tried, obtains test result;The training result of the current probability boosted tree is updated according to the test result.
- 7. according to any described methods of claim 1-6, it is characterised in that the medical image includes:Angiographic image Or heart contrastographic picture;Accordingly, the target location includes vessel seed point or aorta petal.
- A kind of 8. image positioning device, it is characterised in that including:Image collection module, for obtaining medical image, and determine the characteristic information of each pixel of the medical image;Probability determination module, each pixel is clicked through for lifting forest classified device according to the characteristic information and random chance Row classification, and calculate the probability that each pixel belongs to target location;Positioning result determining module, for the positional information for meeting the pixel of predetermined probabilities condition to be defined as into the target position The positioning result put.
- 9. a kind of computer, it is characterised in that can be transported on a memory and on a processor including memory, processor and storage Capable computer program, it is characterised in that can be used for perform claim to require any institutes of 1-7 during the computing device described program The method stated.
- 10. a kind of computer-readable recording medium, is stored thereon with computer program, it is characterised in that the program is by processor During execution any described methods of 1-7 are required available for perform claim.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710757889.1A CN107590510A (en) | 2017-08-29 | 2017-08-29 | A kind of image position method, device, computer and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710757889.1A CN107590510A (en) | 2017-08-29 | 2017-08-29 | A kind of image position method, device, computer and storage medium |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107590510A true CN107590510A (en) | 2018-01-16 |
Family
ID=61050728
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710757889.1A Pending CN107590510A (en) | 2017-08-29 | 2017-08-29 | A kind of image position method, device, computer and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107590510A (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108717538A (en) * | 2018-06-08 | 2018-10-30 | 上海联影医疗科技有限公司 | A kind of tubular structure extracting method and device |
CN109670484A (en) * | 2019-01-16 | 2019-04-23 | 电子科技大学 | A kind of mobile phone individual discrimination method based on bispectrum feature and boosted tree |
CN109816721A (en) * | 2018-12-29 | 2019-05-28 | 上海联影智能医疗科技有限公司 | Image position method, device, equipment and storage medium |
CN110348541A (en) * | 2019-05-10 | 2019-10-18 | 腾讯医疗健康(深圳)有限公司 | Optical fundus blood vessel image classification method, device, equipment and storage medium |
CN110706794A (en) * | 2019-09-26 | 2020-01-17 | 中国科学院深圳先进技术研究院 | Medical image processing system and medical image processing method |
CN111178429A (en) * | 2019-11-25 | 2020-05-19 | 上海联影智能医疗科技有限公司 | System and method for providing medical guidance using patient depth images |
US11367181B2 (en) | 2018-12-29 | 2022-06-21 | Shanghai United Imaging Intelligence Co., Ltd. | Systems and methods for ossification center detection and bone age assessment |
WO2022237746A1 (en) * | 2021-05-11 | 2022-11-17 | 上海宝藤生物医药科技股份有限公司 | Method and apparatus for determining concentration of low-density lipoprotein reagent, and storage medium |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20080240514A1 (en) * | 2007-03-26 | 2008-10-02 | The Hong Kong Polytechnic University | Method of personal recognition using hand-shape and texture |
CN102592133A (en) * | 2010-09-22 | 2012-07-18 | 西门子公司 | Method and system for evaluation using probabilistic boosting trees |
CN105389589A (en) * | 2015-11-06 | 2016-03-09 | 北京航空航天大学 | Random-forest-regression-based rib detection method of chest X-ray film |
-
2017
- 2017-08-29 CN CN201710757889.1A patent/CN107590510A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20080240514A1 (en) * | 2007-03-26 | 2008-10-02 | The Hong Kong Polytechnic University | Method of personal recognition using hand-shape and texture |
CN102592133A (en) * | 2010-09-22 | 2012-07-18 | 西门子公司 | Method and system for evaluation using probabilistic boosting trees |
CN105389589A (en) * | 2015-11-06 | 2016-03-09 | 北京航空航天大学 | Random-forest-regression-based rib detection method of chest X-ray film |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108717538B (en) * | 2018-06-08 | 2021-06-29 | 上海联影医疗科技股份有限公司 | Tubular structure extraction method and device |
CN108717538A (en) * | 2018-06-08 | 2018-10-30 | 上海联影医疗科技有限公司 | A kind of tubular structure extracting method and device |
US11735322B2 (en) | 2018-12-29 | 2023-08-22 | Shanghai United Imaging Intelligence Co., Ltd. | Systems and methods for ossification center detection and bone age assessment |
CN109816721A (en) * | 2018-12-29 | 2019-05-28 | 上海联影智能医疗科技有限公司 | Image position method, device, equipment and storage medium |
US11367181B2 (en) | 2018-12-29 | 2022-06-21 | Shanghai United Imaging Intelligence Co., Ltd. | Systems and methods for ossification center detection and bone age assessment |
CN109816721B (en) * | 2018-12-29 | 2021-07-16 | 上海联影智能医疗科技有限公司 | Image positioning method, device, equipment and storage medium |
CN109670484B (en) * | 2019-01-16 | 2022-03-25 | 电子科技大学 | Mobile phone individual identification method based on bispectrum characteristics and lifting tree |
CN109670484A (en) * | 2019-01-16 | 2019-04-23 | 电子科技大学 | A kind of mobile phone individual discrimination method based on bispectrum feature and boosted tree |
CN110348541B (en) * | 2019-05-10 | 2021-12-10 | 腾讯医疗健康(深圳)有限公司 | Method, device and equipment for classifying fundus blood vessel images and storage medium |
CN110348541A (en) * | 2019-05-10 | 2019-10-18 | 腾讯医疗健康(深圳)有限公司 | Optical fundus blood vessel image classification method, device, equipment and storage medium |
CN110706794A (en) * | 2019-09-26 | 2020-01-17 | 中国科学院深圳先进技术研究院 | Medical image processing system and medical image processing method |
CN111178429A (en) * | 2019-11-25 | 2020-05-19 | 上海联影智能医疗科技有限公司 | System and method for providing medical guidance using patient depth images |
WO2022237746A1 (en) * | 2021-05-11 | 2022-11-17 | 上海宝藤生物医药科技股份有限公司 | Method and apparatus for determining concentration of low-density lipoprotein reagent, and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107590510A (en) | A kind of image position method, device, computer and storage medium | |
CN110047056A (en) | With the cross-domain image analysis and synthesis of depth image to image network and confrontation network | |
US10489907B2 (en) | Artifact identification and/or correction for medical imaging | |
US8938113B2 (en) | Adaptive visualization for direct physician use | |
KR102219890B1 (en) | Apparatus for spatial normalization of medical image using deep learning and method thereof | |
CN113256592B (en) | Training method, system and device of image feature extraction model | |
CN112396606B (en) | Medical image segmentation method, system and device based on user interaction | |
US11933870B2 (en) | Contrast and/or system independent motion detection for magnetic resonance imaging | |
Singhal et al. | Study of deep learning techniques for medical image analysis: A review | |
CN114170212A (en) | False positive detection method and system based on small lung nodule in CT image | |
CN110599444B (en) | Device, system and non-transitory readable storage medium for predicting fractional flow reserve of a vessel tree | |
CN111568451A (en) | Exposure dose adjusting method and system | |
CN108597589B (en) | Model generation method, target detection method and medical imaging system | |
Tiago et al. | A domain translation framework with an adversarial denoising diffusion model to generate synthetic datasets of echocardiography images | |
CN113837179A (en) | Multi-discriminant GAN network construction method, device and system for processing images and storage medium | |
US11408954B2 (en) | Systems and methods of reducing noise and artifacts in magnetic resonance imaging | |
WO2021184195A1 (en) | Medical image reconstruction method, and medical image reconstruction network training method and apparatus | |
US20230377195A1 (en) | Reinforcement learning to perform localization, segmentation, and classification on biomedical images | |
CN115719352A (en) | Improved UNet + + based feature alignment tongue image segmentation method and system | |
Bhattacharjya et al. | A genetic algorithm for intelligent imaging from quantum-limited data | |
Mikhailov et al. | A deep learning-based interactive medical image segmentation framework | |
CN113628230A (en) | Ventricular myocardium segmentation model training method, segmentation method and device in cardiac nuclear magnetic resonance image | |
CN111180046B (en) | Determining a processing sequence for processing an image | |
KR20180006120A (en) | Optimized Image Segmentation Methods and System with DL and PDE | |
Eriksen | A machine learning approach to improve consistency in user-driven medical image analysis |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
CB02 | Change of applicant information |
Address after: 201807 Shanghai City, north of the city of Jiading District Road No. 2258 Applicant after: Shanghai Lianying Medical Technology Co., Ltd Address before: 201807 Shanghai City, north of the city of Jiading District Road No. 2258 Applicant before: SHANGHAI UNITED IMAGING HEALTHCARE Co.,Ltd. |
|
CB02 | Change of applicant information | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180116 |
|
RJ01 | Rejection of invention patent application after publication |