CN106504253A - A kind of processing method of medical imaging photo and system - Google Patents
A kind of processing method of medical imaging photo and system Download PDFInfo
- Publication number
- CN106504253A CN106504253A CN201610899910.7A CN201610899910A CN106504253A CN 106504253 A CN106504253 A CN 106504253A CN 201610899910 A CN201610899910 A CN 201610899910A CN 106504253 A CN106504253 A CN 106504253A
- Authority
- CN
- China
- Prior art keywords
- image
- module
- denoising
- mri
- mri image
- 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
- 238000002059 diagnostic imaging Methods 0.000 title claims abstract description 20
- 238000003672 processing method Methods 0.000 title claims description 6
- 238000000034 method Methods 0.000 claims abstract description 57
- 238000003709 image segmentation Methods 0.000 claims abstract description 22
- 230000011218 segmentation Effects 0.000 claims abstract description 18
- 238000012545 processing Methods 0.000 claims abstract description 17
- 238000009877 rendering Methods 0.000 claims abstract description 17
- 230000009466 transformation Effects 0.000 claims abstract description 10
- 239000000284 extract Substances 0.000 claims abstract description 7
- 230000000694 effects Effects 0.000 claims description 6
- 238000005192 partition Methods 0.000 claims 1
- 230000000875 corresponding effect Effects 0.000 description 8
- 230000008569 process Effects 0.000 description 8
- 238000003745 diagnosis Methods 0.000 description 7
- 238000005516 engineering process Methods 0.000 description 6
- 238000004458 analytical method Methods 0.000 description 5
- 210000004556 brain Anatomy 0.000 description 5
- 238000003384 imaging method Methods 0.000 description 5
- 238000005457 optimization Methods 0.000 description 5
- 238000013507 mapping Methods 0.000 description 4
- 238000003325 tomography Methods 0.000 description 4
- 230000002596 correlated effect Effects 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 3
- 201000010099 disease Diseases 0.000 description 3
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 230000007971 neurological deficit Effects 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 230000004044 response Effects 0.000 description 3
- 238000012800 visualization Methods 0.000 description 3
- 208000024827 Alzheimer disease Diseases 0.000 description 2
- 241000208340 Araliaceae Species 0.000 description 2
- 238000005481 NMR spectroscopy Methods 0.000 description 2
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 2
- 235000003140 Panax quinquefolius Nutrition 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000004195 computer-aided diagnosis Methods 0.000 description 2
- 238000007405 data analysis Methods 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 235000008434 ginseng Nutrition 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000002595 magnetic resonance imaging Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 230000005311 nuclear magnetism Effects 0.000 description 2
- 239000002245 particle Substances 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- 230000002123 temporal effect Effects 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 206010003805 Autism Diseases 0.000 description 1
- 208000020706 Autistic disease Diseases 0.000 description 1
- 208000014644 Brain disease Diseases 0.000 description 1
- 240000007711 Peperomia pellucida Species 0.000 description 1
- 235000012364 Peperomia pellucida Nutrition 0.000 description 1
- 238000012952 Resampling Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 230000011157 brain segmentation Effects 0.000 description 1
- 230000002490 cerebral effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000000739 chaotic effect Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 238000000205 computational method Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000013399 early diagnosis Methods 0.000 description 1
- 230000005484 gravity Effects 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000010191 image analysis Methods 0.000 description 1
- 230000003902 lesion Effects 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 208000018389 neoplasm of cerebral hemisphere Diseases 0.000 description 1
- 210000000056 organ Anatomy 0.000 description 1
- 231100000915 pathological change Toxicity 0.000 description 1
- 230000036285 pathological change Effects 0.000 description 1
- 238000012805 post-processing Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 210000001525 retina Anatomy 0.000 description 1
- 241000894007 species Species 0.000 description 1
- 238000003786 synthesis reaction Methods 0.000 description 1
- 230000002194 synthesizing effect Effects 0.000 description 1
- 238000000844 transformation Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G06T5/70—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2200/00—Indexing scheme for image data processing or generation, in general
- G06T2200/32—Indexing scheme for image data processing or generation, in general involving image mosaicing
-
- 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30016—Brain
Abstract
The present invention proposes a kind of processing system of medical imaging photo, and the processing system of the medical imaging photo is included such as lower module, image segmentation module, image denoising module, image study module, three-dimensional reconstruction module.Image segmentation module, split for MRI image, by different region segmentations in image out, these regions are mutually disjoint to the gray scale, color or geometric propertieses according to image, each region meets its respective feature, and extracts target area wherein interested；Image denoising module, by the way of the regularization of total variation TV, by sparse transformation, and is fitted to MRI image data, realizes denoising；Image study module：Using Bayesian learning models, to denoising after MRI image be modeled；Three-dimensional reconstruction module, is modeled to MRI image using surface renderings method or overall replay method.
Description
Technical field
The present invention relates to image processing field, more particularly to a kind of processing method of medical imaging photo and system.
Background technology
In recent years, computer technology, communication technology, network technology are developed rapidly to computer-aided diagnosises
(ComputerAidedDiagnosis, CAD) brings new life, and CAD make use of computer to medical image analysis process
High efficiency, save substantial amounts of human and material resources, and can not be affected to make diagnosing for objective reality by people's subjective consciousness
As a result, accurately judge so as to coordinate doctor to make the state of an illness.In Medical Image Processing, medical image segmentation is to solve medical science
The precondition that image is clinically applied, it are complicated and challenging tasks, and particularly the mankind are to brain image
Segmentation, brain segmentation has particularly important meaning for diseases such as the early diagnosiss cerebral tumor, Alzheimer's diseases.
Nuclear magnetic resonance (MagneticResonanceImaging, MRI) is a very important medical image imaging
Technology, as it can provide the detail pictures of biological tissue, imaging parameters are more, to the "dead" damage of human body, now should by MRI
Auxiliary diagnosis research for people's disease of brain has received extensive attention.Ambiguity because of border between MRI image interior tissue
With inherent uncertainty so that Fuzzy clustering techniques are more widely used in MRI image segmentation compared with other technology.Mesh
Front most widely used be Fuzzy C-Means Clustering (FuzzyC-Means, FCM) algorithm, it is containing fuzzy ginseng by iteration optimization
Several object functions carries out image segmentation to obtain pixel and belong to all kinds of degrees of membership, but the algorithm is to initial value and noise
Interference is very sensitive, relies on the selection of initial cluster center to a great extent, when initial cluster center substantial deviation global most
During excellent cluster centre, locally optimal solution is absorbed in easily, makes splitting speed and performance all be affected.
In recent years, some scholars propose to carry out using particle cluster algorithm (ParticleSwarmOptimization, PSO)
The Optimization Solution of initial cluster center, the Chaos particle swarm optimization algorithm proposed by the inspiration of chaology
(ChaosParticleSwarmOptimization, CPSO) has been applied to image segmentation, but these documents fail to propose to grain
The judgement of subgroup precocity phenomenon, and the chaotic model for utilizing mostly is Logistic mappings, but Logistic mappings are produced
Raw sequence is extremely uneven, less stable.Also scholar adopts the equally distributed grain of logic self-mapping function chaos intialization
Subgroup, and integrated precocious judgment mechanism, to improve image segmentation speed and precision.
Multiscale analysis is correctly to recognize one of important method of things and phenomenon, is now widely used for medical image point
In analysis.But existing method generally existing computational methods are simplified very much at present, image comparison is easily subject under certain condition
Degree and the impact of brightness flop, if user of service will try to achieve optimal threshold, generally require to carry out fairly large traversal calculating
And variance is calculated, when computationally intensive, efficiency can be very low.Meanwhile, in real image, due to image intensity profile itself with
And the impact of the factor such as noise jamming, method later at present can not make image segmentation obtain satisfied result, though certain
Effect of noise can be eliminated in degree, but the method amount of calculation is quite big, it is difficult to be applied to real-time system.
Due to the human brain nuclear-magnetism that nuclear magnetic resonance (Magnetic Resonance Imaging, abbreviation MRI) equipment is obtained
Image is affected by factors such as the diversity between noise, radio-frequency field inhomogeneities, brain different tissues and partial volume effects, is made
The uniformity of Adult Human Brain nuclear-magnetism image is deteriorated, and the gray-scale intensity information for therefore only relying on image is brought to the classification of accurate brain image
Very big difficulty, if correct cerebral tissue classification will be obtained it may first have to which gray scale is corrected.
In recent years, some image data processing techniques based on the multiscale space of fuzzy clustering are proposed successively, such
The process of method is typically directly to carry out fuzzy clustering method on each yardstick level of scale space images sequence, and upper one
The segmentation result of yardstick level passes through the original state frequently as next yardstick level, due to not having to introduce effectively different yardsticks
Constraint between level is optimizing the segmentation in current scale level, so segmentation result is only obtained in the yardstick level that finally splits
To optimizing, therefore, these methods are not fine for the robustness of degraded image.
Content of the invention
The purpose of the present invention is achieved through the following technical solutions.
The present invention proposes a kind of processing system of medical imaging photo, and the processing system of the medical imaging photo includes
Such as lower module, image segmentation module, image denoising module, image study module, three-dimensional reconstruction module, wherein,
Image segmentation module, for splitting to MRI image, gray scale, color according to image or geometric propertieses will
In image, out, these regions are mutually disjoint to different region segmentations, and each region meets its respective feature, and
Extract target area wherein interested；
Image denoising module, by the way of the regularization of total variation TV, by sparse transformation, and intends to MRI image data
Close, realize denoising；
Image study module：Using Bayesian learning models, to denoising after MRI image be modeled；
Three-dimensional reconstruction module, is modeled to MRI image using surface renderings method or overall reproduction method；
Described surface renderings method is to be fitted body surface to reach reconstruction effect by the splicing of geometric units；
The overall reproduction method is that voxel is projected to display plane by reading volume data to be shown.
The invention allows for a kind of processing method of medical imaging photo, comprises the steps：
Step one, MRI image is split, the gray scale, color or geometric propertieses according to image will be different in image
Region segmentation out, these regions are mutually disjoint, and each region meets its respective feature, and extract and wherein feel
The target area of interest；
Step 2, by the way of the regularization of total variation TV, by sparse transformation, and to MRI image data be fitted, realize
Denoising；
Step 3, adopt Bayesian learning models, to denoising after MRI image be modeled；
Step 4：MRI image is modeled using surface renderings method or overall reproduction method.
The processing system and method for the medical imaging photo of the application, by image segmentation, denoising, deep learning and three-dimensional
Rebuild, substantially increase the identification precision of medical image, reliable support can be provided for subsequently making accurately diagnosis.
Description of the drawings
By reading the detailed description of hereafter preferred implementation, various other advantages and benefit are common for this area
Technical staff will be clear from understanding.Accompanying drawing is only used for the purpose for illustrating preferred implementation, and is not considered as to the present invention
Restriction.And in whole accompanying drawing, it is denoted by the same reference numerals identical part.In the accompanying drawings：
Accompanying drawing 1 shows the processing system schematic diagram of the medical imaging photo according to embodiment of the present invention.
Specific embodiment
The illustrative embodiments of the disclosure are more fully described below with reference to accompanying drawings.Although this public affairs is shown in accompanying drawing
The illustrative embodiments that opens, it being understood, however, that may be realized in various forms the disclosure and the reality that should do not illustrated here
The mode of applying is limited.On the contrary, there is provided these embodiments are able to be best understood from the disclosure, and can be by this public affairs
What the scope opened was complete conveys to those skilled in the art.
According to the embodiment of the present invention, the present invention proposes a kind of processing system of medical imaging photo, the medical treatment
The processing system of image photo is included such as lower module, image segmentation module, image denoising module, image study module, Three-dimensional Gravity
Modeling block, as shown in Figure 1.
Image segmentation module, is split for MRI image, and the gray scale, color or geometric propertieses according to image will be schemed
As in, out, these regions are mutually disjoint to different region segmentations, and each region meets its respective feature, and carries
Take target area wherein interested；
Image denoising module, by the way of the regularization of total variation TV, by sparse transformation, and intends to MRI image data
Close, realize denoising；
Image study module：Using Bayesian learning models, to denoising after MRI image be modeled；
Three-dimensional reconstruction module, is modeled to MRI image using surface renderings method or overall reproduction method.
According to the embodiment of the present invention, it is also proposed that a kind of processing method of medical imaging photo, comprise the steps：
Step one, MRI image is split, the gray scale, color or geometric propertieses according to image will be different in image
Region segmentation out, these regions are mutually disjoint, and each region meets its respective feature, and extract and wherein feel
The target area of interest；
Step 2, by the way of the regularization of total variation TV, by sparse transformation, and to MRI image data be fitted, realize
Denoising；
Step 3, adopt Bayesian learning models, to denoising after MRI image be modeled；
Step 4：MRI image is modeled using surface renderings method or overall reproduction method.
Below the processing procedure of above-mentioned medical imaging photo is described in detail：
In order to preferably process MRI image, first have to split MRI image, and beat in the position of segmentation
Label, to carry out detailed comparison.Image segmentation refers to and divides the image into the part for differing from one another according to certain rule, and
The method and process of target area wherein interested can be extracted.It is will according to the gray scale of image, color or geometric propertieses
In image, out, these regions are mutually disjoint to different region segmentations, and each region meets its respective feature, point
The region for cutting out should meet three conditions simultaneously：1st, the region for splitting has uniformity and connectedness；2nd, adjacent point
Cut；3rd, the border of cut zone should ensure the positioning accurate on border regular and simultaneously
Degree.
According to the concept of mathematical set, following mathematical definition is given to image segmentation：Whole image region is represented with R,
So image segmentation can just be regarded as R is divided into n sub-regions, as R1, R2, and R3 ..., Rn meet following condition：1、R1
It is R altogether to Rn；2nd, Ri is a continuum, i=1,2,3 ..., n；3rd, the common factor of Ri and Rj is empty set, and i is not equal to j.
The region segmentation for meeting conditions above can just be truly realized target area interested is split.
The levels of precision of image segmentation is directly affected or even determines the levels of precision of subsequent treatment.The image of the application point
Cut and specifically include：Edge cuts and scope cutting, discontinuity of the edge cuts using image gray levels, by detection not
Carry out segmentation figure picture with the edge between region；The scope cutting is different with the property of background area according to target area, directly
Extract target area, rather than preferential detected edge points.
MRI image is split completing, and after the position of segmentation labels, need to remove MRI image
Make an uproar process.
MRI is imaged compared with other medical imaging procedures, with harmless, can adopt multiparameter imaging and can
The biochemical character feature of reflection organ or tissue, becomes one of important means that medical science is diagnosed when participating in the cintest.In MRI image, noise
Essentially from hardware circuit and by two aspects of imaging object, mainly thermal noise, other also physiological noise.Quickly resonate into
As low signal to noise ratio and contrast can be caused, therefore, when the temporal resolution of image is improved, picture noise can drop significantly
The quality of low magnetic resonance image (MRI) so that the border of some tissues thickens, and fine structure is difficult to distinguish, increased thin to image
Section identification and the difficulty of analysis, affect medical diagnosiss.Reduce picture noise as much as possible into one of approach of Accurate Diagnosis, because
This, to the post processing denoising of image improving picture quality.
Compressive sensing theory is pointed out：When signal is sparse or compressible in certain transform domain, it is possible to use with conversion
Conversion coefficient linear projection is low-dimensional observation vector by the noncoherent calculation matrix of matrix, while this projection maintains reconstruction letter
Information needed for number, by further solve Sparse Optimization just can from low-dimensional observation vector accurately or high probability essence
Original high dimensional signal really is rebuild.Therefore, we can realize the denoising of MRI image in the way of using compressed sensing.
In compressed sensing (CompressedSensing, CS), one critically important to study a question exactly how by signal
Carry out sparse expression.Under the theme of CS, sparse graph picture can be rebuild using considerably less scanning amount.These method profits
With inherent openness in MRI image；Image can be transformed to and cause under some domains its numerical nonzero by some sparse transformations
Coefficient is considerably less.Very famous example is the regularization of total variation TV, and its implicit hypothesis major part image is with very dilute
Thin gradient.If there is such sparse transformation, be defined as ψ, we preferably can select one most sparse under ψ
Estimation cause the fitting of its scan data consistent.Therefore, by the way of the regularization of total variation TV, by sparse transformation, and
MRI image data is fitted, so as to realize denoising.
After completing denoising, it is possible to carry out deep learning to MRI image.
MRI data is typically stored with the form of three-dimensional data sample, and each sample volume data correspond to
One species, such as patient or normal person, then have the machine learning method of supervision just can be incorporated in MRI data analysis.
But the problem that MRI data analysis is frequently run onto is the feature quantity that usable samples quantity is far smaller than extracted, so as to produce
Over-fitting problem, greatly reduces the precision of prediction of machine learning model.Then, we generally need right before learning machine is set up
Feature carries out dimensionality reduction, to mitigate over-fitting problem.Traditional method passes through Feature Mapping to a lower dimensional space, reduces original
The dimension of beginning feature, but the feature after the shortcoming of the method is to map loses the physical characteristics of primitive character.
We establish Bayesian learning models, are modeled for the data to being obtained.With traditional Bayes
Model is compared, and Bayesian model proposed by the present invention introduces a class Gaussian prior (ARD) so that solution to model has sparse
Property.The Solve problems of model of the present invention are actually four classification problems, the corresponding a quarter chess of each class
The experiment of disk lattice stimulates.Traditional many sorting techniques, such as many classification SVM, its essence is that all classes are set up with a mould between any two
Type simultaneously combines certain strategy, such as temporal voting strategy, realizes classify more.In the present invention, we set up one linearly to each class
Model, after all of model parameter is estimated, for the test sample of any one unknown generic, directly can pass through
The posterior probability of class is predicted.Specifically, we calculate test sample for posterior probability p (C1 | x) of each class, p
(C2 | x), p (C3 | x) and p (C4 | x), thenIt should be noted that due to the grader that is set up
Each class is individually modeled, therefore after parameter estimation, each class correspond to one group of sparse model ginseng
Number, then we can be according to the maximally related feature of these selected each classes of parameter determination out.
After the posterior probability for obtaining class, we can calculate the likelihood function under this posterior probability.For one two
Classification problem, it is contemplated that such a problem, if it is p that an any given sample belongs to the probability of first class,
It is 1-p that it belongs to the probability of second class.Now the probability distribution of class label obeys Bernoulli Jacob's distribution.
Using the management loading model based on ARD priori, it is 91.6% by the nicety of grading that averagely obtains of 4 foldings,
9 correlated characteristics have been picked out altogether, and we return the voxel maps corresponding to which between original Naokong, it has been found that pick out
Correlated characteristic between original Naokong in position meet retina map feature, i.e., select associated voxels are corresponding
A quarter gridiron pattern visual stimulus spatially into handstand opposed relationship.This explanation, the model set up by we can have
Select and the maximally related feature of task to effect.In contrast experiment, all voxels are used as 4 folding of support vector machine of feature
Average nicety of grading is 85.4%, less than the nicety of grading of Bayesian model.Bayesian model eliminates unrelated with classification task
Feature, only retain most correlated characteristic, thus relative support vector machine precision increase.
For same MRI image feature, plurality of target response can be obtained under many circumstances.For example clinician exists
When diagnosing to the disease related to behavioristicss, except providing final diagnostic result, also have a series of behavioristicss and comment
Point, typical such as Alzheimer disease, infantile autism etc..For diagnostic result, typically ill or do not have disease, can with class label come
Represent；For neurological deficit score, then can regard the sampling of continuous variable as.Theoretically, we can set up MRI image spy
The feature learning model with each observed responses is levied, the such as feature learning model of characteristics of image and class label, or image is special
Levy the feature learning model between neurological deficit score.For each model, they are both independently, are associated again
, because diagnostic result can be reflected in neurological deficit score to a certain extent.Each learning model is individually analyzed, though
So we can be carried out feature selection, but have ignored the complementary information between observed responses for model performance may bring
Lifting.Also one example is, current MRI decoded models, be generally all to each tested be individually chosen feature, single
Decoded model is set up solely, group analysis after obtaining statistical result, are carried out again.But it is envisioned that tested with tested between MRI special
Although levy there is difference, general character, i.e. dependency is there is also.
After the deep learning to MRI image is completed, three-dimensional reconstruction is carried out to MRI image just.The task of three-dimensional reconstruction
Seek to, by the object structures information in three-dimensional data that is included in is recovered in a series of two dimensional images, construct sense of reality stronger
3-D graphic.
The three-dimensional reconstruction of medical image generally refers to section piece set for being the collection from imaging device, after synthesizing through computer
Form 3 D graphic data, then through computer visualization process, easily show in tissue physical aspect and
Spatial relationship.The three-dimensional reconstruction result of medical image accurately can determine the size at human lesion position, geometry and
The spatial information of surrounding tissue.This is particularly critical for this faultage image data of CT, MRI, because relative to sectional slice
Image, the 3-D graphic formed after reconstruction more intuitive and accurate can be shown in human body in tissue.
The three-dimensional reconstruction of medical image mainly includes the pretreatment to input picture and the main researchs such as drafting, and which is led
It is exactly to realize that three-dimensional visualization shows to want task, and provides reliable data by subsequent analysis for diagnosing and treating, so as to
Doctor can be aided in be analyzed pathological changes body surrounding tissue, greatly improve accuracy and the science of medical diagnosiss so that
The ability of medical diagnosiss and treatment is greatly enhanced.The three-dimensional reconstruction of the application adopts two ways：Surface renderings method and
Overall replay method.Surface renderings method is to be fitted body surface to rebuild effect to reach by the splicing of geometric units；
And overall replay method is that voxel is projected to display plane by reading volume data to be shown.The base of surface renderings method
This flow process is to extract the surface information of attention object first, constructs middle geometric units by volume data, then passes through geometry list
First interpolation forms body surface, and finally with traditional computer graphics techniques, such as illumination, texture are rendered etc. carries out surface and weigh
Existing.
The surface renderings method is specifically included：Profile is extracted, is that each faultage image is processed, is extracted object
Profile information at the fault plane, is usually used the method for image segmentation to obtain.
Profile is corresponding, when there is more than one profile on a tomography, needs to carry out suitably the profile on each layer
Combination constituting significant object.Profile correspondence problem can essentially be attributed to the optimization with Prescribed Properties
Problem, but the constraint of profile correspondence problem is not strong, so its corresponding result has very big randomness.
Contours connection, is with correspondence profile surface on polygonal structure adjacent layer.Between putting on correspondence profile is determined
After corresponding relation, it is only necessary to splice surface with triangle.
Bifurcated process, is to solve the problems, such as such a：It is flat with one group when bifurcated occurs in three-dimensional body own form
Row plane goes to cut this body, can obtain the face profile that number is not waited on a different plane.If specific method is construction
Profile in dry middle critical state, exactly constructs multiple tomographies so that between original two tomographies between two tomographies
Only exist one a pair two problem, remaining is man-to-man corresponding relation.
Finally, after all corresponding relations are determined, it is possible to use series of parameters patch is fitted, formed most
Whole body surface.
Different from surface renderings method, when overall replay method is directly research and application light passes through object with voxel it
Between principle, by volume data resampling come synthetically produced 3-D view, i.e., do not produce middle geometric units, directly by body number
According to reconstruction.Institute can in this way retain more voxel details, in medical image applications, then can more reflect real human body
Structure.
The overall replay method is specifically included：
1. light projection is all projected through volume data with Ray Of Light to each pixel in image.
2. sampling needs to select some sampled points to carry out the projection for calculating light in light through the part of volume data field
Attribute.In general, volume data is all without aliging with light, thus the position of sampled point typically can all fall voxel it
Between.This is accomplished by us and enters row interpolation to calculate the value of sampled point.The methods such as trilinear interpolation can generally be used.
3. color rendering is next, need to calculating gradient to each sampled point.These Grad illustrate interior of articles
The direction of local surfaces.Then color rendering is carried out to sampled point, for example, strengthens color or reduce color, try their local table
Depending on relation between face direction and the light of light source.
4. synthesis after rendering is combined to the color value of all sampled points along sight line, determines to send this with this
The pixel value of the position of sight line.Calculate from back to front, i.e., from object farthest away from the sampled point of observer to nearest sampled point.This
Individual order ensure that the object parts being blocked do not interfere with final pixel value.
Visualization MRI image based on three-dimensional reconstruction, you can make accurately diagnosis.
The above, the only present invention preferably specific embodiment, but protection scope of the present invention is not limited thereto,
Any those familiar with the art the invention discloses technical scope in, the change or replacement that can readily occur in,
Should all be included within the scope of the present invention.Therefore, protection scope of the present invention should be with the protection model of the claim
Enclose and be defined.
Claims (6)
1. a kind of processing system of medical imaging photo, the processing system of the medical imaging photo are included such as lower module, image
Segmentation module, image denoising module, image study module, three-dimensional reconstruction module, wherein：
Image segmentation module, for splitting to MRI image, gray scale, color or geometric propertieses according to image are by image
Out, these regions are mutually disjoint to middle different region segmentation, and each region meets its respective feature, and extracts
Target area wherein interested；
Image denoising module, by the way of total variation regularization, by sparse transformation, and is fitted to MRI image data, realizes
Denoising；
Image study module：Using Bayesian learning models, to denoising after MRI image be modeled；
Three-dimensional reconstruction module, is modeled to MRI image using surface renderings method or overall replay method；
Described surface renderings method is to be fitted body surface to reach reconstruction effect by the splicing of geometric units；
The overall replay method is that voxel is projected to display plane by reading volume data to be shown.
2. the processing system of medical imaging photo as claimed in claim 1, it is characterised in that：
Described image segmentation module carries out segmentation to MRI image and specifically includes：Edge cuts and scope cutting, the edge cuts
Using the discontinuity of image gray levels, by detecting the edge between zones of different come segmentation figure picture；The scope cuts root
Different with the property of background area according to target area, extracting directly goes out target area, rather than preferential detected edge points.
3. the processing system of medical imaging photo as claimed in claim 1, it is characterised in that：
Image denoising module realizes the denoising of MRI image by the way of compressed sensing.
4. the processing system using medical imaging photo as claimed in claim 1 carries out the processing method of medical imaging photo,
Characterized in that, comprising the steps：
Step one, MRI image is split, the gray scale, color or geometric propertieses according to image is by different areas in image
Out, these regions are mutually disjoint to regional partition, and each region meets its respective feature, and extracts wherein interested
Target area；
Step 2, by the way of total variation regularization, by sparse transformation, and to MRI image data be fitted, realize at denoising
Reason；
Step 3, adopt Bayesian learning models, to denoising after MRI image be modeled；
Step 4：MRI image is modeled using surface renderings method or overall replay method.
5. the method for video identification as claimed in claim 4, it is characterised in that：
Described segmentation is carried out to MRI image specifically include：Edge cuts and scope cutting, the edge cuts utilize gradation of image
The discontinuity of level, by detecting the edge between zones of different come segmentation figure picture；Scope cutting according to target area and
The property of background area is different, and extracting directly goes out target area, rather than preferential detected edge points.
6. video recognition system as claimed in claim 5, it is characterised in that：
By the way of compressed sensing, in the step 2, realize the denoising of MRI image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610899910.7A CN106504253A (en) | 2016-10-14 | 2016-10-14 | A kind of processing method of medical imaging photo and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610899910.7A CN106504253A (en) | 2016-10-14 | 2016-10-14 | A kind of processing method of medical imaging photo and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106504253A true CN106504253A (en) | 2017-03-15 |
Family
ID=58294223
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610899910.7A Pending CN106504253A (en) | 2016-10-14 | 2016-10-14 | A kind of processing method of medical imaging photo and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106504253A (en) |
Cited By (22)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109190699A (en) * | 2018-08-29 | 2019-01-11 | 西安电子科技大学 | A kind of more disease joint measurement methods based on multi-task learning |
| CN110582159A (en) * | 2019-08-30 | 2019-12-17 | 武汉科技大学 | FPC bonding pad surface defect processing system and method |
| CN113349809A (en) * | 2020-03-05 | 2021-09-07 | 高健 | Image reconstruction method of multi-modal imaging system |
| CN113554742A (en) * | 2020-04-26 | 2021-10-26 | 上海联影医疗科技股份有限公司 | Three-dimensional image reconstruction method, device, equipment and storage medium |
| US11403069B2 (en) | 2017-07-24 | 2022-08-02 | Tesla, Inc. | Accelerated mathematical engine |
| US11409692B2 (en) | 2017-07-24 | 2022-08-09 | Tesla, Inc. | Vector computational unit |
| US11487288B2 (en) | 2017-03-23 | 2022-11-01 | Tesla, Inc. | Data synthesis for autonomous control systems |
| US11537811B2 (en) | 2018-12-04 | 2022-12-27 | Tesla, Inc. | Enhanced object detection for autonomous vehicles based on field view |
| US11561791B2 (en) | 2018-02-01 | 2023-01-24 | Tesla, Inc. | Vector computational unit receiving data elements in parallel from a last row of a computational array |
| US11562231B2 (en) | 2018-09-03 | 2023-01-24 | Tesla, Inc. | Neural networks for embedded devices |
| US11567514B2 (en) | 2019-02-11 | 2023-01-31 | Tesla, Inc. | Autonomous and user controlled vehicle summon to a target |
| US11610117B2 (en) | 2018-12-27 | 2023-03-21 | Tesla, Inc. | System and method for adapting a neural network model on a hardware platform |
| US11636333B2 (en) | 2018-07-26 | 2023-04-25 | Tesla, Inc. | Optimizing neural network structures for embedded systems |
| US11665108B2 (en) | 2018-10-25 | 2023-05-30 | Tesla, Inc. | QoS manager for system on a chip communications |
| US11681649B2 (en) | 2017-07-24 | 2023-06-20 | Tesla, Inc. | Computational array microprocessor system using non-consecutive data formatting |
| US11734562B2 (en) | 2018-06-20 | 2023-08-22 | Tesla, Inc. | Data pipeline and deep learning system for autonomous driving |
| US11748620B2 (en) | 2019-02-01 | 2023-09-05 | Tesla, Inc. | Generating ground truth for machine learning from time series elements |
| US11790664B2 (en) | 2019-02-19 | 2023-10-17 | Tesla, Inc. | Estimating object properties using visual image data |
| US11816585B2 (en) | 2018-12-03 | 2023-11-14 | Tesla, Inc. | Machine learning models operating at different frequencies for autonomous vehicles |
| US11841434B2 (en) | 2018-07-20 | 2023-12-12 | Tesla, Inc. | Annotation cross-labeling for autonomous control systems |
| US11893774B2 (en) | 2018-10-11 | 2024-02-06 | Tesla, Inc. | Systems and methods for training machine models with augmented data |
| US11893393B2 (en) | 2017-07-24 | 2024-02-06 | Tesla, Inc. | Computational array microprocessor system with hardware arbiter managing memory requests |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104834943A (en) * | 2015-05-25 | 2015-08-12 | 电子科技大学 | Brain tumor classification method based on deep learning |
| CN105957066A (en) * | 2016-04-22 | 2016-09-21 | 北京理工大学 | CT image liver segmentation method and system based on automatic context model |
-
2016
- 2016-10-14 CN CN201610899910.7A patent/CN106504253A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104834943A (en) * | 2015-05-25 | 2015-08-12 | 电子科技大学 | Brain tumor classification method based on deep learning |
| CN105957066A (en) * | 2016-04-22 | 2016-09-21 | 北京理工大学 | CT image liver segmentation method and system based on automatic context model |
Non-Patent Citations (2)
| Title |
|---|
| 王犁野: "MRI 图像分析中的稀疏特征学习方法研究", 《中国博士学位论文全文数据库-信息科技辑》 * |
| 章新友主编: "《医学图形图像处理》", 30 April 2015 * |
Cited By (26)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11487288B2 (en) | 2017-03-23 | 2022-11-01 | Tesla, Inc. | Data synthesis for autonomous control systems |
| US11893393B2 (en) | 2017-07-24 | 2024-02-06 | Tesla, Inc. | Computational array microprocessor system with hardware arbiter managing memory requests |
| US11681649B2 (en) | 2017-07-24 | 2023-06-20 | Tesla, Inc. | Computational array microprocessor system using non-consecutive data formatting |
| US11403069B2 (en) | 2017-07-24 | 2022-08-02 | Tesla, Inc. | Accelerated mathematical engine |
| US11409692B2 (en) | 2017-07-24 | 2022-08-09 | Tesla, Inc. | Vector computational unit |
| US11797304B2 (en) | 2018-02-01 | 2023-10-24 | Tesla, Inc. | Instruction set architecture for a vector computational unit |
| US11561791B2 (en) | 2018-02-01 | 2023-01-24 | Tesla, Inc. | Vector computational unit receiving data elements in parallel from a last row of a computational array |
| US11734562B2 (en) | 2018-06-20 | 2023-08-22 | Tesla, Inc. | Data pipeline and deep learning system for autonomous driving |
| US11841434B2 (en) | 2018-07-20 | 2023-12-12 | Tesla, Inc. | Annotation cross-labeling for autonomous control systems |
| US11636333B2 (en) | 2018-07-26 | 2023-04-25 | Tesla, Inc. | Optimizing neural network structures for embedded systems |
| CN109190699B (en) * | 2018-08-29 | 2022-12-30 | 西安电子科技大学 | Multi-disease joint determination method based on multi-task learning |
| CN109190699A (en) * | 2018-08-29 | 2019-01-11 | 西安电子科技大学 | A kind of more disease joint measurement methods based on multi-task learning |
| US11562231B2 (en) | 2018-09-03 | 2023-01-24 | Tesla, Inc. | Neural networks for embedded devices |
| US11893774B2 (en) | 2018-10-11 | 2024-02-06 | Tesla, Inc. | Systems and methods for training machine models with augmented data |
| US11665108B2 (en) | 2018-10-25 | 2023-05-30 | Tesla, Inc. | QoS manager for system on a chip communications |
| US11816585B2 (en) | 2018-12-03 | 2023-11-14 | Tesla, Inc. | Machine learning models operating at different frequencies for autonomous vehicles |
| US11537811B2 (en) | 2018-12-04 | 2022-12-27 | Tesla, Inc. | Enhanced object detection for autonomous vehicles based on field view |
| US11908171B2 (en) | 2018-12-04 | 2024-02-20 | Tesla, Inc. | Enhanced object detection for autonomous vehicles based on field view |
| US11610117B2 (en) | 2018-12-27 | 2023-03-21 | Tesla, Inc. | System and method for adapting a neural network model on a hardware platform |
| US11748620B2 (en) | 2019-02-01 | 2023-09-05 | Tesla, Inc. | Generating ground truth for machine learning from time series elements |
| US11567514B2 (en) | 2019-02-11 | 2023-01-31 | Tesla, Inc. | Autonomous and user controlled vehicle summon to a target |
| US11790664B2 (en) | 2019-02-19 | 2023-10-17 | Tesla, Inc. | Estimating object properties using visual image data |
| CN110582159A (en) * | 2019-08-30 | 2019-12-17 | 武汉科技大学 | FPC bonding pad surface defect processing system and method |
| CN113349809A (en) * | 2020-03-05 | 2021-09-07 | 高健 | Image reconstruction method of multi-modal imaging system |
| CN113554742B (en) * | 2020-04-26 | 2024-02-02 | 上海联影医疗科技股份有限公司 | Three-dimensional image reconstruction method, device, equipment and storage medium |
| CN113554742A (en) * | 2020-04-26 | 2021-10-26 | 上海联影医疗科技股份有限公司 | Three-dimensional image reconstruction method, device, equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106504253A (en) | A kind of processing method of medical imaging photo and system | |
| CN110120033A (en) | Based on improved U-Net neural network three-dimensional brain tumor image partition method | |
| Quellec et al. | Three-dimensional analysis of retinal layer texture: identification of fluid-filled regions in SD-OCT of the macula | |
| US4945478A (en) | Noninvasive medical imaging system and method for the identification and 3-D display of atherosclerosis and the like | |
| Mahapatra et al. | Automatic detection and segmentation of Crohn's disease tissues from abdominal MRI | |
| NL1027538C2 (en) | Methods and devices for processing image data to assist in the detection of disease. | |
| Acosta et al. | Automated voxel-based 3D cortical thickness measurement in a combined Lagrangian–Eulerian PDE approach using partial volume maps | |
| Arata et al. | Three-dimensional anatomical model-based segmentation of MR brain images through principal axes registration | |
| EP1780651A1 (en) | Method and system for automatic processing and evaluation of images, particularly diagnostic images | |
| US8634614B2 (en) | System and method for volumetric analysis of medical images | |
| CN109166133A (en) | Soft tissue organs image partition method based on critical point detection and deep learning | |
| EP1315125A2 (en) | Method and system for lung disease detection | |
| CN107230206A (en) | A kind of 3D Lung neoplasm dividing methods of the super voxel sequence lung images based on multi-modal data | |
| Huang et al. | VP-Nets: Efficient automatic localization of key brain structures in 3D fetal neurosonography | |
| CN113610808B (en) | Group brain map individuation method, system and equipment based on individual brain connection diagram | |
| CN110415234A (en) | Brain tumor dividing method based on multi-parameter magnetic resonance imaging | |
| Hu et al. | Segmentation of brain from computed tomography head images | |
| Park et al. | Segmentation of perivascular spaces in 7 T MR image using auto-context model with orientation-normalized features | |
| Colliot et al. | Segmentation of focal cortical dysplasia lesions on MRI using level set evolution | |
| Chen et al. | Automated ventricular systems segmentation in brain CT images by combining low-level segmentation and high-level template matching | |
| Wu et al. | AAR-RT–a system for auto-contouring organs at risk on CT images for radiation therapy planning: principles, design, and large-scale evaluation on head-and-neck and thoracic cancer cases | |
| Govyadinov et al. | Robust tracing and visualization of heterogeneous microvascular networks | |
| Li et al. | A hybrid approach to detection of brain hemorrhage candidates from clinical head ct scans | |
| Ghadimi et al. | Segmentation of scalp and skull in neonatal MR images using probabilistic atlas and level set method | |
| Shah et al. | Efficient T2 brain region extraction algorithm using morphological operation and overlapping test from 2D and 3D MRI images |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170315 |
|
| RJ01 | Rejection of invention patent application after publication |