CN104361318A  Disease diagnosis auxiliary system and disease diagnosis auxiliary method both based on diffusion tensor imaging technology  Google Patents
Disease diagnosis auxiliary system and disease diagnosis auxiliary method both based on diffusion tensor imaging technology Download PDFInfo
 Publication number
 CN104361318A CN104361318A CN201410627555.9A CN201410627555A CN104361318A CN 104361318 A CN104361318 A CN 104361318A CN 201410627555 A CN201410627555 A CN 201410627555A CN 104361318 A CN104361318 A CN 104361318A
 Authority
 CN
 China
 Prior art keywords
 tensor
 image
 dispersion
 projection
 disease
 Prior art date
Links
 201000010099 diseases Diseases 0.000 title claims abstract description 48
 238000003745 diagnosis Methods 0.000 title claims abstract description 40
 238000002598 diffusion tensor imaging Methods 0.000 title abstract description 3
 238000009792 diffusion process Methods 0.000 claims abstract description 33
 238000000605 extraction Methods 0.000 claims abstract description 25
 210000004556 Brain Anatomy 0.000 claims abstract description 15
 210000000278 Spinal Cord Anatomy 0.000 claims abstract description 13
 238000007781 preprocessing Methods 0.000 claims abstract description 11
 239000006185 dispersions Substances 0.000 claims description 77
 239000011159 matrix materials Substances 0.000 claims description 39
 238000005457 optimization Methods 0.000 claims description 14
 238000004458 analytical methods Methods 0.000 claims description 13
 238000000926 separation method Methods 0.000 claims description 12
 238000003909 pattern recognition Methods 0.000 claims description 6
 230000003247 decreasing Effects 0.000 abstract 1
 238000000034 methods Methods 0.000 description 7
 238000000513 principal component analysis Methods 0.000 description 5
 201000000980 schizophrenia Diseases 0.000 description 3
 206010003571 Astrocytoma Diseases 0.000 description 2
 210000003169 Central Nervous System Anatomy 0.000 description 2
 206010028570 Myelopathy Diseases 0.000 description 2
 210000004126 Nerve Fibers Anatomy 0.000 description 2
 206010061367 Spinal cord diseases Diseases 0.000 description 2
 238000009412 basement excavation Methods 0.000 description 2
 230000003210 demyelinating Effects 0.000 description 2
 238000010586 diagrams Methods 0.000 description 2
 238000010801 machine learning Methods 0.000 description 2
 238000002595 magnetic resonance imaging Methods 0.000 description 2
 230000004048 modification Effects 0.000 description 2
 238000006011 modification reactions Methods 0.000 description 2
 230000002265 prevention Effects 0.000 description 2
 210000004885 white matter Anatomy 0.000 description 2
 208000003174 Brain Neoplasms Diseases 0.000 description 1
 206010008190 Cerebrovascular accident Diseases 0.000 description 1
 206010013932 Dyslexia Diseases 0.000 description 1
 208000006011 Stroke Diseases 0.000 description 1
 210000004027 cells Anatomy 0.000 description 1
 125000004435 hydrogen atoms Chemical group data:image/svg+xml;base64,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 data:image/svg+xml;base64,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 [H]* 0.000 description 1
 238000003384 imaging method Methods 0.000 description 1
 201000006417 multiple sclerosis Diseases 0.000 description 1
 230000001537 neural Effects 0.000 description 1
 238000007619 statistical methods Methods 0.000 description 1
 230000001131 transforming Effects 0.000 description 1
Classifications

 G—PHYSICS
 G06—COMPUTING; CALCULATING; COUNTING
 G06K—RECOGNITION OF DATA; PRESENTATION OF DATA; RECORD CARRIERS; HANDLING RECORD CARRIERS
 G06K9/00—Methods or arrangements for reading or recognising printed or written characters or for recognising patterns, e.g. fingerprints
 G06K9/62—Methods or arrangements for recognition using electronic means
 G06K9/6201—Matching; Proximity measures
 G06K9/6202—Comparing pixel values or logical combinations thereof, or feature values having positional relevance, e.g. template matching
 G06K9/6203—Shifting or otherwise transforming the patterns to accommodate for positional errors
 G06K9/6211—Matching configurations of points or features, e.g. constellation matching

 G—PHYSICS
 G06—COMPUTING; CALCULATING; COUNTING
 G06K—RECOGNITION OF DATA; PRESENTATION OF DATA; RECORD CARRIERS; HANDLING RECORD CARRIERS
 G06K9/00—Methods or arrangements for reading or recognising printed or written characters or for recognising patterns, e.g. fingerprints
 G06K9/62—Methods or arrangements for recognition using electronic means
 G06K9/6217—Design or setup of recognition systems and techniques; Extraction of features in feature space; Clustering techniques; Blind source separation
 G06K9/6256—Obtaining sets of training patterns; Bootstrap methods, e.g. bagging, boosting

 G—PHYSICS
 G06—COMPUTING; CALCULATING; COUNTING
 G06K—RECOGNITION OF DATA; PRESENTATION OF DATA; RECORD CARRIERS; HANDLING RECORD CARRIERS
 G06K9/00—Methods or arrangements for reading or recognising printed or written characters or for recognising patterns, e.g. fingerprints
 G06K9/62—Methods or arrangements for recognition using electronic means
 G06K9/6217—Design or setup of recognition systems and techniques; Extraction of features in feature space; Clustering techniques; Blind source separation
 G06K9/626—Selecting classification rules

 G—PHYSICS
 G06—COMPUTING; CALCULATING; COUNTING
 G06K—RECOGNITION OF DATA; PRESENTATION OF DATA; RECORD CARRIERS; HANDLING RECORD CARRIERS
 G06K9/00—Methods or arrangements for reading or recognising printed or written characters or for recognising patterns, e.g. fingerprints
 G06K9/62—Methods or arrangements for recognition using electronic means
 G06K9/6267—Classification techniques
 G06K9/6268—Classification techniques relating to the classification paradigm, e.g. parametric or nonparametric approaches
 G06K9/627—Classification techniques relating to the classification paradigm, e.g. parametric or nonparametric approaches based on distances between the pattern to be recognised and training or reference patterns
 G06K9/6271—Classification techniques relating to the classification paradigm, e.g. parametric or nonparametric approaches based on distances between the pattern to be recognised and training or reference patterns based on distances to prototypes

 G—PHYSICS
 G06—COMPUTING; CALCULATING; COUNTING
 G06K—RECOGNITION OF DATA; PRESENTATION OF DATA; RECORD CARRIERS; HANDLING RECORD CARRIERS
 G06K2209/00—Indexing scheme relating to methods or arrangements for reading or recognising printed or written characters or for recognising patterns, e.g. fingerprints
 G06K2209/05—Recognition of patterns in medical or anatomical images

 G—PHYSICS
 G06—COMPUTING; CALCULATING; COUNTING
 G06K—RECOGNITION OF DATA; PRESENTATION OF DATA; RECORD CARRIERS; HANDLING RECORD CARRIERS
 G06K2209/00—Indexing scheme relating to methods or arrangements for reading or recognising printed or written characters or for recognising patterns, e.g. fingerprints
 G06K2209/05—Recognition of patterns in medical or anatomical images
 G06K2209/053—Recognition of patterns in medical or anatomical images of protuberances, polyps nodules, etc.
Abstract
Description
Technical field
The invention belongs to medical image aided diagnosis technique field, particularly relate to a kind of medical diagnosis on disease backup system based on diffusion tensor technology and method.
Background technology
Diffusion tensor (Diffusion Tensor Imaging, DTI) technology is current unique AT white matter nerve fiber bundles living imaging method, is the special shape of Magnetic resonance imaging (MRI).Being different from Magnetic resonance imaging is follow the trail of the hydrogen atom in hydrone, and diffusion tensor is according to the drawing of hydrone moving direction.Diffusion tensor figure can disclose brain tumor how to affect the nerves cell connect, guide healthcare givers carry out operation on brain.It can also disclose the trickle abnormality change of the regarding brain such as same apoplexy, multiple sclerosis, schizophrenia, Dyslexia and spinal cord.Diffusion tensor data are secondorder tensor structure in essence, and its each voxel contains the threedimensional spatial information of hydrone disperse in white matter nerve fiber bundles.
The effective information judging human body diseases can be extracted from original dispersion tensor image by machine learning method, thus provide strong help for the disease that forecast analysis brain is relevant with spinal cord.But traditional machine learning method is all the algorithm based on vector pattern, such as support vector machine, linear discriminant analysis and neural network etc., some scalar indexs of these algorithms or just a process diffusion tensor data, and the structure space information of dispersion tensor image can not be made full use of; Before analyzing and processing, first tensor is expanded into vector.But this way can bring following problem: 1, destroy the structure of raw data and the inherent correlativity of tensor structured data.2, brain diffusion tensor data are after vectorization, the dimension of its statistical parameter (being typically covariance matrix) generated is very big, the structure of raw data can be destroyed, the inherent correlativity of obliterated data, thus cause high computation complexity and storage cost.
Summary of the invention
The invention provides a kind of medical diagnosis on disease backup system based on diffusion tensor technology and method, be intended to solve the structure space information that existing vector pattern learning algorithm can not make full use of dispersion tensor view data, and in the process of tensor data vector, the structure of raw data can be destroyed, the inherent correlativity of obliterated data, increases the technical matters of computation complexity and storage cost.
The present invention is achieved in that a kind of medical diagnosis on disease backup system based on diffusion tensor technology, comprises image preprocessing module, expertise library module, tensor study module and diagnostic result output module; Described image preprocessing module is used for the accurate and dispersion tensor image characteristics extraction of the collection of dispersion tensor image, dispersion tensor image reprovision; Described expertise library module is for setting up the expert knowledge library of brain and spinal cord relevant disease; Described tensor study module is used for utilizing expert knowledge library and tensor learning algorithm carry out the training of tensor disaggregated model and the optimization of tensor disaggregated model to dispersion tensor image and test; Described diagnostic result output module is for exporting diagnostic result.
The technical scheme that the embodiment of the present invention is taked also comprises: described image preprocessing module comprises image acquisition units, image registration unit and feature extraction unit;
Described image acquisition units is for gathering dispersion tensor image;
Described image registration unit for carrying out the dispersion tensor image registration based on tensor similarity, then carries out the dispersion tensor image registration based on scalar similarity;
Described feature extraction unit is used for the feature extraction and the dimensionality reduction that carry out dispersion tensor image based on the polyteny core principle component analysis method of tensor algebra.
The technical scheme that the embodiment of the present invention is taked also comprises: feature extraction and the dimensionality reduction of described dispersion tensor image are specially: first choose suitable kernel function and raw data is mapped to respective feature space; Again by tensor product by all Feature Mapping to multilinear subspace, make each subspace can capture the amount of variability of most of orthogonal multidimensional; According to minimum alternately square principle, calculate and solve new feature.
The technical scheme that the embodiment of the present invention is taked also comprises: described tensor study module comprises tensor disaggregated model training unit and unit optimized by tensor disaggregated model;
Described tensor disaggregated model training unit is used for utilizing expert knowledge library and optimum projection to support that tensor machine carries out the training of tensor disaggregated model to dispersion tensor image;
Described tensor disaggregated model is optimized unit and is used for carrying out the optimization of tensor disaggregated model and test.
The technical scheme that the embodiment of the present invention is taked also comprises: the training of described tensor disaggregated model is specially: support that tensor machine carries out patternrecognition to dispersion tensor image based on optimum projection, find the optimum projection of image on boundary direction, and calculating optimum projection vector by the ratio of the distribution of matrix in matrix and class between the class of sample after maximizing projection, the projection vector of described calculating optimum specifically comprises: matrix S between compute classes _{b}with matrix S in class _{w}maximum value the max [(V of the ratio of distribution ^{t}s _{b}v)/(V ^{t}s _{w}v)], the optimum projection vector v on first boundary direction is solved; Training sample is projected by v, solves quadratic programming problem, in like manner try to achieve the optimum projection vector on second boundary direction; With scatter matrix definition projection coefficient of dispersion R=Tr (S in scatter matrix between class and class _{b}× S _{b} ^{t})/Tr (S _{w}× S _{w} ^{t}), R represents the degree of separation of training sample after the projection of optimum projection vector between sample, R value larger expression training sample after projecting sample separation from larger.
The technical scheme that the embodiment of the present invention is taked also comprises: the training of described tensor disaggregated model also comprises: support that tensor machine is analyzed two kinds of projection scheme and is based on optimum projection: the training sample of each subclassifier is to same direction projection, or each subclassifier determines different projection vectors, then training sample projects respectively on respective projection vector direction.
Another technical scheme that the embodiment of the present invention is taked is: a kind of medical diagnosis on disease householder method based on diffusion tensor technology, comprising:
Step a: gather dispersion tensor image, and carry out dispersion tensor image reprovision standard and dispersion tensor image characteristics extraction;
Step b: the expert knowledge library setting up brain and spinal cord relevant disease;
Step c: utilize expert knowledge library and tensor learning algorithm carry out the training of tensor disaggregated model and the optimization of tensor disaggregated model to dispersion tensor image and test;
Steps d: export diagnostic result.
The technical scheme that the embodiment of the present invention is taked also comprises: in described step a, and described dispersion tensor image reprovision standard comprises the dispersion tensor image registration based on tensor similarity and the dispersion tensor image registration based on scalar similarity.
The technical scheme that the embodiment of the present invention is taked also comprises: in described step a, and described dispersion tensor image characteristics extraction is: the polyteny core principle component analysis method based on tensor algebra carries out feature extraction and the dimensionality reduction of dispersion tensor image; Specifically comprise: first choose suitable kernel function and raw data is mapped to respective feature space, again by tensor product by all Feature Mapping to multilinear subspace, make each subspace can capture the amount of variability of most of orthogonal multidimensional, according to minimum alternately square principle, calculate and solve new feature.
The technical scheme that the embodiment of the present invention is taked also comprises: in described step c, the training of described tensor disaggregated model is specially: support that tensor machine carries out patternrecognition to dispersion tensor image based on optimum projection, find the optimum projection of image on boundary direction, and calculating optimum projection vector by the ratio of the distribution of matrix in matrix and class between the class of sample after maximizing projection, the projection vector of described calculating optimum specifically comprises: matrix S between compute classes _{b}with matrix S in class _{w}maximum value the max [(V of the ratio of distribution ^{t}s _{b}v)/(V ^{t}s _{w}v)], the optimum projection vector v on first boundary direction is solved; Training sample is projected by v, solves quadratic programming problem, in like manner try to achieve the optimum projection vector on second boundary direction; With scatter matrix definition projection coefficient of dispersion R=Tr (S in scatter matrix between class and class _{b}× S _{b} ^{t})/Tr (S _{w}× S _{w} ^{t}), R represents the degree of separation of training sample after the projection of optimum projection vector between sample, R value larger expression training sample after projecting sample separation from larger.
The medical diagnosis on disease backup system based on diffusion tensor technology of the embodiment of the present invention and method, for the feature more responsive than halftone information of directional information in dispersion tensor image, propose double tolerance method for registering, secondly, feature extraction and dimensionality reduction that a kind of polyteny core principle component analysis method realizes dispersion tensor image are proposed, finally, for the limitation of vector pattern learning algorithm in process tensor structured data, the structural risk minimization principle of combining classification device and existing tensor pattern learning algorithm, and based on optimum projection vector criterion, propose to support tensor machine based on optimum projection, realize the medical diagnosis on disease Real time identification based on tensor pattern algorithm and dispersion tensor image, can under the prerequisite not destroying initial data structure, the original information of abundant excavation image, improve pattern classification precision, greatly can reduce calculated amount and assess the cost, and the realtime of effectively having protected medical diagnosis on disease whole.The present invention is used in the diseases analysis and prediction aspect that modification between astrocytoma, spiritual demyelinating pathology, development of central nervous system exception, schizophrenia, depression, cervical spondylotic myelopathy etc. be associated with brain and spinal cord, and the present invention simultaneously also may be used for the brain of normal person and the prevention of spinal cord relevant disease and health care aspect.
Accompanying drawing explanation
Fig. 1 is the structural representation of the medical diagnosis on disease backup system based on diffusion tensor technology of the embodiment of the present invention;
Fig. 2 is the process flow diagram of the medical diagnosis on disease householder method based on diffusion tensor technology of the embodiment of the present invention.
Embodiment
In order to make object of the present invention, technical scheme and advantage clearly understand, below in conjunction with drawings and Examples, the present invention is further elaborated.Should be appreciated that specific embodiment described herein only in order to explain the present invention, be not intended to limit the present invention.
Referring to Fig. 1, is the structural representation of the medical diagnosis on disease backup system based on diffusion tensor technology of the embodiment of the present invention.The medical diagnosis on disease backup system based on diffusion tensor technology of the embodiment of the present invention comprises image preprocessing module, expertise library module, tensor study module and diagnostic result output module; Wherein, the collection of image preprocessing module for dispersion tensor image, the standard of the dispersion tensor image reprovision based on tensor pattern and the dispersion tensor image characteristics extraction based on polyteny core principle component analysis method; Expertise library module is for setting up the expert knowledge library of brain and spinal cord relevant disease; Tensor study module is used for utilizing expert knowledge library and tensor learning algorithm carry out the training of tensor disaggregated model and the optimization of tensor disaggregated model to dispersion tensor image and test; Diagnostic result output module is for exporting diagnostic result.
Particularly, image preprocessing module comprises image acquisition units, image registration unit and feature extraction unit,
Image acquisition units is for gathering dispersion tensor image;
Image registration unit is used for, for the feature more responsive than halftone information of directional information in dispersion tensor image, first carrying out the dispersion tensor image registration based on tensor similarity, then carrying out the dispersion tensor image registration based on scalar similarity.
Feature extraction unit is used for the feature extraction and the dimensionality reduction that realize dispersion tensor image based on the polyteny core principle component analysis method of tensor algebra, be specially: first choose suitable kernel function and raw data is mapped to respective feature space, again by tensor product by all Feature Mapping to multilinear subspace, make each subspace can capture the amount of variability of most of orthogonal multidimensional, according to minimum alternately square principle, calculate and solve new feature.Wherein, principal component analysis (PCA) (Principal Component Analysis, PCA) method be by multiple variable by linear transformation to select a kind of Multielement statistical analysis method of less number significant variable, also known as principal component analysis.
Tensor study module comprises tensor disaggregated model training unit and unit optimized by tensor disaggregated model,
Tensor disaggregated model training unit is used for utilizing expert knowledge library and optimum projection to support that tensor machine (OPSTM) carries out the training of tensor disaggregated model to dispersion tensor image; Be specially: utilize Fisher criterion, support that tensor machine carries out patternrecognition to dispersion tensor image based on optimum projection, find the optimum projection of image on boundary direction, and calculate optimum projection vector by the ratio of the distribution of matrix in matrix and class between the class of sample after maximizing projection; Algorithm is divided into two steps to realize: matrix S between (1), compute classes _{b}with matrix S in class _{w}maximum value the max [(V of the ratio of distribution ^{t}s _{b}v)/(V ^{t}s _{w}v)], the optimum projection vector v on first boundary direction is solved.(2), training sample projected by v, solve quadratic programming problem, in like manner can try to achieve the optimum projection vector on second boundary direction.The present invention to utilize between class scatter matrix definition projection coefficient of dispersion R=Tr (S in scatter matrix and class _{b}× S _{b} ^{t})/Tr (S _{w}× S _{w} ^{t}), R represents the degree of separation of training sample after the projection of optimum projection vector between sample, R value larger expression training sample after projecting sample separation from larger.To support that tensor machine is generalized to many classification from two classification based on optimum projection, analyze two kinds of projection scheme: 1, the training sample of each subclassifier is to same direction projection.2, each subclassifier determines different projection vectors, and then training sample projects respectively on respective projection vector direction.Due to training sample is projected in tensor subspace, inhomogeneous training sample can be separated from each other, and similar training sample can be close to each other.This makes the scheme discrimination choosing same projection vector higher than the scheme choosing nonsame projection vector.
Tensor disaggregated model is optimized unit and is used for carrying out the optimization of tensor disaggregated model and test; Be specially: in conjunction with multilinear algebra operation rule, for flow data feature, optimum projection is supported that tensor machine is generalized to online form, and the inequality constrain in optimization problem is converted into equality constraint, concerning diffusion tensor data (secondorder tensor), only need to solve two systems of linear equations, carry out simple matrix inversion operation, significantly can reduce calculated amount.In objective function optimization method, take into full account the feature of flow data, utilize stochastic gradient descent method to complete Lagrange multiplier iteration.Assess the cost to keep the prototype structure information of flow data and reduce, utilize the CP of tensor to decompose auxiliary inner product of tensor computing, this while the natural structure information keeping original tensor, can reduce calculated amount and saves storage space.
Referring to Fig. 2, is the process flow diagram of the medical diagnosis on disease householder method based on diffusion tensor technology of the embodiment of the present invention.The medical diagnosis on disease householder method based on diffusion tensor technology of the embodiment of the present invention comprises the following steps:
Step 100: gather dispersion tensor image;
Step 200: registration is carried out to the dispersion tensor image of described collection.Specifically:
For the feature that directional information in dispersion tensor image is more responsive than halftone information, first carry out the dispersion tensor image registration based on tensor similarity, then carry out the dispersion tensor image registration based on scalar similarity
Step 300: the polyteny core principle component analysis method based on tensor algebra carries out feature extraction and the dimensionality reduction of dispersion tensor image;
In step 300, feature extraction and the dimensionality reduction of dispersion tensor image are specially: first choose suitable kernel function and raw data is mapped to respective feature space, again by tensor product by all Feature Mapping to multilinear subspace, make each subspace can capture the amount of variability of most of orthogonal multidimensional, according to minimum alternately square principle, calculate and solve new feature.
Step 400: the expert knowledge library setting up brain and spinal cord relevant disease;
Step 500: find the optimum projection of abovementioned dispersion tensor image on boundary direction, and calculate optimum projection vector by the ratio of the distribution of matrix in matrix and class between the class of sample after maximizing projection.Specifically:
Be divided into two steps to realize: matrix S between (1), compute classes _{b}with matrix S in class _{w}maximum value the max [(V of the ratio of distribution ^{t}s _{b}v)/(V ^{t}s _{w}v)], the optimum projection vector v on first boundary direction is solved.(2), training sample projected by v, solve quadratic programming problem, in like manner can try to achieve the optimum projection vector on second boundary direction.
Step 600: utilize expert knowledge library and optimum projection to support that tensor machine carries out the training of tensor disaggregated model;
In step 600, the training of tensor disaggregated model is specially: utilize Fisher criterion, supports that tensor machine carries out patternrecognition to dispersion tensor image based on optimum projection.The present invention to utilize between class scatter matrix definition projection coefficient of dispersion R=Tr (S in scatter matrix and class _{b}× S _{b} ^{t})/Tr (S _{w}× S _{w} ^{t}), R represents the degree of separation of training sample after the projection of optimum projection vector between sample, R value larger expression training sample after projecting sample separation from larger.To support that tensor machine is generalized to many classification from two classification based on optimum projection, analyze two kinds of projection scheme: 1, the training sample of each subclassifier is to same direction projection.2, each subclassifier determines different projection vectors, and then training sample projects respectively on respective projection vector direction.Due to training sample is projected in tensor subspace, inhomogeneous training sample can be separated from each other, and similar training sample can be close to each other.This makes the scheme discrimination choosing same projection vector higher than the scheme choosing nonsame projection vector.
Step 700: carry out the optimization of tensor disaggregated model and test;
In step 700, the optimization of tensor disaggregated model and test are specially: in conjunction with multilinear algebra operation rule, for flow data feature, optimum projection is supported that tensor machine is generalized to online form, and the inequality constrain in optimization problem is converted into equality constraint, concerning diffusion tensor data (secondorder tensor), only need to solve two systems of linear equations, carry out simple matrix inversion operation, significantly can reduce calculated amount.In objective function optimization method, take into full account the feature of flow data, utilize stochastic gradient descent method to complete Lagrange multiplier iteration.Assess the cost to keep the prototype structure information of flow data and reduce, utilize the CP of tensor to decompose auxiliary inner product of tensor computing, this while the natural structure information keeping original tensor, can reduce calculated amount and saves storage space.
Step 800: export diagnostic result.
The medical diagnosis on disease backup system based on diffusion tensor technology of the embodiment of the present invention and method, for the feature more responsive than halftone information of directional information in dispersion tensor image, propose double tolerance method for registering, secondly, feature extraction and dimensionality reduction that a kind of polyteny core principle component analysis method realizes dispersion tensor image are proposed, finally, for the limitation of vector pattern learning algorithm in process tensor structured data, the structural risk minimization principle of combining classification device and existing tensor pattern learning algorithm, and based on optimum projection vector criterion, propose to support tensor machine based on optimum projection, realize the medical diagnosis on disease Real time identification based on tensor pattern algorithm and dispersion tensor image, can under the prerequisite not destroying initial data structure, the original information of abundant excavation image, improve pattern classification precision, greatly can reduce calculated amount and assess the cost, and the realtime of effectively having protected medical diagnosis on disease whole.The present invention is used in the diseases analysis and prediction aspect that modification between astrocytoma, spiritual demyelinating pathology, development of central nervous system exception, schizophrenia, depression, cervical spondylotic myelopathy etc. be associated with brain and spinal cord, and the present invention simultaneously also may be used for the brain of normal person and the prevention of spinal cord relevant disease and health care aspect.
The foregoing is only preferred embodiment of the present invention, not in order to limit the present invention, all any amendments done within the spirit and principles in the present invention, equivalent replacement and improvement etc., all should be included within protection scope of the present invention.
Claims (10)
Priority Applications (1)
Application Number  Priority Date  Filing Date  Title 

CN201410627555.9A CN104361318B (en)  20141110  20141110  A kind of medical diagnosis on disease accessory system based on diffusion tensor technology 
Applications Claiming Priority (1)
Application Number  Priority Date  Filing Date  Title 

CN201410627555.9A CN104361318B (en)  20141110  20141110  A kind of medical diagnosis on disease accessory system based on diffusion tensor technology 
Publications (2)
Publication Number  Publication Date 

CN104361318A true CN104361318A (en)  20150218 
CN104361318B CN104361318B (en)  20180206 
Family
ID=52528576
Family Applications (1)
Application Number  Title  Priority Date  Filing Date 

CN201410627555.9A CN104361318B (en)  20141110  20141110  A kind of medical diagnosis on disease accessory system based on diffusion tensor technology 
Country Status (1)
Country  Link 

CN (1)  CN104361318B (en) 
Cited By (8)
Publication number  Priority date  Publication date  Assignee  Title 

CN104715260A (en) *  20150305  20150617  中南大学  Multimodal fusion image sorting method based on RLSELM 
CN104866713A (en) *  20150512  20150826  南京霁云信息科技有限公司  Kawasaki disease and fever diagnosis system based on embedding of incremental local discrimination subspace 
CN105184794A (en) *  20150907  20151223  中国科学院深圳先进技术研究院  CSM assistant analysis system and method based on tensor image 
CN105654110A (en) *  20151204  20160608  深圳先进技术研究院  Supervised learning optimization method under tensor mode and system thereof 
WO2017092022A1 (en) *  20151204  20170608  深圳先进技术研究院  Optimization method and system for supervised tensor learning 
CN107330267A (en) *  20170628  20171107  首都医科大学宣武医院  Utilize the white matter fiber brain map construction method of diffusion tensor medical image 
CN107590806A (en) *  20170919  20180116  陈烨  A kind of detection method and system based on brain medical imaging 
TWI639830B (en) *  20170317  20181101  長庚大學  Method for identifying neurological diseases using magnetic resonance imaging images 
Citations (3)
Publication number  Priority date  Publication date  Assignee  Title 

CN103383716A (en) *  20130626  20131106  北京理工大学  Senile dementia computeraided diagnosis method based on 18FFDG PET image 
CN103425999A (en) *  20130827  20131204  西安电子科技大学  Brain cognitive state judgment method based on nonnegative tensor projection operator decomposition algorithm 
US20140293069A1 (en) *  20130402  20141002  Microsoft Corporation  Realtime image classification and automated image content curation 

2014
 20141110 CN CN201410627555.9A patent/CN104361318B/en active IP Right Grant
Patent Citations (3)
Publication number  Priority date  Publication date  Assignee  Title 

US20140293069A1 (en) *  20130402  20141002  Microsoft Corporation  Realtime image classification and automated image content curation 
CN103383716A (en) *  20130626  20131106  北京理工大学  Senile dementia computeraided diagnosis method based on 18FFDG PET image 
CN103425999A (en) *  20130827  20131204  西安电子科技大学  Brain cognitive state judgment method based on nonnegative tensor projection operator decomposition algorithm 
NonPatent Citations (2)
Title 

李文: "磁共振扩散张量图像配准的研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * 
牛少波: "基于张量学习的目标识别技术研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * 
Cited By (11)
Publication number  Priority date  Publication date  Assignee  Title 

CN104715260A (en) *  20150305  20150617  中南大学  Multimodal fusion image sorting method based on RLSELM 
CN104866713A (en) *  20150512  20150826  南京霁云信息科技有限公司  Kawasaki disease and fever diagnosis system based on embedding of incremental local discrimination subspace 
CN104866713B (en) *  20150512  20180213  南京霁云信息科技有限公司  Locally differentiate the Kawasaki disease and fever diagnostic system of subspace insertion based on increment 
CN105184794A (en) *  20150907  20151223  中国科学院深圳先进技术研究院  CSM assistant analysis system and method based on tensor image 
CN105184794B (en) *  20150907  20180417  中国科学院深圳先进技术研究院  A kind of CSM Computer Aided Analysis Systems and method based on tensor image 
CN105654110A (en) *  20151204  20160608  深圳先进技术研究院  Supervised learning optimization method under tensor mode and system thereof 
WO2017092022A1 (en) *  20151204  20170608  深圳先进技术研究院  Optimization method and system for supervised tensor learning 
US10748080B2 (en)  20151204  20200818  Shenzhen Institutes Of Advanced Technology  Method for processing tensor data for pattern recognition and computer device 
TWI639830B (en) *  20170317  20181101  長庚大學  Method for identifying neurological diseases using magnetic resonance imaging images 
CN107330267A (en) *  20170628  20171107  首都医科大学宣武医院  Utilize the white matter fiber brain map construction method of diffusion tensor medical image 
CN107590806A (en) *  20170919  20180116  陈烨  A kind of detection method and system based on brain medical imaging 
Also Published As
Publication number  Publication date 

CN104361318B (en)  20180206 
Similar Documents
Publication  Publication Date  Title 

Giovannucci et al.  CaImAn an open source tool for scalable calcium imaging data analysis  
Graving et al.  DeepPoseKit, a software toolkit for fast and robust animal pose estimation using deep learning  
Wang et al.  Single slice based detection for Alzheimer’s disease via wavelet entropy and multilayer perceptron trained by biogeographybased optimization  
Wu et al.  Scalable highperformance image registration framework by unsupervised deep feature representations learning  
Parisot et al.  Spectral graph convolutions for populationbased disease prediction  
Mwangi et al.  A review of feature reduction techniques in neuroimaging  
Davison et al.  Brain network adaptability across task states  
Zhang et al.  Classification of Alzheimer disease based on structural magnetic resonance imaging by kernel support vector machine decision tree  
Liu et al.  Relationship induced multitemplate learning for diagnosis of Alzheimer’s disease and mild cognitive impairment  
Tang et al.  Shape abnormalities of subcortical and ventricular structures in mild cognitive impairment and Alzheimer's disease: detecting, quantifying, and predicting  
WO2016192612A1 (en)  Method for analysing medical treatment data based on deep learning, and intelligent analyser thereof  
Mur et al.  Human objectsimilarity judgments reflect and transcend the primateIT object representation  
Jenatton et al.  Multiscale mining of fMRI data with hierarchical structured sparsity  
Sui et al.  A review of multivariate methods for multimodal fusion of brain imaging data  
Hjelm et al.  Restricted Boltzmann machines for neuroimaging: an application in identifying intrinsic networks  
Mazziotta et al.  A probabilistic atlas and reference system for the human brain: International Consortium for Brain Mapping (ICBM)  
Michel et al.  Total variation regularization for fMRIbased prediction of behavior  
US8799202B2 (en)  Predicting states of subjects  
US10366491B2 (en)  Deep imagetoimage recurrent network with shape basis for automatic vertebra labeling in largescale 3D CT volumes  
Thompson et al.  A framework for computational anatomy  
Durrleman et al.  Statistical models of sets of curves and surfaces based on currents  
Wernick et al.  Machine learning in medical imaging  
Sarraf et al.  Deep learningbased pipeline to recognize Alzheimer's disease using fMRI data  
Agrawal et al.  Pixels to voxels: Modeling visual representation in the human brain  
Casanova et al.  High dimensional classification of structural MRI Alzheimer’s disease data based on large scale regularization 
Legal Events
Date  Code  Title  Description 

PB01  Publication  
C06  Publication  
SE01  Entry into force of request for substantive examination  
C10  Entry into substantive examination  
CB03  Change of inventor or designer information 
Inventor after: Wang Shuqiang Inventor after: Tan Weiqi Inventor after: Shen Yanyan Inventor after: Hu Jinxing Inventor after: Yin Ling Inventor after: Zeng Chunxia Inventor before: Wang Shuqiang Inventor before: Hu Yong Inventor before: Tan Weiqi Inventor before: Shen Yanyan Inventor before: Hu Jinxing Inventor before: Yin Ling Inventor before: Zeng Chunxia 

C53  Correction of patent for invention or patent application  
COR  Change of bibliographic data 
Free format text: CORRECT: INVENTOR; FROM: WANG SHUQIANG HU YONG TAN WEIQI SHEN YANYAN HU JINXING YIN LING CENG CHUNXIA TO: WANG SHUQIANG TAN WEIQI SHEN YANYAN HU JINXING YIN LING CENG CHUNXIA 

GR01  Patent grant  
GR01  Patent grant 