CN110400300B - Pathological blood vessel accurate detection method based on block matching self-adaptive weight sparse representation - Google Patents

Pathological blood vessel accurate detection method based on block matching self-adaptive weight sparse representation Download PDF

Info

Publication number
CN110400300B
CN110400300B CN201910669052.0A CN201910669052A CN110400300B CN 110400300 B CN110400300 B CN 110400300B CN 201910669052 A CN201910669052 A CN 201910669052A CN 110400300 B CN110400300 B CN 110400300B
Authority
CN
China
Prior art keywords
blood vessel
sparse representation
scale
block
information block
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.)
Active
Application number
CN201910669052.0A
Other languages
Chinese (zh)
Other versions
CN110400300A (en
Inventor
胡鑫
初佃辉
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Harbin Institute of Technology Weihai
Original Assignee
Harbin Institute of Technology Weihai
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Harbin Institute of Technology Weihai filed Critical Harbin Institute of Technology Weihai
Priority to CN201910669052.0A priority Critical patent/CN110400300B/en
Publication of CN110400300A publication Critical patent/CN110400300A/en
Application granted granted Critical
Publication of CN110400300B publication Critical patent/CN110400300B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/22—Matching criteria, e.g. proximity measures
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/20—Image enhancement or restoration by the use of local operators
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0012—Biomedical image inspection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/60—Analysis of geometric attributes
    • G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/60—Analysis of geometric attributes
    • G06T7/68—Analysis of geometric attributes of symmetry
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20024—Filtering details
    • 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
    • G06T2207/30101—Blood vessel; Artery; Vein; Vascular
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
    • Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation

Abstract

The invention discloses a lesion blood vessel accurate detection method based on block matching self-adaptive weight sparse representation, which is used for generating a multi-scale blood vessel training set; extracting information blocks based on the axis of the blood vessel; based on multi-scale dictionary learning of sparse representation, self-adaptively capturing local lesion vascular features from a block information base through a sparse representation algorithm; and carrying out accurate detection on the lesion blood vessel by adopting a self-adaptive weight sparse representation classification algorithm. The beneficial effects of the invention are as follows: aiming at the difficulty that the pathological blood vessel cannot be accurately detected, the invention provides a block-matching-based adaptive weight sparse representation detection algorithm, the detection precision of the pathological blood vessel is greatly improved by utilizing redundant information contained in a blood vessel information block, and the calculation amount of the sparse representation detection algorithm is reduced by adopting the adaptive weight algorithm and a multi-scale training set.

Description

Pathological blood vessel accurate detection method based on block matching self-adaptive weight sparse representation
Technical Field
The invention belongs to the technical field of medicine, and relates to a lesion blood vessel accurate detection method based on block matching self-adaptive weight sparse representation.
Background
At present, the blood vessel segmentation technology obtains higher segmentation precision for healthy blood vessels in a complex environment, but higher false detection rate and missed detection rate can occur when the blood vessels are faced with pathological changes, so that the detection result lacks clinical guidance significance. Redundant information contained in the multi-scale training set is a key for realizing accurate detection of lesion blood vessels by the adaptive weight sparse representation method. However, the huge amount of computation brought by the sparse representation method severely reduces the convergence of the detection algorithm.
Disclosure of Invention
The invention aims to provide a lesion blood vessel accurate detection method based on block matching self-adaptive weight sparse representation, which has the beneficial effects that: aiming at the difficulty that a lesion blood vessel in a CT image cannot be accurately detected, a block-matching-based adaptive weight sparse representation detection algorithm is provided, the detection precision of the lesion blood vessel is greatly improved by utilizing redundant information contained in a blood vessel information block, and the calculation amount of the sparse representation detection algorithm is reduced by adopting the adaptive weight algorithm and a multi-scale training set.
The technical scheme adopted by the invention is carried out according to the following steps:
step 1: generating a multi-scale blood vessel training set;
step 2: extracting information blocks based on the axis of the blood vessel;
step 3: based on multi-scale dictionary learning of sparse representation, self-adaptively capturing local lesion vascular features from a block information base through a sparse representation algorithm;
step 4: and carrying out accurate detection on the lesion blood vessel by adopting a self-adaptive weight sparse representation classification algorithm.
Further, step 1 is to consider the vascular structure as a linear structure in the human body, construct a multidimensional linear structure filter based on Gaussian convolution by utilizing a hessian matrix, and convert the vascular training set into a training set related to the vascular scale by applying the constructed multidimensional linear structure filter on the basis of a traditional semi-automatic segmentation vascular training set.
Further, step 2 is to extract the central axis of the lesion blood vessel by using a multiscale hierarchical blood vessel tracking algorithm, take the maximum scale of the blood vessel as the radius along the section direction of the axis, extracting blood vessel information blocks from a multi-scale blood vessel training set, constructing a multi-scale block information base, the size of the information block is 7 x 7 voxels, the scale of which is determined by the scale of the target vessel being extracted.
Further, the procedure of step 3 is as follows:
Figure GDA0004238635320000021
wherein d is s The scale is s, beta is a sparse representation coefficient vector, p is an information block in a training set, and d s The optimal solution of beta is given by a sparse representation algorithm, the residual error of the sparse representation determines the characterizability of the information block, and the information block with high characterizability is put into a multi-scale dictionary to realize the lesion bloodCapturing local features of the tube. Multi-scale dictionary learning (step 3) and sparse representation optimization solution (step 4) are performed iteratively. Setting a residual error threshold value, and when the sparse representation residual error of a block in the information base is larger than the threshold value, putting the block into the dictionary to realize the updating learning of the multi-scale dictionary.
Further, the procedure of step 4 is as follows:
Figure GDA0004238635320000022
wherein p is a target blood vessel information block to be detected; representing matrix element multiplication; w represents a block information base d s Is used for the adaptive weight vector of the (a); w is a weight-dependent mapping matrix; a, a i And a j For block information base d s Any two information blocks; lambda (lambda) 1 、λ 2 And lambda (lambda) 3 Is a non-negative influence factor parameter;
Figure GDA0004238635320000024
for normalization of the mapping matrix W, the block information base d may be reduced s Characteristic distance of the medium similar information block; according to the sparse representation residual error of the target information block p, defining the similarity between p and a blood vessel, wherein the similarity is shown in the following formula:
Figure GDA0004238635320000023
wherein beta is a sparse representation coefficient obtained by optimizing the formula (2), and when the similarity between the target information block and the blood vessel is high, the target information block p is considered as a part of the blood vessel, so that the classification of the target information block is realized; the collecting process of the target information block p to be detected is consistent with the collecting process of the training set information block, namely along the section direction of the axis of the blood vessel. Thus, the target information block forms a sequence { p } to be detected 0 ,...,p n-1 ,p n ,...}. Due to the proximity of the information blocks to be detected, the similarity between adjacent information blocks is p n-1 Can guide p n Thus, the optimization solution of equation (2)The process comprises the following five steps: 1) At p n-1 Scale s of (2) n-1 Vicinity search p n Is the optimal scale s of (2) n The method comprises the steps of carrying out a first treatment on the surface of the 2) At parameter w n-1 、W n-1 Sum s n Under certain conditions, the solving parameter beta is optimized n In this case, the formula (2) is converted into a simple conventional sparse representation optimization solution problem, and an optimal solution can be obtained easily:
Figure GDA0004238635320000031
wherein->
Figure GDA0004238635320000032
G is X n Is a diagonal matrix of (a); 3) At parameter W n-1 、β n Sum s n Under the determined condition, optimizing the solving parameter w n In this case, the Lagrangian function of equation (2) satisfies the Karush-Kuhn-Tucker synchronous convergence condition, thereby obtaining w n Is the optimal solution of (a):
Figure GDA0004238635320000033
wherein->
Figure GDA0004238635320000034
ε m-l+1 M-l+1 element of ε, m is β n The number of elements in (1) is beta n Number of non-zero elements->
Figure GDA0004238635320000035
For a full 1 matrix, the matrix size is consistent with the weight matrix wn; 4) At parameter w n 、β n Sum s n Under the determined condition, optimizing the solving parameter W n In this case, equation (2) becomes an unconstrained optimization problem, which can be solved using a conventional CG Schmidt (2005) optimization operator; 5) At parameter w n 、W n Sum s n Under the determined condition, the optimization solving process of the second step is repeated, and the updating parameter beta is optimized n 。
Detailed Description
The present invention will be described in detail with reference to the following embodiments.
The invention relates to a lesion blood vessel accurate detection method based on block matching self-adaptive weight sparse representation, which comprises the following steps:
step 1: a multi-scale vascular training set is generated. The vascular structure is regarded as a linear structure in the human body, and a multi-dimensional linear structure filter based on Gaussian convolution is constructed by utilizing a hessian matrix, wherein the standard deviation parameter of the filter is related to the vascular scale. Based on the traditional semi-automatic segmentation blood vessel training set, the constructed multidimensional linear structure filter is applied to convert the blood vessel training set into a training set related to blood vessel scale.
Step 2: and extracting information blocks based on the axis of the blood vessel. And extracting the central axis of the lesion blood vessel by using a multi-scale hierarchical blood vessel tracking algorithm. And (3) extracting a blood vessel information block by taking the maximum dimension of the blood vessel as the radius along the axial section direction to construct a multi-scale block information base. The size of the information block is 7 x 7 voxels, the scale of which is determined by the scale of the target vessel being extracted.
The multi-scale block information base can capture local characteristics of changeable lesion blood vessels and low robustness, and unification of detection precision and algorithm time complexity is realized.
Step 3: sparse representation-based multi-scale dictionary learning. Through a sparse representation algorithm, the local lesion vascular characteristics are adaptively captured from a block information base, and the specific process is as follows:
Figure GDA0004238635320000041
wherein d is s And (3) a block information base with a scale of s, wherein beta is a sparse representation coefficient vector, and p is an information block in a training set. d, d s And the optimization solution of beta is given by a sparse representation algorithm, the residual error of the sparse representation determines the characterizations of the information blocks, and the information blocks with high characterizations are put into a multi-scale dictionary, so that the capture of the local characteristics of the lesion blood vessels is realized. Multi-scale dictionary learning (step 3) and sparse representation optimization solution (step 4) are performed iteratively. Setting a residual error threshold value, when the sparse representation residual error of the block in the information base is greater than the threshold value, the block is putAnd entering a dictionary to realize the updating learning of the multi-scale dictionary.
Step 4: and carrying out accurate detection on the lesion blood vessel by adopting a self-adaptive weight sparse representation classification algorithm. Traditional sparse representation algorithm minimizes residual error
Figure GDA0004238635320000045
) The sparse representation vector of the target blood vessel is searched, and local characteristics of variable lesion blood vessels and low robustness bring huge calculation amount to the minimization of residual errors, so that the traditional sparse representation algorithm cannot converge when detecting the lesion blood vessels. Aiming at the difficulty, the invention guides the optimization solving process of the sparse representation vector by adding the self-adaptive weight related to the characteristic distance to each information block, so that the use of the multi-scale dictionary is more efficient, and the accurate detection of the lesion blood vessel is realized. The process of the adaptive weight sparse representation is as follows:
Figure GDA0004238635320000042
wherein p is a target blood vessel information block to be detected; representing matrix element multiplication; w represents a block information base d s Is used for the adaptive weight vector of the (a); w is a weight-dependent mapping matrix; a, a i And a j For block information base d s Any two information blocks; lambda (lambda) 1 、λ 2 And lambda (lambda) 3 Is a non-negative influence factor parameter;
Figure GDA0004238635320000043
for normalization of the mapping matrix W, the block information base d may be reduced s Characteristic distance of similar information blocks.
According to the sparse representation residual error of the target information block p, defining the similarity between p and a blood vessel, wherein the similarity is shown in the following formula:
Figure GDA0004238635320000044
wherein beta is a sparse representation coefficient obtained by optimizing the formula (2). When the similarity between the target information block and the blood vessel is high, the target information block p is considered as a part of the blood vessel, and classification of the target information block is realized. So far, the detection problem of the lesion blood vessel is reduced to the optimization solving problem of the formula (2).
The collecting process of the target information block p to be detected is consistent with the collecting process of the training set information block, namely along the section direction of the axis of the blood vessel, so that the target information block forms a sequence { p to be detected 0 ,...,p n-1 ,p n ,...}. Due to the proximity of the information blocks to be detected, the similarity between adjacent information blocks is p n-1 Can guide p n Is described. Therefore, the optimization solving process of the formula (2) is divided into the following five steps: 1) At p n-1 Scale s of (2) n-1 Vicinity search p n Is the optimal scale s of (2) n The method comprises the steps of carrying out a first treatment on the surface of the 2) At parameter w n-1 、W n-1 Sum s n Under certain conditions, the solving parameter beta is optimized n . In this case, the formula (2) is converted into a simple conventional sparse representation optimization solution problem, and an optimal solution can be easily obtained:
Figure GDA0004238635320000051
wherein->
Figure GDA0004238635320000052
G is X n Is a diagonal matrix of (a); 3) At parameter W n-1 、β n Sum s n Under the determined condition, optimizing the solving parameter w n . In this case, the Lagrangian function of equation (2) satisfies the Karush-Kuhn-Tucker synchronous convergence condition, thereby obtaining w n Is the optimal solution of (a):
Figure GDA0004238635320000053
wherein->
Figure GDA0004238635320000054
ε m-l+1 M-l+1 element of ε, m is β n The number of elements in (1) is beta n The number of non-zero elements in the matrix; 4) At the position ofParameter w n 、β n Sum s n Under the determined condition, optimizing the solving parameter W n . In this case, equation (2) becomes an unconstrained optimization problem, which can be solved using a conventional CG Schmidt (2005) optimization operator; 5) At parameter w n 、W n Sum s n Under the determined condition, the optimization solving process of the second step is repeated, and the updating parameter beta is optimized n 。
By adding the self-adaptive weight related to the feature distance to each information block in the library, the optimization solving process of the sparse representation vector is guided, so that the use of the multi-scale dictionary is more efficient, and the accurate detection of the lesion blood vessel is realized.
The method adopts a blood vessel axis guiding constraint segmentation strategy, adopts a multi-scale training set in the training process, and realizes accurate segmentation while considering time complexity. Experimental results show that the accuracy rate of the method for segmenting the lesion blood vessels is up to 91%, and the method is an accurate and efficient lesion blood vessel segmentation method.
The foregoing description is only of the preferred embodiments of the present invention, and is not intended to limit the invention in any way, and any simple modification, equivalent variation and modification made to the above embodiments according to the technical substance of the present invention falls within the scope of the technical solution of the present invention.

Claims (2)

1. The pathological blood vessel accurate detection method based on block matching self-adaptive weight sparse representation is characterized by comprising the following steps of:
step 1: generating a multi-scale vascular training set: regarding the vascular structure as a linear structure in a human body, constructing a multidimensional linear structure filter based on Gaussian convolution by utilizing a hessian matrix, and converting the vascular training set into a training set related to the vascular scale by applying the constructed multidimensional linear structure filter on the basis of a traditional semi-automatic segmentation vascular training set;
step 2: extracting the central axis of a lesion blood vessel by utilizing a multi-scale hierarchical blood vessel tracking algorithm, extracting a blood vessel information block along the section direction of the axis by taking the maximum scale of the blood vessel as a radius, and constructing a multi-scale block information base, wherein the size of the information block is 7 multiplied by 7 voxels, and the scale of the information block is determined by the scale of the extracted target blood vessel;
step 3: based on multi-scale dictionary learning of sparse representation, self-adaptively capturing local lesion vascular features from a block information base through a sparse representation algorithm;
step 4: and carrying out accurate detection on lesion blood vessels by adopting a self-adaptive weight sparse representation classification algorithm:
Figure FDA0004238635310000011
wherein p is a target blood vessel information block to be detected; representing matrix element multiplication; w represents a block information base d s Is used for the adaptive weight vector of the (a); w is a weight-dependent mapping matrix; a, a i And a j For block information base d s Any two information blocks; lambda (lambda) 1 、λ 2 And lambda (lambda) 3 Is a non-negative influence factor parameter;
Figure FDA0004238635310000012
for normalization of the mapping matrix W, the block information base d may be reduced s Characteristic distance of the medium similar information block; according to the sparse representation residual error of the target information block p, defining the similarity between p and a blood vessel, wherein the similarity is shown in the following formula:
Figure FDA0004238635310000013
wherein beta is a sparse representation coefficient obtained by optimizing the formula (2), and when the similarity between the target information block and the blood vessel is high, the target information block p is considered as a part of the blood vessel, so that the classification of the target information block is realized; the collecting process of the target information block p to be detected is consistent with the collecting process of the training set information block, namely along the section direction of the axis of the blood vessel, so that the target information block forms a sequence { p to be detected 0 ,...,p n-1 ,p n ,., proximity due to proximity in the location of the information block to be detectedThe similarity between the information blocks is such that p n-1 Can guide p n Therefore, the optimization solving process of the formula (2) is divided into the following five steps: 1) At p n-1 Scale s of (2) n-1 Vicinity search p n Is the optimal scale s of (2) n The method comprises the steps of carrying out a first treatment on the surface of the 2) At parameter w n-1 、W n-1 Sum s n Under certain conditions, the solving parameter beta is optimized n In this case, the formula (2) is converted into a simple conventional sparse representation optimization solution problem, and an optimal solution can be obtained easily:
Figure FDA0004238635310000021
wherein->
Figure FDA0004238635310000022
G is X n Is a diagonal matrix of (a); 3) At parameter W n-1 、β n Sum s n Under the determined condition, optimizing the solving parameter w n In this case, the Lagrangian function of equation (2) satisfies the Karush-Kuhn-Tucker synchronous convergence condition, thereby obtaining w n Is the optimal solution of (a):
Figure FDA0004238635310000023
wherein->
Figure FDA0004238635310000024
ε m-l+1 M-l+1 element of ε, m is β n The number of elements in (1) is beta n Number of non-zero elements->
Figure FDA0004238635310000025
For a full 1 matrix, the matrix size is consistent with the weight matrix wn; 4) At parameter w n 、β n Sum s n Under the determined condition, optimizing the solving parameter W n In this case, equation (2) becomes an unconstrained optimization problem, which can be solved using a conventional CG Schmidt (2005) optimization operator; 5) At parameter w n 、W n Sum s n In the case of determination, repeat the secondOptimizing and solving process and optimizing updating parameter beta n 。
2. The accurate detection method for the lesion blood vessel based on the block matching self-adaptive weight sparse representation is characterized by comprising the following steps of: the process of the step 3 is as follows:
Figure FDA0004238635310000026
wherein d is s The scale is s, beta is a sparse representation coefficient vector, p is an information block in a training set, and d s And the optimization solution of beta is given by a sparse representation algorithm, the residual error of the sparse representation determines the characterizations of the information blocks, and the information blocks with high characterizations are put into a multi-scale dictionary, so that the capture of the local characteristics of the lesion blood vessels is realized.
CN201910669052.0A 2019-07-24 2019-07-24 Pathological blood vessel accurate detection method based on block matching self-adaptive weight sparse representation Active CN110400300B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910669052.0A CN110400300B (en) 2019-07-24 2019-07-24 Pathological blood vessel accurate detection method based on block matching self-adaptive weight sparse representation

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910669052.0A CN110400300B (en) 2019-07-24 2019-07-24 Pathological blood vessel accurate detection method based on block matching self-adaptive weight sparse representation

Publications (2)

Publication Number Publication Date
CN110400300A CN110400300A (en) 2019-11-01
CN110400300B true CN110400300B (en) 2023-06-23

Family

ID=68325802

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910669052.0A Active CN110400300B (en) 2019-07-24 2019-07-24 Pathological blood vessel accurate detection method based on block matching self-adaptive weight sparse representation

Country Status (1)

Country Link
CN (1) CN110400300B (en)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112184840B (en) * 2020-09-22 2022-09-02 上海交通大学 3D point cloud compression system based on multi-scale structured dictionary learning
LU102391B1 (en) * 2021-01-07 2022-04-28 Harbin Inst Technology Weihai Method for accurate detection of blood vessels with pathological changes based on adaptive weighted sparse representation of block matching

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105678343A (en) * 2015-08-25 2016-06-15 浙江工业大学 Adaptive-weighted-group-sparse-representation-based diagnosis method for noise abnormity of hydroelectric generating set
CN109492570A (en) * 2018-11-02 2019-03-19 电子科技大学 A kind of SAR image target recognition method based on multiple dimensioned rarefaction representation

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
GB0305315D0 (en) * 2003-03-07 2003-04-09 Weber Martin Image processing system
US9734601B2 (en) * 2014-04-04 2017-08-15 The Board Of Trustees Of The University Of Illinois Highly accelerated imaging and image reconstruction using adaptive sparsifying transforms
WO2016154495A2 (en) * 2015-03-25 2016-09-29 Oregon Health & Science University Systems and methods of choroidal neovascularization detection using optical coherence tomography angiography

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105678343A (en) * 2015-08-25 2016-06-15 浙江工业大学 Adaptive-weighted-group-sparse-representation-based diagnosis method for noise abnormity of hydroelectric generating set
CN109492570A (en) * 2018-11-02 2019-03-19 电子科技大学 A kind of SAR image target recognition method based on multiple dimensioned rarefaction representation

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
王阳 ; 夏利民 ; 王嘉 ; .基于加权分块稀疏表示的步态识别.计算机工程与设计.(第08期),全文. *
薛文渲 ; 刘建霞 ; 刘然 ; 袁晓辉 ; .改进U型网络的眼底视网膜血管分割方法.光学学报.(第12期),全文. *

Also Published As

Publication number Publication date
CN110400300A (en) 2019-11-01

Similar Documents

Publication Publication Date Title
Wang et al. An improved dice loss for pneumothorax segmentation by mining the information of negative areas
Wu et al. Handwritten mathematical expression recognition via paired adversarial learning
Da Nobrega et al. Lung nodule malignancy classification in chest computed tomography images using transfer learning and convolutional neural networks
Keetha et al. U-Det: A modified U-Net architecture with bidirectional feature network for lung nodule segmentation
Alex et al. Semisupervised learning using denoising autoencoders for brain lesion detection and segmentation
CN115171838B (en) Training method of medical report generation model based on cross-modal fusion
CN110400300B (en) Pathological blood vessel accurate detection method based on block matching self-adaptive weight sparse representation
Cong et al. Boundary guided semantic learning for real-time COVID-19 lung infection segmentation system
Chung et al. Deeply self-supervised contour embedded neural network applied to liver segmentation
Yu et al. Break: Bronchi reconstruction by geodesic transformation and skeleton embedding
Sargin et al. Variable length open contour tracking using a deformable trellis
Wang et al. A 3D tubular flux model for centerline extraction in neuron volumetric images
Date et al. Brain image segmentation algorithm using K-means clustering
Howe Inkball models for character localization and out-of-vocabulary word spotting
Liu et al. U2F-GAN: weakly supervised super-pixel segmentation in thyroid ultrasound images
Das et al. AWOLSE: Adaptive weight optimized level set evolution-based blood cell segmentation
Meng et al. Research on the ROI registration algorithm of the cardiac CT image time series
CN114358191A (en) Gene expression data clustering method based on depth automatic encoder
Ding et al. Deep Multi-Instance Learning with Adaptive Recurrent Pooling for Medical Image Classification
Sun et al. 3D segmentation of lungs with juxta-pleural tumor using the improved active shape model approach
Zhao et al. Research on automatic detection algorithm of pulmonary nodules based on deep learning
Chai et al. Automatic computer aided analysis algorithms and system for adrenal tumors on CT images
CN114240844B (en) Unsupervised key point positioning and target detection method in medical image
Liu Coarse to Fine Framework for Kidney Tumor Segmentation
Wang et al. Relevant shape contour snippet extraction with metadata supported hidden markov models

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant