CN116580008B - Biomedical marking method based on local augmentation space geodesic - Google Patents
Biomedical marking method based on local augmentation space geodesic Download PDFInfo
- Publication number
- CN116580008B CN116580008B CN202310554646.3A CN202310554646A CN116580008B CN 116580008 B CN116580008 B CN 116580008B CN 202310554646 A CN202310554646 A CN 202310554646A CN 116580008 B CN116580008 B CN 116580008B
- Authority
- CN
- China
- Prior art keywords
- key point
- geodesic
- point
- radius
- matrix
- 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
Links
- 238000000034 method Methods 0.000 title claims abstract description 20
- 230000003416 augmentation Effects 0.000 title claims abstract description 11
- 210000004204 blood vessel Anatomy 0.000 claims abstract description 24
- 230000002792 vascular Effects 0.000 claims abstract description 12
- 210000001210 retinal vessel Anatomy 0.000 claims abstract description 11
- 230000003190 augmentative effect Effects 0.000 claims abstract 5
- 238000010586 diagram Methods 0.000 claims description 70
- 230000002207 retinal effect Effects 0.000 claims description 34
- 239000011159 matrix material Substances 0.000 claims description 29
- 238000012360 testing method Methods 0.000 claims description 24
- 210000001525 retina Anatomy 0.000 claims description 22
- 238000012549 training Methods 0.000 claims description 9
- 238000013507 mapping Methods 0.000 claims description 7
- 238000012795 verification Methods 0.000 claims description 4
- GNFTZDOKVXKIBK-UHFFFAOYSA-N 3-(2-methoxyethoxy)benzohydrazide Chemical compound COCCOC1=CC=CC(C(=O)NN)=C1 GNFTZDOKVXKIBK-UHFFFAOYSA-N 0.000 claims description 3
- FGUUSXIOTUKUDN-IBGZPJMESA-N C1(=CC=CC=C1)N1C2=C(NC([C@H](C1)NC=1OC(=NN=1)C1=CC=CC=C1)=O)C=CC=C2 Chemical compound C1(=CC=CC=C1)N1C2=C(NC([C@H](C1)NC=1OC(=NN=1)C1=CC=CC=C1)=O)C=CC=C2 FGUUSXIOTUKUDN-IBGZPJMESA-N 0.000 claims description 3
- 230000009471 action Effects 0.000 claims description 3
- 239000008280 blood Substances 0.000 claims description 3
- 210000004369 blood Anatomy 0.000 claims description 3
- 239000003550 marker Substances 0.000 claims description 3
- 238000005259 measurement Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 abstract description 5
- 238000013528 artificial neural network Methods 0.000 abstract description 2
- 230000006870 function Effects 0.000 description 6
- 238000013527 convolutional neural network Methods 0.000 description 4
- 238000001514 detection method Methods 0.000 description 4
- 238000000605 extraction Methods 0.000 description 3
- 230000007547 defect Effects 0.000 description 2
- 238000003709 image segmentation Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000010200 validation analysis Methods 0.000 description 2
- 230000001174 ascending effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 210000000481 breast Anatomy 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000003759 clinical diagnosis Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Classifications
-
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
- G06N3/0455—Auto-encoder networks; Encoder-decoder networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
- G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/776—Validation; Performance evaluation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- 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
- G06T2207/30041—Eye; Retina; Ophthalmic
-
- 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
-
- 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
- G06T2207/30204—Marker
-
- 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
- G06T2207/30242—Counting objects in image
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Medical Informatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Mathematical Physics (AREA)
- General Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Geometry (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Quality & Reliability (AREA)
- Eye Examination Apparatus (AREA)
Abstract
A geodesic biomedical marking method based on local augmentation space introduces a geodesic voting algorithm on the basis of a U-Net neural network, ensures the integrity and accuracy of detected key points, and increases the radius dimension in the geodesic voting algorithm so as to extract more accurate vessel center lines. The vascular biomedical mark calculation method provided by the invention can use a small amount of data sets to extract key points of retinal blood vessels through U-Net, then use a local radius augmented geodesic voting algorithm to track the blood vessel structure, more accurately detect the key points and calculate other vascular biomedical marks.
Description
Technical Field
The invention relates to the field of computer vision, in particular to a local augmentation space earth wire based biomedical marking method.
Background
With the widespread use of biomedical images, particularly in research and clinical diagnosis of vascular structures, there is an increasing need for automatic calculation of vascular biomedical markers. However, conventional manual marking and calculation methods are generally time consuming, labor intensive and subject to subjective factors from the operator, and thus a computer-based method is needed to improve efficiency and accuracy.
In recent years, deep learning techniques, particularly neural network-based methods, have achieved significant results in the field of medical image processing. Harry Pratt et al (ref. P.Harry, W.Bryan, K.Jae, et al Automatic Detection and Distinction of Retinal Vessel Bifurcations and Crossings in Colour Fundus Photography [ J ]. Journal of Imaging,2017,4 (1): 4) propose a method of classification by supervised learning using CNN convolutional neural networks for automatic detection and classification of bifurcation intersections for fundus images. However, the extraction method based on CNN only focuses on the local features of the image, and ignores the global features of the image. And only by means of the feature points extracted based on CNN, leakage problems are easy to occur, and the accuracy of image segmentation is unstable. The network of the Unet (ref. O.Ronneberger, P.Fischer, and T. Brox, "U-net: convolutional networks for biomedical image segmentation," in International Conference on Medical Image Computing and Computer-Assisted Intervention,2015, pp. 234-241.) was proposed by Ronneberger et al in 2016, which is a very lightweight network and can fuse different levels of features enabling end-to-end automatic feature learning and prediction on samples. The immobilized structure and the small sample nature of the medical image make the U-Net network become the best model for extracting the characteristics of the medical image, but the U-Net network may cause the defect of key points during detection due to the problems of image noise and interference during extraction.
Furthermore, Y.Rouchdy and L.D.Cohen propose geodesic voting algorithms (reference: Y.Rouchdy and L.D.Cohen, "Geodesic voting for the automatic extraction of tree structures. Methods and applications," in Computer Vision and Image Understanding, volume,2013,117 (10): 1453-1467) as an efficient method for extracting a tree-like tubular structure, which requires manual source points in application, and thus has low stability in fine branch detection.
Disclosure of Invention
In order to overcome the defects of the technology, the invention provides a method for detecting key points more accurately and calculating other vascular biomedical markers.
The technical scheme adopted for overcoming the technical problems is as follows:
a biomedical marking method based on local augmentation space geodesic comprises the following steps:
a) A retinal fundus image dataset T is acquired,wherein T is i For the ith retinal fundus image, i e {1, 2., n. 1 },n 1 The total number of retinal fundus images;
b) Performing data enhancement operation on the retina fundus image dataset T to obtain an enhanced retina fundus image dataset T',T i ' is the retina fundus image after the ith enhancement,i∈{1,2,...,n 2 },n 2 The total number of retina fundus images after enhancement;
c) Dividing the enhanced retina fundus image data set T' into a training set train, a verification set val and a test set test;
d) The enhanced retinal fundus image in the training set train is input into a U-Net network, the U-Net network is trained to obtain a trained U-Net network, and the first enhanced retinal fundus image T in the test set test is input into the U-Net network l ' input into a trained U-Net network to obtain a vascular key point diagram I, l epsilon {1,2,., n 3 },n 3 Obtaining a key point set D and a key point coordinate set from the key point diagram I for the total number of the retina fundus images after enhancement in the test set test,D={x 1 ,x 2 ,...,x j ,...,x N X, where x j For the jth key point in the blood vessel key point diagram I, j epsilon {1, 2.,. N }, N is the total number of key points in the key point diagram I, and +.>Wherein (1)>Is the j-th key point x j Coordinates of->Is the j-th key point x j Is>Is the j-th key point x j Ordinate of>Omega is the image domain;
e) Inputting the key point set D as a source point set into a local radius augmentation geodesic voting algorithm, and calculating to obtain a geodesic lineDiameter set { R 1 ,R 2 ,...,R j ,...,R N };
f) Calculate the total geodesic density mu t (x i ,r);
g) According to the total geodesic density mu t (x i R) calculating the geodesic space densityAnd maximum value of total geodesic density +.>
h) Calculating the radius of retinal blood vessels in the key point diagram IAnd vessel centerline->
i) Calculating to obtain a blood vessel end point set C and a bifurcation cross point set
j) Obtaining the radius of retinal blood vesselVascular centerline->Blood vessel endpoint set C and bifurcation intersection set +.>The constituted retinal vascular biomedical marker.
Preferably, in step a), 40 retinal fundus images are selected from the blood vessel keypoint disclosure dataset DRIVE to form a retinal fundus image dataset T.
Further, in the step b), the OpenCV packet in python is used to sequentially perform operations of flipping, rotating and contrast changing on the retinal fundus image dataset T, so as to obtain an enhanced retinal fundus image dataset. Preferably, in step c), the enhanced retinal fundus image dataset T' is divided into a training set train, a validation set val, and a test set test in a ratio of 7:2:1.
Further, step e) comprises the steps of:
e-1) establishing an ith pixel point x in the image domain omega in the key point diagram I i Taking a sphere with r as a radius as a center, I epsilon {1, 2.. The number of pixels of the key point diagram I is equal to Q, and r epsilon [0, r ] max ],r max Is the maximum radius;
e-2) passing through the formulaCalculated to obtain a potential function P (x i R), wherein ω is a real normal number weight, λ 1 And lambda is 2 Are all positive real weights, beta is a real positive constant, m (x i R) is the pixel point x in the image domain omega of the key point diagram I i As center, average value, sigma of sphere with radius r 2 (x i R) is the pixel point x in the image domain omega of the key point diagram I i The j-th key point x is taken as the center and the variance of the sphere with r as the radius j As the source point, m 0 To use the source point x in the image domain omega of the key point diagram I j Centering on the mean value of the sphere with radius r, +.>In the critical point diagram I image domain omega, the source point x is used j As a center, the variance of a sphere with r as a radius;
e-3) establishing an endpoint set V, V= { x of all M points on the boundary of the key point diagram I 1 ,x 2 ,...,x k ,...,x M },x k K e {1, 2..m } for the kth point on the boundary of key point map I;
e-4) passing through the formulaCalculated to obtainMinimal action figure->In the formula, I and II are norms, and the formula is>Is a gradient;
e-5) calculating the kth point x in the endpoint set V using a fast-marching algorithm k To the source point x j Is measured by the ground wire path y k Obtaining a source point x j Corresponding geodesic path set R j ={y 1 ,y 2 ,...,y k ,...,y M };
e-6) repeating the step e-5) until all the key points in the key point set D are used as source points to obtain a geodesic path set { R } 1 ,R 2 ,...,R j ,...,R N }。
Preferably, in step e-1) r max =15。
Further, in step f), the formula is passedCalculating the ground wire density mu j (x i R) if the geodesic path y k And the ith pixel point x in the image domain omega of the key point diagram I i As a center, spheres with radius r intersect, then +.>Geodesic path y k And the ith pixel point x in the image domain omega of the key point diagram I i As a center, spheres with radius r do not intersect, then by the formula +.>By the formula->Calculating to obtain an ith pixel point x in the image domain omega of the key point diagram I i As the center, the total geodesic density mu in the sphere with r as the radius t (x i ,r)。
Further, step g) comprises the steps of:
g-1) is represented by the formulaCalculating to obtain an ith pixel point x in the image domain omega of the key point diagram I i For the center, the space density of the geodesic in the sphere with r as radius is +.>
g-2) is represented by the formulaCalculating to obtain an ith pixel point x in the image domain omega of the key point diagram I i As center, maximum value +.f of total geodesic density in sphere with radius r>
Further, step h) comprises the steps of:
h-1) when the total geodesic density is at maximumAbove threshold Th, by the formulaCalculating to obtain the radius of the retinal blood vessel in the key point diagram I>Wherein
h-2) maximum value of geodesic density according to the sum of measurements on the key point diagram IMarking total geodesic densityMaximum value->Pixels larger than threshold Th, each marked pixel constitutes vessel center line +.>Further, step i) comprises the steps of:
i-1) constructing a matrix H with the same size as the key point diagram I, and collecting the key point coordinatesMapping each coordinate in the matrix to a matrix H one by one, setting the value of all mapping coordinates to 0.5, and setting the rest value to 0;
i-2) constructing a matrix of the same size as the key point diagram IIn the key dot map I image domain Ω, a maximum value according to the total geodesic density +.>Pixel point x greater than threshold Th i Coordinate of (2) pixel point x i Density of each space atMapping one to matrix->In matrix->Pixel point x is divided internally i The values on the other coordinates are set to 0 except the corresponding coordinates;
i-3) matrixThe values on are mapped to [0,1 ] by normalization]Between, get matrix->
i-4) by the formulaCalculating to obtain matrix->When matrix->When the value of the element in the matrix is more than or equal to 0.5 and less than 1, the matrix is +.>The pixel points in the key point diagram I corresponding to the elements in the middle are used as the retinal blood vessel end points to obtain a blood vessel end point set C, and the matrix is +.>When the value of the element in the matrix is more than or equal to 1, the matrix is +.>The pixel points in the key point diagram I corresponding to the elements in the middle are taken as bifurcation points or crossing points of blood vessels to obtain a bifurcation crossing point set +.>
The beneficial effects of the invention are as follows: the biomedical mark of the blood vessel is automatically calculated and extracted in the biomedical image, so that the subjective factors of traditional manual work and the workload of operators are reduced, the calculation and marking efficiency is improved, and the more accurate and rapid calculation of the biomedical mark of the blood vessel is realized. The U-Net network is utilized to perform feature learning and prediction, end-to-end automatic processing can be performed on the retina blood vessel sample, the spatial geodesic voting algorithm with local radius augmentation can accurately position key points of a blood vessel structure and track blood vessels, and the accuracy and stability of vascular biomedical markers are improved. The invention is suitable for various biomedical images, such as retina images, cerebrovascular images, breast images and the like, and has wide application prospect and potential clinical application value.
Drawings
FIG. 1 is a flow chart of the method of the present invention.
Detailed Description
The invention is further described with reference to fig. 1.
A biomedical marking method based on local augmentation space geodesic comprises the following steps:
a) A retinal fundus image dataset T is acquired,wherein T is i For the ith retinal fundus image, i e {1, 2., n. 1 },n 1 Is the total number of retinal fundus images.
b) Performing data enhancement operation on the retina fundus image dataset T to obtain an enhanced retina fundus image dataset T',T i ' is the retina fundus image after the i-th enhancement, i e {1,2,., n 2 },n 2 To the total number of retinal fundus images after enhancement.
c) The enhanced retinal fundus image dataset T' is divided into a training set train, a validation set val, and a test set test.
d) The enhanced retinal fundus image in the training set train is input into a U-Net network, the U-Net network is trained to obtain a trained U-Net network, and the first enhanced retinal fundus image T in the test set test is input into the U-Net network l ' input into a trained U-Net network to obtain a vascular key point diagram I, l epsilon {1,2,., n 3 },n 3 Obtaining a key point set D and a key point coordinate set from the key point diagram I for the total number of the retina fundus images after enhancement in the test set test,D={x 1 ,x 2 ,...,x j ,...,x N X, where x j For the jth key point in the blood vessel key point diagram I, j epsilon {1, 2.,. N }, N is the total number of key points in the key point diagram I, and +.>Wherein (1)>Is the j-th key point x j Coordinates of->Is the j-th key point x j Is>Is the j-th key point x j Ordinate of>Omega is the image domain.
e) Inputting the key point set D as a source point set into a local radius augmentation geodesic voting algorithm, and calculating to obtain a geodesic path set { R } 1 ,R 2 ,...,R j ,…,R N }。
f) Calculate the total geodesic density mu t (x i ,r)。
g) According to the total geodesic density mu t (x i R) calculating the geodesic space densityAnd maximum value of total geodesic density +.>
h) Calculating the radius of retinal blood vessels in the key point diagram IAnd vessel centerline->。
i) Calculating to obtain a blood vessel end point set C and a bifurcation cross point set。
j) Obtaining the radius of retinal blood vesselVascular centerline->Blood vessel endpoint set C and bifurcation intersection set +.>The constituted retinal vascular biomedical marker.
Example 1:
in the step a), 40 retinal fundus images are selected from a blood vessel key point open data set DRIVE to form a retinal fundus image data set T. And b), sequentially performing flipping, rotation and contrast change operations on the retina fundus image dataset T by using an OpenCV package in python to obtain an enhanced retina fundus image dataset.
Example 2:
in the step c), the enhanced retina fundus image data set T' is divided into a training set train, a verification set val and a test set test according to the proportion of 7:2:1.
Example 3:
step e) comprises the steps of:
e-1) in order to better obtain biomedical marking information, the potential function is used for completing geodesic voting, the traditional potential function is calculated point by point in the image domain, and the ith pixel point x in the image domain omega of the key point diagram I is established by utilizing the ascending dimension thought i A sphere centered on r, i e {1,2, …, Q },q is the total number of pixel points of the key point diagram I, r is E [0, r max ],r max At the maximum radius x i e.OMEGA. The radius r of the sphere is integrated as an additional dimension into the potential function.
e-2) passing through the formulaCalculated to obtain a potential function P (x i R), wherein ω is a real normal number weight, λ 1 And lambda is 2 Are all positive real weights, beta is a real positive constant, m (x i R) is the key point x in the image domain omega of the key point diagram I i As center, average value, sigma of sphere with radius r 2 (x i R) is the key point x in the image domain omega of the key point diagram I i The j-th key point x is taken as the center and the variance of the sphere with r as the radius j As the source point, m 0 To use the source point x in the image domain omega of the key point diagram I j Centering on the mean value of the sphere with radius r, +.>In the critical point diagram I image domain omega, the source point x is used j As a center, the variance of a sphere with radius r. Potential function P (x i R) the key point diagram I can be divided into pixel points x i The values of all pixels within the sphere centered at radius r are merged together.
e-3) establishing an endpoint set V, V= { x of all M points on the boundary of the key point diagram I 1 ,x 2 ,...,x k ,...,x M },x k K e {1,2, …, M } is the kth point on the boundary of key point diagram I.
e-4) passing through the formulaCalculating to obtain the minimum action figure->Wherein ||· | is the norm ++>Is a gradient.
e-5) calculating the kth point x in the endpoint set V using a fast-marching algorithm k To the source point x j Is measured by the ground wire path y k Obtaining a source point x j Corresponding geodesic path set R j ={y 1 ,y 2 ,…,y k ,…,y M }。
e-6) repeating the step e-5) until all the key points in the key point set D are used as source points to obtain a geodesic path set { R } 1 ,R 2 ,...,R j ,...,R N }。
In this embodiment, it is preferred that r in step e-1) max =15。
Example 4:
in step f) by the formulaCalculating the ground wire density mu j (x i R) if the geodesic path y k And the ith pixel point x in the image domain omega of the key point diagram I i As a center, spheres with radius r intersect, delta xi,r (y k ) =1, geodesic path y k And the ith pixel point x in the image domain omega of the key point diagram I i Centering on, the spheres with radius r do not intersect, delta xi,r (y k ) =0. Wherein, there are N source points in the key point set D. Thus passing through the formulaCalculating to obtain an ith pixel point x in the image domain omega of the key point diagram I i As the center, the total geodesic density mu in the sphere with r as the radius t (x i ,r)。
Example 5:
step g) comprises the steps of:
g-1) is represented by the formulaCalculating to obtain an ith pixel point x in the image domain omega of the key point diagram I i For the center, the space density of the geodesic in the sphere with r as radius is +.>
g-2) is represented by the formulaCalculating to obtain an ith pixel point x in the image domain omega of the key point diagram I i As center, maximum value +.f of total geodesic density in sphere with radius r>
Example 6:
step h) comprises the steps of:
h-1) when the total geodesic density is at maximumAbove threshold Th, by the formulaCalculating to obtain the radius of the retinal blood vessel in the key point diagram I>Wherein
h-2) maximum value of geodesic density according to the sum of measurements on the key point diagram IMark total geodesic density maximum +.>Pixels larger than threshold Th, each marked pixel constitutes vessel center line +.>
Step i) comprises the steps of:
i-1) constructing a matrix H with the same size as the key point diagram I, and collecting the key point coordinatesThe coordinates of the map are mapped to the matrix H one by one, the values of all the mapped coordinates are set to 0.5, and the remaining values are set to 0.
I-2) constructing a matrix of the same size as the key point diagram IIn the key dot map I image domain Ω, a maximum value according to the total geodesic density +.>Pixel point x greater than threshold Th i Coordinate of (2) pixel point x i Density of each space atMapping one to matrix->In matrix->Pixel point x is divided internally i The values on the remaining coordinates are set to 0 except for the corresponding coordinates.
i-3) matrixThe values on are mapped to [0,1 ] by normalization]Between, get matrix->
i-4) by the formulaCalculating to obtain matrix->When matrix->When the value of the element in the matrix is more than or equal to 0.5 and less than 1, the matrix is +.>The pixel points in the key point diagram I corresponding to the elements in the middle are used as the retinal blood vessel end points to obtain a blood vessel end point set C, and the matrix is +.>When the value of the element in the matrix is more than or equal to 1, the matrix is +.>The pixel points in the key point diagram I corresponding to the elements in the middle are taken as bifurcation points or crossing points of blood vessels to obtain a bifurcation crossing point set +.>
Finally, it should be noted that: the foregoing description is only a preferred embodiment of the present invention, and the present invention is not limited thereto, but it is to be understood that modifications and equivalents of some of the technical features described in the foregoing embodiments may be made by those skilled in the art, although the present invention has been described in detail with reference to the foregoing embodiments. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (5)
1. The biomedical marking method based on local augmentation space geodesic is characterized by comprising the following steps:
a) A retinal fundus image dataset T is acquired,wherein T is i For the ith retinal fundus image, i e {1, 2., n. 1 },n 1 The total number of retinal fundus images;
b) Performing data enhancement operation on the retina fundus image dataset T to obtain an enhanced retina fundus image dataset T',T i ' is the retina fundus image after the i-th enhancement, i e {1,2,., n 2 },n 2 The total number of retina fundus images after enhancement;
c) Dividing the enhanced retina fundus image data set T' into a training set train, a verification set val and a test set test;
d) The enhanced retinal fundus image in the training set train is input into a U-Net network, the U-Net network is trained to obtain a trained U-Net network, and the first enhanced retinal fundus image T in the test set test is input into the U-Net network l ' input into a trained U-Net network to obtain a vascular key point diagram I, l epsilon {1,2,., n 3 },n 3 Obtaining a key point set D and a key point coordinate set from the key point diagram I for the total number of the retina fundus images after enhancement in the test set testD={x 1 ,x 2 ,…,x j ,…,x N X, where x j For the jth key point in the blood vessel key point diagram I, j epsilon {1,2, …, N }, N is the total number of key points in the key point diagram I, { I }>Wherein (1)>Is the j-th key point x j Coordinates of->Is the j-th key point x j Is>Is the j-th key point x j Ordinate of>Omega is the image domain;
e) Inputting the key point set D as a source point set into a local radius augmentation geodesic voting algorithm, and calculating to obtain a geodesic path set { R } 1 ,R 2 ,...,R j ,...,R N };
f) Calculate the total geodesic density mu t (x i ,r);
g) According to the total geodesic density mu t (x i R) calculating the geodesic space densityAnd maximum value of total geodesic density +.>
h) Calculating the radius of retinal blood vessels in the key point diagram IAnd vessel centerline->
i) Calculating to obtain a blood vessel end point set C and a bifurcation cross point set
j) Obtaining the radius of retinal blood vesselVascular centerline->Blood vessel end point set C and bifurcation intersection point set +.>A constituted retinal vascular biomedical marker;
step e) comprises the steps of:
e-1) establishing an ith pixel point x in the image domain omega in the key point diagram I i Taking r as a radius sphere as a center, I epsilon {1,2, …, Q }, Q as the total number of pixel points of the key point diagram I, r epsilon [0, r ] max ],r max Is the maximum radius;
e-2) passing through the formulaCalculated to obtain a potential function P (x i R), wherein ω is a real normal number weight, λ 1 And lambda is 2 Are all positive real weights, beta is a real positive constant, m (x i R) is the pixel point x in the image domain omega of the key point diagram I i As center, average value, sigma of sphere with radius r 2 (x i R) is the pixel point x in the image domain omega of the key point diagram I i The j-th key point x is taken as the center and the variance of the sphere with r as the radius j As the source point, m 0 To use the source point x in the image domain omega of the key point diagram I j Centering on the mean value of the sphere with radius r, +.>In the critical point diagram I image domain omega, the source point x is used j As a center, the variance of a sphere with r as a radius;
e-3) establishing an endpoint set V, V= { x of all M points on the boundary of the key point diagram I 1 ,x 2 ,…,x k ,…,x M },x k K is {1,2, …, M };
e-4) passing through the formulaCalculating to obtain the minimum action figure->Wherein ||· | is the norm ++>Is a gradient;
e-5) calculating the kth point x in the endpoint set V using a fast-marching algorithm k To the source point x j Is measured by the ground wire path y k Obtaining a source point x j Corresponding geodesic path set R j ={y 1 ,y 2 ,...,y k ,...,y M };
e-6) repeating the step e-5) until all the key points in the key point set D are used as source points to obtain a geodesic path set { R } 1 ,R 2 ,...,R j ,...,R N };
In step f) by the formulaCalculating the ground wire density mu j (x i R) if the geodesic path y k And the ith pixel point x in the image domain omega of the key point diagram I i As a center, the spheres with r as radius intersectGeodesic path y k And the ith pixel point x in the image domain omega of the key point diagram I i As a center, spheres with radius r do not intersect, then +.>By the formula->Calculating to obtain an ith pixel point x in the image domain omega of the key point diagram I i As the center, the total geodesic density mu in the sphere with r as the radius t (x i ,r);
Step g) comprises the steps of:
g-1) is represented by the formulaCalculating to obtain an ith pixel point x in the image domain omega of the key point diagram I i For the center, the space density of the geodesic in the sphere with r as radius is +.>
g-2) is represented by the formulaCalculating to obtain an ith pixel point x in the image domain omega of the key point diagram I i As center, maximum value +.f of total geodesic density in sphere with radius r>
Step h) comprises the steps of:
h-1) when the total geodesic density is at maximumAbove threshold Th, by the formulaCalculating to obtain the radius of the retinal blood vessel in the key point diagram I>Wherein the method comprises the steps of
h-2) at the key pointMaximum value of geodesic density according to total measurement on graph IMarking the total geodesic density maximumPixels larger than threshold Th, each marked pixel constitutes vessel center line +.>Step i) comprises the steps of:
i-1) constructing a matrix H with the same size as the key point diagram I, and collecting the key point coordinatesMapping each coordinate in the matrix to a matrix H one by one, setting the value of all mapping coordinates to 0.5, and setting the rest value to 0;
i-2) constructing a matrix of the same size as the key point diagram IIn the key point map I image field Ω, a maximum value according to the total geodesic density +.>Pixel point x greater than threshold Th i Coordinate of (2) pixel point x i The respective spatial density of the sites->Mapping one to matrix->In matrix->Pixel point x is divided internally i Corresponding seatOutside the label, the values on the rest coordinates are set to 0;
i-3) matrixThe values on are mapped to [0,1 ] by normalization]Between, get matrix->
i-4) by the formulaCalculating to obtain matrix->When matrix->When the value of the element in the matrix is more than or equal to 0.5 and less than 1, the matrix is +.>The pixel points in the key point diagram I corresponding to the elements in the middle are used as the retinal blood vessel end points to obtain a blood vessel end point set C, and the matrix is +.>When the value of the element in the matrix is more than or equal to 1, the matrix is +.>The pixel points in the key point diagram I corresponding to the elements in the middle are taken as bifurcation points or crossing points of blood vessels to obtain a bifurcation crossing point set +.>
2. The locally augmented space geodesic biomedical marking method according to claim 1, wherein: in the step a), 40 retinal fundus images are selected from a blood vessel key point open data set DRIVE to form a retinal fundus image data set T.
3. The locally augmented space geodesic biomedical marking method according to claim 1, wherein: in the step b), the OpenCV packet in python is used to sequentially perform operations of flipping, rotating and contrast changing on the retinal fundus image dataset T, so as to obtain an enhanced retinal fundus image dataset T'.
4. The locally augmented space geodesic biomedical marking method according to claim 1, wherein: in the step c), the enhanced retina fundus image data set T' is divided into a training set train, a verification set val and a test set test according to the proportion of 7:2:1.
5. The locally augmented space geodesic biomedical marking method according to claim 1, wherein: r in step e-1) max =15。
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310554646.3A CN116580008B (en) | 2023-05-16 | 2023-05-16 | Biomedical marking method based on local augmentation space geodesic |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310554646.3A CN116580008B (en) | 2023-05-16 | 2023-05-16 | Biomedical marking method based on local augmentation space geodesic |
Publications (2)
Publication Number | Publication Date |
---|---|
CN116580008A CN116580008A (en) | 2023-08-11 |
CN116580008B true CN116580008B (en) | 2024-01-26 |
Family
ID=87540880
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310554646.3A Active CN116580008B (en) | 2023-05-16 | 2023-05-16 | Biomedical marking method based on local augmentation space geodesic |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116580008B (en) |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110276381A (en) * | 2019-05-24 | 2019-09-24 | 中山大学 | A kind of Data Dimensionality Reduction Algorithm decomposed based on collimation geodesic curve with low-dimensional insertion |
CN111815562A (en) * | 2020-06-10 | 2020-10-23 | 三峡大学 | Retinal vessel segmentation method combining U-Net and self-adaptive PCNN |
CN112348826A (en) * | 2020-10-26 | 2021-02-09 | 陕西科技大学 | Interactive liver segmentation method based on geodesic distance and V-net |
CN112651933A (en) * | 2020-12-21 | 2021-04-13 | 山东省人工智能研究院 | Blood vessel segmentation method based on geodesic distance graph and engineering function equation |
WO2021253939A1 (en) * | 2020-06-18 | 2021-12-23 | 南通大学 | Rough set-based neural network method for segmenting fundus retinal vascular image |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
ES2806351T3 (en) * | 2017-08-08 | 2021-02-17 | Adas3D Medical Sl | Computer-Implemented Procedure to Calculate Indicative Values for the Local Spatial Structure of Conductive Properties of Cardiac Muscle Tissue and Associated Computer Programs |
US11380061B2 (en) * | 2020-06-11 | 2022-07-05 | Samsung Electronics Co., Ltd. | Method and apparatus for three-dimensional (3D) object and surface reconstruction |
-
2023
- 2023-05-16 CN CN202310554646.3A patent/CN116580008B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110276381A (en) * | 2019-05-24 | 2019-09-24 | 中山大学 | A kind of Data Dimensionality Reduction Algorithm decomposed based on collimation geodesic curve with low-dimensional insertion |
CN111815562A (en) * | 2020-06-10 | 2020-10-23 | 三峡大学 | Retinal vessel segmentation method combining U-Net and self-adaptive PCNN |
WO2021253939A1 (en) * | 2020-06-18 | 2021-12-23 | 南通大学 | Rough set-based neural network method for segmenting fundus retinal vascular image |
CN112348826A (en) * | 2020-10-26 | 2021-02-09 | 陕西科技大学 | Interactive liver segmentation method based on geodesic distance and V-net |
CN112651933A (en) * | 2020-12-21 | 2021-04-13 | 山东省人工智能研究院 | Blood vessel segmentation method based on geodesic distance graph and engineering function equation |
Non-Patent Citations (4)
Title |
---|
Da Chen 等.Geodesic Paths for Image Segmentation With Implicit Region-Based Homogeneity Enhancement.《IEEE Transactions on Image Processing 》.2021,30第5138 - 5153页. * |
Gangming Zhao.Graph Convolution Based Cross-Network Multiscale Feature Fusion for Deep Vessel Segmentation.《IEEE Transactions on Medical Imaging》.2023,第42卷(第1期),第183-195页. * |
刘丽.医学图像分割算法研究及应用.《中国博士学位论文全文数据库 医药卫生科技辑》.2021,(第1期),第E060-3页. * |
陈达.活动轮廓模型算法研究及其生物医学图像应用.《中国优秀硕士学位论文全文数据库 信息科技辑》.2013,(2),第I138-1489页. * |
Also Published As
Publication number | Publication date |
---|---|
CN116580008A (en) | 2023-08-11 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111553236B (en) | Road foreground image-based pavement disease target detection and example segmentation method | |
CN107609525B (en) | Remote sensing image target detection method for constructing convolutional neural network based on pruning strategy | |
CN106920243B (en) | Improved ceramic material part sequence image segmentation method of full convolution neural network | |
CN107506761B (en) | Brain image segmentation method and system based on significance learning convolutional neural network | |
He et al. | Fully convolutional boundary regression for retina OCT segmentation | |
CN107169974A (en) | It is a kind of based on the image partition method for supervising full convolutional neural networks more | |
CN109903284B (en) | HER2 immunohistochemical image automatic discrimination method and system | |
CN107665492B (en) | Colorectal panoramic digital pathological image tissue segmentation method based on depth network | |
WO2021203795A1 (en) | Pancreas ct automatic segmentation method based on saliency dense connection expansion convolutional network | |
CN108257135A (en) | The assistant diagnosis system of medical image features is understood based on deep learning method | |
CN104063713B (en) | A kind of semi-autonomous on-line study method based on random fern grader | |
CN110796661B (en) | Fungal microscopic image segmentation detection method and system based on convolutional neural network | |
CN108537751B (en) | Thyroid ultrasound image automatic segmentation method based on radial basis function neural network | |
CN108875794A (en) | Image visibility detection method based on transfer learning | |
CN108052886A (en) | A kind of puccinia striiformis uredospore programming count method of counting | |
CN113763340B (en) | Automatic grading method based on multitask deep learning ankylosing spondylitis | |
CN108596038A (en) | Erythrocyte Recognition method in the excrement with neural network is cut in a kind of combining form credit | |
CN109284779A (en) | Object detection method based on deep full convolution network | |
CN108564085A (en) | A kind of method of automatic reading pointer type instrument reading | |
CN108205806A (en) | A kind of automatic analytic method of pyramidal CT image three-dimensional cranio-orbital tumor | |
CN112396619B (en) | Small particle segmentation method based on semantic segmentation and internally complex composition | |
CN110992314B (en) | Pavement defect detection method, device and storage medium | |
CN110349168B (en) | Femoral head CT image segmentation method | |
CN109035227A (en) | The system that lung tumors detection and diagnosis is carried out to CT image | |
CN115546605A (en) | Training method and device based on image labeling and segmentation model |
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 |