CN110210578A - Cervical cancer tissues pathology micro-image clustering system based on graph theory - Google Patents
Cervical cancer tissues pathology micro-image clustering system based on graph theory Download PDFInfo
- Publication number
- CN110210578A CN110210578A CN201910531040.1A CN201910531040A CN110210578A CN 110210578 A CN110210578 A CN 110210578A CN 201910531040 A CN201910531040 A CN 201910531040A CN 110210578 A CN110210578 A CN 110210578A
- Authority
- CN
- China
- Prior art keywords
- cluster
- cervical cancer
- cancer tissues
- image
- system based
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/69—Microscopic objects, e.g. biological cells or cellular parts
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/69—Microscopic objects, e.g. biological cells or cellular parts
- G06V20/695—Preprocessing, e.g. image segmentation
Abstract
The invention discloses a kind of cervical cancer tissues pathology micro-image clustering system based on graph theory, comprising: step 1 acquires cervical cancer tissues micro-image data, carries out first stage cluster；Step 2, by the node of skeletonizing come the distribution of approximate representation nucleus；Step 3, using the skeleton of each connection as a region, by the skeleton joint structure in each region at minimum spanning tree；Step 4: different statistical values is calculated according to minimum spanning tree structure chart, as graphic feature；Step 5: graphic feature and global characteristics based on extraction carry out the cluster of second stage.The present invention provides a kind of cervical cancer tissues pathology micro-image clustering system based on graph theory, makes full use of the node generated in skeletonizing processing to carry out the distribution of approximate representation nucleus, improves Clustering Effect；Different types of tissue is expressed by way of secondary cluster, embodies the special construction and complexity in tissue.
Description
Technical field
The present invention relates to the image analysis technologies of image medical treatment field of auxiliary.It is more particularly related to a kind of base
In the cervical cancer tissues pathology micro-image clustering system of graph theory.
Background technique
The prior art is when clustering cervical cancer tissues micro-image, usually using flow processing as shown in Figure 2
Method is split the micro-image to cervical cancer tissues, the specific steps of which are as follows:
(a) the original microsection image of cervical cancer tissues is obtained；
(b) original image is changed into gray level image；
(c) image is reconstructed by operator operation, prominent nucleus and cytoplasmic difference；
(d) use Morphological scale-space, filter unwanted part, using automatic threshold to the nucleus in organization chart picture into
Row positioning；
(e) graph structure is generated with the position of nucleus.
The analysis based on region further is carried out to the image after segmentation, obtains corresponding cluster result as shown in Figure 3.
It can be seen in figure 3 that (a) is the graph structure of normal tissue, (b) (c) (d) respectively represents tri- ranks of CIN1, CIN2, CIN3
Section, wherein the cluster of yellow mark Y1, Y2, Y3, Y4 represent basal layer, and the cluster of green mark G1, G2, G3, G4 represent intermediate
The cluster of layer, blue mark B1, B2, B3, B4 represents superficial layer, for analyzing the relationship of the classification of CIN and the variation of figure.
And for using such flow and method and then cluster, it will usually have the following problems:
(1) in cluster process, histocyte usually will appear large area adhesion and overlapping and cause to be difficult to the problem of dividing,
Using existing cutting techniques carry out processing accurately identifying for cell nuclear location is not achieved, and can generate over-segmentation or
Divide insufficient situation, generates certain difference with cell actual form, influence the judgement for structure.
(2) prior art is distinguished using according to the triangle area of graph structure, labeled as different colors, is used
Indicate the space structure between cell, this method has specificity, different zones and different shape to tissue it is thin
Born of the same parents' discrimination is not high, and since organization chart picture has the feature of complexity and particularity, existing method is only capable of representing cell sky
Between on density, can not illustrate the difference in structure, this is very unfavorable for the differentiation of cancer.
Summary of the invention
It is excellent it is an object of the invention to solve at least the above problems and/or defect, and provide at least to will be described later
Point.
It is a still further object of the present invention to provide a kind of, and the cervical cancer tissues pathology micro-image based on graph theory clusters system
System makes full use of the node generated in skeletonizing processing to come the distribution of approximate representation nucleus, the knowledge to nucleus space structure
It is more inaccurate, improve Clustering Effect；Further different types of tissue is expressed by way of secondary cluster, is gathered
Class result will be obvious that special construction and complexity in tissue.
In order to realize these purposes and other advantages according to the present invention, a kind of cervical cancer tissues based on graph theory are provided
Pathology micro-image clustering system, comprising:
Step 1 acquires cervical cancer tissues micro-image data, carries out first stage cluster；
Step 2 carries out skeletonizing processing to the result of first stage cluster, with the node by skeletonizing come approximate table
Show the distribution of nucleus；
Node is carried out separating treatment by step 3, using the skeleton of each connection as a region, and then by each
Skeleton joint structure in region is at minimum spanning tree；
Step 4: according to the minimum spanning tree structure chart of generation, different statistical values is calculated, is indicated as graphic feature
Different tissues；
Step 5: based on graphic feature and global characteristics are extracted, the cluster of second stage is carried out.
Preferably, wherein in the first stage cluster process of step 1, using rgb pixel value as color characteristic,
Cervical tissue pathological image is subjected to first cluster operation by application k-means algorithm；
Wherein, the K value in k-means algorithm is set as 2, and cluster result represents nuclear area by using white
Domain, black represents the region other than core, and then distinguishes to nucleus and cytoplasm, intercellular substance.
Preferably, wherein further include by Sobel edge detection, Canny edge detection, Otsu in step 1
Thresholding, watershed transform result merged, to improve Clustering Effect, and increase cluster by using morphological operation
Precision.
Preferably, wherein in the skeletonizing treatment process in step 2, need superfluous to being generated in skeletonizing processing
Remaining node is deleted, and then the distribution of approximate representation nucleus.
Preferably, wherein in step 4, extracted graphic feature is configured to include edge lengths in each region
And mean value, variance, the degree of bias, the kurtosis of angle；
In step 5, extracted global characteristics be configured to include tissue perimeter and each tissue interior nodes
The parameter value being independently fitted.
Preferably, wherein in step 5, the second stage cluster is configured as by using k-means algorithm
To realize secondary cluster；
Wherein, the k value in secondary cluster k-means algorithm is configured as 3, to pass through more detailed cluster result for tissue
Structure is divided into three-level, effectively distinguishes the complexity and particularity for finishing structure, and then predict the risk of cancer of tissue.
Preferably, wherein further include step 6, analysis assessment is carried out to cluster result by outline value；
Wherein, the range of outline value profile, which is set, is negative 1 to 1, and value is higher, then illustrates that the clustering method used more closes
Suitable, value is lower or is negative value, then shows that cluster is too many or very little.
The present invention is include at least the following beneficial effects: first, the present invention is by improving existing tissue clustering technique,
The node generated in skeletonizing processing is made full use of to solve the problems, such as the prior art because cutting techniques deficiency causes, and passes through bone
The node of frame carrys out the distribution of approximate representation nucleus, more accurate to the identification of nucleus space structure, with the practical shape of cell
State difference is obviously reduced, and can be conducive to reduce error, the judgement conducive to the later period to its structure improves Clustering Effect；
Second, the present invention expresses different types of tissue by way of secondary cluster, compared with the existing technology
In triangle expression way for, cluster result will be obvious that special construction and complexity in tissue, can be effective
Ground illustrates the difference in structure, and the differentiation for cancer is highly beneficial.
Further advantage, target and feature of the invention will be partially reflected by the following instructions, and part will also be by this
The research and practice of invention and be understood by the person skilled in the art.
Detailed description of the invention
Fig. 1 is the cervical cancer tissues pathology micro-image clustering system based on graph theory in one embodiment of the present of invention
Program flow chart；
Fig. 2 is the segmentation flow chart of cervical cancer tissues in the prior art；
Fig. 3 is the cluster result figure of cervical cancer tissues in the prior art；
Fig. 4 is the cervical cancer tissues pathology micro-image clustering system based on graph theory the in one embodiment of the present of invention
Tissue Clustering Effect figure after secondary cluster；
Fig. 5 is that the cervical cancer tissues pathology micro-image clustering system based on graph theory exists in one embodiment of the present of invention
Carry out the effect picture when assessment of outline value；
Fig. 6 is the comparison signal of original image and first cluster, second of cluster, two Stage Clustering results in the present invention
Figure.
Specific embodiment
Present invention will be described in further detail below with reference to the accompanying drawings, to enable those skilled in the art referring to specification text
Word can be implemented accordingly.
It should be appreciated that such as " having ", "comprising" and " comprising " term used herein do not allot one or more
The presence or addition of a other elements or combinations thereof.
Fig. 1 shows a kind of cervical cancer tissues pathology micro-image clustering system based on graph theory according to the present invention
Way of realization, comprising:
Step 1 acquires cervical cancer tissues micro-image data, carries out first stage cluster；
Step 2 carries out skeletonizing processing to the result of first stage cluster, with the node by skeletonizing come approximate table
The distribution for showing nucleus, can be seen that sparse nucleus from first stage cluster result can cluster well, but high glutinous
The nucleus of degree cell (overlapping cell and attached cell) is difficult to identify.In addition, shadow of the cluster result by acutance and colouring method
Sound is larger, therefore a kind of method that this programme proposes node for generating skeletonizing carrys out the distribution of approximate representation core；
Node is carried out separating treatment by step 3, using the skeleton of each connection as a region, this isolated place
Reason can promote the differentiation of different type tissue, and then by the skeleton joint structure in each region at minimum spanning tree,
Minimum spanning tree can also be replaced using triangulation, calculate the figures features such as side length, triangle number and carry out image most
Optimization cluster；Or minimum spanning tree is replaced using Thiessen polygon algorithm, it calculates the figures features such as side number, area and carries out image
Optimize cluster；
Step 4: according to the minimum spanning tree structure chart of generation, different statistical values is calculated, is indicated as graphic feature
Different tissues is that seed point prim algorithm generates minimum spanning tree, calculates the length of spanning tree, the average distance of line segment
And variance distribution, it is distributed according to length, distance and variance to classify；
Step 5: based on graphic feature (statistical value) and global characteristics (geometry value) are extracted, carrying out the cluster of second stage,
Its cluster process for being used to lead to second stage obtains more detailed as a result, being used to predict the risk of cancer of tissue.This programme uses
Cervical tissue pathological image cluster task is solved based on the unsupervised learning method of figure, and institutional framework is divided into three-level, is had
Effect has distinguished the complexity and particularity of structure.
In another embodiment, in the first stage cluster process of step 1, using rgb pixel value as color spy
Cervical tissue pathological image is carried out first cluster operation by application k-means algorithm by sign；
Wherein, the K value in k-means algorithm is set as 2, and cluster result represents nuclear area by using white
Domain, black represents the region other than core, and then distinguishes to nucleus and cytoplasm, intercellular substance.In this scheme
Cluster result is indicated with black and white, has distinguished foreground object (nucleus) and background object (cytoplasm and iuntercellular object
Matter), algorithm can be by replacing the clustering algorithm merged with GrabCut algorithm, and the first stage for carrying out image optimizes cluster.
In another embodiment, in step 1, further include by Sobel edge detection, Canny edge detection,
Otsu thresholding, watershed transform result merged, to improve Clustering Effect, the nucleus that k-means algorithm identifies
There are many quantity, and close to actual value, but impurity can be relatively more；And the profile that sobel edge detection identifies is accurate, this calculation
Method is affected by noise bigger, and nucleus discrimination unconspicuous for boundary is low；The partial region that canny edge detection obtains
It is interrupted, incomplete marginal information, but detects speed quickly, it is same affected by noise big；Point that threshold method obtains
It is clear to cut figure, calculate simple and is not influenced by picture luminance and contrast variation, has a wide range of application, but part nucleus
Profile is unobvious, and computational efficiency is not high；Dividing ridge method accurate positioning, however there are problems that over-segmentation in identifying.In order to
Better segmentation result is obtained, above a variety of methods are warm, so that fused algorithm is accurately higher, nucleus profile is bright
It is aobvious, while over-segmentation will not be generated, have precision high in cervical cancer tissues pathology micro image analysis, applicability is extensive
Advantage.
And increase the precision of cluster by using morphological operation.It is thin due to what is obtained after Threshold segmentation in this scheme
Born of the same parents' bianry image generally has noise, and cell edges are not mellow and full enough, be directly used as cytoskeletonization processing can generate it is biggish
Therefore error carries out Morphological scale-space to the bianry image obtained after Threshold segmentation, so that the essence of later period skeletonizing processing
Du Genggao, morphologic operation concrete operations are as follows: by eight connectivity background area pixels area in each cell bianry image
Region less than the first preset value is set as foreground area.Since intracellular hole is all background element, it can be achieved that intracellular
The filling of hole."ON" operation is carried out to each cell bianry image, for removing the noise of some cell edges and some
Small noise, and the cell of some slight adhesions is separated, "ON" operation is the prior art, and details are not described herein again.By each
Eight connectivity foreground area elemental area is set as background area less than the region of the second preset value in cell bianry image.By upper
The cell bianry image for stating processing, which can still have some biggish impurity noises, not can be removed, since these impurity noises are all
Foreground elements realize the removal of impurity noise.
In another embodiment, it in the skeletonizing treatment process in step 2, needs to generate in skeletonizing processing
Redundant node deleted, and then the distribution of approximate representation nucleus is tied in this scheme firstly, clustering to the first stage
Fruit carries out skeletonizing processing.Then, micronization processes are carried out to skeletonizing result.It specifically, is quick parallel using Zhang
Thinning algorithm carries out cytoskeleton processing to the cell bianry image after each Morphological scale-space；To having in an image
Have 46 nucleus pending tissue pathological image carry out skeletonizing processing, can be subdivided into first to the image to be processed into
The processing of row gray processing then carries out image segmentation to the gray processing treated image and obtains cell bianry image, then to cell
Bianry image carries out Morphological scale-space, further carries out cytoskeleton processing to it, passes through the Node distribution of its skeletonizing
It is found that according to the cytoskeletonization, treated that image middle skeleton number of nodes is consistent with number of cells, and accuracy can reach
93.5%, and under normal circumstances, the skeleton of each cell can generate two nodes, the spatial distribution of skeleton node and nucleus
It is similar.Finally, deleting redundant node, remaining node is used to the distribution of approximate representation nucleus.Experiment shows practical nucleus
Reach 90% with the Duplication of skeleton node location, therefore can carry out approximate table to nucleus by way of skeletonizing node
It reaches.
In another embodiment, in step 4, extracted graphic feature is configured to include each region inner edge
The mean value of length and angle, variance, the degree of bias, kurtosis；
In step 5, extracted global characteristics be configured to include tissue perimeter and each tissue interior nodes
The parameter value being independently fitted, by the cooperations of global characteristics and graphic feature in this scheme, jointly to each in image
Expression is described in tissue, so that the precision of its secondary cluster is higher, effect is more preferable.
In another embodiment, in step 5, the second stage cluster is configured as calculating by using k-means
Method is to realize secondary cluster；
Wherein, the k value in secondary cluster k-means algorithm is configured as 3, to pass through more detailed cluster result for tissue
Structure is divided into three-level, effectively distinguishes the complexity and particularity for finishing structure, and then predict the risk of cancer of tissue.In this programme
According to the feature of figure, we gather tissue for three classes, because experiment shows that identification effect is best when k=3.Histiocytic cluster
As a result as shown in figure 4, three kinds of tissue cluster results have apparent difference on morphosis and complexity.From (a) to (c) is opened up
Structure change is flutterred it is obvious that showing the outstanding discriminating power of figure feature.
In another embodiment, further include step 6, analysis assessment, outline value are carried out to cluster result by outline value
The quality for measuring classification, obtain be each data point in data set a distance value, this value is each sample
With the dissimilar degree of samples other in same category, and a relation value with the dissimilar degree of sample in other classifications, more
It is big better；
Wherein, the range of outline value profile, which is set, is negative 1 to 1, and value is higher, then illustrates that the clustering method used more closes
Suitable, value is lower or is negative value, then shows that cluster is too many or very little.In this scheme, in order to further analyze cluster result,
By the assessment of outline value for illustrating between other data points in each data point and adjacent cluster in a cluster
Relationship is a kind of extremely effective numerical Evaluation method.Specifically, it is commented as shown in figure 5, giving outline value profile of the present invention
It is estimating as a result, can be seen that as used in the present invention from its result is practical medical data, different types of tissue quantity
Differ larger, the uniformity of cluster not can guarantee, and structure is more complicated, and quantity is fewer, and difference is also bigger.Three kinds when k=3
The outline average value of cluster is respectively 92%, 87%, 71%.First order cluster and second level cluster are compared.Such as Fig. 6 institute
Show, (a) represents original structure image, (b) represents first stage cluster result, (c) indicate corresponding second stage cluster result,
Wherein blue B5, green G5 and red R 1 respectively represent the tissues of three types in (c) figure, therefore can be bright according to cluster result
The aobvious special construction and complexity found out in tissue.
Compared to the prior art, the present invention forms complete cluster process, and is made that assessment.Utilize characteristics of graph theory
Cervical cancer tissues pathological image is divided into according to the space structure of core different classes of, this can be applied in the daily of histologist
In practice, risk of cancer prediction field show huge potentiality, doctor can be assisted to judge, accelerate diagnosis when
Between, improve the accuracy of diagnosis.The architectural characteristic of different cluster results is related from the tumour of different brackets and different risks, with this
As according to presence, rank, risk and the result that can be used to assess cancer.
It is a kind of explanation of preferred embodiments using above scheme, however, it is not limited to this.It in carrying out the present invention, can be with
Replacement and/or modification appropriate are carried out according to user's demand.
Number of devices and treatment scale described herein are for simplifying explanation of the invention.To of the invention based on figure
The application of the cervical cancer tissues pathology micro-image clustering system of opinion, modifications and variations are to one skilled in the art
Obviously.
Although the embodiments of the present invention have been disclosed as above, but its is not only in the description and the implementation listed
With.It can be applied to various suitable the field of the invention completely.It for those skilled in the art, can be easily
Realize other modification.Therefore without departing from the general concept defined in the claims and the equivalent scope, the present invention is simultaneously unlimited
In specific details and legend shown and described herein.
Claims (7)
1. a kind of cervical cancer tissues pathology micro-image clustering system based on graph theory characterized by comprising
Step 1 acquires cervical cancer tissues micro-image data, carries out first stage cluster；
Step 2 carries out skeletonizing processing to the result of first stage cluster, thin come approximate representation with the node by skeletonizing
The distribution of karyon；
Node is carried out separating treatment by step 3, using the skeleton of each connection as a region, and then passes through each region
In skeleton joint structure at minimum spanning tree；
Step 4: according to the minimum spanning tree structure chart of generation, calculating different statistical values, indicates different as graphic feature
Tissue；
Step 5: based on graphic feature and global characteristics are extracted, the cluster of second stage is carried out.
2. the cervical cancer tissues pathology micro-image clustering system based on graph theory as described in claim 1, which is characterized in that
It, will by application k-means algorithm using rgb pixel value as color characteristic in the first stage cluster process of step 1
Cervical tissue pathological image carries out first cluster operation；
Wherein, the K value in k-means algorithm is set as 2, and cluster result represents nuclear area by using white, it is black
Color represents the region other than core, and then distinguishes to nucleus and cytoplasm, intercellular substance.
3. the cervical cancer tissues pathology micro-image clustering system based on graph theory as described in claim 1, which is characterized in that
In step 1, further include by Sobel edge detection, Canny edge detection, Otsu thresholding, watershed transform knot
Fruit is merged, and to improve Clustering Effect, and increases by using morphological operation the precision of cluster.
4. the cervical cancer tissues pathology micro-image clustering system based on graph theory as described in claim 1, which is characterized in that
In skeletonizing treatment process in step 2, need to delete the redundant node generated in skeletonizing processing, and then close
Like the distribution for indicating nucleus.
5. the cervical cancer tissues pathology micro-image clustering system based on graph theory as described in claim 1, which is characterized in that
In step 4, extracted graphic feature be configured to include the mean value of edge lengths and angle in each region, variance, partially
Degree, kurtosis；
In step 5, extracted global characteristics be configured to include tissue perimeter and it is each tissue interior nodes it is only
The parameter value of vertical fitting.
6. the cervical cancer tissues pathology micro-image clustering system based on graph theory as described in claim 1, which is characterized in that
In step 5, the second stage cluster is configured as by using k-means algorithm to realize secondary cluster；
Wherein, the k value in secondary cluster k-means algorithm is configured as 3, with by more detailed cluster result by institutional framework
It is divided into three-level, effectively distinguishes the complexity and particularity for finishing structure, and then predict the risk of cancer of tissue.
7. the cervical cancer tissues pathology micro-image clustering system based on graph theory as described in claim 1, which is characterized in that
Further include step 6, analysis assessment is carried out to cluster result by outline value；
Wherein, the range of outline value profile, which is set, is negative 1 to 1, and value is higher, then illustrates that the clustering method used is more suitable,
It is worth lower or is negative value, then shows that cluster is too many or very little.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910531040.1A CN110210578B (en) | 2019-06-19 | 2019-06-19 | Cervical cancer histopathology microscopic image clustering system based on graph theory |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910531040.1A CN110210578B (en) | 2019-06-19 | 2019-06-19 | Cervical cancer histopathology microscopic image clustering system based on graph theory |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110210578A true CN110210578A (en) | 2019-09-06 |
| CN110210578B CN110210578B (en) | 2021-11-16 |
Family
ID=67793473
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910531040.1A Active CN110210578B (en) | 2019-06-19 | 2019-06-19 | Cervical cancer histopathology microscopic image clustering system based on graph theory |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110210578B (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100172567A1 (en) * | 2007-04-17 | 2010-07-08 | Prokoski Francine J | System and method for using three dimensional infrared imaging to provide detailed anatomical structure maps |
| US20140227682A1 (en) * | 2011-09-13 | 2014-08-14 | Koninklijke Philips N.V. | System and method for the detection of abnormalities in a biological sample |
| CN104282026A (en) * | 2014-10-24 | 2015-01-14 | 上海交通大学 | Distribution uniformity assessment method based on watershed algorithm and minimum spanning tree |
| CN105719294A (en) * | 2016-01-21 | 2016-06-29 | 中南大学 | Breast cancer pathology image mitosis nucleus automatic segmentation method |
| CN109389594A (en) * | 2018-10-09 | 2019-02-26 | 东北大学 | A kind of cervical cancer tissues micro image analysis method based on graph theory |
| CN109685783A (en) * | 2018-12-18 | 2019-04-26 | 东北大学 | A kind of method for cell count based on skeletal extraction |
| CN111739026A (en) * | 2020-05-28 | 2020-10-02 | 数坤（北京）网络科技有限公司 | Blood vessel center line-based adhesion cutting method and device |
-
2019
- 2019-06-19 CN CN201910531040.1A patent/CN110210578B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100172567A1 (en) * | 2007-04-17 | 2010-07-08 | Prokoski Francine J | System and method for using three dimensional infrared imaging to provide detailed anatomical structure maps |
| US20140227682A1 (en) * | 2011-09-13 | 2014-08-14 | Koninklijke Philips N.V. | System and method for the detection of abnormalities in a biological sample |
| CN104282026A (en) * | 2014-10-24 | 2015-01-14 | 上海交通大学 | Distribution uniformity assessment method based on watershed algorithm and minimum spanning tree |
| CN105719294A (en) * | 2016-01-21 | 2016-06-29 | 中南大学 | Breast cancer pathology image mitosis nucleus automatic segmentation method |
| CN109389594A (en) * | 2018-10-09 | 2019-02-26 | 东北大学 | A kind of cervical cancer tissues micro image analysis method based on graph theory |
| CN109685783A (en) * | 2018-12-18 | 2019-04-26 | 东北大学 | A kind of method for cell count based on skeletal extraction |
| CN111739026A (en) * | 2020-05-28 | 2020-10-02 | 数坤（北京）网络科技有限公司 | Blood vessel center line-based adhesion cutting method and device |
Non-Patent Citations (5)
| Title |
|---|
| JOYCE JIYOUNG WHANG等: "Scalable and Memory-Efficient Clustering of Large-Scale Social Networks", 《2012 IEEE 12TH INTERNATIONAL CONFERENCE ON DATA MINING》 * |
| RAMESWAR PANDA等: "Scalable Video Summarization using Skeleton Graph and Random Walk", 《2014 22ND INTERNATIONAL CONFERENCE ON PATTERN RECOGNITION》 * |
| WEI ZHANG等: "Network-based machine learning and graph theory algorithms for precision oncology", 《NPJ PRECISION ONCOLOGY》 * |
| 山坡坡上的蜗牛: "聚类评估算法-轮廓系数（Silhouette Coefficient", 《HTTPS://BLOG.CSDN.NET/WANGXIAOPENG0329/ARTICLE/DETAILS/53542606》 * |
| 梅林: "基于K-均值聚类及数学形态学的细胞图像自动分割方法研究", 《中国优秀硕士学位论文全文数据库 (信息科技辑)》 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110210578B (en) | 2021-11-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US9971931B2 (en) | Systems and methods for segmentation and processing of tissue images and feature extraction from same for treating, diagnosing, or predicting medical conditions | |
| Veta et al. | Automatic nuclei segmentation in H&E stained breast cancer histopathology images | |
| US10229488B2 (en) | Method and system for determining a stage of fibrosis in a liver | |
| Tosta et al. | Segmentation methods of H&E-stained histological images of lymphoma: A review | |
| EP3175389B1 (en) | Automatic glandular and tubule detection in histological grading of breast cancer | |
| US20070019854A1 (en) | Method and system for automated digital image analysis of prostrate neoplasms using morphologic patterns | |
| Jia et al. | Multi-layer segmentation framework for cell nuclei using improved GVF Snake model, Watershed, and ellipse fitting | |
| Shaker et al. | Automatic detection and segmentation of sperm head, acrosome and nucleus in microscopic images of human semen smears | |
| CN110517273B (en) | Cytology image segmentation method based on dynamic gradient threshold | |
| CN111402267A (en) | Segmentation method, device and terminal for epithelial cell nucleus in prostate cancer pathological image | |
| CN114067114A (en) | Area nucleus segmentation counting method based on algae cell morphology | |
| Song et al. | New morphological features for grading pancreatic ductal adenocarcinomas | |
| WO2006122251A2 (en) | Method and system for automated digital image analysis of prostrate neoplasms using morphologic patterns | |
| Saxena et al. | Study of Computerized Segmentation & Classification Techniques: An Application to Histopathological Imagery | |
| CN111210449A (en) | Automatic segmentation method for gland cavity in prostate cancer pathological image | |
| Jadhav et al. | Quantitative analysis of histopathological features of precancerous lesion and condition using image processing technique | |
| Touil et al. | A new conditional region growing approach for microcalcification delineation in mammograms | |
| CN110210578A (en) | Cervical cancer tissues pathology micro-image clustering system based on graph theory | |
| Gonzalez et al. | Automatic marker determination algorithm for watershed segmentation using clustering | |
| CN111815613B (en) | Liver cirrhosis disease stage identification method based on envelope line morphological feature analysis | |
| Guatemala-Sanchez et al. | Nuclei segmentation on histopathology images of breast carcinoma | |
| Acharya et al. | Segmentation of pap smear images to diagnose cervical cancer types and stages | |
| Siddique et al. | Effective Segmentation of Liver CT images using Marker Controlled Watershed Algorithm | |
| Touil et al. | A new conditional region growing approach for an accurate detection of microcalcifications from mammographic images | |
| Ravi et al. | Machine Learning-based Classification and Analysis of Breast Cancer Pathological Images |
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 |