CN105654111B - Pathological image swept-focus classification method based on k-means cluster - Google Patents

Pathological image swept-focus classification method based on k-means cluster Download PDF

Info

Publication number
CN105654111B
CN105654111B CN201510970704.6A CN201510970704A CN105654111B CN 105654111 B CN105654111 B CN 105654111B CN 201510970704 A CN201510970704 A CN 201510970704A CN 105654111 B CN105654111 B CN 105654111B
Authority
CN
China
Prior art keywords
center
focus
mass point
th
distance
Prior art date
Application number
CN201510970704.6A
Other languages
Chinese (zh)
Other versions
CN105654111A (en
Inventor
刘炳宪
谢菊元
王焱辉
王克惠
丁科迪
Original Assignee
宁波江丰生物信息技术有限公司
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 宁波江丰生物信息技术有限公司 filed Critical 宁波江丰生物信息技术有限公司
Priority to CN201510970704.6A priority Critical patent/CN105654111B/en
Publication of CN105654111A publication Critical patent/CN105654111A/en
Application granted granted Critical
Publication of CN105654111B publication Critical patent/CN105654111B/en

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING; COUNTING
    • G06KRECOGNITION OF DATA; PRESENTATION OF DATA; RECORD CARRIERS; HANDLING RECORD CARRIERS
    • G06K9/00Methods or arrangements for reading or recognising printed or written characters or for recognising patterns, e.g. fingerprints
    • G06K9/62Methods or arrangements for recognition using electronic means
    • G06K9/6217Design or setup of recognition systems and techniques; Extraction of features in feature space; Clustering techniques; Blind source separation
    • G06K9/6218Clustering techniques
    • G06K9/622Non-hierarchical partitioning techniques
    • G06K9/6221Non-hierarchical partitioning techniques based on statistics
    • G06K9/6223Non-hierarchical partitioning techniques based on statistics with a fixed number of clusters, e.g. K-means clustering
    • GPHYSICS
    • G06COMPUTING; CALCULATING; COUNTING
    • G06KRECOGNITION OF DATA; PRESENTATION OF DATA; RECORD CARRIERS; HANDLING RECORD CARRIERS
    • G06K9/00Methods or arrangements for reading or recognising printed or written characters or for recognising patterns, e.g. fingerprints
    • G06K9/62Methods or arrangements for recognition using electronic means
    • G06K9/6267Classification techniques
    • G06K9/6268Classification techniques relating to the classification paradigm, e.g. parametric or non-parametric approaches
    • G06K9/627Classification techniques relating to the classification paradigm, e.g. parametric or non-parametric approaches based on distances between the pattern to be recognised and training or reference patterns
    • G06K9/6271Classification techniques relating to the classification paradigm, e.g. parametric or non-parametric approaches based on distances between the pattern to be recognised and training or reference patterns based on distances to prototypes
    • G06K9/6272Classification techniques relating to the classification paradigm, e.g. parametric or non-parametric approaches based on distances between the pattern to be recognised and training or reference patterns based on distances to prototypes based on distances to cluster centroïds

Abstract

The invention discloses the pathological image swept-focus classification method clustered based on k-means, this method provides a pathological section, includes the following steps: that S1. obtains K focus from the pathological section;S2. the K focuses are obtained into N number of center of mass point using iteration self-organizing data analysis method one by one;S3. classified according to N number of center of mass point of acquisition to the K focuses, to divide class belonging to each focus, each center of mass point is corresponding a kind of;S4. the position of the center of mass point of every class is adjusted according to the corresponding focus of every class after division, is terminated.The present invention sorts to classification of focal spot to be calculated using iteration self-organizing data analysis method, effectively accelerates focusing speed, improves focusing precision, improves the applicability of focal plane.

Description

Pathological image swept-focus classification method based on k-means cluster

Technical field

The present invention relates to technical field of microscopy more particularly to a kind of pathological image swept-focus based on k-means cluster Classification method.

Background technique

Digital slices scanner is the Medical Devices quickly grown in recent years, and digital slices scanning system can be complete by glass slide Information, comprehensive quick scanning make the glass slide of conventional matter become new-generation digital pathological section, digital slices scanning Instrument is to realize epoch-making change to pathological diagnosis technology.A digital slices scanner essential ring during the scanning process Section is exactly to focus, i.e., is cooperated by moving up and down for Z axis with camera grabgraf, finds the slice a certain number of focuses of different location Afterwards, the focal plane for calculating slice, for preparing for slice scanning.There are mainly two types of existing focusing methods, and one is one Secondary property is to complete all focuses, but the error focused in this way is big, and scan image is fuzzy;Another method is to complete one Part focus scanning a part, i.e. dynamic focusing, this method there occur a problem that when being sliced larger or tissue be island When, the focal plane calculated for the first time may be not suitable for farther away focus, cause farther away focus focusing difficult.

Summary of the invention

The above problem existing for focusing method for existing digital slices scanner, now providing that one kind aims at can It calculates the first time focal plane of dynamic focusing and is suitable for all focuses to be calculated, accelerate focusing speed, improve focusing precision Pathological image swept-focus classification method based on k-means cluster.

Specific technical solution is as follows:

Based on the pathological image swept-focus classification method of k-means cluster, a pathological section, including following steps are provided It is rapid:

S1. K focus is obtained from the pathological section;

S2. the K focuses are obtained into N number of center of mass point using iteration self-organizing data analysis method one by one;

S3. classified according to N number of center of mass point of acquisition to the K focuses, to divide each focus institute The class of category, each center of mass point are corresponding a kind of;

S4. the position of the center of mass point of every class is adjusted according to the corresponding focus of every class after division, is terminated.

Preferably, the detailed process of the step S2 are as follows:

S21. choose a focus as the i-th=1 center of mass point, obtain respectively the K-1 focuses with it is described The distance D of i-th of center of mass pointi(j), and by K-1 distance Di(j) it adds up to obtain total distance sum (Di),

Wherein, i, j are positive integer, Di(j) the distance between j-th of focus and i-th of center of mass point are indicated, 1≤i≤K,1≤j≤K,i≠j;

S22. i+1 center of mass point is obtained according to formula (1):

Random-Di(j) < 0 (1),

Wherein, Random is less than or equal to sum (Di) integer, when Random meets formula (1), the distance Di (j) corresponding j-th focus is i-th=i+1 center of mass point;

S23. judge whether i is equal to N, if so, executing step S24;If it is not, executing step S3;

S24. the distance D of K-1 the focuses and described i-th center of mass point is obtained respectivelyi(j), and by K-1 A distance Di(j) it adds up to obtain the total distance sum (Di), it returns, executes step S22.

Preferably, the distance that each focus arrives each center of mass point respectively is obtained one by one in the step S3, The focus is classified as shortest apart from the corresponding class of the corresponding center of mass point in corresponding distance.

Preferably, each focus is obtained according to formula (2) in the step S3 arrive each center of mass point respectively Distance:

Cj=argmin ‖ Xj-Yi2(2),

Wherein, XjIndicate the position coordinates of j-th of focus, YiIndicate the position coordinates of i-th of center of mass point, Cj Indicate described j-th focus to i-th of center of mass point distance.

Preferably, being averaged for the position coordinates for all focuses for belonging to every class is obtained one by one in the step S4 Coordinate, using the average coordinates as the position coordinates of the center of mass point corresponding with the class.

Preferably, it is sat in the step S4 according to the position that formula (3) obtain all focuses for belonging to every class Target average coordinates:

Wherein, M indicates all focuses belonged in the corresponding class of i-th of center of mass point, QiIndicate i-th of institute State the average coordinates of the center of mass point of the corresponding class of center of mass point.

Above-mentioned technical proposal the utility model has the advantages that

In the technical scheme, it is sorted using iteration self-organizing data analysis method to classification of focal spot to be calculated, it is effective to add Fast focusing speed, improves focusing precision, improves the applicability of focal plane.

Detailed description of the invention

Fig. 1 is a kind of implementation of the pathological image swept-focus classification method of the present invention based on k-means cluster The flow chart of example.

Specific embodiment

Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on Embodiment in the present invention, those of ordinary skill in the art without creative labor it is obtained it is all its His embodiment, shall fall within the protection scope of the present invention.

It should be noted that in the absence of conflict, the feature in embodiment and embodiment in the present invention can phase Mutually combination.

The present invention will be further explained below with reference to the attached drawings and specific examples, but not as the limitation of the invention.

As shown in Figure 1, the pathological image swept-focus classification method based on k-means cluster, provides a pathological section, wrap Include following step:

S1. K focus is obtained from pathological section;

S2. K focus is obtained into N number of center of mass point using iteration self-organizing data analysis method one by one;

S3. classified according to N number of center of mass point of acquisition to K focus, to divide class belonging to each focus, Mei Gezhi Heart point is corresponding a kind of;

S4. the position of the center of mass point of every class is adjusted according to the corresponding focus of every class after division, is terminated.

In the present embodiment, it is sorted using iteration self-organizing data analysis method to classification of focal spot to be calculated, it is effective to accelerate Focusing speed improves focusing precision, improves the applicability of focal plane.

In a preferred embodiment, the detailed process of step S2 are as follows:

S21. a focus is chosen as the i-th=1 center of mass point, obtains K-1 focus respectively at a distance from i-th of center of mass point Di(j), and by K-1 distance Di(j) it adds up to obtain total distance sum (Di),

Wherein, i, j are positive integer, Di(j) expression the distance between j-th of focus and i-th of center of mass point, 1≤i≤K, 1≤j≤K,i≠j;

S22. i+1 center of mass point is obtained according to formula (1):

Random-Di(j) < 0 (1),

Wherein, Random is less than or equal to sum (Di) integer, when Random meets formula (1), distance Di(j) right J-th of the focus answered is i-th=i+1 center of mass point;

S23. judge whether i is equal to N, if so, executing step S24;If it is not, executing step S3;

S24. K-1 focus and i-th of center of mass point distance D are obtained respectivelyi(j), and by K-1 distance Di(j) add up with Obtain total distance sum (Di), it returns, executes step S22.

In the present embodiment, the selection number of center of mass point and position will have a direct impact on the quality of cluster result, therefore use Iteration self-organizing data analysis method algorithm calculates center of mass point number, and focusing precision can be improved.

In a preferred embodiment, the distance that each focus arrives each center of mass point respectively is obtained one by one in step s3, it will Focus is classified as shortest apart from the corresponding class of corresponding center of mass point in corresponding distance.

Further, the distance that each focus arrives each center of mass point respectively is obtained according to formula (2) in step s3:

Cj=argmin ‖ Xj-Yi2(2),

Wherein, XjIndicate the position coordinates of j-th of focus, YiIndicate the position coordinates of i-th of center of mass point, CjIt indicates j-th Distance of the focus to i-th of center of mass point.

In a preferred embodiment, in step s 4 one by one obtain belong to every class institute focal position coordinates put down Equal coordinate, using average coordinates as the position coordinates of center of mass point corresponding with class.

Further, in step s 4 according to formula (3) obtain belong to every class institute focal position coordinates put down Equal coordinate:

Wherein, M indicates all focuses belonged in the corresponding class of i-th of center of mass point, QiIndicate that i-th of center of mass point is corresponding Class center of mass point average coordinates.

In the present embodiment, using average coordinates as the position coordinates of center of mass point corresponding with class, after completing focus cluster, By the sequence to inhomogeneity focus, efficient focus identification can be realized.

The foregoing is merely preferred embodiments of the present invention, are not intended to limit embodiments of the present invention and protection model It encloses, to those skilled in the art, should can appreciate that all with made by description of the invention and diagramatic content Equivalent replacement and obviously change obtained scheme, should all be included within the scope of the present invention.

Claims (5)

1. the pathological image swept-focus classification method based on k-means cluster, which is characterized in that provide a pathological section, packet Include following step:
S1. K focus is obtained from the pathological section;
S2. the K focuses are obtained into N number of center of mass point using iteration self-organizing data analysis method one by one;
S3. classified according to N number of center of mass point of acquisition to the K focuses, to divide belonging to each focus Class, each center of mass point are corresponding a kind of;
S4. the position of the center of mass point of every class is adjusted according to the corresponding focus of every class after division, is terminated;
The detailed process of the step S2 are as follows:
S21. a focus is chosen as the i-th=1 center of mass point, obtains the K-1 focuses and described i-th respectively The distance D of the center of mass pointi(j), and by K-1 distance Di(j) it adds up to obtain total distance sum (Di),
Wherein, i, j are positive integer, Di(j) the distance between j-th of focus and i-th of center of mass point, 1≤i are indicated ≤K,1≤j≤K,i≠j;
S22. i+1 center of mass point is obtained according to formula (1):
Random-Di(j) < 0 (1),
Wherein, Random is less than or equal to sum (Di) integer, when Random meets formula (1), the distance Di(j) right J-th answered the focus is i-th=i+1 center of mass point;
S23. judge whether i is equal to N, if so, executing step S24;If it is not, executing step S3;
S24. the distance D of K-1 the focuses and described i-th center of mass point is obtained respectivelyi(j), and by K-1 institute State distance Di(j) it adds up to obtain the total distance sum (Di), it returns, executes step S22.
2. the pathological image swept-focus classification method as described in claim 1 based on k-means cluster, which is characterized in that It obtains the distance that each focus arrives each center of mass point respectively one by one in the step S3, the focus is classified as phase It is shortest apart from the corresponding class of the corresponding center of mass point in the distance answered.
3. the pathological image swept-focus classification method as claimed in claim 2 based on k-means cluster, which is characterized in that The distance that each focus arrives each center of mass point respectively is obtained according to formula (2) in the step S3:
Cj=argmin | | Xj-Yi||2(2),
Wherein, XjIndicate the position coordinates of j-th of focus, YiIndicate the position coordinates of i-th of center of mass point, CjIt indicates Distance of j-th of the focus to i-th of center of mass point.
4. the pathological image swept-focus classification method as claimed in claim 3 based on k-means cluster, which is characterized in that Obtain the average coordinates for belonging to the position coordinates of all focuses of every class one by one in the step S4, it will be described average Position coordinates of the coordinate as the center of mass point corresponding with the class.
5. the pathological image swept-focus classification method as claimed in claim 4 based on k-means cluster, which is characterized in that The average coordinates for belonging to the position coordinates of all focuses of every class are obtained according to formula (3) in the step S4:
Wherein, M indicates all focuses belonged in the corresponding class of i-th of center of mass point, QiIndicate i-th of the matter The average coordinates of the center of mass point of the corresponding class of heart point, CjIndicate described j-th focus to i-th center of mass point Distance.
CN201510970704.6A 2015-12-21 2015-12-21 Pathological image swept-focus classification method based on k-means cluster CN105654111B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201510970704.6A CN105654111B (en) 2015-12-21 2015-12-21 Pathological image swept-focus classification method based on k-means cluster

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201510970704.6A CN105654111B (en) 2015-12-21 2015-12-21 Pathological image swept-focus classification method based on k-means cluster

Publications (2)

Publication Number Publication Date
CN105654111A CN105654111A (en) 2016-06-08
CN105654111B true CN105654111B (en) 2019-03-29

Family

ID=56477539

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201510970704.6A CN105654111B (en) 2015-12-21 2015-12-21 Pathological image swept-focus classification method based on k-means cluster

Country Status (1)

Country Link
CN (1) CN105654111B (en)

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104390981A (en) * 2014-12-12 2015-03-04 宁波江丰生物信息技术有限公司 Method for rapidly and accurately focusing and scanning pathological section tissue on basis of image collection device
CN105118088A (en) * 2015-08-06 2015-12-02 曲阜裕隆生物科技有限公司 3D imaging and fusion method based on pathological slice scanning device

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20070031056A1 (en) * 2005-08-02 2007-02-08 Perz Cynthia B System for and method of focusing in automated microscope systems

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104390981A (en) * 2014-12-12 2015-03-04 宁波江丰生物信息技术有限公司 Method for rapidly and accurately focusing and scanning pathological section tissue on basis of image collection device
CN105118088A (en) * 2015-08-06 2015-12-02 曲阜裕隆生物科技有限公司 3D imaging and fusion method based on pathological slice scanning device

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
基于分层聚类的k-means算法;金微等;《河海大学常州分校学报》;20070331;第21卷(第1期);第8页
杨坤等.基于视场边界最大最小聚类分析的快速扫描方法.《光 电 子 •激 光》.2014,第25卷(第4期),

Also Published As

Publication number Publication date
CN105654111A (en) 2016-06-08

Similar Documents

Publication Publication Date Title
Tang et al. Differential evolution with an individual-dependent mechanism
Mihos et al. Diffuse light in the Virgo Cluster
Punchi‐Manage et al. Effects of topography on structuring local species assemblages in a S ri L ankan mixed dipterocarp forest
CN104536249B (en) The method and apparatus of regulation projector focal length
CN1151464C (en) Method of reading characters and method of reading postal addresses
CN1284373C (en) Methods of and units for motion or depth estimation and image processing apparatus provided with such motion estimation unit
Lutzoni et al. Major fungal lineages are derived from lichen symbiotic ancestors
KR101860506B1 (en) A system, a method and a computer program product for cad­based registration
CN105308944A (en) Classifying objects in images using mobile devices
Ray et al. Recovering the geographic origin of early modern humans by realistic and spatially explicit simulations
US10443029B2 (en) Cell evaluation device, cell evaluation method, and cell evaluation program
JP6706788B2 (en) Image recognition method, image recognition device and program
KR101691563B1 (en) Cell counting method, cell counting device, and cell counting program
US9892504B2 (en) Image inspection method and inspection region setting method
JP2008077677A (en) Measurement of mitotic activity
JP2005316068A (en) Optical system
US9996764B2 (en) Image matching method based on cascaded binary encoding
CN1411289A (en) Positioning system, positioning server, base station and terminal position estimating method
CN1176566A (en) Directivity control antenna apparatus for shaping radiation pattern of antenna of base station in mobile communication system in accordance with estimated directions or positions of mobile stations
EP3234550A1 (en) Method for identifying biological particles using stacks of defocused holographic images
CN102884397B (en) Structured-light measuring method and system
WO2016182612A1 (en) Systems and methods for reducing a plurality of bounding regions
US20170061618A1 (en) Cell evaluation device, cell evaluation method, and cell evaluation program
CN103455811B (en) Indoor wireless locating fingerprint collection method and system
CN103973993A (en) Image composition apparatus and method

Legal Events

Date Code Title Description
PB01 Publication
C06 Publication
SE01 Entry into force of request for substantive examination
C10 Entry into substantive examination
GR01 Patent grant
GR01 Patent grant