CN108537767A - Cell image segmentation method and Methods of Segmentation On Cell Images device - Google Patents
Cell image segmentation method and Methods of Segmentation On Cell Images device Download PDFInfo
- Publication number
- CN108537767A CN108537767A CN201710121108.XA CN201710121108A CN108537767A CN 108537767 A CN108537767 A CN 108537767A CN 201710121108 A CN201710121108 A CN 201710121108A CN 108537767 A CN108537767 A CN 108537767A
- Authority
- CN
- China
- Prior art keywords
- image
- graph
- seed pixel
- pixel points
- network
- 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.)
- Pending
Links
- 230000011218 segmentation Effects 0.000 title claims abstract description 54
- 238000003709 image segmentation Methods 0.000 title claims abstract description 43
- 238000000034 method Methods 0.000 title claims abstract description 36
- 210000004027 cell Anatomy 0.000 claims abstract description 121
- 238000013507 mapping Methods 0.000 claims description 29
- 210000000805 cytoplasm Anatomy 0.000 claims description 18
- 210000003855 cell nucleus Anatomy 0.000 claims description 8
- 210000004940 nucleus Anatomy 0.000 claims description 8
- 230000001413 cellular effect Effects 0.000 claims description 4
- 210000000265 leukocyte Anatomy 0.000 abstract description 19
- 238000009499 grossing Methods 0.000 description 22
- 230000008569 process Effects 0.000 description 12
- 230000000694 effects Effects 0.000 description 10
- 238000012545 processing Methods 0.000 description 10
- 238000010586 diagram Methods 0.000 description 6
- 230000032823 cell division Effects 0.000 description 4
- 230000001464 adherent effect Effects 0.000 description 3
- 238000001514 detection method Methods 0.000 description 3
- 238000003745 diagnosis Methods 0.000 description 3
- 210000004369 blood Anatomy 0.000 description 2
- 239000008280 blood Substances 0.000 description 2
- 210000003743 erythrocyte Anatomy 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 241000270295 Serpentes Species 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 239000008186 active pharmaceutical agent Substances 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 210000000601 blood cell Anatomy 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 239000003550 marker Substances 0.000 description 1
- 238000007431 microscopic evaluation Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012706 support-vector machine 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10056—Microscopic image
- G06T2207/10061—Microscopic image from scanning electron microscope
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
A kind of cell image segmentation method and device, the method for one embodiment include：Demarcate the foreground seeds pixel and background sub-pixel point in the pending cell image obtained；Using background sub-pixel point as source point, foreground seeds pixel be meeting point, other points are that pending image is mapped as network by node of graph；Each node of graph is respectively connected to source point and meeting point, determine node of graph to source point while the first weight, node of graph to meeting point while the second weight and determine data item；According to the neighbouring relations between each cell pixel, the connection between two node of graph as third weight when connecting and is determined into smooth item in the punishment for not forming continuous boundary；The energy function figure of network is determined according to data item and smooth item；The energy function figure of network is solved and obtains minimal cut solving result；The cell segmentation result of pending cell image is determined according to minimal cut solving result.Embodiment improves the accuracys and splitting speed of Leukocyte Image segmentation.
Description
Technical Field
The present invention relates to the field of image processing technologies, and in particular, to a cell image segmentation method and a cell image segmentation apparatus.
Background
Leukocytes are a very important blood cell in human blood, and are abnormal in high and unstable quantity, and in high or low quantity. The detection of the white blood cells is one of the most basic examination items in medical detection, and the detection of the white blood cells can provide important diagnosis basis for medical diagnosis, the microscopic analysis of the white blood cells has high value for the diagnosis of diseases, and the segmentation of the white blood cells is an important step in the analysis, which lays a foundation for the subsequent characteristic parameter extraction. However, the current automatic segmentation effect still cannot meet the clinical medical needs due to the complexity of the leukocyte types and the interference factors, and improving the accuracy of leukocyte segmentation is a hot spot of the current medical image processing research. In the process of detecting the white blood cells, the segmentation and extraction of the white blood cell image become an important content.
At present, various methods are presented for segmenting and extracting a white blood cell image, for example, algorithms such as a watershed algorithm, a support vector machine, a GVF Snake and a rough set are most applied at present, however, the methods only improve the white blood cell segmentation effect, and the problem of poor segmentation effect of overlapped cells and adhered cells is not effectively solved. Some processing modes also appear for the segmentation of the adherent cells, but the processing modes are sensitive to defined parameters, complex in steps, large in calculation amount, poor in effect of segmenting and extracting the white blood cell image, low in accuracy rate and poor in segmentation effect.
Disclosure of Invention
In view of the above, an object of the embodiments of the present invention is to provide a cell image segmentation method and a cell image segmentation apparatus, so as to improve the accuracy and the segmentation speed of the white blood cell image segmentation.
In order to achieve the purpose, the embodiment of the invention adopts the following technical scheme:
a method of cellular image segmentation, comprising the steps of:
acquiring a cell image to be processed;
calibrating foreground seed pixel points and background seed pixel points in the cell image to be processed;
mapping the to-be-processed image into a network map by taking the background seed pixel points as source points of the network map and the foreground seed pixel points as sink points of the network map, wherein points in the to-be-processed cell image except the background seed pixel points and the foreground seed pixel points are map nodes of the network map;
respectively connecting each graph node to a source point and a sink point, determining a first weight of an edge from the graph node to the source point according to the probability that the graph node and the source point are similar pixels, determining a second weight of the edge from the graph node to the sink point according to the probability that the graph node and the sink point are similar pixels, and determining a data item according to the first weight and the second weight;
determining a connection edge with a connection relation between graph nodes in the network graph according to the adjacent relation between the cell pixel points in the cell image to be processed, taking a penalty that the connection edge between two graph nodes does not form a continuous edge as a third weight of the connection edge, and determining a smooth item according to the third weight;
determining an energy function graph of the network graph according to the data items and the smooth items;
solving the energy function graph of the network graph by adopting a maximum flow minimum cut algorithm to obtain a minimum cut solving result;
and determining a cell segmentation result of the cell image to be processed according to the minimum segmentation solution result.
A cell image segmentation apparatus comprising:
the image acquisition module is used for acquiring a cell image to be processed;
the calibration module is used for calibrating foreground seed pixel points and background seed pixel points in the cell image to be processed;
the map mapping module is used for mapping the to-be-processed image into a network map by taking the background seed pixel points as source points of the network map and the foreground seed pixel points as sink points of the network map, wherein points in the to-be-processed cell image except the background seed pixel points and the foreground seed pixel points are map nodes of the network map;
the data item determining module is used for respectively connecting each graph node to a source point and a sink point, determining a first weight of an edge from the graph node to the source point according to the probability that the graph node and the source point are similar pixels, determining a second weight of the edge from the graph node to the sink point according to the probability that the graph node and the sink point are similar pixels, and determining a data item according to the first weight and the second weight;
the smooth item determining module is used for determining a connecting edge with a connecting relation between graph nodes in the network graph according to the adjacent relation between the pixel points of each cell in the cell image to be processed, taking a penalty that the connecting edge between two graph nodes does not form a continuous edge as a third weight of the connecting edge, and determining a smooth item according to the third weight;
the energy function determining module is used for determining an energy function graph of the network graph according to the data item and the smooth item;
the graph solving module is used for solving the energy function graph of the network graph by adopting a maximum flow minimum cut algorithm to obtain a minimum cut solving result;
and the image segmentation module is used for determining the cell segmentation result of the cell image to be processed according to the minimal segmentation solution result.
According to the scheme of the embodiment, after the foreground seed pixel points and the background seed pixel points in the cell image to be processed are calibrated, the cell image to be processed is mapped into the network map according to the calibrated foreground seed pixel points and the calibrated background seed pixel points, the data items and the smooth items are determined according to the network map, then the appropriate energy function map is established, then the maximum flow minimum cut algorithm is adopted for solving the energy function map, and therefore the cell segmentation result is obtained.
Drawings
FIG. 1 is a schematic flow chart diagram of a cell image segmentation method in one embodiment;
FIG. 2 is a schematic illustration of mapping a red blood cell image to a network map in one particular example;
FIG. 3 is a diagram illustrating a segmentation result obtained after a cell image is segmented in a specific application example;
fig. 4 is a schematic structural diagram of a cell image segmentation apparatus in one embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the detailed description and specific examples, while indicating the scope of the invention, are intended for purposes of illustration only and are not intended to limit the scope of the invention.
Fig. 1 is a schematic flow chart of a cell image segmentation method in an embodiment, and as shown in fig. 1, the cell image segmentation method in this embodiment includes the following steps:
step S101: acquiring a cell image to be processed;
step S102: calibrating foreground seed pixel points and background seed pixel points in the cell image to be processed;
step S103: mapping the to-be-processed image into a network map by taking the background seed pixel points as source points of the network map and the foreground seed pixel points as sink points of the network map, wherein points in the to-be-processed cell image except the background seed pixel points and the foreground seed pixel points are map nodes of the network map;
step S104: respectively connecting each graph node to a source point and a sink point, determining a first weight of an edge from the graph node to the source point according to the probability that the graph node and the source point are similar pixels, determining a second weight of the edge from the graph node to the sink point according to the probability that the graph node and the sink point are similar pixels, and determining a data item according to the first weight and the second weight;
step S105: determining a connection edge with a connection relation between graph nodes in the network graph according to the adjacent relation between the cell pixel points in the cell image to be processed, taking a penalty that the connection edge between two graph nodes does not form a continuous edge as a third weight of the connection edge, and determining a smooth item according to the third weight;
step S106: determining an energy function graph of the network graph according to the data items and the smooth items;
step S107: solving the energy function graph of the network graph by adopting a maximum flow minimum cut algorithm to obtain a minimum cut solving result;
step S108: and determining a cell segmentation result of the cell image to be processed according to the minimum segmentation solution result.
According to the scheme of the embodiment, after the foreground seed pixel points and the background seed pixel points in the cell image to be processed are calibrated, the cell image to be processed is mapped into the network map according to the calibrated foreground seed pixel points and the calibrated background seed pixel points, the data items and the smooth items are determined according to the network map, then the appropriate energy function map is established, then the maximum flow minimum cut algorithm is adopted for solving the energy function map, and therefore the cell segmentation result is obtained.
The cell image segmentation method in this embodiment can be used for image segmentation of a white blood cell image. Based on the characteristics of the cells, the foreground seed pixel points can include cytoplasm seed pixel points and nucleus seed pixel points. At this time, in the step S103, the cytoplasm seed pixel point is taken as a sink of the network map, the background seed pixel point is taken as a source point of the network map, and the cytoplasm seed pixel point is taken as a sink of the network map, and the image to be processed is mapped into the network map. Thus, the cell division result obtained in step S108 is a cell division result for cytoplasm. After the cell segmentation result is obtained in step S108, the cell nucleus seed pixel point may be used as a sink of the network map, and the process returns to step S103, where the background seed pixel point is used as a source point of the network map, and the foreground seed pixel point is used as a sink of the network map, and the image to be processed is mapped into the network map, so that the above process is executed once again to obtain the cell segmentation result for the cell nucleus.
The selected foreground seed pixel points may include more than two. At this time, in the step S103, the to-be-processed image may be mapped into the network map by using the background seed pixel point as a source point of the network map and one of the foreground seed pixel points as a sink point of the network map. After the graph segmentation result is obtained in step S108, the next foreground seed pixel point may be used as a sink of the network graph, and the process returns to step S103, where the background seed pixel point is used as a source point of the network graph, and the next foreground seed pixel point is used as a sink of the network graph, and the image to be processed is mapped into the network graph. Therefore, the process is executed for each foreground seed pixel point, and the graph segmentation result for each foreground seed pixel point is obtained in an iterative execution mode.
After the cell image to be processed is obtained in step S101, the cell image to be processed may be grayed into a grayscale image, so that in step S102, foreground seed pixel points and background seed pixel points of the grayscale image are calibrated. After the gray image is converted, only one gray value is needed to calibrate one pixel point, so that the subsequent processing process can be simplified.
Based on the scheme of the embodiment described above, the following example is described in conjunction with one of the examples, and it should be noted that the following example is only an example, and the description is not intended to limit the embodiment of the present invention.
In performing image segmentation of a leukocyte image, it is first necessary to obtain an image of a cell to be processed, which may be obtained in any manner, such as a cell image obtained by a microscope, a read stored cell image, and the like.
After obtaining the cell image to be processed, the cell image to be processed is converted into a gray image so as to simplify the subsequent processing process. It can be understood that, in the case where the obtained cell image to be processed is a gray-scale image, the subsequent processing may be performed without performing the graying processing. In the following description, for the sake of convenience, the following process of directly processing a cell image to be processed will be described as an example.
And then, calibrating the foreground seed pixel points and the background seed pixel points in the cell image to be processed, wherein the calibrated foreground seed pixel points can be multiple in calibration, and the calibrated background seed pixel points can also be multiple in calibration. It can be understood that the foreground seed pixel points refer to pixel points belonging to cytoplasm or nucleus, that is, the calibrated foreground seed pixel points may include cytoplasm seed pixel points and nucleus seed pixel points, and the background seed pixel points refer to pixel points that are neither cytoplasm nor nucleus. When calibration is performed, calibration can be performed in any possible manner, for example, by clicking a mouse, and the embodiment of the present invention does not specifically limit the calibration manner, as long as foreground seed pixel points and background seed pixel points can be calibrated.
After the foreground seed pixel points and the background seed pixel points are calibrated, the background seed pixel points can be used as source points of a network graph, the foreground seed pixel points are used as sink points of the network graph, and points except the background seed pixel points and the foreground seed pixel points in a cell image to be processed are used as graph nodes of the network graph, and the image to be processed is mapped into the network graph.
A schematic diagram of mapping a red blood cell image into a network map in one specific example is shown in fig. 2. As shown in fig. 2, the background seed pixel is a source point S, and the foreground seed pixel is a sink point T. And other pixel points are regarded as graph nodes except the background seed pixel points and the foreground seed pixel points. As shown in FIG. 2, connecting each graph node to a source and sink may characterize how similar the graph nodes are to the source/sink by determining the weight of its connecting edges, and how similar the graph nodes are to the source/sink by determining the weight of the connecting lines between the two graph nodes. Therefore, an energy function can be determined based on the two similarity degrees, and a corresponding energy function map can be determined.
In one example, the graph nodes are respectively connected to a source point and a sink point, a first weight of an edge from the graph node to the source point is determined according to the probability that the graph node and the source point are similar pixels, a second weight of the edge from the graph node to the sink point is determined according to the probability that the graph node and the sink point are similar pixels, and the data item is determined according to the first weight and the second weight.
It can be seen that the vertices of the network graph correspond to the pixel points of the cell image to be processed, and each vertex has two edges: t-links connecting the source (S) and sink (T) (reflecting the degree of preference for each marker) and n-links connecting the neighborhoods (reflecting the smoothing term, indicating a discontinuity between vertices).
Thus, the constructed network graph may be defined as:
G＝<V,E>
wherein,
V＝{s,t}∪P，E＝N∪p∈P{{p,s},{p,t}}。
in this embodiment, each graph node is connected to a sink (target node, i.e., foreground seed pixel point) and a source (background node) by the data item, and as shown in fig. 2, the edge weight Rp(I) Representing a first weight from the graph node P to the source point S, namely that similar pixels P exist in the background pixel region; edge weight Rp(O) second weight representing the edge of graph node P to sink T, i.e. there are similar pixels P in the foreground pixel area, where Rp(O) and Rp(I) Are all non-negative weights, i.e., are all non-negative values. Specifically, the first weight may be determined according to the similarity between the graph node P and the source point S in any possible manner, where the similarity between the graph node P and the source point S may be determined in any manner capable of representing the phase velocity between the pixel points, for example, the difference between the pixel values of the pixel points of the graph node P and the source point S is used as the similarity between the graph node P and the source point S. The manner of determining the second weight of the graph node P to the sink T according to the similarity of the graph node P to the sink T may be handled in a similar manner.
Then, according to the adjacent relation between each cell pixel point in the cell image to be processed, determining a connection edge with connection relation between each graph node in the network graph, taking the punishment that the connection edge between two graph nodes does not form a continuous edge as the third weight of the connection edge, and determining a smooth item according to the third weight. The method for determining the smooth item by specifically determining that the penalty of a continuous edge is not formed on a connecting edge between two graph nodes may be performed in any possible manner, which is not specifically limited in this embodiment.
As can be seen, the smoothing term implements a connection for each pair of adjacent pixels (p, q), the non-negative edge weight Bp,qAnd determining the punishment of edge discontinuity.
Subsequently, an energy function map of the network map may be determined from the data items and the smoothing items.
By combining the above, it can be determined that the energy function graph is mainly determined by establishing the data item and the smooth item, and by establishing the appropriate data item and the smooth item, a more accurate segmentation result can be obtained in the subsequent segmentation. Therefore, the following is an exemplary description of the process of establishing an energy function map in connection with determining data items, smoothing items.
Recording the set of each pixel point of the cell image to be processed as a pixel data set P, wherein A is { A ═ A }1…AP…A|P|Is a two-dimensional vector, element APIs the pixel P, A allocated in PPThe vector a can be segmented by representing either "target (foreground)" or "background", and the segmentation energy formula is:
E(A)＝λ·R(A)+B(A) (1)
wherein,
r (A) is a data item, and B (A) is a smoothing item.
Assuming that the data item R (A) is the punishment of assigning the pixel P to the target and the background, the punishment R assigned to the target is obtained respectivelyp("obj") and a penalty R assigned to the backgroundp("bkg")：
Rp("obj")＝-lnpr(Ip|"obj")
Rp("bkg")＝-lnpr(Ip|"bkg")
The smoothing term B (A) contains the boundary property for A segmentation, the coefficient B of whichp,q≧ 0 is considered a penalty for the discontinuity between p and q:where dist (p, q) denotes the distance between p and q, the manner in which dist (p, q) is calculated may be performed in any possible manner.
When Ip-IqWhen | is less than sigma, the similarity strength of function punishment pixel discontinuity is larger, and when | Ip-IqIf i > σ, i.e. the pixels are very different, then this penalty is small.
In combination with the network graph obtained by mapping in fig. 2, the foreground seed pixel is marked as a set "O", and the background seed pixel is marked as a set "B", so that the weight assignment of the network graph shown in fig. 2 can be shown in table 1 below.
TABLE 1
In the above table 1, the first and second,
in the segmentation of white blood cell images, white blood can be segmentedCell image segmentation is seen as a binary label problem, i.e. foreground and background constitute a typical label set. Denote the data item as Dp(.) to obtain formula (4)
Dp(fp)＝|Ip-I′p| (4)
Wherein,
I＝{Ipl P ∈ P } is true value, i.e. actual pixel value of pixel point in the cell image to be processed, I ═ I'pI | P ∈ P } is an expected value, that is, an expected pixel value of a pixel point in the cell image to be processed, in this embodiment, the expected pixel value may be a pixel value of a foreground seed pixel point, or a pixel statistic value obtained by performing statistics on each pixel point of the cell image to be processed, and P is a pixel data set.
After the foreground seed pixel points and the background seed pixel points are calibrated, the I 'can be obtained in an interactive designation or estimation mode'p. Can be obtained by obtaining I '═ I'p| P ∈ P } determines the data item in the energy function.
The smooth term represents the continuous relationship between the pixels, which may be the sum of the adjacent interaction functions of all the adjacent pixels. Let the smoothness term be V{p,q}The smoothing term V may be determined in conjunction with the selected smoothing type, e.g., everywhere smoothing, piecewise constant, piecewise smoothing, etc{p,q}In a specific form.
Assuming the type of smoothing term is set to smooth everywhere, for example, smoothing term V{p,q}Can be expressed by the following formula (5):
V{p,q}(fp,fq)＝u{p,q}|fp-fq| (5)
assuming that the type of smoothing term is set to a piecewise constant, the smoothing term V{p,q}Can be expressed by the following formula (6):
V{p,q}(fp,fq)＝u{p,q}δ(fp≠fq) (6)
assuming that the type of smoothing term is set to piecewise smoothing, smoothing term V{p,q}Can be expressed by the following formula (7):
wherein u is{p,q}And may be specifically set according to a specific problem. When u is{p,q}The smoothing energy C in the formula (7) may be Potts energy when it is constant. It can be seen that the smooth term gives punishment when two adjacent pixels are not pixel points of the same type, for example, if one pixel point is a pixel point belonging to the cell nucleus, and the other adjacent pixel point does not belong to the cell nucleus, the value of the corresponding punishment term is larger, otherwise, the value of the corresponding punishment term is smaller.
After the data item and the smoothing item have been created, an energy function can thus be determined, and the determination of the energy function on the basis of the data item and the smoothing item can be carried out in any possible way known at present. In the embodiment of image segmentation of the leukocyte image, the determined energy function can be represented by the following Potts model:
wherein k is{p,q}I.e. the smoothing term V mentioned above{p,q}T (-) is an indicator function that equals 1 when the value in the parenthesis is true, and equals 0 otherwise.
After the network graph G is established, the network graph G may be solved by using a graph theory method, in this embodiment, the network graph G may be solved by using a maximum flow minimum cut algorithm to obtain a minimum cut of the network graph G, and the cell image is segmented based on the minimum cut of the network graph G, so that a best target is separated from the cell image to be processed to implement image segmentation. The network graph G is solved by the max-flow min-cut algorithm in any way that is currently available and that may later appear.
To further verify the effect of the embodiment of the present invention, in a specific application example, a segmentation experiment is performed on 58 cell images, fig. 3 shows segmentation results of some representative cell images when the 58 cell images are segmented, for conveniently observing the segmentation effect, cytoplasm and nucleus are represented in a binary diagram in fig. 3, a first row represents a raw cell image, and a second row represents the segmentation results. In FIG. 3, the cells in the first and second rows have no adhesion and the segmentation effect is good; the latter three columns all have sticking and the segmentation results are also good. Therefore, the cell image segmentation method in the scheme of the embodiment has a good effect on segmentation of the adherent cells, solves the problem that some common cell image segmentation algorithms have poor effect on segmentation of the adherent cells, and greatly improves the speed and the accuracy.
Based on the same idea as the method, an embodiment of the present invention further provides a cell image segmentation apparatus, and fig. 4 shows a schematic structural diagram of the cell image segmentation apparatus in an embodiment.
As shown in fig. 4, the apparatus in this embodiment includes:
an image acquisition module 401, configured to acquire an image of a cell to be processed;
a calibration module 402, configured to calibrate foreground seed pixel points and background seed pixel points in the cell image to be processed;
a map mapping module 403, configured to map the to-be-processed image into a network map by using the background seed pixel points as source points of the network map and the foreground seed pixel points as sink points of the network map, where points in the to-be-processed cell image other than the background seed pixel points and the foreground seed pixel points are map nodes of the network map;
a data item determining module 404, configured to connect each graph node to a source point and a sink point respectively, determine a first weight of an edge from the graph node to the source point according to a probability that the graph node and the source point are similar pixels, determine a second weight of the edge from the graph node to the sink point according to the probability that the graph node and the sink point are similar pixels, and determine a data item according to the first weight and the second weight;
a smooth term determining module 405, configured to determine, according to an adjacent relationship between cell pixel points in the cell image to be processed, a connection edge having a connection relationship between the graph nodes in the network graph, use a penalty that a connection edge between two graph nodes does not form a continuous edge as a third weight of the connection edge, and determine a smooth term according to the third weight;
an energy function determination module 406, configured to determine an energy function map of the network map according to the data item and the smoothing item;
the graph solving module 407 is configured to solve the energy function graph of the network graph by using a maximum flow minimum cut algorithm, and obtain a minimum cut solution result;
and a map segmentation module 408, configured to determine a cell segmentation result of the cell image to be processed according to the minimal segmentation solution result.
According to the scheme of the embodiment, after the foreground seed pixel points and the background seed pixel points in the cell image to be processed are calibrated, the cell image to be processed is mapped into the network map according to the calibrated foreground seed pixel points and the calibrated background seed pixel points, the data items and the smooth items are determined according to the network map, then the appropriate energy function map is established, then the maximum flow minimum cut algorithm is adopted for solving the energy function map, and therefore the cell segmentation result is obtained.
It is to be understood that the first weight, the second weight, and the third weight are all non-negative weights.
The cell image segmentation apparatus in this embodiment can be mainly used for image segmentation of a white blood cell image. Based on the characteristics of the cells, the foreground seed pixel points can include cytoplasm seed pixel points and nucleus seed pixel points. At this time, the map mapping module 403 maps the image to be processed into the network map by using the cytoplasm seed pixel point as the sink of the network map, using the background seed pixel point as the source point of the network map, and using the cytoplasm seed pixel point as the sink of the network map. Thus, the cell division result obtained by the map division module 408 is a cell division result for cytoplasm. At this time, the map mapping module 403 further takes the cell nucleus seed pixel point as a sink of the network map, takes the background seed pixel point as a source point of the network map, and takes the foreground seed pixel point as a sink of the network map, and maps the image to be processed into the network map after the map segmentation module 408 obtains the map segmentation result. Thereby, the above process is executed again to obtain the cell segmentation result aiming at the cell nucleus.
The selected foreground seed pixel points may include more than two. At this time, the map mapping module 403 may map the image to be processed into the network map by using the background seed pixel point as a source point of the network map and one of the foreground seed pixel points as a sink point of the network map. In addition, after the graph segmentation module 408 obtains the graph segmentation result, the graph mapping module 403 maps the image to be processed into the network graph by using the next foreground seed pixel point as a sink of the network graph, using the background seed pixel point as a source point of the network graph, and using the next foreground seed pixel point as a sink of the network graph. Therefore, the process is executed for each foreground seed pixel point, and the graph segmentation result for each foreground seed pixel point is obtained in an iterative execution mode.
In an application example, as shown in fig. 4, the apparatus in this embodiment may further include:
and the graying module 4012 is configured to gray the cell image to be processed into a grayscale image.
At this time, the calibration module 402 may calibrate foreground seed pixel points and background seed pixel points in the grayscale image. After the gray image is converted, only one gray value is needed to calibrate one pixel point, so that the subsequent processing process can be simplified.
The technical features of the embodiments described above may be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the embodiments described above are not described, but should be considered as being within the scope of the present specification as long as there is no contradiction between the combinations of the technical features.
The above-mentioned embodiments only express several embodiments of the present invention, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the inventive concept, which falls within the scope of the present invention. Therefore, the protection scope of the present patent shall be subject to the appended claims.
Claims (10)
1. A cell image segmentation method is characterized by comprising the following steps:
acquiring a cell image to be processed;
calibrating foreground seed pixel points and background seed pixel points in the cell image to be processed;
mapping the to-be-processed image into a network map by taking the background seed pixel points as source points of the network map and the foreground seed pixel points as sink points of the network map, wherein points in the to-be-processed cell image except the background seed pixel points and the foreground seed pixel points are map nodes of the network map;
respectively connecting each graph node to a source point and a sink point, determining a first weight of an edge from the graph node to the source point according to the probability that the graph node and the source point are similar pixels, determining a second weight of the edge from the graph node to the sink point according to the probability that the graph node and the sink point are similar pixels, and determining a data item according to the first weight and the second weight;
determining a connection edge with a connection relation between graph nodes in the network graph according to the adjacent relation between the cell pixel points in the cell image to be processed, taking a penalty that the connection edge between two graph nodes does not form a continuous edge as a third weight of the connection edge, and determining a smooth item according to the third weight;
determining an energy function graph of the network graph according to the data items and the smooth items;
solving the energy function graph of the network graph by adopting a maximum flow minimum cut algorithm to obtain a minimum cut solving result;
and determining a cell segmentation result of the cell image to be processed according to the minimum segmentation solution result.
2. The cell image segmentation method according to claim 1, wherein the foreground seed pixels comprise cytoplasm seed pixels and nucleus seed pixels;
the step of mapping the image to be processed into the network map by taking the background seed pixel points as the source points of the network map and the foreground seed pixel points as the sink points of the network map comprises the following steps: taking cytoplasm seed pixel points as sinks of the network graph, taking the background seed pixel points as source points of the network graph and the cytoplasm seed pixel points as sinks of the network graph, and mapping the image to be processed into the network graph;
and after obtaining the image segmentation result, taking the cell nucleus seed pixel points as the sinks of the network image, returning to the step of taking the background seed pixel points as the source points of the network image and the foreground seed pixel points as the sinks of the network image, and mapping the image to be processed into the network image.
3. The cell image segmentation method according to claim 1, wherein the foreground seed pixel points include two or more;
the step of mapping the image to be processed into the network map by taking the background seed pixel points as the source points of the network map and the foreground seed pixel points as the sink points of the network map comprises the following steps: mapping the image to be processed into a network map by taking the background seed pixel points as source points of the network map and one of the foreground seed pixel points as a sink point of the network map;
and after obtaining the graph segmentation result, taking the next foreground seed pixel point as a sink of the network graph, returning to the step of taking the background seed pixel point as a source point of the network graph and the next foreground seed pixel point as a sink of the network graph, and mapping the image to be processed into the network graph.
4. The cell image segmentation method according to any one of claims 1 to 3, wherein after the cell image to be processed is obtained, before foreground seed pixel points and background seed pixel points in the cell image to be processed are calibrated, the method further comprises the steps of:
and graying the cell image to be processed into a grayscale image.
5. The cellular image segmentation method according to any one of claims 1 to 3, wherein the first weight, the second weight, and the third weight are non-negative weights.
6. A cell image segmentation apparatus, comprising:
the image acquisition module is used for acquiring a cell image to be processed;
the calibration module is used for calibrating foreground seed pixel points and background seed pixel points in the cell image to be processed;
the map mapping module is used for mapping the to-be-processed image into a network map by taking the background seed pixel points as source points of the network map and the foreground seed pixel points as sink points of the network map, wherein points in the to-be-processed cell image except the background seed pixel points and the foreground seed pixel points are map nodes of the network map;
the data item determining module is used for respectively connecting each graph node to a source point and a sink point, determining a first weight of an edge from the graph node to the source point according to the probability that the graph node and the source point are similar pixels, determining a second weight of the edge from the graph node to the sink point according to the probability that the graph node and the sink point are similar pixels, and determining a data item according to the first weight and the second weight;
the smooth item determining module is used for determining a connecting edge with a connecting relation between graph nodes in the network graph according to the adjacent relation between the pixel points of each cell in the cell image to be processed, taking a penalty that the connecting edge between two graph nodes does not form a continuous edge as a third weight of the connecting edge, and determining a smooth item according to the third weight;
the energy function determining module is used for determining an energy function graph of the network graph according to the data item and the smooth item;
the graph solving module is used for solving the energy function graph of the network graph by adopting a maximum flow minimum cut algorithm to obtain a minimum cut solving result;
and the image segmentation module is used for determining the cell segmentation result of the cell image to be processed according to the minimal segmentation solution result.
7. The cell image segmentation apparatus according to claim 6, wherein the foreground seed pixels include cytoplasm seed pixels and nucleus seed pixels;
the map mapping module is used for mapping the image to be processed into the network map by taking cytoplasm seed pixel points as sinks of the network map, taking the background seed pixel points as source points of the network map and the cytoplasm seed pixel points as sinks of the network map;
and the map mapping module is also used for mapping the image to be processed into the network map by taking the cell nucleus seed pixel points as the sinks of the network map, taking the background seed pixel points as the source points of the network map and taking the foreground seed pixel points as the sinks of the network map after the map segmentation module obtains the map segmentation result.
8. The cell image segmentation apparatus according to claim 6, wherein the foreground seed pixel points include two or more;
the image mapping module is used for mapping the image to be processed into a network map by taking the background seed pixel points as source points of the network map and one of the foreground seed pixel points as a sink point of the network map;
and the image mapping module is also used for mapping the image to be processed into the network map by taking the next foreground seed pixel point as a sink of the network map, taking the background seed pixel point as a source point of the network map and the next foreground seed pixel point as a sink of the network map after the image segmentation module obtains the image segmentation result.
9. The cellular image segmentation apparatus according to any one of claims 6 to 8, wherein:
further comprising: the graying module is used for graying the cell image to be processed into a grayscale image;
the calibration module is used for calibrating foreground seed pixel points and background seed pixel points in the gray level image.
10. The cellular image segmentation apparatus according to any one of claims 6 to 8, wherein the first weight, the second weight, and the third weight are non-negative weights.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710121108.XA CN108537767A (en) | 2017-03-02 | 2017-03-02 | Cell image segmentation method and Methods of Segmentation On Cell Images device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710121108.XA CN108537767A (en) | 2017-03-02 | 2017-03-02 | Cell image segmentation method and Methods of Segmentation On Cell Images device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108537767A true CN108537767A (en) | 2018-09-14 |
Family
ID=63489045
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710121108.XA Pending CN108537767A (en) | 2017-03-02 | 2017-03-02 | Cell image segmentation method and Methods of Segmentation On Cell Images device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108537767A (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080056577A1 (en) * | 2006-08-29 | 2008-03-06 | Leo Grady | Seed segmentation using L-infinity minimization |
| CN102163327A (en) * | 2011-04-22 | 2011-08-24 | 陈宇珂 | Medical cardiac CT (computed tomography) image segmentation method |
| CN102831607A (en) * | 2012-08-08 | 2012-12-19 | 深圳市迈科龙生物技术有限公司 | Method for segmenting cervix uteri liquid base cell image |
| CN104036497A (en) * | 2014-05-26 | 2014-09-10 | 西安电子科技大学 | Graph cut interactive image segmentation algorithm based on local coefficient of variation |
| CN104463885A (en) * | 2014-12-18 | 2015-03-25 | 江苏大学 | Partition method for multiple-sclerosis damage area |
| CN105389821A (en) * | 2015-11-20 | 2016-03-09 | 重庆邮电大学 | Medical image segmentation method based on combination of cloud module and image segmentation |
| US20160104287A1 (en) * | 2014-10-08 | 2016-04-14 | Samsung Electronics Co., Ltd. | Image processing apparatus, method of controlling image processing apparatus and medical imaging apparatus |
-
2017
- 2017-03-02 CN CN201710121108.XA patent/CN108537767A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080056577A1 (en) * | 2006-08-29 | 2008-03-06 | Leo Grady | Seed segmentation using L-infinity minimization |
| CN102163327A (en) * | 2011-04-22 | 2011-08-24 | 陈宇珂 | Medical cardiac CT (computed tomography) image segmentation method |
| CN102831607A (en) * | 2012-08-08 | 2012-12-19 | 深圳市迈科龙生物技术有限公司 | Method for segmenting cervix uteri liquid base cell image |
| CN104036497A (en) * | 2014-05-26 | 2014-09-10 | 西安电子科技大学 | Graph cut interactive image segmentation algorithm based on local coefficient of variation |
| US20160104287A1 (en) * | 2014-10-08 | 2016-04-14 | Samsung Electronics Co., Ltd. | Image processing apparatus, method of controlling image processing apparatus and medical imaging apparatus |
| CN104463885A (en) * | 2014-12-18 | 2015-03-25 | 江苏大学 | Partition method for multiple-sclerosis damage area |
| CN105389821A (en) * | 2015-11-20 | 2016-03-09 | 重庆邮电大学 | Medical image segmentation method based on combination of cloud module and image segmentation |
Non-Patent Citations (4)
| Title |
|---|
| LING ZHANG ET.AL: ""Segmentation of cytoplasm and nuclei of abnormal cells in cervicalcytology using global and local graph cuts"", 《COMPUTERIZED MEDICAL IMAGING AND GRAPHICS》 * |
| MAEDEH BEHESHTI ET.AL: ""Bio-Cell Image Segmentation using Bayes Graph-Cut Model"", 《2015 INTERNATIONAL CONFERENCE ON DIGITAL IMAGE COMPUTING: TECHNIQUES AND APPLICATIONS》 * |
| YURI BOYKOV ET AL.: ""An Experimental Comparison of Min-Cut/Max-Flow Algorithms for Energy Minimization in Vision"", 《IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE》 * |
| YURI Y.BOYKOV ET AL.: ""Interactive Graph Cuts for Optimal Boundary & Region Segmentation of Objects in N-D Images"", 《PROCEEDINGS EIGHTH IEEE INTERNATIONAL CONFERENCE ON COMPUTER VISION》 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109146948B (en) | Vision-based analysis method of crop growth phenotype parameter quantification and yield correlation | |
| CN112396619B (en) | A Small Particle Segmentation Method Based on Semantic Segmentation with Internal Complex Composition | |
| CN106340016B (en) | A kind of DNA quantitative analysis method based on microcytoscope image | |
| EP3807839B1 (en) | Deformity edge detection | |
| CN106780522B (en) | A kind of bone marrow fluid cell segmentation method based on deep learning | |
| CN104715251B (en) | A kind of well-marked target detection method based on histogram linear fit | |
| CN107680110B (en) | Three-dimensional level set segmentation method of inner ear based on statistical shape model | |
| CN110197206B (en) | Image processing method and device | |
| EP3971762A1 (en) | Method, device and system for processing image | |
| CN104951765B (en) | Remote Sensing Target dividing method based on shape priors and visual contrast | |
| CN106780503A (en) | Remote sensing images optimum segmentation yardstick based on posterior probability information entropy determines method | |
| CN105389821B (en) | It is a kind of that the medical image cutting method being combined is cut based on cloud model and figure | |
| Che et al. | The effect of growth stage and plant counting accuracy of maize inbred lines on LAI and biomass prediction | |
| CN104751461B (en) | A kind of leukocyte cell core dividing method based on histogram thresholding and low-rank representation | |
| CN107315999A (en) | A kind of tobacco plant recognition methods based on depth convolutional neural networks | |
| Bernier et al. | Automatized evaluation of the left ventricular ejection fraction from echocardiographic images using graph cut | |
| Arifin et al. | Detection of overlapping teeth on dental panoramic radiograph | |
| CN118982516B (en) | A method and system for monitoring pyrenoid algae based on binocular images | |
| Zhang et al. | Generalizability and quality control of deep learning-based 2D echocardiography segmentation models in a large clinical dataset | |
| CN108564564A (en) | Based on the medical image cutting method for improving fuzzy connectedness and more seed points | |
| CN106485714B (en) | 3-dimensional image segmenting system and its dividing method | |
| CN108537767A (en) | Cell image segmentation method and Methods of Segmentation On Cell Images device | |
| Mahender et al. | Multi-Level Thresholding for Image Segmentation on Medical Images Using Multi Otsu and Sine Cosine Optimization Algorithm. | |
| Zhao et al. | An automatic counting method of maize ear grain based on image processing | |
| CN111581308A (en) | Agricultural remote sensing monitoring data multidimensional fusion method and system |
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 |