CN109554432A - A kind of cell type analysis method, analytical equipment and electronic equipment - Google Patents
A kind of cell type analysis method, analytical equipment and electronic equipment Download PDFInfo
- Publication number
- CN109554432A CN109554432A CN201811456650.1A CN201811456650A CN109554432A CN 109554432 A CN109554432 A CN 109554432A CN 201811456650 A CN201811456650 A CN 201811456650A CN 109554432 A CN109554432 A CN 109554432A
- Authority
- CN
- China
- Prior art keywords
- area
- region
- image
- cell
- continuum
- 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
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
- G01N33/48—Biological material, e.g. blood, urine; Haemocytometers
- G01N33/50—Chemical analysis of biological material, e.g. blood, urine; Testing involving biospecific ligand binding methods; Immunological testing
- G01N33/5005—Chemical analysis of biological material, e.g. blood, urine; Testing involving biospecific ligand binding methods; Immunological testing involving human or animal cells
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Immunology (AREA)
- Molecular Biology (AREA)
- Urology & Nephrology (AREA)
- Biomedical Technology (AREA)
- Chemical & Material Sciences (AREA)
- Hematology (AREA)
- Food Science & Technology (AREA)
- Tropical Medicine & Parasitology (AREA)
- Microbiology (AREA)
- Cell Biology (AREA)
- Biotechnology (AREA)
- Medicinal Chemistry (AREA)
- Physics & Mathematics (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- General Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Pathology (AREA)
- Investigating Or Analysing Biological Materials (AREA)
Abstract
The present invention provides a kind of cell type analysis methods, comprising: chooses the region to be analyzed in blood sample image;Determine the nuclear area in the region to be analyzed;It is partitioned into the cell compartment image including individual cells image;And the cell compartment image is inputted into first nerves network model, the type of individual cells is identified by the first nerves network model.It is realized by the method for artificial intelligence and extracts cell from blood automatically and identify its type, to count the quantity of the various types of cells in blood.Existing artificial range estimation is replaced using artificial intelligence, can significantly improve working efficiency, also, artificial intelligence, under the support of a large amount of sample data, accuracy and consistency with higher improve accurately and effectively data foundation for diagnosis disease.
Description
Technical field
The present invention relates to field of medical image processing, and in particular to a kind of cell type analysis method, analytical equipment and electricity
Sub- equipment.
Background technique
The quantity of various types of cells and form can provide very valuable information in blood for diagnosis, such as sentence
When some serious disease (such as leukaemia) of breaking, the quantity of certain specific cells in blood will appear apparent difference,
Therefore, there is good diagnostic value by the statistics to various types of cells quantity in blood.It is each in existing statistics blood
The method of class cell quantity is with the naked eye to go to count and judge one by one the number amount and type of various types of cells under the microscope, in this way
Method it is not only time-consuming and laborious, workload is very heavy, and the professional people that the statistical work needs to have suitable medical knowledge
Member, the workload of each staff has just been further aggravated in this, in addition, since the growth change procedure of cell is one continuous
Process, thus be difficult accurately to judge its type by the form of cell, this will also result in different staff and has one
Fixed judgement difference.It is therefore, existing that by the method that manually counts, there are time-consuming and laborious and accuracy and consistency be not high.
Summary of the invention
It is set in view of this, the embodiment of the present application is dedicated to providing a kind of cell type analysis method, analytical equipment and electronics
It is standby, it solves the problems, such as above-mentioned manually to count time-consuming and laborious and accuracy and consistency is not high.
According to an aspect of the present invention, a kind of cell type analysis method that one embodiment of the invention provides, comprising: choose
Region to be analyzed in blood sample image;Determine the nuclear area in the region to be analyzed;It is partitioned into including single thin
The cell compartment image of born of the same parents' image, wherein the individual cells image includes the nuclear area and cytosolic domain;And
The cell compartment image is inputted into first nerves network model, individual cells are identified by the first nerves network model
Type.
In one embodiment, described to be partitioned into the cell compartment image including individual cells image and comprise determining that comprising institute
The cell boundaries of nuclear area are stated to be partitioned into the cell compartment image including individual cells image.
In one embodiment, the determination includes that the cell boundaries of the nuclear area are comprised determining that comprising described thin
The region to be split in karyon region;The image in the region to be split is inputted into nervus opticus network model, passes through described second
Neural network model determines the cell boundaries comprising the nuclear area.
In one embodiment, the region to be split is centered on the center of gravity of the nuclear area.
In one embodiment, the big 10%-15% of area of nuclear area described in the area ratio in the region to be split.
In one embodiment, the nuclear area in the determination region to be analyzed includes: and chooses to meet default face
The continuum of vitta part is the nuclear area in the region to be analyzed.
In one embodiment, the continuum for meeting pre-set color condition include: the continuum it is red, green,
Blue three Color Channel component values are respectively in preset threshold range.
In one embodiment, the nuclear area in the determination region to be analyzed includes: selection same color
Continuum;Judge whether the continuum meets pre-set color condition;When the continuum meets the pre-set color
When condition, judge whether the area of the continuum is greater than the first area threshold and is less than second area threshold value;And work as institute
When stating the area of continuum and being greater than the first area threshold and be less than second area threshold value, choose the continuum be it is described to
The nuclear area in analyzed area.
In a further embodiment, the nuclear area in the determination region to be analyzed further include: when described
When the area of continuum is greater than or equal to the second area threshold value, the continuum is divided into multiple nucleus
Region.
It is in one embodiment, described that the continuum is divided into multiple nuclear areas includes: by equal ratios
Example reduces the image of the continuum, and the continuum is divided into multiple nuclear areas.
In a further embodiment, the nuclear area in the determination region to be analyzed further include: when described
When the area of continuum is less than or equal to first area threshold, judgement includes the default third area of the continuum
Meet the pre-set color condition with the presence or absence of other in threshold range and area is respectively less than or is equal to first area threshold
Continuum;Other meet the pre-set color condition if it exists and area is respectively less than or equal to first area threshold
Meet the pre-set color condition and area within the scope of the default third area threshold is respectively less than or is equal to by continuum
All continuums of first area threshold are merged into the nuclear area.
In one embodiment, the nuclear area in the determination region to be analyzed further include: with described to be analyzed
The direction on orthogonal two boundaries in region is respectively that reference axis establishes coordinate system;When in the determining nuclear area
At least there is a point when in the reference axis, abandons the subsequent processing the nuclear area.
In one embodiment, the nuclear area in the determination region to be analyzed includes: by the area to be analyzed
The image in domain inputs third nerve network model, determines the institute in the region to be analyzed by the third nerve network model
State nuclear area.
In one embodiment, the region to be analyzed includes that blood sample on glass slide scatters the middle portion at direction both ends
Point;Preferably, the width of the middle section part is that the blood sample scatters the 1/3 of width.
In one embodiment, the region to be analyzed chosen in blood sample image includes: successively to choose multiple visuals field
One in region is used as region to be analyzed.
In one embodiment, length of the overlapping region of two adjacent area of visual field on the moving direction of the visual field is big
In pre-set length threshold;Preferably, the pre-set length threshold is 4-5 microns.
In one embodiment, one successively chosen in multiple area of visual field includes: to institute as region to be analyzed
The image for stating multiple area of visual field carries out coordinate label, is successively chosen in the multiple area of visual field according to coordinate label
One is used as the region to be analyzed.
In one embodiment, it is described by the cell compartment image input first nerves network model include: delete segmentation
Other images in the cell compartment image out in addition to the individual cells image will delete the institute of other images
It states cell compartment image and inputs the first nerves network model.
In one embodiment, the method also includes: multiple cell compartment images of coincidence are merged into an institute
State cell compartment image.
In one embodiment, the establishment step of the first nerves network model include: will be comprising the nuclear area
First sample image input nervus opticus network model, by the nervus opticus network model determine include the nucleus
The cell boundaries in region, to be partitioned into the second sample image of determining cell boundaries;The nervus opticus network model is divided
The training set of the first nerves network model is added as sample for second sample image and corresponding cell type out
In.
In one embodiment, second sample image that the nervus opticus network model is partitioned into and correspondence
Cell type to be added in the training set of the first nerves network model as sample include: to delete the nervus opticus network
Other images in second sample image that model is partitioned into addition to the individual cells image, will delete it is described other
The training of the first nerves network model is added as sample for second sample image and corresponding cell type of image
It concentrates.
In one embodiment, second sample image that the nervus opticus network model is partitioned into and correspondence
Cell type to be added in the training set of the first nerves network model as sample include: described in different color marks
The individual cells area image in second sample image that nervus opticus network model is partitioned into, will mark color
Second sample image and corresponding cell type are added in the training set of the first nerves network model as sample.
In one embodiment, the establishment step of the nervus opticus network model include: will be comprising the nuclear area
First sample image and the second sample image of corresponding determining cell boundaries the nervus opticus network is added as sample
In the training set of model.
In one embodiment, the type that individual cells are identified by the first nerves network model includes: difference
The value of the confidence that individual cells are various cell types is calculated, and determines the type of the individual cells according to the value of the confidence.
In one embodiment, the type that the individual cells are determined according to the value of the confidence includes: that maximum ought set
When letter value is greater than default confidence threshold, determine that the individual cells are type corresponding to the maximum the value of the confidence.
In one embodiment, the type that the individual cells are determined according to the value of the confidence includes: that maximum ought set
When letter value is less than or equal to default confidence threshold, maximum at least two type of the value of the confidence and corresponding the value of the confidence are marked.
In one embodiment, described to mark maximum at least two type of the value of the confidence and corresponding the value of the confidence includes: use
Different colours mark maximum at least two type of the value of the confidence and corresponding the value of the confidence.
According to another aspect of the present invention, a kind of cell type analytical equipment that one embodiment of the invention provides, comprising: choosing
Modulus block is configured to choose the region to be analyzed in blood sample image;Nucleus determining module is configured to determine described wait divide
Analyse the nuclear area in region;Cell compartment divides module, is configured to be partitioned into the cell compartment including individual cells image
Image, wherein the individual cells image includes the nuclear area and cytosolic domain;And cell type identifies mould
Block is configured to the cell compartment image inputting first nerves network model, be identified by the first nerves network model
The type of individual cells.
In one embodiment, the cell compartment segmentation module is further configured to: being determined comprising the nuclear area
Cell boundaries to be partitioned into the cell compartment image including individual cells image.
In one embodiment, the cell compartment segmentation module is further configured to: being determined comprising the nuclear area
Region to be split;The image in the region to be split is inputted into nervus opticus network model, passes through the nervus opticus network
Model determines the cell boundaries comprising the nuclear area.
In one embodiment, the region to be split is centered on the center of gravity of the nuclear area.
In one embodiment, the big 10%-15% of area of nuclear area described in the area ratio in the region to be split.
In one embodiment, the nucleus determining module is further configured to: choosing the company for meeting pre-set color condition
Continuous region is the nuclear area in the region to be analyzed.
In one embodiment, the nucleus determining module is further configured to: the red, green, blue three of the continuum
A Color Channel component value is respectively in preset threshold range.
In one embodiment, the nucleus determining module includes: selection unit, is configured to the continuous of selection same color
Region;First judging unit is configured to judge that the continuum demonstration meets pre-set color condition;Second judgment unit, structure
It makes to judge whether the area of the continuum is greater than the first face when the continuum meets the pre-set color condition
It accumulates threshold value and is less than second area threshold value;And nucleus selection unit, it is configured to be greater than the when the area of the continuum
One area threshold and be less than second area threshold value when, choose the continuum be the region to be analyzed in the nucleus
Region.
In a further embodiment, the second judgment unit is further configured to: when the area of the continuum
When more than or equal to the second area threshold value, the continuum is divided into multiple nuclear areas.
In one embodiment, the nucleus determining module is further configured to: by continuum described in scaled down
The continuum is divided into multiple nuclear areas by the image in domain.
In a further embodiment, the nucleus determining module further comprises: third judging unit is configured to work as
When the area of the continuum is respectively less than or is equal to first area threshold, judgement comprising the continuum default the
Meet the pre-set color condition with the presence or absence of other within the scope of three area thresholds and area is respectively less than or is equal to first face
The continuum of product threshold value;Other meet the pre-set color condition if it exists and area is respectively less than or is equal to first area
The continuum of threshold value, by within the scope of the default third area threshold meet the pre-set color condition and area is respectively less than
Or the nuclear area is merged into equal to all continuums of first area threshold.
In one embodiment, the nucleus determining module is further configured to: mutual with the region to be analyzed is hung down
The direction on two straight boundaries is respectively that reference axis establishes coordinate system;When in the determining nuclear area at least exist one
Point abandons the subsequent processing the nuclear area when in the reference axis.
In one embodiment, the nucleus determining module is further configured to: the image in the region to be analyzed is defeated
Enter third nerve network model, the nuclear area in the region to be analyzed is determined by the third nerve network model
Domain.
In one embodiment, the region to be analyzed includes that blood sample on glass slide scatters the middle portion at direction both ends
Point;Preferably, the width of the middle section part is that the blood sample scatters the 1/3 of width.
In one embodiment, the module of choosing is further configured to: successively choosing a work in multiple area of visual field
For region to be analyzed.
In one embodiment, length of the overlapping region of two adjacent area of visual field on the moving direction of the visual field is big
In pre-set length threshold;Preferably, the pre-set length threshold is 4-5 microns.
In one embodiment, the selection module is further configured to: being sat to the image of the multiple area of visual field
Mark label, one successively chosen in the multiple area of visual field according to coordinate label are used as the region to be analyzed.
In one embodiment, the cell type identification module is further configured to: deleting the cellular regions being partitioned into
Other images in area image in addition to the individual cells image will delete the cell compartment image of other images
Input the first nerves network model.
In one embodiment, described device further include: duplicate checking module is configured to the multiple cell compartment figures that will be overlapped
As merging into the cell compartment image.
In one embodiment, described device further include: first nerves network model establishes module, and being configured to will be comprising described
The first sample image of nuclear area inputs nervus opticus network model, includes by the nervus opticus network model determination
The cell boundaries of the nuclear area, to be partitioned into the second sample image of determining cell boundaries;By the nervus opticus net
The first nerves network mould is added as sample in second sample image and corresponding cell type that network model is partitioned into
In the training set of type.
In one embodiment, the first nerves network model is established module and is further configured to: deleting second mind
Other images in second sample image being partitioned into through network model in addition to the individual cells image, by deleting
Second sample image and corresponding cell type for stating other images are as the sample addition first nerves network model
Training set in.
In one embodiment, the first nerves network model is established module and is further configured to: with different color marks
The individual cells area image in second sample image that the nervus opticus network model is partitioned into is infused, will be marked
The training of the first nerves network model is added as sample for second sample image and corresponding cell type of color
It concentrates.
In one embodiment, described device further include: nervus opticus network model establishes module, and being configured to will be comprising described
The first sample image of nuclear area and the corresponding cell compartment image are as the sample addition nervus opticus network
In the training set of model.
In one embodiment, the cell type identification module includes: the value of the confidence computing unit, is configured to calculate single
A cell is the value of the confidence of various cell types, and the type of the individual cells is determined according to the value of the confidence.
In one embodiment, the cell type identification module further comprises: type determining units, is configured to when maximum
The value of the confidence when being greater than default confidence threshold, determine that the individual cells are type corresponding to the maximum the value of the confidence.
In one embodiment, the cell type identification module further comprises: the value of the confidence marking unit, being configured to ought be most
When big the value of the confidence is less than or equal to default confidence threshold, maximum at least two type of the value of the confidence and corresponding confidence are marked
Value.
In one embodiment, the value of the confidence marking unit is further configured to: marking the value of the confidence most with different colours
Big at least two types and corresponding the value of the confidence.
According to another aspect of the present invention, a kind of electronic equipment that one embodiment of the invention provides, comprising: processor;It deposits
Reservoir;And the computer program instructions of storage in the memory, the computer program instructions are by the processor
The processor is made to execute as above described in any item methods when operation.
According to another aspect of the present invention, a kind of computer program product that one embodiment of the invention provides, including calculate
Machine program instruction, the computer program instructions execute the processor described in any one as above
Method.
According to another aspect of the present invention, a kind of computer readable storage medium that one embodiment of the invention provides, thereon
The step of being stored with computer program, as above any one the method realized when the computer program is executed by processor.
Cell type analysis method provided by the embodiments of the present application, by choosing the area to be analyzed in blood sample image
Domain determines the nuclear area in the region to be analyzed, determines the cell boundaries comprising the nuclear area to be partitioned into
Cell compartment image including individual cells image, and cell compartment image input first nerves network model is identified
The type of individual cells.Artificial intelligence is realized through the above steps and replaces artificial segmentation and identification cell, greatly improved point
The working efficiency of analysis, and being learnt by a large amount of real data, can step up analysis accuracy and result it is consistent
Property, reliable data, which are provided, for diagnosis supports.
Detailed description of the invention
Fig. 1 show the schematic diagram of the blood sample image of analysis of the embodiment of the present invention.
Fig. 2 show a kind of flow chart of cell type analysis method of one embodiment of the invention offer.
Fig. 3 show another embodiment of the present invention provides a kind of cell type analysis method flow chart.
Fig. 4 show the flow chart of the determination nuclear area method of one embodiment of the invention offer.
Fig. 5 show another embodiment of the present invention provides a kind of cell type analysis method flow chart.
Fig. 6 show another embodiment of the present invention provides a kind of cell type analysis method flow chart.
Fig. 7 show the flow chart of the determination cell boundaries method of one embodiment of the invention offer.
Fig. 8 show the flow chart of the identification cell type method of one embodiment of the invention offer.
Fig. 9 show another embodiment of the present invention provides a kind of cell type analysis method flow chart.
Figure 10 show a kind of flow chart of Establishment of Neural Model method of one embodiment of the invention offer.
Figure 11 show another embodiment of the present invention provides a kind of Establishment of Neural Model method flow chart.
Figure 12 show another embodiment of the present invention provides a kind of Establishment of Neural Model method flow chart.
Figure 13 show another embodiment of the present invention provides a kind of Establishment of Neural Model method flow chart.
Figure 14 show another embodiment of the present invention provides a kind of cell type analysis method flow chart.
Figure 15 show a kind of structural schematic diagram of cell type analytical equipment of one embodiment of the invention offer.
Figure 16 show another embodiment of the present invention provides a kind of cell type analytical equipment structural schematic diagram.
Figure 17 show another embodiment of the present invention provides a kind of cell type analytical equipment structural schematic diagram.
Figure 18 show another embodiment of the present invention provides a kind of cell type analytical equipment structural schematic diagram.
Figure 19 show the structural schematic diagram of the electronic equipment of one embodiment of the invention offer.
Specific embodiment
Below in conjunction with the attached drawing in the embodiment of the present application, technical solutions in the embodiments of the present application carries out clear, complete
Site preparation description, it is clear that the described embodiment is only a part of the embodiment of the present invention, instead of all the embodiments.Based on this
Embodiment in invention, every other reality obtained by those of ordinary skill in the art without making creative efforts
Example is applied, shall fall within the protection scope of the present invention.
Blood cell is roughly divided into three classes: red blood cell, leucocyte and blood platelet, and wherein leucocyte is again by lymphocyte, list
Nucleus, neutrophil leucocyte (being subdivided into band form neutrophilic granulocyte and neutrophilic segmented granulocyte), eosinophil, basophilic
Property granulocyte five subclass composition.It is white thin for different subclass in biological analysis field, especially clinical examination analysis field
Born of the same parents carry out accurately classification for diagnosing and research has very important effect, and the variation of each subclass leucocyte content exists
Clinically it is used as the important indicator of diagnosing and treating.For example, acute bacterial infections will cause neutrophilic leukocytosis, it is parasitic
Insect infection and anaphylactia etc. will cause eosinophilia.
Currently, white blood cell count(WBC) method is usually to estimate counting method, range estimation counting method is to be counted on tally with microscopy
Several methods can directly be counted without processing, will first can also be counted again after haemocyte nuclear targeting with coloring agent.
However, range estimation counting method needs the personnel of profession to complete, and therefore, heavy workload, low efficiency, and result is inaccurate, consistent
Property is not high.
For the disadvantages mentioned above for solving range estimation counting method, the embodiment of the present application provide a kind of cell type analysis method,
Analytical equipment and electronic equipment extract leucocyte automatically by the method realization of artificial intelligence from blood and identify its type,
To count the quantity of all kinds of leucocytes in blood.Existing artificial range estimation is replaced by artificial intelligence, it can be significantly
Working efficiency is improved, also, artificial intelligence is under the support of a large amount of sample data, accuracy and consistency with higher,
Accurately and effectively data foundation is improved for diagnosis disease.
After describing the basic principle of the application, carry out the various non-limits for specifically introducing the application below with reference to the accompanying drawings
Property embodiment processed.
Fig. 1 show the schematic diagram of the blood sample image of analysis of the embodiment of the present invention, wherein 1 is region to be analyzed, 2
It is nuclear area for individual cells image, 3,4 is cell compartment image.It should be appreciated that Fig. 1 is in order to more explicit
The illustrative blood sample image for illustrating the implementation of the embodiment of the present application and providing, not blood sample in practical application
True picture.
Fig. 2 show a kind of flow chart of cell type analysis method of one embodiment of the invention offer.As shown in Fig. 2,
The cell type analysis method includes the following steps:
Step 110: choosing the region to be analyzed 1 in blood sample image.Choose the blood sample being placed on glass slide
Image in region to be analyzed 1, analyze the type and quantity of cell in region 1 to be analyzed.
Region 1 to be analyzed can be the blood sample on glass slide and scatter the middle section part at direction both ends;Preferably, middle section
Partial width is that blood sample scatters the 1/3 of width.Blood sample can cause to scatter when scattering on glass slide because of extruding
The cell quantity of the initial segment in direction is less and the accumulation of the cell quantity of concluding paragraph is more, and the cell quantity of middle section part is moderate
And it is more uniform to scatter, and therefore, can choose middle portion and be allocated as preferably reflecting blood sample for region 1 to be analyzed
In cell distribution situation.
It should be appreciated that the embodiment of the present application can according to the demand of practical application scene choose blood sample whole or
Partial region is as region to be analyzed, as long as selected region to be analyzed can accurately reflect cell in blood sample
Respectively state, the embodiment of the present application for region to be analyzed specific location without limitation.
Region to be analyzed 1 in blood sample image can also be by successively choosing a conduct in multiple area of visual field
Region 1 to be analyzed.It, can be directly by the blood under microscope since the image that blood sample is showed under the microscope is bigger
Sample image scanning after choose region to be analyzed (such as above-mentioned middle section part) or first choose region to be analyzed scan again it is selected
The image in region to be analyzed, and treat analyzed area image and analyzed;Blood sample to be analyzed can also will be chosen to divide
For multiple area of visual field, one in multiple area of visual field is successively chosen as region to be analyzed, and to region to be analyzed
Image is analyzed, and the result in multiple regions to be analyzed is aggregated to form to final analysis result.By by blood sample image
It is divided into multiple regions to be analyzed, it is possible to reduce the data volume analyzed every time reduces the present processes to realization device
It is required that.It should be appreciated that can be selected using any one of the above under the premise of not considering the configuration and the speed of service of realization device
The method for taking region to be analyzed, the embodiment of the present application is for choosing the specific method in region to be analyzed without limitation.
In order to preferably realize the analysis of multiple area of visual field, coordinate label can be carried out to multiple field-of-view images, according to
Coordinate marks one successively chosen in multiple area of visual field to be used as region 1 to be analyzed.It is marked by setting coordinate, it can be accurate
The multiple area of visual field of realization continuous analysis, avoid the omission in partial visual field region and part cell.
Optionally, length of the overlapping region of two adjacent area of visual field on the moving direction of the visual field is greater than preset length
Threshold value;Preferably, pre-set length threshold is 4-5 microns.The diameter of usual leucocyte is 4-5 microns, therefore, is in order to prevent
The omission of cell between two adjacent area of visual field, the overlapping region of two adjacent area of visual field at least can include one
A cell, i.e., length of the overlapping region of two adjacent area of visual field on the moving direction of the visual field are greater than the straight of a cell
Diameter.
Step 120: determining the nuclear area 3 in region 1 to be analyzed.Cell is determined in selected region to be analyzed 1
The cell of core region 3, cell-free core will not generate canceration, and therefore, only can need to pay close attention in the diseases such as analysis cancer has core
Cell.
In the embodiment of the present application, as shown in figure 3, step 120 may include sub-step:
Step 1201: selection meets the continuum of pre-set color condition for the nuclear area 3 in region 1 to be analyzed.
In order to distinguish the various pieces in blood sample, it will usually be dyed to blood sample, using various pieces in coloring agent
Lower the shown different colours of effect can distinguish the various pieces such as nucleus, cytoplasm, cell, impurity in blood sample.
Pre-set color condition is arranged according to the color that stained cells core is shown in the embodiment of the present application, and selection meets the pre-set color item
The continuum of part is nuclear area.Staining method can be include the existing colouring method such as Wright's staining, the application is real
Example is applied for staining method without limitation.
Optionally, pre-set color condition may is that red (R), green (G), blue (B) three colors of the continuum of selection are logical
Road component value is respectively in preset threshold range.Rgb color mode is a kind of color standard of industry, is by red
(R), green (G), the variation of blue (B) three Color Channels and their mutual superpositions obtain miscellaneous color
, i.e., each color can be uniquely determined by it in the component value of red (R), green (G), blue (B) three Color Channels, because
This, can choose color area required for our by the way that the threshold range of red (R), green (G), blue (B) three component values is arranged
Domain, i.e. nuclear area.
Optionally, as shown in figure 4, step 120 can also include following sub-step:
Step 1202: selecting the continuum of same color.
Step 1203: judging whether continuum meets pre-set color condition, if so, going to step 1204, otherwise, abandon
Subsequent step to the continuum simultaneously goes to step 1202.
Step 1204: the area of continuum and the size relation of the first area threshold, second area threshold value are judged, when even
When the area in continuous region is greater than the first area threshold and is less than second area threshold value, 1205 are gone to step;When the area of continuum
When more than or equal to second area threshold value, 1206 are gone to step;Otherwise, 1207 are gone to step.
Step 1205: choosing continuum is the nuclear area in region to be analyzed.When the color of a continuum
When meeting pre-set color condition, it can determine that the continuum is nuclear area, but due to selected region to be analyzed
Image is flat image, and multiple cells are likely to result in that multiple nucleus are selected or the leaflet core of cell may when being laminated
It can be chosen for different multiple nuclear areas.In order to avoid above-mentioned erroneous judgement, the embodiment of the present application is to the default face of satisfaction
The continuum of vitta part is further to be judged, by judging whether the area of continuum meets the big of individual cells core
Small, i.e., whether the area of continuum is greater than the first area threshold and is less than second area threshold value, if meeting the preset condition,
Determine that the continuum is single nuclear area.
Step 1206: continuum is divided into multiple nuclear areas.When the area of continuum is greater than or equal to the
When two area thresholds, i.e. the area of continuum has been more than the area of nucleus, then it is assumed that have in the continuum it is multiple (at least
Two) therefore the continuum is divided into multiple nuclear areas, and to multiple nuclear areas point by the nucleus of stacking
It is not analyzed.
The method that continuum is divided into multiple nuclear areas be may is that into the figure by scaled down continuum
Continuum is divided into multiple nuclear areas by picture.It, can will be more in image by the image of scaled down continuum
The width of the join domain of a nucleus gradually reduces, until join domain disconnect, that is, realize continuum is divided into it is more
A nuclear area.
It should be appreciated that the embodiment of the present application can choose other modes for continuum point according to practical application scene
Multiple nuclear areas are cut into, continuum for being divided into the concrete mode of multiple nuclear areas not by the embodiment of the present application
It limits.
Step 1207: judgement is default comprising meeting within the scope of the default third area threshold of continuum with the presence or absence of other
Color condition and area are respectively less than or otherwise, turn step if so, going to step 1208 equal to the continuum of the first area threshold
Rapid 130.
Step 1208: meet pre-set color condition and the area within the scope of default third area threshold are respectively less than or are equal to
All continuums of first area threshold are merged into a nuclear area.When a certain range (default third area threshold
Be worth range) in multiple continuums area be respectively less than or be equal to the first area threshold when, then it is assumed that multiple continuum
For a leaflet nuclear area (cell B as shown in Figure 1), a nuclear area is merged into multiple continuum.
Wherein, the area value that third area threshold can be leaflet nucleus is preset.
Optionally, as shown in figure 5, step 120 can also include sub-step:
Step 1209: being analysed to the image input third nerve network model in region 1, pass through third nerve network model
Determine the nuclear area 3 in region 1 to be analyzed.By establishing third nerve network model, the image for being analysed to region is defeated
Enter third nerve network model, determines nuclear area using third nerve network model.It can be with by a large amount of sample training
The accuracy for effectively improving determining nuclear area, avoid because dyeing effect is bad cause determine nuclear area certainty
The problem of reduction.
Optionally, third nerve network model can be convolutional neural networks model.It should be appreciated that third nerve network mould
Type can also be other neural network models, the embodiment of the present application to the type of third nerve network model without limitation.
Step 130: being partitioned into the cell compartment image 4 including individual cells image 2, wherein individual cells image includes
Nuclear area and cytosolic domain (cell A as shown in Figure 1).After determining nuclear area 3, determine to include the nucleus
The cell boundaries of the cell in region 3 obtain individual cells image 2, to be partitioned into the cell compartment including individual cells image 2
Image 4.Wherein, cell compartment image 4 may include individual cells image 2 and be greater than individual cells image 2, cell compartment image
4 can also be exactly individual cells image 2, as long as cell compartment image 4 includes complete individual cells image 2, the application
Embodiment for cell compartment image concrete shape without limitation.
Optionally, as shown in fig. 6, step 130 may include sub-step:
Step 1301: determining the cell boundaries comprising nuclear area 3.By determining that cell boundaries are available complete
Individual cells image, so as to be partitioned into the cell compartment image including individual cells image.
Optionally, as shown in fig. 7, step 130 may include following sub-step:
Step 1302: determining the region to be split comprising nuclear area.
Step 1303: the image in region to be split being inputted into nervus opticus network model, passes through nervus opticus network model
Determine the cell boundaries comprising nuclear area.Determine one include nuclear area region to be split, and by area to be split
The image in domain inputs nervus opticus network model, determines the cell side comprising nuclear area using nervus opticus network model
Boundary.Wherein, region to be split can be above-mentioned cell compartment, may not be, as long as region to be split includes complete single
A cell, the embodiment of the present application for region to be split concrete shape without limitation.
Optionally, region to be split can be centered on the center of gravity of nuclear area 3.Small as far as possible includes cell in order to obtain
The center in the region to be split of core region, region to be split can be set as the center of gravity of nuclear area, i.e., region to be split is
Equal proportion expands on the basis of nuclear area.Certainly, region to be split may be the other shapes such as rectangular.
Optionally, the area in region to be split can be 10%-15% bigger than the area of nuclear area.According to usual cell
Area and its nucleus area corresponding relationship, the area in region to be split can be set to the face than nuclear area
The big 10%-15% of product, to ensure that region to be split includes nuclear area.
Step 140: cell compartment image 4 being inputted into first nerves network model, is identified by first nerves network model
The type of individual cells.It is after being partitioned into including the cell compartment image 4 of individual cells image 2, the cell compartment image 4 is defeated
Enter first nerves network model, the type of the individual cells is identified using first nerves network model.
Optionally, as shown in figure 8, step 140 may include sub-step:
Step 1401: deleting other images in the cell compartment image 4 being partitioned into addition to individual cells image 2.
Step 1402: the cell compartment image 4 for deleting other images is inputted into first nerves network model.
Other images (including impurity and other cells etc.) in cell compartment image 4 in addition to individual cells image are deleted
Except being in order to exclude interference of other images to the type of identification individual cells, to improve the accuracy and identification of identification
Efficiency.
It optionally, is respectively that reference axis establishes coordinate system with the direction on orthogonal two boundaries in region 1 to be analyzed
(as shown in Figure 1);When at least there is a point in certain nucleus region 3 in reference axis, nuclear area can be removed
Domain 3 (cell C as shown in Figure 1).I.e. when the nuclear area of cell C and the contour connection in region to be analyzed, then show to wrap
Cell C containing the nucleus some (do not include complete cell in the region to be analyzed in the region 1 to be analyzed
C), therefore, the subsequent processing to the nuclear area is abandoned, chooses a new nuclear area, again to avoid because of cell
Not exclusively lead to the problem of identification inaccuracy.
It is realized by the method for artificial intelligence and extracts leucocyte from blood automatically and identify its type, to count bleeding
The quantity of all kinds of leucocytes in liquid.Existing artificial range estimation is replaced using artificial intelligence, can significantly improve working efficiency,
Also, artificial intelligence is under the support of a large amount of sample data, accuracy and consistency with higher, is diagnosis disease
Improve accurately and effectively data foundation.
Fig. 9 show another embodiment of the present invention provides a kind of cell type analysis method flow chart.Such as Fig. 9 institute
Show, the present embodiment may also include the steps of: on the basis of the above embodiments
Step 135: multiple cell compartment images of coincidence are merged into a cell compartment image.When the cell being partitioned into
When there are multiple cell compartment picture registrations in region, then illustrate that multiple cells that multiple cell compartment image is included are one
Cell, wherein the nucleus of the cell is leaflet core, then multiple cell compartment image is merged into a cell compartment figure
Picture prevents from repeating to count the cell and cause final statistical result.
Figure 10 show a kind of flow chart of Establishment of Neural Model method of one embodiment of the invention offer.Such as Figure 10
It is shown, the establishment step of first nerves network model can include:
Step 210: the first sample image comprising nuclear area being inputted into nervus opticus network model, passes through the second mind
The cell boundaries comprising nuclear area 3 are determined through network model, to be partitioned into the second sample image of determining cell boundaries.It is logical
It crosses nervus opticus network to go out to determine the second sample image of cell boundaries by first sample image segmentation, that is, passes through nervus opticus net
The segmentation of network model realization individual cells.
Step 220: the second sample image and corresponding cell type that nervus opticus network model is partitioned into are as sample
In the training set of this addition first nerves network model.Preamble mould of the nervus opticus network model as first nerves network model
Type, the second sample image that nervus opticus network is partitioned into is as the input of first nerves network model, corresponding cell class
Output of the type as first nerves network model, using the input-output as the training set of first nerves network model in sample
This, the identification of individual cells is realized by first nerves network model.
By the way that the segmentation of individual cells and identification is complete by nervus opticus network model and first nerves network model respectively
At the complexity and calculation amount of each neural network model being reduced, to can also improve the efficiency of analysis.
Figure 11 show another embodiment of the present invention provides a kind of Establishment of Neural Model method flow chart.Such as figure
Shown in 11, above-mentioned steps 220 can also include:
Step 221: deleting in the second sample image that nervus opticus network model is partitioned into addition to individual cells image 2
Other images.
Step 222: being added first using the second sample image for deleting other images and corresponding cell type as sample
In the training set of neural network model.
By other image-erasings in the second sample image in addition to individual cells image, and will delete except other images
Second sample image and corresponding cell type are added in the training set of first nerves network model as sample, eliminate other
Interference of the image to the type of identification individual cells, to improve the accuracy of identification and the efficiency of identification.
Figure 12 show another embodiment of the present invention provides a kind of Establishment of Neural Model method flow chart.Such as figure
Shown in 12, above-mentioned steps 220 can also include:
Step 223: described in the second sample image 4 being partitioned into different color mark nervus opticus network models
Individual cells area image.
Step 224: first nerves are added using the second sample image for marking color and corresponding cell type as sample
In the training set of network model.
The individual cells area image in the second sample image is marked using different colors, can clearly be distinguished different
Boundary between individual cells region (the individual cells region especially closed on), avoids interfering with each other, to improve identification
Accuracy.
Optionally, first nerves network model can be convolutional neural networks model.It should be appreciated that first nerves network mould
Type can also be other neural network models, the embodiment of the present application to the type of first nerves network model without limitation.
Figure 13 show another embodiment of the present invention provides a kind of Establishment of Neural Model method flow chart.Such as figure
Shown in 13, the establishment step of nervus opticus network model can include:
Step 310: by the first sample image comprising nuclear area 3 and the second sample of corresponding determining cell boundaries
Image is added in the training set of nervus opticus network model as sample.
Step 320: the second sample image and corresponding cell type that nervus opticus network model is partitioned into are as sample
In the training set of this addition first nerves network model.
By establishing nervus opticus network model and first nerves network model, the segmentation of individual cells and identification are distinguished
It is completed by nervus opticus network model and first nerves network model, reduces the complexity and meter of each neural network model
Calculation amount, to can also improve the efficiency of analysis.
Optionally, nervus opticus network model can be convolutional neural networks model.It should be appreciated that nervus opticus network mould
Type can also be other neural network models, the embodiment of the present application to the type of nervus opticus network model without limitation.
Optionally, first sample image and/or the second sample image can be obtained by way of manually demarcating.Due to
The acquisition modes of one sample image and the second sample image are simple compared to the sample acquisition mode in first nerves network model,
Therefore, the mode manually obtained can also be chosen in the embodiment of the present invention.
Figure 14 show another embodiment of the present invention provides a kind of cell type analysis method flow chart.Such as Figure 14 institute
Show, the present embodiment may also include that on the basis of Fig. 2 corresponding embodiment
Step 150: calculating separately the value of the confidence that individual cells are various cell types, and determining single thin according to the value of the confidence
The type of born of the same parents.To each individual cells, the value of the confidence that it is various cell types is calculated separately, and determine according to each the value of the confidence
The type of individual cells.It should be appreciated that existing probability can be used by calculating the value of the confidence that individual cells are certain cell type
Any one method in statistics, the embodiment of the present application to calculate the value of the confidence specific method without limitation.Usual the value of the confidence is one
A percent value, to describe the determination degree that individual cells are this kind of cell type.
Optionally, when maximum the value of the confidence is greater than default confidence threshold, it may be determined that individual cells are maximum the value of the confidence
Corresponding type.Confidence threshold (such as can be set to 90%-95%) is preset, after calculating each the value of the confidence,
If maximum the value of the confidence is greater than the confidence threshold, then it is assumed that determine that the individual cells are that maximum the value of the confidence is corresponding thin substantially
Born of the same parents' type.
Optionally, when maximum the value of the confidence is less than or equal to default confidence threshold, it is maximum extremely that the value of the confidence can be marked
Few two types and corresponding the value of the confidence.When maximum the value of the confidence is less than or equal to default confidence threshold, then it is assumed that this is single
The type of cell is unable to judge accurately, and maximum at least two type of the value of the confidence and corresponding the value of the confidence are marked, by
It is artificial to determine, it avoids judging by accident and influencing final statistical result.It should be appreciated that the embodiment of the present application can mark preset quantity
Type can also preset a threshold value (such as 40%), and the value of the confidence is only marked to be greater than the type of the threshold value, the embodiment of the present application pair
The quantity and mode of label are without limitation.
Preferably, maximum at least two type of the value of the confidence and corresponding the value of the confidence can also be marked with different colours.
Cell type and the value of the confidence are marked by different colours, more convenient can manually be checked.
Figure 15 show a kind of structural schematic diagram of cell type analytical equipment of one embodiment of the invention offer.Such as Figure 15
Shown, which includes:
Module 81 is chosen, is configured to choose the region to be analyzed in blood sample image;Nucleus determining module 82, construction
For the nuclear area in determination region to be analyzed;Cell compartment divides module 83, is configured to be partitioned into including individual cells figure
The cell compartment image of picture, wherein the individual cells image includes the nuclear area and cytosolic domain;And cell
Type identification module 84 is configured to cell compartment image inputting first nerves network model, passes through first nerves network model
Identify the type of individual cells.
It is realized by the method for artificial intelligence and extracts leucocyte from blood automatically and identify its type, to count bleeding
The quantity of all kinds of leucocytes in liquid.Existing artificial range estimation is replaced using artificial intelligence, can significantly improve working efficiency,
Also, artificial intelligence is under the support of a large amount of sample data, accuracy and consistency with higher, is diagnosis disease
Improve accurately and effectively data foundation.
In one embodiment, choosing module 81 can be further configured to:
Region 1 to be analyzed can be the blood sample on glass slide and scatter the middle section part at direction both ends;Preferably, middle section
Partial width is that blood sample scatters the 1/3 of width.Blood sample can cause to scatter when scattering on glass slide because of extruding
The cell quantity of the initial segment in direction is less and the accumulation of the cell quantity of concluding paragraph is more, and the cell quantity of middle section part is moderate
And it is more uniform to scatter, and therefore, can choose middle portion and be allocated as preferably reflecting blood sample for region 1 to be analyzed
In cell distribution situation.
Region to be analyzed 1 in blood sample image can also be by successively choosing a conduct in multiple area of visual field
Region 1 to be analyzed.Directly region to be analyzed will can be chosen after the blood sample image scanning under microscope or first choose wait divide
The image in selected region to be analyzed is scanned in analysis region again, and treats analyzed area image and analyzed;Can also will choose to
The blood sample of analysis is divided into multiple area of visual field, successively chooses one in multiple area of visual field as area to be analyzed
Domain, and area image to be analyzed is analyzed, the result in multiple regions to be analyzed is aggregated to form to final analysis result.
By the way that blood sample image is divided into multiple regions to be analyzed, it is possible to reduce the data volume analyzed every time reduces the application
Requirement of the method to realization device.
Preferably, in order to preferably realize the analyses of multiple area of visual field, coordinate mark can be carried out to multiple field-of-view images
Note, one successively chosen in multiple area of visual field according to coordinate label are used as region 1 to be analyzed.
Optionally, length of the overlapping region of two adjacent area of visual field on the moving direction of the visual field is greater than preset length
Threshold value;Preferably, pre-set length threshold is 4-5 microns.The diameter of usual leucocyte is 4-5 microns, therefore, is in order to prevent
The omission of cell between two adjacent area of visual field, the overlapping region of two adjacent area of visual field at least can include one
A cell, i.e., length of the overlapping region of two adjacent area of visual field on the moving direction of the visual field are greater than the straight of a cell
Diameter.
In one embodiment, nucleus determining module 82 can be further configured to:
Selection meets the continuum of pre-set color condition for the nuclear area 3 in region 1 to be analyzed.The application is implemented
Pre-set color condition is arranged according to the color that stained cells core is shown in example, chooses the continuum for meeting the pre-set color condition
Domain is nuclear area.
Optionally, pre-set color condition may is that red (R), green (G), blue (B) three colors of the continuum of selection are logical
Road component value is respectively in preset threshold range.
Optionally, nucleus determining module 82 may further include: selection unit, for selecting the continuous of same color
Region;First judging unit, for judging whether continuum meets pre-set color condition;Second judgment unit, for when company
When continuous region meets pre-set color condition, judge whether the area of continuum is greater than the first area threshold and is less than second area
Threshold value;And nucleus selection unit, it is greater than the first area threshold for the area when continuum and is less than second area threshold
When value, choosing continuum is the nuclear area in region to be analyzed.In order to avoid above-mentioned erroneous judgement, the embodiment of the present application
It is further to the continuum for meeting pre-set color condition to be judged.
Optionally, second judgment unit can be further configured to: when the area of continuum is greater than or equal to the second face
When product threshold value, continuum is divided into multiple nuclear areas.When the area of continuum is greater than or equal to second area threshold
When value, i.e. the area of continuum has been more than the area of nucleus, then it is assumed that has multiple (at least two) layers in the continuum
Therefore the continuum is divided into multiple nuclear areas, and divide respectively multiple nuclear areas by folded nucleus
Analysis.
The method that continuum is divided into multiple nuclear areas be may is that into the figure by scaled down continuum
Continuum is divided into multiple nuclear areas by picture.
Optionally, nucleus determining module 82 may further include: third judging unit, for working as the face of continuum
When product is less than or equal to first area threshold, whether judgement within the scope of the default third area threshold of continuum comprising depositing
Other meet pre-set color condition and area be respectively less than or equal to the first area threshold continuum;Other meet if it exists
The pre-set color condition and area be respectively less than or equal to first area threshold continuum, by default third area threshold
Meet pre-set color condition and area within the scope of value are respectively less than or close equal to all continuums of first area threshold
And at a nuclear area.
In one embodiment, nucleus determining module 82 can be further configured to:
Direction with orthogonal two boundaries in region 1 to be analyzed is respectively that reference axis establishes coordinate system (such as Fig. 1 institute
Show);When at least there is a point in certain nucleus region 3 in reference axis, can abandon to the nuclear area
Subsequent processing (cell C as shown in Figure 1) chooses a nuclear area again, avoids causing to identify because cell is incomplete
The problem of inaccuracy.
In one embodiment, nucleus determining module 82 can also be further configured to: directly be analysed to the figure in region 1
As input third nerve network model, the nuclear area 3 in region 1 to be analyzed is determined by third nerve network model.It is logical
Crossing a large amount of sample training can effectively improve the accuracy of determining nuclear area, avoid causing really because dyeing effect is bad
The problem of determining the certainty of nuclear area reduces.
Optionally, third nerve network model can be convolutional neural networks model.
In one embodiment, cell compartment segmentation module 83 can also be further configured to:
Determine the cell boundaries comprising nuclear area 3.By determining the available complete individual cells of cell boundaries
Image, so as to be partitioned into the cell compartment image including individual cells image.
Optionally it is determined that the region to be split comprising nuclear area;The image in region to be split is inputted into nervus opticus
Network model determines the cell boundaries comprising nuclear area by nervus opticus network model.Determine that includes a nucleus
The region to be split in region, and the image in region to be split is inputted into nervus opticus network model, utilize nervus opticus network mould
Type determines the cell boundaries comprising nuclear area.
Optionally, region to be split can be centered on the center of gravity of nuclear area 3.Small as far as possible includes cell in order to obtain
The center in the region to be split of core region, region to be split can be set as the center of gravity of nuclear area, i.e., region to be split is
Equal proportion expands on the basis of nuclear area.
Optionally, the area in region to be split can be 10%-15% bigger than the area of nuclear area.According to usual cell
Area and its nucleus area corresponding relationship, the area in region to be split can be set to the face than nuclear area
The big 10%-15% of product, to ensure that region to be split includes nuclear area.
In one embodiment, cell type identification module 84 can be further configured to:
Other images in the cell compartment image 4 being partitioned into addition to individual cells image 2 are deleted, other figures will be deleted
The cell compartment image 4 of picture inputs first nerves network model.By its in cell compartment image 4 in addition to individual cells image
His image-erasing is the interference in order to exclude other images to the type of identification individual cells, to improve the accuracy of identification
And the efficiency of identification.
In one embodiment, cell type identification module 84 may include: the value of the confidence computing unit, for calculating separately list
A cell is the value of the confidence of various cell types, and the type of individual cells is determined according to the value of the confidence.To each individual cells, divide
The value of the confidence that it is various cell types is not calculated, and the type of individual cells is determined according to each the value of the confidence.
Optionally, cell type identification module 84 may further include: type determining units, for working as maximum confidence
When value is greater than default confidence threshold, determine that individual cells are type corresponding to maximum the value of the confidence.Preset confidence threshold
(such as can be set to 90%-95%), after calculating each the value of the confidence, if maximum the value of the confidence is greater than the confidence threshold
When, then it is assumed that determine that the individual cells are the corresponding cell type of maximum the value of the confidence substantially.
Optionally, cell type identification module 84 further comprises: the value of the confidence marking unit, for working as maximum the value of the confidence
When less than or equal to default confidence threshold, maximum at least two type of the value of the confidence and corresponding the value of the confidence can be marked.When most
When big the value of the confidence is less than or equal to default confidence threshold, then it is assumed that the type of the individual cells is unable to judge accurately, and will be set
Maximum at least two type of letter value and corresponding the value of the confidence are marked, and by manually determining, avoid judging by accident and influencing finally
Statistical result.
Preferably, the value of the confidence marking unit is also configured as: marking the value of the confidence maximum at least two with different colours
Seed type and corresponding the value of the confidence.Cell type and the value of the confidence are marked by different colours, more convenient can manually be checked.
Figure 16 show another embodiment of the present invention provides a kind of cell type analytical equipment structural schematic diagram.Such as figure
Shown in 16, which can also include:
Duplicate checking module 85 is configured to the multiple cell compartment images being overlapped merging into a cell compartment image.When point
When there are multiple cell compartment picture registrations in the cell compartment cut out, then illustrate that multiple cell compartment image is included multiple
Cell is a cell, wherein the nucleus of the cell is leaflet core, then merges into multiple cell compartment image one thin
Born of the same parents' area image prevents from repeating to count the cell and cause final statistical result.
Figure 17 show another embodiment of the present invention provides a kind of cell type analytical equipment structural schematic diagram.Such as figure
Shown in 17, which can also include:
First nerves network model establishes module 86, and being configured to will be comprising the first sample image input the of nuclear area
Two neural network models determine the cell boundaries comprising nuclear area 3 by nervus opticus network model, to be partitioned into determination
Second sample image of cell boundaries;And the second sample image and corresponding cell for being partitioned into nervus opticus network model
Type is added in the training set of first nerves network model as sample.
In one embodiment, first nerves network model is established module 86 and can also be further configured to:
Delete other figures in the second sample image that nervus opticus network model is partitioned into addition to individual cells image 2
Picture;And first nerves network mould is added using the second sample image for deleting other images and corresponding cell type as sample
In the training set of type.
By other image-erasings in the second sample image in addition to individual cells image, and will delete except other images
Second sample image and corresponding cell type are added in the training set of first nerves network model as sample, eliminate other
Interference of the image to the type of identification individual cells, to improve the accuracy of identification and the efficiency of identification.
In one embodiment, first nerves network model is established module 86 and can also be further configured to:
The individual cells in the second sample image that nervus opticus network model is partitioned into are marked with different colors
Area image;And first nerves network is added using the second sample image for marking color and corresponding cell type as sample
In the training set of model.
The individual cells area image in the second sample image is marked using different colors, can clearly be distinguished different
Boundary between individual cells region (the individual cells region especially closed on), avoids interfering with each other, to improve identification
Accuracy.
Optionally, first nerves network model can be convolutional neural networks model.
Figure 18 show another embodiment of the present invention provides a kind of cell type analytical equipment structural schematic diagram.Such as figure
Shown in 18, which can also include:
Nervus opticus network model establishes module 87, and being configured to will be comprising the first sample image of nuclear area 3 and right
Second sample image of the determination cell boundaries answered is added in the training set of nervus opticus network model as sample;And by
First nerves network mould is added as sample in the second sample image and corresponding cell type that two neural network models are partitioned into
In the training set of type.
By establishing nervus opticus network model and first nerves network model, the segmentation of individual cells and identification are distinguished
It is completed by nervus opticus network model and first nerves network model, reduces the complexity and meter of each neural network model
Calculation amount, to can also improve the efficiency of analysis.
Optionally, nervus opticus network model can be convolutional neural networks model.
Figure 19 show the structural schematic diagram of the electronic equipment of one embodiment of the invention offer.As shown in figure 19, the electronics
Equipment can be the online electronic equipment of such as medicine detector device equipped with cell type analytical equipment etc thereon, can also
Capable of being communicated with online electronic equipment to transmit the offline electronic equipment of trained machine learning model to it.
Figure 19 illustrates the block diagram of the electronic equipment according to the embodiment of the present application.
As shown in figure 19, electronic equipment 9 includes one or more processors 91 and memory 92.
Processor 91 can be central processing unit (CPU) or have data-handling capacity and/or instruction execution capability
Other forms processing unit, and can control the other assemblies in electronic equipment 9 to execute desired function.
Memory 92 may include one or more computer program products, and the computer program product may include each
The computer readable storage medium of kind form, such as volatile memory and/or nonvolatile memory.The volatile storage
Device for example may include random access memory (RAM) and/or cache memory (cache) etc..It is described non-volatile to deposit
Reservoir for example may include read-only memory (ROM), hard disk, flash memory etc..It can be deposited on the computer readable storage medium
One or more computer program instructions are stored up, processor 91 can run described program instruction, to realize this Shen described above
The area marking method of each embodiment please and/or other desired functions.In the computer readable storage medium
In can also store the various contents such as blood sample image information, location information, target area, training sample.
In one example, electronic equipment 9 can also include: input unit 93 and output device 94, these components pass through
The interconnection of bindiny mechanism's (not shown) of bus system and/or other forms.
For example, the input unit 93 can be image device, for acquiring blood sample image information, acquired image
Information can be stored in memory 92 for other assemblies use.It is of course also possible to using other it is integrated or it is discrete at
The blood sample image information is acquired as device, and is sent to electronic equipment 9.In addition, the input equipment 93 may be used also
To include such as keyboard, mouse and communication network and its remote input equipment connected etc..
The output device 94 can export various information, including determination to external (for example, user or machine learning model)
Cell type, cell quantity, training sample out etc..The output equipment 94 may include such as display, loudspeaker, printing
Machine and communication network and its remote output devices connected etc..
Certainly, to put it more simply, illustrated only in Figure 19 it is some in component related with the application in the electronic equipment 9,
The component of such as bus, input/output interface etc. is omitted.In addition to this, according to concrete application situation, electronic equipment 9 is also
It may include any other component appropriate.
Other than the above method and equipment, embodiments herein can also be computer program product comprising meter
Calculation machine program instruction, the computer program instructions execute the processor in this specification to retouch
The step in the area marking method according to the various embodiments of the application stated.
The computer program product can be write with any combination of one or more programming languages for holding
The program code of row the embodiment of the present application operation, described program design language includes object oriented program language, such as
Java, C++ etc. further include conventional procedural programming language, such as " C " language or similar programming language.Journey
Sequence code can be executed fully on the user computing device, partly execute on a user device, be independent soft as one
Part packet executes, part executes on a remote computing or completely in remote computing device on the user computing device for part
Or it is executed on server.
In addition, embodiments herein can also be computer readable storage medium, it is stored thereon with computer program and refers to
It enables, it is described in this specification according to this that the computer program instructions execute the processor
Apply for the step in the area marking method of various embodiments.
The computer readable storage medium can be using any combination of one or more readable mediums.Readable medium can
To be readable signal medium or readable storage medium storing program for executing.Readable storage medium storing program for executing for example can include but is not limited to electricity, magnetic, light, electricity
Magnetic, the system of infrared ray or semiconductor, device or device, or any above combination.Readable storage medium storing program for executing it is more specific
Example (non exhaustive list) includes: the electrical connection with one or more conducting wires, portable disc, hard disk, random access memory
Device (RAM), read-only memory (ROM), erasable programmable read only memory (EPROM or flash memory), optical fiber, portable compact disc
Read-only memory (CD-ROM), light storage device, magnetic memory device or above-mentioned any appropriate combination.
The basic principle of the application is described in conjunction with specific embodiments above, however, it is desirable to, it is noted that in this application
The advantages of referring to, advantage, effect etc. are only exemplary rather than limitation, must not believe that these advantages, advantage, effect etc. are the application
Each embodiment is prerequisite.In addition, detail disclosed above is merely to exemplary effect and the work being easy to understand
With, rather than limit, it is that must be realized using above-mentioned concrete details that above-mentioned details, which is not intended to limit the application,.
Device involved in the application, device, equipment block diagram only as illustrative example and be not intended to require
Or hint must be attached in such a way that box illustrates, arrange, configure.As the skilled person will recognize,
It can be connected by any way, arrange, configure these devices, device, equipment, system.Such as "include", "comprise", " having "
Etc. word be open vocabulary, refer to " including but not limited to ", and can be used interchangeably with it.Vocabulary "or" used herein above
Refer to vocabulary "and/or" with "and", and can be used interchangeably with it, unless it is not such that context, which is explicitly indicated,.It is used herein above
Vocabulary " such as " refers to phrase " such as, but not limited to ", and can be used interchangeably with it.
It may also be noted that each component or each step are can to decompose in the device of the application, device and method
And/or reconfigure.These decompose and/or reconfigure the equivalent scheme that should be regarded as the application.
The above description of disclosed aspect is provided so that any person skilled in the art can make or use this
Application.Various modifications in terms of these are readily apparent to those skilled in the art, and are defined herein
General Principle can be applied to other aspect without departing from scope of the present application.Therefore, the application is not intended to be limited to
Aspect shown in this, but according to principle disclosed herein and the consistent widest range of novel feature.
In order to which purpose of illustration and description has been presented for above description.In addition, this description is not intended to the reality of the application
It applies example and is restricted to form disclosed herein.Although multiple embodiment aspect and embodiment already discussed above, this field
Its certain modifications, modification, change, addition and sub-portfolio will be recognized in technical staff.
Claims (10)
1. a kind of cell type analysis method characterized by comprising
Choose the region to be analyzed in blood sample image;
Determine the nuclear area in the region to be analyzed;
It is partitioned into the cell compartment image including individual cells image, wherein the individual cells image includes the nuclear area
Domain and cytosolic domain;And
The cell compartment image is inputted into first nerves network model, it is single thin by first nerves network model identification
The type of born of the same parents.
2. the method according to claim 1, wherein described be partitioned into the cell compartment including individual cells image
Image includes:
Determine the cell boundaries comprising the nuclear area to be partitioned into the cell compartment image including individual cells image.
3. according to the method described in claim 2, it is characterized in that, the determination includes the cell boundaries of the nuclear area
Include:
Determine the region to be split comprising the nuclear area;
The image in the region to be split is inputted into nervus opticus network model, is determined and is wrapped by the nervus opticus network model
Cell boundaries containing the nuclear area.
4. the method according to claim 1, wherein the nuclear area in the determination region to be analyzed
Include:
Selection meets the continuum of pre-set color condition for the nuclear area in the region to be analyzed.
5. the method according to claim 1, wherein the nuclear area in the determination region to be analyzed
Include:
Select the continuum of same color;
Judge whether the continuum meets pre-set color condition;
When the continuum meets the pre-set color condition, judge whether the area of the continuum is greater than the first face
It accumulates threshold value and is less than second area threshold value;And
When the area of the continuum is greater than the first area threshold and is less than second area threshold value, the continuum is chosen
For the nuclear area in the region to be analyzed.
6. according to the method described in claim 5, it is characterized in that, nuclear area in the determination region to be analyzed
Further include:
When the area of the continuum is greater than or equal to the second area threshold value, the continuum is divided into multiple
The nuclear area.
7. according to the method described in claim 6, it is characterized in that, described be divided into multiple cells for the continuum
Core region includes:
By the image of continuum described in scaled down, the continuum is divided into multiple nuclear areas.
8. according to the method described in claim 5, it is characterized in that, nuclear area in the determination region to be analyzed
Further include:
When the area of the continuum is less than or equal to first area threshold, judgement is pre- comprising the continuum
If meeting the pre-set color condition and area with the presence or absence of other within the scope of third area threshold is respectively less than or is equal to described the
The continuum of one area threshold;
If it exists other meet the pre-set color condition and area be respectively less than or equal to first area threshold continuum
Meet the pre-set color condition and area within the scope of the default third area threshold is respectively less than or is equal to described the by domain
All continuums of one area threshold are merged into the nuclear area.
9. the method according to claim 1, wherein the nuclear area in the determination region to be analyzed
Further include:
Direction with orthogonal two boundaries in the region to be analyzed is respectively that reference axis establishes coordinate system;
When at least there is a point in the determining nuclear area in the reference axis, abandon the nucleus
The subsequent processing in region.
10. a kind of cell type analytical equipment characterized by comprising
Module is chosen, is configured to choose the region to be analyzed in blood sample image;
Nucleus determining module is configured to determine the nuclear area in the region to be analyzed;
Cell compartment divides module, is configured to be partitioned into the cell compartment image including individual cells image, wherein described single
Cell image includes the nuclear area and cytosolic domain;And
Cell type identification module is configured to the cell compartment image inputting first nerves network model, passes through described the
The type of one neural network model identification individual cells.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811456650.1A CN109554432A (en) | 2018-11-30 | 2018-11-30 | A kind of cell type analysis method, analytical equipment and electronic equipment |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811456650.1A CN109554432A (en) | 2018-11-30 | 2018-11-30 | A kind of cell type analysis method, analytical equipment and electronic equipment |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109554432A true CN109554432A (en) | 2019-04-02 |
Family
ID=65868360
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811456650.1A Pending CN109554432A (en) | 2018-11-30 | 2018-11-30 | A kind of cell type analysis method, analytical equipment and electronic equipment |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109554432A (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
RU2732895C1 (en) * | 2019-05-27 | 2020-09-24 | Общество с ограниченной ответственностью "ПЛАТФОРМА ТРЕТЬЕ МНЕНИЕ" | Method for isolating and classifying blood cell types using deep convolution neural networks |
CN112215790A (en) * | 2019-06-24 | 2021-01-12 | 杭州迪英加科技有限公司 | KI67 index analysis method based on deep learning |
CN112233060A (en) * | 2020-09-04 | 2021-01-15 | 广州金域医学检验中心有限公司 | Method and device for screening abnormal samples of digital pathological images, computer equipment and storage medium |
CN113178228A (en) * | 2021-05-25 | 2021-07-27 | 郑州中普医疗器械有限公司 | Cell analysis method based on nuclear DNA analysis, computer device, and storage medium |
CN116824579A (en) * | 2023-06-27 | 2023-09-29 | 长沙金域医学检验实验室有限公司 | Method and device for detecting yarrowia pneumocystis based on direct immunofluorescence staining |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103473739A (en) * | 2013-08-15 | 2013-12-25 | 华中科技大学 | White blood cell image accurate segmentation method and system based on support vector machine |
CN108021903A (en) * | 2017-12-19 | 2018-05-11 | 南京大学 | The error calibrating method and device of artificial mark leucocyte based on neutral net |
CN108885681A (en) * | 2015-12-18 | 2018-11-23 | 雅培实验室 | For assessing the method and system of cellular morphology |
-
2018
- 2018-11-30 CN CN201811456650.1A patent/CN109554432A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103473739A (en) * | 2013-08-15 | 2013-12-25 | 华中科技大学 | White blood cell image accurate segmentation method and system based on support vector machine |
CN108885681A (en) * | 2015-12-18 | 2018-11-23 | 雅培实验室 | For assessing the method and system of cellular morphology |
CN108021903A (en) * | 2017-12-19 | 2018-05-11 | 南京大学 | The error calibrating method and device of artificial mark leucocyte based on neutral net |
Non-Patent Citations (7)
Title |
---|
姚力等: "基于细胞神经网络的图像分割及其在医学图像中的应用", 《中国科学》 * |
宁正元: "《计算机在生物科学研究中的应用》", 30 November 2006, 厦门大学出版社 * |
杨嘉麟等: "基于改进BP神经网络的白细胞图像分割", 《微计算机信息》 * |
杨嘉麟等: "基于改进的BP神经网络的血液细胞图像分割", 《微型电脑应用》 * |
林璇: "基于spiking神经网络的白细胞图像边缘检测", 《应用研究》 * |
王文亮等: "细胞神经网络在图像处理中的应用研究", 《中国会议》 * |
蒋英等: "神经网络在白血病图像识别中的应用", 《福建电脑》 * |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
RU2732895C1 (en) * | 2019-05-27 | 2020-09-24 | Общество с ограниченной ответственностью "ПЛАТФОРМА ТРЕТЬЕ МНЕНИЕ" | Method for isolating and classifying blood cell types using deep convolution neural networks |
WO2020242341A1 (en) * | 2019-05-27 | 2020-12-03 | Общество С Ограниченной Ответственностью "Лаб Кмд" | Method for selecting and classifying blood cell types by means of deep convolutional neural networks |
CN112215790A (en) * | 2019-06-24 | 2021-01-12 | 杭州迪英加科技有限公司 | KI67 index analysis method based on deep learning |
CN112233060A (en) * | 2020-09-04 | 2021-01-15 | 广州金域医学检验中心有限公司 | Method and device for screening abnormal samples of digital pathological images, computer equipment and storage medium |
CN112233060B (en) * | 2020-09-04 | 2024-03-29 | 广州金域医学检验中心有限公司 | Screening method, device, equipment and medium for digital pathological image abnormal samples |
CN113178228A (en) * | 2021-05-25 | 2021-07-27 | 郑州中普医疗器械有限公司 | Cell analysis method based on nuclear DNA analysis, computer device, and storage medium |
CN116824579A (en) * | 2023-06-27 | 2023-09-29 | 长沙金域医学检验实验室有限公司 | Method and device for detecting yarrowia pneumocystis based on direct immunofluorescence staining |
CN116824579B (en) * | 2023-06-27 | 2023-12-22 | 长沙金域医学检验实验室有限公司 | Method and device for detecting yarrowia pneumocystis based on direct immunofluorescence staining |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109636782A (en) | A kind of cell type analysis model training method, device and analysis method | |
CN109554432A (en) | A kind of cell type analysis method, analytical equipment and electronic equipment | |
CN109507406A (en) | A kind of cellular morphology intelligent extract method, device and analysis method | |
CN111325103B (en) | Cell labeling system and method | |
Chandradevan et al. | Machine-based detection and classification for bone marrow aspirate differential counts: initial development focusing on nonneoplastic cells | |
Dimauro et al. | Nasal cytology with deep learning techniques | |
CN103930762B (en) | The system and method with checking are shown for sample | |
CN105210083B (en) | For checking and the system and method for analytical cytology sample | |
US6246785B1 (en) | Automated, microscope-assisted examination process of tissue or bodily fluid samples | |
Bjornsson et al. | Associative image analysis: a method for automated quantification of 3D multi-parameter images of brain tissue | |
CN103345654B (en) | Method for differential counting of white blood cells based on morphology | |
CN104094118B (en) | Automatically hematoblastic method and apparatus in whole blood sample is identified by micro-image | |
CN103827657B (en) | Blood analyser is calibrated and assessed | |
CN110084150A (en) | A kind of Automated Classification of White Blood Cells method and system based on deep learning | |
KR20220062332A (en) | Systems and methods for processing images of slides for digital pathology | |
JP3916395B2 (en) | Test system with sample pretreatment function | |
CN109580458A (en) | Fluidic cell intelligent immunity classifying method, device and electronic equipment | |
CN110414317B (en) | Full-automatic leukocyte classification counting method based on capsule network | |
CN108447047A (en) | Acid-fast bacilli detection method and device | |
CN112001315A (en) | Bone marrow cell classification and identification method based on transfer learning and image texture features | |
Lv et al. | High-accuracy morphological identification of bone marrow cells using deep learning-based Morphogo system | |
Priyankara et al. | An extensible computer vision application for blood cell recognition and analysis | |
Wang et al. | Experimental evaluation of deep learning method in reticulocyte enumeration in peripheral blood | |
Nakamura et al. | Evaluation of two semi-supervised learning methods and their combination for automatic classification of bone marrow cells | |
EP3563342A1 (en) | Automated system and method for creating and executing a scoring guide to assist in the analysis of tissue specimen |
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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190402 |