CN109035216A - Handle the method and device of cervical cell sectioning image - Google Patents
Handle the method and device of cervical cell sectioning image Download PDFInfo
- Publication number
- CN109035216A CN109035216A CN201810733952.2A CN201810733952A CN109035216A CN 109035216 A CN109035216 A CN 109035216A CN 201810733952 A CN201810733952 A CN 201810733952A CN 109035216 A CN109035216 A CN 109035216A
- Authority
- CN
- China
- Prior art keywords
- cell
- cervical
- cervical cell
- sectioning image
- database
- 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
- 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
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30024—Cell structures in vitro; Tissue sections in vitro
Abstract
The present invention provides a kind of method and device for handling cervical cell sectioning image, wherein method includes: step A: acquiring single free non-overlap cell material and slice background material from the true cervical cell sectioning image of subsidiary cell rank markup information, establishes single cell database and slice background database;Step B: the emulation cervical cell sectioning image of subsidiary cell example rank mark is synthesized according to single cell database and slice background database;Step C: cervical cell example segmentation depth model is trained using emulation cervical cell sectioning image, then cervical cell segmentation is carried out to true cervical cell sectioning image to be detected using trained cervical cell example segmentation depth model, obtains cervical cell segmentation result.
Description
Technical field
The present invention relates to a kind of computer and software technology fields, and in particular to a kind of processing cervical cell sectioning image
Method and device.
Background technique
With the development of artificial intelligence, the work of many traditional artificial completions can be completed gradually by computer.Such as
Neural network model is trained using deep learning method, then with trained model realization to thin in cervical cell slice
The segmentation of born of the same parents' example, and can further be partitioned into the nucleus of individual cells example.
During training neural network model, it is desirable to provide the training sample set of significant amount amount high quality just can guarantee training
As a result accuracy.However, in practical application, since cell overlap degree is relatively high in most of cervical cell sectioning images
(the especially usual high aggregation of abnormal cell), medical profession are manually difficult to be partitioned into each cervical cell edge and standard
Really calibration cell type.So in existing technical solution, training sample be mainly the cell demarcated through medical profession from
It dissipating distribution, stacks few cervical cell sectioning image, these training samples numbers are limited, there is gap with practical application example, because
This accuracy based on the cervical cell segmentation depth model that the training of these training samples obtains is not high enough.
Summary of the invention
In view of this, the present invention provides a kind of method and device for handling cervical cell sectioning image, it is able to solve existing
The not high enough technical problem of the accuracy of the cervical cell segmentation depth model of technology.
To achieve the above object, according to an aspect of the invention, there is provided a kind of processing cervical cell sectioning image
Method, comprising: step A: acquired from the true cervical cell sectioning image of subsidiary cell rank markup information it is single dissociate it is non-
It is overlapped cell material and slice background material, establishes single cell database and slice background database;Step B: according to described
Single cell database and slice background database synthesize the emulation cervical cell sectioning image of subsidiary cell example rank mark;
Step C: cervical cell example segmentation depth model is trained using the emulation cervical cell sectioning image, is then used
The trained cervical cell example segmentation depth model carries out cervical cell to true cervical cell sectioning image to be detected
Segmentation, obtains cervical cell segmentation result.
Optionally, the single free non-overlap cell material include NILM cell material, LSIL cell material and
HSIL cell material, the step B includes: step B1: it is thin to randomly select multiple LSIL from the single cell database
Born of the same parents' material and/or the HSIL cell material are successively combined, and obtain the first foreground image, and first is met in anabolic process about
Beam condition: it represents the friendship of cell material overlapping degree and is greater than first threshold than parameter and is less than second threshold;Step B2: from
Multiple single free non-overlap cell materials are randomly selected in the single cell database, are carried out with first foreground image
Combination, obtains the second foreground image, the second constraint condition is met in anabolic process: representing the friendship of cell material overlapping degree and ratio
Parameter is less than 1;Step B3: from the slice background database choose slice background material, with second foreground image into
Row combination, obtains the emulation cervical cell sectioning image.
Optionally, the first threshold is 0.1, and the second threshold is 0.6.
Optionally, the cervical cell example segmentation depth model is MASK-RCNN.
Optionally, further includes: step D: the true cervical cell to be detected being cut using nucleus segmentation depth model
Picture carries out nuclear area regional partition, obtains nucleus segmentation result;Step E: from the cervical cell segmentation result and institute
It states and extracts feature in nucleus segmentation result, then according to the feature to each uterine neck in the cervical cell segmentation result
Cell is classified.
To achieve the above object, according to another aspect of the present invention, it is also proposed that a kind of processing cervical cell sectioning image
Device, comprising: acquisition module, it is single for being acquired from the true cervical cell sectioning image of subsidiary cell rank markup information
One free non-overlap cell material and slice background material, establish single cell database and slice background database;Emulation is closed
At module, for synthesizing the imitative of subsidiary cell example rank mark according to the single cell database and slice background database
True cervical cell sectioning image;First segmentation module, for real to cervical cell using the emulation cervical cell sectioning image
Example segmentation depth model is trained, then using the trained cervical cell example segmentation depth model to be detected true
Real cervical cell sectioning image carries out cervical cell segmentation, obtains cervical cell segmentation result.
Optionally, the single free non-overlap cell material include NILM cell material, LSIL cell material and
HSIL cell material, the emulation synthesis module are used for: it is thin to randomly select multiple LSIL from the single cell database
Born of the same parents' material and/or the HSIL cell material are successively combined, and obtain the first foreground image, and first is met in anabolic process about
Beam condition: it represents the friendship of cell material overlapping degree and is greater than first threshold than parameter and is less than second threshold;From the list
Multiple single free non-overlap cell materials are randomly selected in one cell database, are combined with first foreground image,
The second foreground image is obtained, the second constraint condition is met in anabolic process: representing the friendship of cell material overlapping degree and compares parameter
Less than 1;Slice background material is chosen from the slice background database, is combined, is obtained with second foreground image
The emulation cervical cell sectioning image.
Optionally, the first threshold is 0.1, and the second threshold is 0.6.
Optionally, the cervical cell example segmentation depth model is MASK-RCNN.
Optionally, further includes: the second segmentation module, for dividing depth model to described to be detected true using nucleus
Cervical cell sectioning image carries out nuclear area regional partition, obtains nucleus segmentation result;Categorization module is used for from the uterine neck
Feature is extracted in cell segmentation result and the nucleus segmentation result, then the cervical cell is divided according to the feature
As a result each cervical cell in is classified.
According to the technique and scheme of the present invention, a large amount of simulation training sample with cell grade mark can be synthesized, then
Based on simulation training sample training model, then application model carries out detection segmentation application.Technical solution of the present invention at least has
It has the advantages that: high degree of automation, by machine synthesis of artificial sample, without the sectioning image high to cell overlap degree
It is manually demarcated, time saving and energy saving, the cell edges and nominal data in sample are accurate and reliable.Further, model accurately may be used
It leans on, testing result is also accurate and reliable.
Detailed description of the invention
Attached drawing for a better understanding of the present invention, does not constitute an undue limitation on the present invention.Wherein:
Fig. 1 is the flow diagram of the method for processing cervical cell sectioning image according to an embodiment of the invention;
Fig. 2 is the flow diagram of the method for processing cervical cell sectioning image according to another embodiment of the present invention;
Fig. 3 is the structural schematic diagram of the device of processing cervical cell sectioning image according to an embodiment of the invention;
Fig. 4 is the structural schematic diagram of the device of processing cervical cell sectioning image in accordance with another embodiment of the present invention;
Fig. 5 is the exemplary diagram of a NILM cell;
Fig. 6 is the exemplary diagram of a LSIL cell;
Fig. 7 is the exemplary diagram of a HSIL cell;
Fig. 8 is the exemplary diagram of a slice background;
Fig. 9 is the exemplary diagram of an emulation cervical cell sectioning image;
Figure 10 is the exemplary diagram of a true cervical cell slice to be measured;
Figure 11 is the exemplary diagram of the cervical cell segmentation result of Figure 10.
Specific embodiment
Below in conjunction with attached drawing, an exemplary embodiment of the present invention will be described, including the various of the embodiment of the present invention
Details should think them only exemplary to help understanding.Therefore, those of ordinary skill in the art should recognize
It arrives, it can be with various changes and modifications are made to the embodiments described herein, without departing from scope and spirit of the present invention.Together
Sample, for clarity and conciseness, descriptions of well-known functions and structures are omitted from the following description.
To more fully understand those skilled in the art, the relevant knowledge of lower different cervical cells is described in detail below.
Cervical cell is usually medically divided into NILM (Negative for Intraepithlial Lesion or
Maliganancy has no intraepithelial lesions cell or pernicious) cell, LSIL
(Low-grade Squamous Intraepithelial Lesion, i.e., low lesion in scaly epithelium) cell,
HSIL (High-grade Squamous Intraepithelial Lesion, that is, scaly epithelium inner height lesion) cell.
Wherein LSIL cell explanation is likely to occur cervical atypical hyperplasia, and HSIL explanation is likely to be uterine neck precancerosis
Become, generally refers to the precancerous lesion of scaly epithelium, also known as intraepithelial neoplasia (cin) (CIN) here.Medical field is based on observation sectioning image
It was found that: most of LSIL cells and HSIL cell arrangement are close, but also do not stack completely.Furthermore also random in image to dissipate
A small amount of NILM cell, LSIL cell and HSIL cell are fallen, is in loose discrete distribution.
Fig. 1 is the flow diagram of the method for processing cervical cell sectioning image according to an embodiment of the invention.Such as
Shown in Fig. 1, this method may include following step A to step C.
Step A: acquired from the true cervical cell sectioning image of subsidiary cell rank markup information it is single dissociate it is non-heavy
Folded cell material and slice background material, establish single cell database and slice background database.Wherein, it is single dissociate it is non-heavy
Folded cell material includes NILM cell material, LSIL cell material and HSIL cell material.
Step B: according to single cell database and slice background database synthesis of artificial cervical cell sectioning image.
Optionally step B specifically includes following step B1 to B3.
Step B1: multiple LSIL cell materials and/or HSIL cell material are randomly selected successively from single cell database
It is combined, obtains the first foreground image, the first constraint condition is met in anabolic process: representing the friendship of cell material overlapping degree
And it is greater than first threshold than parameter and is less than second threshold.It should be noted that handing over and than (Intersection-over-
Union, IoU) it is a concept used in target detection, it is the candidate frame (candidate bound) generated and former label
The overlapping rate of frame (ground truth bound), the i.e. ratio of their intersection and union.Ratio is 1 when completely overlapped, complete
Ratio is 0 when not being overlapped entirely.Optionally, first threshold 0.1, second threshold 0.6.It was proved that when handing over and comparing ∈
Overlapping degree is best suitable for reality when (0.1,0.6), and experimental result is preferable.
Step B2: multiple single free non-overlap cell materials are randomly selected from single cell database, before first
Scape image is combined, and obtains the second foreground image, the second constraint condition is met in anabolic process: being represented cell material and is overlapped journey
The friendship of degree and than parameter less than 1.
Step B3: slice background material is chosen from slice background database, is combined, obtains with the second foreground image
Emulate cervical cell sectioning image.
Step C: cervical cell example segmentation depth model is trained using emulation cervical cell sectioning image, then
Cervical cell is carried out to true cervical cell sectioning image to be detected using trained cervical cell example segmentation depth model
Segmentation, obtains cervical cell segmentation result.
Optionally, cervical cell example segmentation depth model is MASK-RCNN.(maskrcnn is introduced: MaskRCNN is language
One combination of justice segmentation and target detection, has become the sharp weapon of example segmentation at present.It is proposed that the paper of MaskRCNN obtains
Computer vision top meeting iccv Best Paper Award in 2017 was obtained, and has original performance in the contests such as coco.
The method of processing cervical cell sectioning image according to an embodiment of the present invention, can synthesize largely with cell grade
The simulation training sample of mark is then based on simulation training sample training model, and then application model carries out detection segmentation application.
The method of processing cervical cell sectioning image of the invention at least has the following beneficial effects: high degree of automation, by machine
Synthesis of artificial sample, time saving and energy saving without manually being demarcated to the high sectioning image of cell overlap degree, the cell side in sample
Edge and nominal data are accurate and reliable.Further, model is accurate and reliable, and testing result is also accurate and reliable.
Fig. 2 is the flow diagram of the method for processing cervical cell sectioning image in accordance with another embodiment of the present invention.
As shown in Fig. 2, this method may include still further comprising step D and step E.
Step D: divide depth model using nucleus, realize nuclear area in true cervical cell sectioning image to be detected
The accurate segmentation in domain, obtains nucleus segmentation result.
Step E: feature is extracted from cervical cell segmentation result and nucleus segmentation result, then according to feature to uterine neck
Each cervical cell in cell segmentation result is classified.Selected feature may include the long term voyage of individual cells, pulp noodle
Product, karyoplasmic ratio, nucleus Average pixel intensity, cytoplasm Average pixel intensity etc. use the algorithm training based on decision tree
Disaggregated model classifies to cell, so realizes the judgement to single cervical cell.
The method of the processing cervical cell sectioning image of embodiment shown in Fig. 2, can further determine that cell type, have
Have the advantages that accurate and reliable.
Fig. 3 is the structural schematic diagram of the device of processing cervical cell sectioning image according to an embodiment of the invention.Such as
Shown in Fig. 3, which may include acquisition module 301, emulation synthesis module 302 and the first segmentation module 303.
Acquisition module 301 is single for acquiring from the true cervical cell sectioning image of subsidiary cell rank markup information
Free non-overlap cell material and slice background material, establish single cell database and slice background database.Wherein, single
Free non-overlap cell material includes NILM cell material, LSIL cell material and HSIL cell material.
Synthesis module 302 is emulated to be used to synthesize subsidiary cell example according to single cell database and slice background database
The emulation cervical cell sectioning image of rank mark.
Specifically, emulation synthesis module 302 can be used for: randomly select multiple LSIL cytokines from single cell database
Material and/or HSIL cell material are successively combined, and obtain the first foreground image, the first constraint condition is met in anabolic process:
It represents the friendship of cell material overlapping degree and is greater than first threshold than parameter and is less than second threshold;From single cell database
In randomly select multiple single free non-overlap cell materials, be combined with the first foreground image, obtain the second foreground image,
Meet the second constraint condition in anabolic process: representing the friendship of cell material overlapping degree and than parameter less than 1;From slice background number
It according to slice background material is chosen in library, is combined with the second foreground image, obtains emulation cervical cell sectioning image.It is optional
Ground, first threshold 0.1, second threshold 0.6.It is MASK-RCNN that cervical cell example, which divides depth model,.
First segmentation module 303 is used to divide depth model to cervical cell example using emulation cervical cell sectioning image
It is trained, then using trained cervical cell example segmentation depth model to true cervical cell sectioning image to be detected
Cervical cell segmentation is carried out, cervical cell segmentation result is obtained.
Fig. 4 is the structural schematic diagram of the device of processing cervical cell sectioning image in accordance with another embodiment of the present invention.
As shown in figure 4, the device 40 may include 403, second points of module of segmentation of acquisition module 401, emulation synthesis module 402, first
Cut module 404 and categorization module 405.
Second segmentation module 404 is used for using nucleus segmentation depth model to true cervical cell sectioning image to be detected
Nuclear area regional partition is carried out, nucleus segmentation result is obtained.
Categorization module 405 is used to extract feature from cervical cell segmentation result and nucleus segmentation result, then basis
Feature classifies to each cervical cell in cervical cell segmentation result.
The method of the processing cervical cell sectioning image of embodiment shown in Fig. 4, can further determine that cell type, have
Have the advantages that accurate and reliable.
To more fully understand those skilled in the art, the following detailed description of a specific embodiment.
Firstly, acquiring NILM cell material from the true cervical cell sectioning image of subsidiary cell rank markup information
(as shown in Figure 5), LSIL cell material (as shown in Figure 6), HSIL cell material (as shown in Figure 7) and slice background material are (such as
Shown in Fig. 8).
Then these materials are respectively put into this four files of HSIL, LSIL, ALL and BG.Wherein HSIL and
Two files of LSIL only store the cell material of corresponding name type;It is (i.e. three kinds thin that ALL file stores all types of cells
Born of the same parents' material is all put into);Slice background material of the BG file storage from true slice interception.
It randomly chooses HSIL or LSIL file (such as currently having selected HSIL file), therefrom repeats to extract 5-20
Cell successively overlays on patch to be produced (segment).It is overlapped constraint condition are as follows: about overlapping be not present of bundle cell completely includes,
It is only overlapping in marginal portion, it hands over and ratio is set as 0.1 to 0.6.
Then, it is placed on patch to be produced from 10 to 30 cells of selection in ALL file.Here it is constrained to from all
Obtaining cell cannot be required to hand over and ratio is less than 1 by the region comprising or included in previous step generation.Meet this width
The cell chosen after loose constraint condition can be put in patch at random.
Then a background element is extracted from BG file at random again and the patch of front has synthesized emulation uterine neck
Cell section image (as shown in Figure 9).Individually being taken out due to all elements in the emulation cervical cell sectioning image is all
Known markup information, so the emulation cervical cell sectioning image that synthesis obtains also possesses the other mark of cell grade.
Using MASK-RCNN technology, divide depth mould according to a large amount of emulation uterine neck sectioning image training cervical cell examples
Type.Then it is applied with the model after training.Figure 10 is true cervical cell sectioning image to be detected, and Figure 11 is Figure 10 pairs
The cervical cell segmentation result answered.
Describe basic principle of the invention in conjunction with specific embodiments above, in the apparatus and method of the present invention, it is clear that
Each component or each step can be decomposed and/or be reconfigured.These decompose and/or reconfigure should be regarded as it is of the invention etc.
Efficacious prescriptions case.Also, the step of executing above-mentioned series of processes can execute according to the sequence of explanation in chronological order naturally, still
It does not need centainly to execute sequentially in time.Certain steps can execute parallel or independently of one another.
Above-mentioned specific embodiment, does not constitute a limitation on the scope of protection of the present invention.Those skilled in the art should be bright
It is white, design requirement and other factors are depended on, various modifications, combination, sub-portfolio and substitution can occur.It is any
Made modifications, equivalent substitutions and improvements etc. within the spirit and principles in the present invention, should be included in the scope of the present invention
Within.
Claims (10)
1. a kind of method for handling cervical cell sectioning image characterized by comprising
Step A: it is thin from the true cervical cell sectioning image of subsidiary cell rank markup information to acquire single free non-overlap
Born of the same parents' material and slice background material, establish single cell database and slice background database;
Step B: the imitative of subsidiary cell example rank mark is synthesized according to the single cell database and slice background database
True cervical cell sectioning image;
Step C: cervical cell example segmentation depth model is trained using the emulation cervical cell sectioning image, then
Uterine neck is carried out to true cervical cell sectioning image to be detected using the trained cervical cell example segmentation depth model
Cell segmentation obtains cervical cell segmentation result.
2. the method according to claim 1, wherein the single free non-overlap cell material includes NILM thin
Born of the same parents' material, LSIL cell material and HSIL cell material, the step B include:
Step B1: multiple LSIL cell materials and/or the HSIL cell are randomly selected from the single cell database
Material is successively combined, and obtains the first foreground image, the first constraint condition is met in anabolic process: it is overlapping to represent cell material
The friendship of degree is simultaneously greater than first threshold than parameter and is less than second threshold;
Step B2: randomly selecting multiple single free non-overlap cell materials from the single cell database, with described
One foreground image is combined, and obtains the second foreground image, the second constraint condition is met in anabolic process: representing the friendship of cell material
The friendship of folded degree and than parameter less than 1;
Step B3: choosing slice background material from the slice background database, be combined with second foreground image,
Obtain the emulation cervical cell sectioning image.
3. according to the method described in claim 2, the second threshold is 0.6 it is characterized in that, the first threshold is 0.1.
4. the method according to claim 1, wherein cervical cell example segmentation depth model is MASK-
RCNN。
5. the method according to claim 1, wherein further include:
Step D: nuclear area is carried out to the true cervical cell sectioning image to be detected using nucleus segmentation depth model
Regional partition obtains nucleus segmentation result;
Step E: feature is extracted from the cervical cell segmentation result and the nucleus segmentation result, then according to the spy
Sign classifies to each cervical cell in the cervical cell segmentation result.
6. a kind of device for handling cervical cell sectioning image characterized by comprising
Acquisition module, for acquired from the true cervical cell sectioning image of subsidiary cell rank markup information it is single dissociate it is non-
It is overlapped cell material and slice background material, establishes single cell database and slice background database;
Synthesis module is emulated, for synthesizing subsidiary cell instance-level according to the single cell database and slice background database
The emulation cervical cell sectioning image not marked;
First segmentation module, for using the emulation cervical cell sectioning image to cervical cell example segmentation depth model into
Row training, then using the trained cervical cell example segmentation depth model to true cervical cell slice map to be detected
As carrying out cervical cell segmentation, cervical cell segmentation result is obtained.
7. device according to claim 6, which is characterized in that the single free non-overlap cell material includes NILM thin
Born of the same parents' material, LSIL cell material and HSIL cell material, the emulation synthesis module are used for:
Multiple LSIL cell materials and/or the HSIL cell material are randomly selected successively from the single cell database
It is combined, obtains the first foreground image, the first constraint condition is met in anabolic process: representing the friendship of cell material overlapping degree
And it is greater than first threshold than parameter and is less than second threshold;
Multiple single free non-overlap cell materials are randomly selected from the single cell database, with first foreground picture
As being combined, the second foreground image is obtained, the second constraint condition is met in anabolic process: representing cell material overlapping degree
It hands over and than parameter less than 1;
Slice background material is chosen from the slice background database, is combined with second foreground image, is obtained institute
State emulation cervical cell sectioning image.
8. device according to claim 7, which is characterized in that the first threshold is 0.1, and the second threshold is 0.6.
9. device according to claim 6, which is characterized in that the cervical cell example segmentation depth model is MASK-
RCNN。
10. device according to claim 6, which is characterized in that further include:
Second segmentation module, for using nucleus segmentation depth model to the true cervical cell sectioning image to be detected into
Row nuclear area regional partition, obtains nucleus segmentation result;
Categorization module, for extracting feature from the cervical cell segmentation result and the nucleus segmentation result, then root
Classify according to the feature to each cervical cell in the cervical cell segmentation result.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810733952.2A CN109035216A (en) | 2018-07-06 | 2018-07-06 | Handle the method and device of cervical cell sectioning image |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810733952.2A CN109035216A (en) | 2018-07-06 | 2018-07-06 | Handle the method and device of cervical cell sectioning image |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109035216A true CN109035216A (en) | 2018-12-18 |
Family
ID=64640721
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810733952.2A Pending CN109035216A (en) | 2018-07-06 | 2018-07-06 | Handle the method and device of cervical cell sectioning image |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109035216A (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109886179A (en) * | 2019-02-18 | 2019-06-14 | 深圳视见医疗科技有限公司 | The image partition method and system of cervical cell smear based on Mask-RCNN |
CN110060244A (en) * | 2019-04-15 | 2019-07-26 | 深圳市麦迪普科技有限公司 | The system and method for cell detection and segmentation based on deep learning neural network |
CN110633651A (en) * | 2019-08-26 | 2019-12-31 | 武汉大学 | Abnormal cell automatic identification method based on image splicing |
CN113066121A (en) * | 2019-12-31 | 2021-07-02 | 深圳迈瑞生物医疗电子股份有限公司 | Image analysis system and method for identifying repeat cells |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1670764A (en) * | 2004-03-19 | 2005-09-21 | 中国科学院计算技术研究所 | Genetic algorithm based human face sample generating method |
US20130097103A1 (en) * | 2011-10-14 | 2013-04-18 | International Business Machines Corporation | Techniques for Generating Balanced and Class-Independent Training Data From Unlabeled Data Set |
CN107256558A (en) * | 2017-05-18 | 2017-10-17 | 深思考人工智能机器人科技(北京)有限公司 | The cervical cell image automatic segmentation method and system of a kind of unsupervised formula |
CN107492084A (en) * | 2017-07-06 | 2017-12-19 | 哈尔滨理工大学 | Typical packed cell core image combining method based on randomness |
CN107609549A (en) * | 2017-09-20 | 2018-01-19 | 北京工业大学 | The Method for text detection of certificate image under a kind of natural scene |
-
2018
- 2018-07-06 CN CN201810733952.2A patent/CN109035216A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1670764A (en) * | 2004-03-19 | 2005-09-21 | 中国科学院计算技术研究所 | Genetic algorithm based human face sample generating method |
US20130097103A1 (en) * | 2011-10-14 | 2013-04-18 | International Business Machines Corporation | Techniques for Generating Balanced and Class-Independent Training Data From Unlabeled Data Set |
CN107256558A (en) * | 2017-05-18 | 2017-10-17 | 深思考人工智能机器人科技(北京)有限公司 | The cervical cell image automatic segmentation method and system of a kind of unsupervised formula |
CN107492084A (en) * | 2017-07-06 | 2017-12-19 | 哈尔滨理工大学 | Typical packed cell core image combining method based on randomness |
CN107609549A (en) * | 2017-09-20 | 2018-01-19 | 北京工业大学 | The Method for text detection of certificate image under a kind of natural scene |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109886179A (en) * | 2019-02-18 | 2019-06-14 | 深圳视见医疗科技有限公司 | The image partition method and system of cervical cell smear based on Mask-RCNN |
CN109886179B (en) * | 2019-02-18 | 2023-03-28 | 深圳视见医疗科技有限公司 | Image segmentation method and system of cervical cell smear based on Mask-RCNN |
CN110060244A (en) * | 2019-04-15 | 2019-07-26 | 深圳市麦迪普科技有限公司 | The system and method for cell detection and segmentation based on deep learning neural network |
CN110060244B (en) * | 2019-04-15 | 2023-07-14 | 深圳市麦迪普科技有限公司 | Cell detection and segmentation system and method based on deep learning neural network |
CN110633651A (en) * | 2019-08-26 | 2019-12-31 | 武汉大学 | Abnormal cell automatic identification method based on image splicing |
CN110633651B (en) * | 2019-08-26 | 2022-05-13 | 武汉大学 | Abnormal cell automatic identification method based on image splicing |
CN113066121A (en) * | 2019-12-31 | 2021-07-02 | 深圳迈瑞生物医疗电子股份有限公司 | Image analysis system and method for identifying repeat cells |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109035216A (en) | Handle the method and device of cervical cell sectioning image | |
US11842556B2 (en) | Image analysis method, apparatus, program, and learned deep learning algorithm | |
US10474874B2 (en) | Applying pixelwise descriptors to a target image that are generated by segmenting objects in other images | |
Sharma et al. | Crowdsourcing for chromosome segmentation and deep classification | |
JP6338674B2 (en) | Perspective inspection system and method for automatically classifying and recognizing cargo | |
CN112215117A (en) | Abnormal cell identification method and system based on cervical cytology image | |
Ziamtsov et al. | Machine learning approaches to improve three basic plant phenotyping tasks using three-dimensional point clouds | |
EP3023910A1 (en) | Method and system for learning pixel visual context from object characteristics to generate rich semantic images | |
CN105893925A (en) | Human hand detection method based on complexion and device | |
CN107292314A (en) | A kind of lepidopterous insects species automatic identification method based on CNN | |
Casanova et al. | Texture analysis using fractal descriptors estimated by the mutual interference of color channels | |
CN112069985A (en) | High-resolution field image rice ear detection and counting method based on deep learning | |
CN114550169A (en) | Training method, device, equipment and medium for cell classification model | |
Xiao et al. | Citrus greening disease recognition algorithm based on classification network using TRL-GAN | |
CN104573701B (en) | A kind of automatic testing method of Tassel of Corn | |
Mostafa et al. | Leveraging guided backpropagation to select convolutional neural networks for plant classification | |
Williams et al. | Leaf only SAM: a segment anything pipeline for zero-shot automated leaf segmentation | |
Irshad et al. | Spectral band selection for mitosis detection in histopathology | |
Khoshdeli et al. | Deep learning models delineates multiple nuclear phenotypes in h&e stained histology sections | |
CN108038472A (en) | A kind of exchange method of bone marrow examination domain software fast finding abnormal cell | |
US11222194B2 (en) | Automated system and method for creating and executing a scoring guide to assist in the analysis of tissue specimen | |
CN108229514A (en) | Object detecting method, device and electronic equipment | |
Li et al. | Persistent homology demarcates a leaf morphospace | |
Kao et al. | A novel deep learning architecture for testis histology image classification | |
Xu et al. | Band selection for hyperspectral images based on particle swarm optimization and differential evolution algorithms with hybrid encoding |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20181218 |