CN106340016B - A kind of DNA quantitative analysis method based on microcytoscope image - Google Patents
A kind of DNA quantitative analysis method based on microcytoscope image Download PDFInfo
- Publication number
- CN106340016B CN106340016B CN201610769611.1A CN201610769611A CN106340016B CN 106340016 B CN106340016 B CN 106340016B CN 201610769611 A CN201610769611 A CN 201610769611A CN 106340016 B CN106340016 B CN 106340016B
- Authority
- CN
- China
- Prior art keywords
- classifier
- image
- region
- tree
- candidate region
- 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.)
- Active
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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10024—Color image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10056—Microscopic image
-
- 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 invention discloses a kind of DNA quantitative analysis methods based on microcytoscope image, comprising the following steps: step 1: pretreatment；Step 2: candidate region being extracted using MSER algorithm, obtains several single tine trees or individual node；Step 3: extracting the simple feature training classifier C1 in region based on training set, and calculate a possibility that it belongs to interested cell, single tine tree obtained in step 2 is further simplified based on the classifier；Step 4: outside the simple feature for extracting region based on training set, also extracting statistics, shape and textural characteristics training classifier C2, and further classified to all candidate regions that classifier C1 output is positive sample with the classifier；Step 5: calculating the DI value of nucleus.The present invention from image using maximum extreme value stability region algorithm to extracting a large amount of candidate regions, and two classifier C1 and C2 composition cascade classifiers of training complete the fast and accurately identification to these candidate regions.
Description
Technical field
The present invention relates to technical field of image processing more particularly to a kind of DNA based on microcytoscope image quantitatively to divide
Analysis method.
Background technique
Cytolgical examination is the main path of the exemplary cancers early screening such as current cervical carcinoma, carcinoma of mouth, is used in early days
Doctor blade technique is because cheap and universal, but its rate of missed diagnosis is high, although the liquid-based tabletting technology being widely used since 21 century changes
It has been apt to production effect, but has been provided after needing veteran pathologist directly to observe under the microscope the form of cell
Diagnostic result, low efficiency, and be easy to receive the influence of subjectivity, repeatability is poor.DNA ploidy body analytical technology is then automatically to thin
Nucleus in born of the same parents' image is analyzed, and then quantitative measurment goes out chromosome or DNA content in nucleus, and provides judgement automatically
As a result.Compared with conventional cell method, which has the characteristics that sensibility is high, high-efficient and favorable repeatability, Ke Yiyou
Effect solves base and lacks experienced this contradiction of pathologist, is with a wide range of applications.
It is DNA ploidy body analytical technology that fast and accurately diploid cell core in cell image, which is split and is identified,
It is crucial.Existing method and system are extracted in segmentation using simple images dividing methods such as Threshold segmentation, morphological methods
Candidate region, various shape, Texture eigenvalue and the trained classifier for then extracting candidate region complete oneself of nucleus type
Dynamic identification.However such methods are when handling more intensive and nucleus the intensity profile variation of nucleus distribution greatly
A large amount of diploid cell can be missed, and often extracts all features in each candidate region in identification, in candidate region number
Time-consuming in the case that amount is big.
Summary of the invention
The present invention is calculated using maximum extreme value stability region (Maximally Stable Extremal Regions, MSER)
Method from image to extracting a large amount of candidate regions, and two classifier C1 and C2 composition cascade classifiers of training are completed to these times
The fast and accurately identification of favored area.Pass through the parameter being rationally arranged first, candidate region caused by MSER algorithm contains
Almost all of effective cell region, recall rate (Recall) is high, but can also generate the region of other a large amount of acellulars, essence simultaneously
Exactness (Precision) is low, and a cell often corresponds to multiple candidate regions, and there are bulk redundancies.Then using cascade
Classifier step by step screens these candidate regions, first order classifier C1 only with the area of candidate region, perimeter and
The simple features such as circularity can quickly exclude a large amount of acellular region；In order to eliminate redundant area, according to the extracted time of MSER
There is only comprising (nested) or non-intersecting (non-overlapping) two kinds of relationships between favored area, establish
Hierarchical MSER tree is indicated it and decomposes to it, obtains a series of single tine tree or individual node (can
Be considered as depth be 1 single tine tree), and based on classifier C1 output using non-maximum value inhibition method by every single tine tree into
Row is reduced to individual node.Remaining candidate region further inputs into second level classifier C2 and is identified.Classifier C2 is adopted
With the complex feature such as region shape, texture, calculating is complex, but precision is high.Due to a large amount of invalid candidate region
Device C1 exclusion, the candidate region negligible amounts of classifier C2 processing are classified, therefore can realize the quick, accurate of whole system
Identification.Technical solution of the invention is as follows:
A kind of DNA quantitative analysis method based on microcytoscope image, comprising the following steps:
Step 1: original RGB color image being pre-processed, including gray processing processing, background correction, Gaussian smoothing
With background pixel removal etc.；
Step 2: using maximum extreme value stability region (Maximally Stable Extremal Regions, MSER) side
Method extracts a large amount of candidate regions from image, on this basis according between the extracted candidate region MSER there is only comprising
(nested) or non-intersecting (non-overlapping) two kinds of relationships, corresponding Hierarchical MSER tree is established, is gone forward side by side
One step is broken down into several single tine trees or individual node；
Step 3: extracting the feature training first order classifier C1 of one 3 dimension of each candidate region, quickly judge each
Whether candidate region is interested cell, and calculates a possibility that it belongs to interested cell (or probability), is based on this point
Class device a possibility that it belongs to cells of interest to the Node extraction feature calculation of every single tine tree, and inhibited using non-maximum value
It is individual node by every single tine tree reduction, finally deletes all nodes (candidate region) that classifier C1 is judged as negative sample；
Step 4: extracting various forms feature, textural characteristics and luminosity (Photometric) for each candidate region
Feature is combined into the feature of one 80 dimension to indicate the candidate region, and training second level classifier C2, using the classifier
Further classified to all candidate regions that classifier C1 output is positive sample, is subdivided into diploid, lymph, grain
Cell and one kind in other equal four classes；
Step 5: calculate all diploids, lymph, granulocyte integral optical density (IOD), and according to reference nucleus
Average integral optical density calculates DNA index (DI) value, completes the DNA quantitative measurment of cell.
Preferably, in the step 1, when carrying out background correction according to given black background BB (x, y) and white background figure
As WB (x, y), the image I after background correction is corrected is carried out to image I (x, y)c(x, y):
When carrying out background pixel removal, 5 × 5 space template approximate Gaussians are used firstConvolution is carried out to image and obtains smoothed out image Is(x, y), after then calculating overturning
Image Ii(x, y)=255-Is(x, y), mean μ and standard deviation δ, can acquire+0.5 δ of global threshold t=μ after, traverse image
Each pixel, by following formula calculate image label figure M (x, y)
It can basis using the label figure M (x, y)After removal background pixel is calculated
Image
Preferably, it in the step 2, is used first when extracting candidate region according to MSER algorithm all in 0-255
Number is used as threshold value pairCarry out binaryzation, and calculate the change rate (i) of some connected region Q when threshold value is i=| Qi-Qi-Δ
|/|Qi-Δ|, wherein Δ is the change of gray value, | | the size for indicating set, as the variation (i) of threshold value obtains pole
Those of small value connected region is used as candidate region, here parameter, Δ=10, since the size and intensity profile of nucleus are deposited
In certain rule, region of the connected region area less than 100 and greater than 10000 is deleted here, while also deleting those (i)
The region of < 0.25；
Due to these regions there is only comprising or non-intersecting two kinds of relationships, the mode that tree can be used unified table is carried out to it
Show, establishes Hierarchical MSER tree, each candidate region one node of corresponding tree, when will increase threshold value at various locations
On the candidate region that occurs first as leaf node, with the candidate region that the increase of threshold value obtains usually contain two or
Multiple regions, then directly using the candidate region newly obtained as the father node of institute's inclusion region, iteration is until reach max-thresholds
Until.
Preferably, in the step 3, after obtaining training set using the method for semi-automatic mark first, each region is extracted
Area, perimeter and circularity as feature, training gradient promotes decision tree classifier as first order classifier C1 for candidate regions
Domain is divided into two classes, and objective function when classifier C1 is trained to be made of loss item L and regular terms Ω, and loss item L is using as follows
Logistic loss function:
Here yi∈ { 0,1 } indicates the true classification of i-th of sample, yiIndicate current classifier to its predicted value, instruction
The depth capacity of every regression tree is 2 when practicing, and maximum tree of regression tree is 5；
Every Hierarchical MSER tree further can be carried out by letter using non-maximum suppression method according to classifier C1
Individual node is turned to, decision is then carried out to all simplified nodes using classifier C1, only retaining C1 output is positive sample
Part candidate region.
Preferably, in the step 4, after obtaining training set using the method for semi-automatic mark first, one GBDT of training
Classifier as second level classifier C2, by candidate region be divided into diploid, granulocyte, lymphocyte and other etc. four classes, this
In each region use one 80 dimension vector representation, including 52 dimension morphological feature, 11 dimension photometric features and 17 dimension
Textural characteristics train objective function when classifier C2 to be still made of loss item L and regular terms Ω, and loss item L is using as follows
Softmax loss function:
Here 1 { } was indicator function, i.e., 1 { true }=1.1 { false }=0, yi∈ { 0,1,2,3 } indicates the
The true classification of i sample,Indicate that current classifier is judged as the predicted value of jth class；Regular terms Ω is then by two
It is grouped as, andT is the number of leaf node, w ∈ R among theseTFor leaf node weight vector, regularization
Parameter γ and λ is determined that the depth of regression tree is 8, using 100 regression trees by 10 times of cross validations by grid search.
Detailed description of the invention
Fig. 1 is a kind of flow chart of the DNA quantitative analysis method based on microcytoscope image proposed by the present invention；
Fig. 2 is a kind of original image of the DNA quantitative analysis method based on microcytoscope image proposed by the present invention；
Fig. 3 is to scheme after a kind of pretreatment of the DNA quantitative analysis method based on microcytoscope image proposed by the present invention
Picture；
Fig. 4 is to be obtained in a kind of DNA quantitative analysis method based on microcytoscope image proposed by the present invention according to MSER
The candidate region arrived；
Fig. 5 is in a kind of DNA quantitative analysis method based on microcytoscope image proposed by the present invention according to the first order
The candidate region obtained after classifier and Hierarchical MSER tree reduction；
Fig. 6 is in a kind of DNA quantitative analysis method based on microcytoscope image proposed by the present invention by the second level
Result after classifier identification；
Fig. 7 is to mark by hand in a kind of DNA quantitative analysis method based on microcytoscope image proposed by the present invention
As a result.
Specific embodiment
Combined with specific embodiments below the present invention is made further to explain.
Referring to Fig.1-7, a kind of DNA quantitative analysis method based on microcytoscope image proposed by the present invention, including with
Lower step:
Step 1: the r (x, y) of original color image, g (x, y), b (x, y are extracted in pretreatment, first progress gray processing processing
Mean value is calculated behind channel, and as gray level image I (x, y) after being overturn, i.e. I (x, y)=(r (x, y)+b (x, y)+g (x,
Y))/3, background correction is then carried out, i.e., according to given black background BB (x, y) and white background image WB (x, y), to image I
(x, y) carries out the image I after background correction is correctedc(x, y):
Further image is denoised, using Gaussian filterTo Ic(x, y) carries out Gauss
Smoothly, i.e. Is(x, y)=Ic(x, y) * h (x, y) uses 5 × 5 space template approximate Gaussian h (x, y), wherein σ in practice
=1.0；It finally calculates global threshold and removes background pixel, image is overturn, i.e. Ii(x, y)=255-Is(x, y), then
Calculate separately image IiThe mean μ and standard deviation δ of (x, y), and then+0.5 δ of global threshold t=μ can be acquired, then traverse image
Each pixel, by following formula calculate image label figure M (x, y)
Image after removal background pixel can be calculated according to the following formula according to the label figure
Step 2: using MSER algorithm extract candidate region, MSER algorithm image is successively used different threshold values ([0,
255] binaryzation) is carried out to image, a series of connected regions are obtained, unlike traditional MSER algorithm, here according under
Formula calculates some connected region Q when threshold value is iiChange rate q (i) (i)=| Qi-qi-Δ|/|Qi-Δ|, wherein Δ is gray value
Change, | | the size for indicating set, as the variation (i) of threshold value obtains the connected region, that is, conduct of those of minimum
Candidate region, parameter, Δ=10 here, since the size and intensity profile of nucleus are there are certain rule, the company of deleting here
Logical region of the region area less than 100 and greater than 10000, while also deleting the region of those (i) < 0.25；Then according to MSER
Segmentation result generate Hierarchical MSER tree, since MSER algorithm extracts candidate region by the way of incremental threshold value,
Therefore obtained region is between each other there is only comprising (nested) or non-intersecting (non-overlapping) two kinds of relationships,
The mode that tree can be used is indicated it, each candidate region one node of corresponding tree, when will increase threshold value at various locations
On the candidate region that occurs first as leaf node, with the candidate region that the increase of threshold value obtains usually contain two or
Multiple regions, then directly using the candidate region newly obtained as the father node of institute's inclusion region, iteration is until reach max-thresholds
Until, a series of Hierarchical MSER trees finally can be obtained；
Step 3: the training of classifier C1 chooses cell distribution dispersion, the simple 200 width cell image of background, to every width
Image uses Otsu algorithm calculating global threshold to divide to obtain several candidate regions, then manually by these regions it into property
Cell and two class of acellular are labeled as training set training classifier C1, decision tree (Gradient is promoted using gradient here
Boosted Decision Trees, GBDT) classifier, extract the area (area) in each region, perimeter (perimeter) and
Circularity (circularity=4 π × area/perimeter2) etc. input of the features as classifier, GBDT classifier is by more
Regression tree composition completes decision as final differentiation result by the sum of each regression tree response, using ladder when training jointly
The mode of degree decline iterative learning, each iteration learn a regression tree using greedy algorithm and go to approach the ladder of current goal function
Degree trains objective function when classifier C1 to be made of loss item L and regular terms Ω, and loss item L is damaged using following Logistic
Lose function:
Here yi∈ { 0,1 } indicates the true classification of i-th of sample,Indicate current classifier to its predicted value, instruction
The depth capacity of every regression tree is 2 when practicing, and maximum tree of regression tree is 5；
In order to eliminate the redundancy of candidate region, further Hierarchical MSER tree is decomposed, first against
It is defeated after each node region extracts the normalization of going forward side by side property of the features such as its area, perimeter and circularity in Hierarchical MSER tree
Enter to classifier C1 and calculate its response, and completes whether may be the classification of cell compartment；Then from each leaf section of tree
Point sets out, and successively accesses its father node, until the root node of the tree, a series of single tine trees can be obtained；According to the sound of classifier C1
The method for using non-maximum value to inhibit should be worth by these single tine tree reductions for individual node region, i.e., only retain and classify in single tine tree
That maximum node region of device C1 response；Finally remaining candidate region is sieved according to the classification results of classifier C1
Choosing only retains classification results and is the candidate region of positive sample, while removing repeat region；
Step 4: the training of classifier C2 randomly selects 5000 width cell images from clinical data, and uses aforementioned side
Method obtains several candidate regions, by the way of artificial by these candidate regions be labeled as diploid, granulocyte, lymphocyte and
Other training sets of equal four classes as classifier C2, classifier C2 still uses GBDT classifier here, is extracted various forms
Feature, textural characteristics and luminosity (Photometric) feature are learned, following 80 dimensional feature is screened out from it:
52 dimension morphological features, comprising: region area and perimeter, circularity, the point on zone boundary to regional barycenter most
Greatly, minimum range and average distance and its variance, convexity, inertia, eccentricity, region main axis length and length ratio, 32 dimension Fu
Hu not bending moment etc. is tieed up in vertical leaf shape description, 7；
11 dimension photometric features, comprising: integral optical density IOD, OD maximum and variance, the area grayscale mean value in region
With variance, the gray average of zone boundary and variance, optical density and the skewness and kurtosis of gray scale etc.；
17 dimension textural characteristics, comprising: 9 dimension invariable rotary uniform local binary pattern (LBP) features, 8
Tie up gray level co-occurrence matrixes feature etc.；
Objective function in training classifier C2 is made of loss item L and regular terms Ω, and loss item L is using as follows
Softmax loss function:
Here 1 { } was indicator function, i.e., 1 { true }=1,1 { false }=0, yi∈ { 0,1,2,3 } indicates the
The true classification of i sample,Indicate that current classifier is judged as the predicted value of jth class；Regular terms Ω is then by two
It is grouped as, andT is the number of leaf node, w ∈ R among theseTFor leaf node weight vector, regularization
Parameter γ and λ is determined that the depth of regression tree is 8, using 100 regression trees by 10 times of cross validations by grid search；
Step 5: calculating the DI value of nucleus, belong to positive sample after Hierarchical MSER tree is decomposed to all
Candidate region carry out feature extraction after be input to classifier C2 and classify, and each nucleus is calculated according to average IOD value
DI value (=IOD/ be averaged IOD).
The foregoing is only a preferred embodiment of the present invention, but scope of protection of the present invention is not limited thereto,
Anyone skilled in the art in the technical scope disclosed by the present invention, according to the technique and scheme of the present invention and its
Inventive concept is subject to equivalent substitution or change, should be covered by the protection scope of the present invention.
Claims (5)
1. a kind of DNA quantitative analysis method based on microcytoscope image, which comprises the following steps:
Step 1: original RGB color image being pre-processed, including gray processing processing, background correction, Gaussian smoothing and back
The removal of scene element；
Step 2: a large amount of candidate regions being extracted from image using maximum extreme value stability region method, on this basis according to MSER
Between extracted candidate region there is only comprising or non-intersecting two kinds of relationships, establish corresponding Hierarchical MSER tree,
And further it is broken down into several single tine trees or individual node；
Step 3: extracting the feature training first order classifier C1 of one 3 dimension of each candidate region, quickly judge each candidate
Whether region is interested cell, and calculates a possibility that it belongs to interested cell, based on the classifier to every list
Its a possibility that belonging to cells of interest of the Node extraction feature calculation of tree is pitched, and is inhibited using non-maximum value by every single tine tree
It is reduced to individual node, finally deletes all nodes that classifier C1 is judged as negative sample；
Step 4: extracting morphological feature, textural characteristics and photometric features for each candidate region, be combined into one 80 dimension
Feature exports the sample that is positive to classifier C1 to indicate the candidate region, and training second level classifier C2, using the classifier
This all candidate regions are further classified, be subdivided into diploid, lymph, granulocyte and other；
Step 5: calculate all diploids, lymph, granulocyte integral optical density, and according to the average product of reference nucleus be divided
Density calculates DNA index value, completes the DNA quantitative measurment of cell.
2. a kind of DNA quantitative analysis method based on microcytoscope image according to claim 1, which is characterized in that
In the step 1, when carrying out background correction according to given black background BB (x, y) and white background image WB (x, y), to image
I (x, y) carries out the image I after background correction is correctedc(x, y):
When carrying out background pixel removal, 5 × 5 space template approximate Gaussians are used firstConvolution is carried out to image and obtains smoothed out image Is(x, y), after then calculating overturning
Image Ii(x, y)=255-Is(x, y), mean μ and standard deviation δ, can acquire+0.5 δ of global threshold t=μ after, traverse image
Each pixel, by following formula calculate image label figure M (x, y)
It can basis using the label figure M (x, y)Image after removal background pixel is calculated
3. a kind of DNA quantitative analysis method based on microcytoscope image according to claim 1, which is characterized in that
In the step 2, use number all in 0-255 as threshold value pair first according to MSER algorithm when extracting candidate regionCarry out binaryzation, and calculate the change rate (i) of some connected region Q when threshold value is i=| Qi-Qi-Δ)|/|Qi-Δ|, wherein
Δ is the change of gray value, | | the size for indicating set, as the variation (i) of threshold value obtains those of minimum even
Logical region is used as candidate region, here parameter, Δ=10, since the size and intensity profile of nucleus are there are certain rule,
Here region of the connected region area less than 100 and greater than 10000 is deleted, while also deleting the region of those (i) < 0.25；
Due to these regions there is only comprising or non-intersecting two kinds of relationships, the mode that tree can be used unified representation is carried out to it, is built
Vertical Hierarchical MSER tree, each candidate region one node of corresponding tree, will increase threshold value when at various locations on first
The candidate region of appearance is as leaf node, as the candidate region that the increase of threshold value obtains usually contains two or more areas
Domain, then directly using the candidate region newly obtained as the father node of institute's inclusion region, iteration is until reaching max-thresholds.
4. a kind of DNA quantitative analysis method based on microcytoscope image according to claim 1, which is characterized in that
In the step 3, after obtaining training set using the method for semi-automatic mark first, the area, perimeter and circle in each region are extracted
Degree is used as feature, and training gradient promotes decision tree classifier as first order classifier C1 and candidate region is divided into two classes, training
Objective function when classifier C1 is made of loss item L and regular terms Ω, and loss item L uses following Logistic loss function:
Here yi∈ { 0,1 } indicates the true classification of i-th of sample, yiCurrent classifier is indicated to its predicted value, when training
The depth capacity of every regression tree is 2, and maximum tree of regression tree is 5；
Further every Hierarchical MSER tree can be reduced to using non-maximum suppression method according to classifier C1
Then individual node carries out decision to all simplified nodes using classifier C1, only retain C1 output is positive sample that
Segment candidate region.
5. a kind of DNA quantitative analysis method based on microcytoscope image according to claim 1, which is characterized in that
In the step 4, after obtaining training set using the method for semi-automatic mark first, one GBDT classifier of training is as the second level
Classifier C2, by candidate region be divided into diploid, granulocyte, lymphocyte and other, here each region using one 80 dimension
Vector representation, including 52 dimension morphological feature, 11 dimension photometric features and 17 dimension textural characteristics, training classifier C2 when
Objective function be still made of loss item L and regular terms Ω, loss item L use following Softmax loss function:
Here 1 { } was indicator function, i.e., 1 { true }=1,1 { false }=0, yi∈ { 0,1,2,3 } indicates i-th of sample
This true classification,Indicate that current classifier is judged as the predicted value of jth class；Regular terms Ω is then by two parts group
At, andT is the number of leaf node, w ∈ R among theseTFor leaf node weight vector, regularization parameter
γ and λ is determined that the depth of regression tree is 8, using 100 regression trees by 10 times of cross validations by grid search.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610769611.1A CN106340016B (en) | 2016-08-31 | 2016-08-31 | A kind of DNA quantitative analysis method based on microcytoscope image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610769611.1A CN106340016B (en) | 2016-08-31 | 2016-08-31 | A kind of DNA quantitative analysis method based on microcytoscope image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106340016A CN106340016A (en) | 2017-01-18 |
| CN106340016B true CN106340016B (en) | 2019-03-22 |
Family
ID=57823309
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610769611.1A Active CN106340016B (en) | 2016-08-31 | 2016-08-31 | A kind of DNA quantitative analysis method based on microcytoscope image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106340016B (en) |
Families Citing this family (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108198171A (en) * | 2017-12-27 | 2018-06-22 | 大连理工大学 | A kind of method of the operating status diagnosis based on server surface Warm status |
| CN108074245B (en) * | 2018-01-03 | 2022-03-25 | 深圳北航新兴产业技术研究院 | Method and device for segmenting and detecting microscopic cell image |
| CN109300111B (en) * | 2018-08-27 | 2020-05-12 | 杭州德适生物科技有限公司 | Chromosome recognition method based on deep learning |
| US10748290B2 (en) * | 2018-10-31 | 2020-08-18 | Fei Company | Smart metrology on microscope images |
| CN109540890A (en) * | 2018-11-27 | 2019-03-29 | 湖南品胜生物技术有限公司 | A kind of DNA quantitative analysis method based on microcytoscope image |
| CN110335267A (en) * | 2019-07-05 | 2019-10-15 | 华侨大学 | A kind of cervical lesions method for detecting area |
| CN110795995A (en) * | 2019-09-16 | 2020-02-14 | 平安科技（深圳）有限公司 | Data processing method, device and computer readable storage medium |
| CN110929347A (en) * | 2019-10-25 | 2020-03-27 | 东北大学 | Hot continuous rolling strip steel convexity prediction method based on gradient lifting tree model |
| CN111161332A (en) * | 2019-12-30 | 2020-05-15 | 上海研境医疗科技有限公司 | Homologous pathology image registration preprocessing method, device, equipment and storage medium |
| WO2021135300A1 (en) * | 2019-12-30 | 2021-07-08 | 歌尔股份有限公司 | Product defect detection method, device, and system |
| CN112396583B (en) * | 2020-11-18 | 2024-01-26 | 深思考人工智能机器人科技（北京）有限公司 | DNA ploidy quantitative analysis method and system based on Papanicolaou staining mode |
| CN112750493B (en) * | 2020-12-22 | 2024-05-03 | 深思考人工智能机器人科技（北京）有限公司 | DNA ploid quantitative analysis method and system based on Papanicolaou staining mode |
| CN112651368B (en) * | 2020-12-30 | 2023-03-28 | 宁波江丰生物信息技术有限公司 | DNA ploidy analysis method for large-size cell microscopic image |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101492740A (en) * | 2009-02-24 | 2009-07-29 | 武汉兰丁医学高科技有限公司 | Correct measurement method for nucleus DNA matter content in cell quantitative investigation |
| CN102147523A (en) * | 2011-03-24 | 2011-08-10 | 姚斌 | Biological digital microscope with double ccd (charge coupled device) light sensitive elements and photographic image processing method thereof |
| CN105550651A (en) * | 2015-12-14 | 2016-05-04 | 中国科学院深圳先进技术研究院 | Method and system for automatically analyzing panoramic image of digital pathological section |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| ITTO20130940A1 (en) * | 2013-11-20 | 2015-05-21 | St Microelectronics Srl | KIT FOR BIOCHEMICAL ANALYSIS AND METHOD TO PERFORM A IMPROVED BIOCHEMICAL PROCESS |
-
2016
- 2016-08-31 CN CN201610769611.1A patent/CN106340016B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101492740A (en) * | 2009-02-24 | 2009-07-29 | 武汉兰丁医学高科技有限公司 | Correct measurement method for nucleus DNA matter content in cell quantitative investigation |
| CN102147523A (en) * | 2011-03-24 | 2011-08-10 | 姚斌 | Biological digital microscope with double ccd (charge coupled device) light sensitive elements and photographic image processing method thereof |
| CN105550651A (en) * | 2015-12-14 | 2016-05-04 | 中国科学院深圳先进技术研究院 | Method and system for automatically analyzing panoramic image of digital pathological section |
Non-Patent Citations (2)
| Title |
|---|
| Learning to detect and track cells for quantitative analysis of time-lapse microscopic image sequences;Pedro D. Kostelec et al.;《2015 IEEE 12th International Symposium on Biomedical Imaging (ISBI)》;20150723;全文 |
| 基于计算机视觉的肿瘤早期诊断方法研究;庞宝川;《华中科技大学博士学位论文 2009》;20111130;全文 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106340016A (en) | 2017-01-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106340016B (en) | A kind of DNA quantitative analysis method based on microcytoscope image | |
| Versari et al. | Long-term tracking of budding yeast cells in brightfield microscopy: CellStar and the Evaluation Platform | |
| Rodriguez et al. | Automated three-dimensional detection and shape classification of dendritic spines from fluorescence microscopy images | |
| CN111079620B (en) | White blood cell image detection and identification model construction method and application based on transfer learning | |
| JP2008077677A (en) | Measurement of mitotic activity | |
| CN113537182B (en) | Automatic identification method and system for metaphase mitosis microscopic image of chromosome | |
| CN105389821B (en) | It is a kind of that the medical image cutting method being combined is cut based on cloud model and figure | |
| CN112215790A (en) | KI67 index analysis method based on deep learning | |
| CN112819821B (en) | Cell nucleus image detection method | |
| CN110853070A (en) | Underwater sea cucumber image segmentation method based on significance and Grabcut | |
| CN109871900A (en) | The recognition positioning method of apple under a kind of complex background based on image procossing | |
| Harun et al. | Unsupervised segmentation technique for acute leukemia cells using clustering algorithms | |
| CN110874835B (en) | Crop leaf disease resistance identification method and system, electronic equipment and storage medium | |
| CN115294377A (en) | System and method for identifying road cracks | |
| CN114581451A (en) | Scattering map neural network-based brain magnetic resonance image segmentation method | |
| CN116523898A (en) | Tobacco phenotype character extraction method based on three-dimensional point cloud | |
| He et al. | An automated three-dimensional detection and segmentation method for touching cells by integrating concave points clustering and random walker algorithm | |
| CN112183237A (en) | Automatic white blood cell classification method based on color space adaptive threshold segmentation | |
| CN110648312A (en) | Method for identifying wool and cashmere fibers based on scale morphological characteristic analysis | |
| Ding et al. | Classification of chromosome karyotype based on faster-rcnn with the segmatation and enhancement preprocessing model | |
| Rathore et al. | CBISC: a novel approach for colon biopsy image segmentation and classification | |
| Zhao et al. | An automatic counting method of maize ear grain based on image processing | |
| Guatemala-Sanchez et al. | Nuclei segmentation on histopathology images of breast carcinoma | |
| CN113889233A (en) | Cell positioning and counting method based on manifold regression network and application | |
| CN113344008A (en) | High-throughput extraction method of stalk tissue anatomical characteristic parameters based on deep learning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| PE01 | Entry into force of the registration of the contract for pledge of patent right |
Denomination of invention: DNA quantitative analysis method based on cell microscope image Effective date of registration: 20200310 Granted publication date: 20190322 Pledgee: Changsha Furong sub branch of China Construction Bank Co., Ltd Pledgor: HUNAN PINXIN BIOENGINEERING Co.,Ltd. Registration number: Y2020980000563 |
|
| PE01 | Entry into force of the registration of the contract for pledge of patent right | ||
| PC01 | Cancellation of the registration of the contract for pledge of patent right |
Date of cancellation: 20210511 Granted publication date: 20190322 Pledgee: Changsha Furong sub branch of China Construction Bank Co.,Ltd. Pledgor: HUNAN PINXIN BIOENGINEERING Co.,Ltd. Registration number: Y2020980000563 |
|
| PC01 | Cancellation of the registration of the contract for pledge of patent right | ||
| PE01 | Entry into force of the registration of the contract for pledge of patent right |
Denomination of invention: A DNA quantitative analysis method based on cell microscope image Effective date of registration: 20210616 Granted publication date: 20190322 Pledgee: Changsha Furong sub branch of China Construction Bank Co.,Ltd. Pledgor: HUNAN PINXIN BIOENGINEERING Co.,Ltd. Registration number: Y2021430000019 |
|
| PE01 | Entry into force of the registration of the contract for pledge of patent right |