CN108764329A - A kind of construction method of lung cancer pathology image data set - Google Patents
A kind of construction method of lung cancer pathology image data set Download PDFInfo
- Publication number
- CN108764329A CN108764329A CN201810510872.0A CN201810510872A CN108764329A CN 108764329 A CN108764329 A CN 108764329A CN 201810510872 A CN201810510872 A CN 201810510872A CN 108764329 A CN108764329 A CN 108764329A
- Authority
- CN
- China
- Prior art keywords
- data set
- cancer cell
- lung cancer
- pathologic
- pathology image
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—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
- G06T2207/30061—Lung
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Quality & Reliability (AREA)
- Investigating Or Analysing Biological Materials (AREA)
Abstract
The present invention relates to a kind of construction methods of lung cancer pathology image data set, include the following steps：Step S1：Build the reference standard data set of pathologic image；Step S2：Build cancer cell assay model；Step S3：Build lung cancer pathology image sample data collection.It the advantage is that, lung cancer pathology image reference standard data set is built by manually marking a small amount of initial data, then reference standard data set is trained to build the cancer cell assay model based on low volume data collection, the detection for being carried out cancer cell to pathologic original image data using model is identified, lung cancer pathology image sample data collection is generated；Artificial sampling verification is carried out to sample data set, lung cancer pathology image reference standard data set is added in the data verified, training then is iterated to data set and model, improves the accuracy rate of lung cancer pathology image sample data collection；Structure speed is effectively improved, solves the problems, such as structure lung cancer pathology image data set heavy workload, longevity of service.
Description
Technical field
The present invention relates to image identification technical field more particularly to a kind of construction methods of lung cancer pathology image data set.
Background technology
Depth learning technology is quickly grown in recent years, obtains huge success in many fields, some researches show that depth
Study has been achieved for certain achievement in terms of the detection identification in medical imaging field, and it is latent to have shown huge application
Power is expected to reach the precision of pathological diagnosis expert in lung cancer pathology image recognition, realizes speed more faster than pathological diagnosis expert
Degree.
The deep learning model haveing excellent performance is built to need to rely on the sample data set of extensive high quality, however at present originally
Field lacks extensive lung cancer pathology image sample data collection, and there are the pathologic slice of a large amount of true cases in many hospitals
Image data carries out sample mark to these image datas and needs a large amount of manpower and time, and according to statistics, pathological diagnosis expert examines
The time of disconnected every image needs about 3 minutes, and extensive lung cancer pathology image pattern number can not be built based entirely on artificial mark
According to collection, so needing a kind of construction method of extensive lung cancer pathology image data set.
Therefore, it can build that identifying processing speed is fast, accuracy rate is high, the method for the big data set of database there is an urgent need for a kind of.
Invention content
The purpose of the present invention is being directed to deficiency in the prior art, a kind of structure side of lung cancer pathology image data set is provided
Method.
To achieve the above object, the technical solution adopted by the present invention is that：
A kind of construction method of lung cancer pathology image data set, includes the following steps：
Step S1：Build the reference standard data set of pathologic image
N pathologic images are arbitrarily chosen in N pathologic images, and to choosing obtained n (n < N)
Zhang Suoshu pathologic images are labeled, and use the pathologic picture construction reference standard data after mark
Collection；
Step S2：Build cancer cell assay model
The cancer cell assay model is built based on deep neural network, and using the reference of step S1 structures
Cancer cell assay model described in normal data set pair is trained；
Step S3：Build lung cancer pathology image sample data collection
Using the cancer cell assay model after step S2 training to the remaining pathologic image
It is detected identification, builds lung cancer pathology image sample data collection.
Preferably, further include：
Step S4：Update reference standard data set
Arbitrary selection m (m < N) lung tissue is concentrated in the lung cancer pathology image sample data of step S3 structures
Pathological image, and the m pathologic images are assessed, and the m after assessment is opened into the pathologic images
The reference standard data set is added to update the reference standard data set.
Preferably, further include：
Step S5：Optimize cancer cell assay model
Using the cancer cell assay mould built to step S2 through the updated reference standard data sets of step S4
Type is trained, and obtains the cancer cell assay model of performance optimization.
Preferably, further include：
Step S6：Update lung cancer pathology image sample data collection
The lung cancer pathological image that step S3 is built using the cancer cell assay model after step S5 optimizations
Sample data set is detected identification, updates the lung cancer pathology image sample data collection.
Preferably, the cancer cell assay model includes cancer cell detection model and cancer cell identification model.
Preferably, Resnet-101 deep neural networks are based on and build the cancer cell detection model.
Preferably, R-FCN deep neural networks are based on and build the cancer cell identification model.
Preferably, the step S4 further includes：
To the pathologic identified by the cancer cell assay model errors in the m pathologic images
Image is marked again.
Preferably, it is described be labeled as include：
The pathologic image is labeled with the presence or absence of cancer cell；
To there are the pathologic images of cancer cell to carry out cancer cell position mark；
To there are the pathologic images of cancer cell to carry out cancer cell-types mark.
Preferably, the step S1 further includes：
Cross validation is carried out to the pathologic image after mark.
The present invention is had the following technical effect that compared with prior art using above technical scheme：
The construction method of a kind of lung cancer pathology image data set of the present invention, by manually marking a small amount of initial data come structure
Lung cancer pathology image reference standard data set is built, then reference standard data set is trained and is based on low volume data collection to build
Machine sort and detection identification model, include cancer-free cell disaggregated model and lung carcinoma cell detection identification model, use
The two models carry out the classification for having cancer-free cell to pathologic original image data and the detection of cancer cell identifies, to
Generate lung cancer pathology image sample data collection；Artificial sampling verification is carried out to sample data set by pathological diagnosis expert, will be verified
Lung cancer pathology image reference standard data set is added in the data crossed, and is then iterated training to data set to improve above-mentioned two
The accuracy rate of lung cancer pathology image sample data collection is continuously improved in the performance of model；Effectively improve extensive lung cancer pathological image
The structure speed of data set solves to build well extensive lung cancer pathology image data set heavy workload, longevity of service
Problem.
Description of the drawings
Fig. 1 is the method flow diagram of a preferred embodiment of the present invention.
Fig. 2 is the method detailed flow chart of a preferred embodiment of the present invention.
Fig. 3 is the network structure of the cancer cell detection model of a preferred embodiment of the present invention.
Fig. 4 is the network structure of the cancer cell identification model of a preferred embodiment of the present invention.
Specific implementation mode
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation describes, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, those of ordinary skill in the art obtained under the premise of not making creative work it is all its
His embodiment, shall fall within the protection scope of the present invention.
It should be noted that in the absence of conflict, the feature in embodiment and embodiment in the present invention can phase
Mutually combination.
The invention will be further described in the following with reference to the drawings and specific embodiments, but not as limiting to the invention.
The preferred embodiment of the present invention, as shown in Fig. 1~2, a kind of structure side of lung cancer pathology image data set
Method, including step S1~S3, it is specific as described below.
Step S1：Build the reference standard data set of pathologic image
N pathologic original image datas are obtained, and are arbitrarily chosen from N pathologic original image datas
N pathologic image datas, and the n pathologic images that selection obtains are labeled, use the n after mark
Pathologic image data, which is used as, refers to standard data set.
In choosing data procedures, n is less than N, and preferably n is less than or equal to the 5% of N, more preferably n be less than or
Equal to the 1% of N.
It is to be diagnosed by several pathological diagnosis experts and manually marked to the method that pathologic image is labeled.
The marked content being labeled to pathologic image includes：With the presence or absence of cancer cell, cancer cell position and
Cancer cell-types.
Its diagnostic process is as follows：
The first, pathological diagnosis expert diagnoses pathologic image with the presence or absence of cancer cell, by pathologic
Image is divided into the pathologic image of cancer-free cell and has the pathologic image of cancer cell and do corresponding mark；
The second, pathological diagnosis expert is to there is the pathologic image of cancer cell to carry out cancer cell position identification and cancer cell
Type identification, and accordingly marked.
In order to improve the accuracy of mark, the pathologic image after mark can be verified.
After pathological diagnosis expert diagnoses and marks to pathologic image, by other pathological diagnosis expert to mark
Pathologic image afterwards carries out cross validation, to ensure the accuracy of pathologic image labeling.
In general, pathologic image is diagnosed and is marked by an at least pathological diagnosis expert, by least another
One pathology is made an arbitrary dicision expert and is confirmed to the pathologic image after mark.
In order to improve efficiency and accuracy rate, in the present embodiment, preferred method is that a pathological diagnosis expert is to lung group
It knits pathological image to be diagnosed and marked, two pathological diagnosis experts to the pathologic image after mark intersect really
Recognize.
Step S2：Build cancer cell assay model
Cancer cell assay model, and the reference standard data set pair obtained using step S1 are built based on deep neural network
Cancer cell assay model is trained.
Cancer cell assay model includes cancer cell detection model and cancer cell identification model, and cancer cell detection model is for examining
It surveys and whether there is cancer cell in pathologic image, cancer cell identification model is for knowing the position of cancer cell and type
Not.
Deep neural network can there are many types, in the present invention it is preferred that depth residual error network (Resnet)
With the full convolutional network (R-FCN) based on region.
Further, cancer cell detection model is built using Resnet-101 deep neural networks, utilizes R-FCN depth
Neural network builds cancer cell identification model.
Such as Fig. 3 institutes are preferably implemented using one of the cancer cell detection model of Resnet-101 deep neural networks structure
Show, flow is as follows：Pathologic image is inputted, and first time convolution operation is carried out to pathologic image, after convolution
The output with multiple Feature Mappings is obtained as a result, and carrying out the operation of first time pondization to the output result；To carrying out for the first time
Output result behind pond carries out second of convolution, and the output with multiple Feature Mappings is obtained as a result, and defeated to this after convolution
Go out result and carries out second of pondization operation；To carry out second of pond after output result enter full articulamentum multilayer perceptron into
Row operation obtains final output result.
Using R-FCN deep neural networks structure cancer cell identification model a preferred embodiment as shown in figure 4, its
Flow is as follows：Pathologic image is inputted, and first time convolution operation is carried out to pathologic image, obtains output knot
Fruit；Convolution again is carried out to the output result after first time convolution, region is respectively obtained and suggests network (RPN) and territorial classification net
Network (RoI) suggests that network extracts area-of-interest using region, and carries out pond to the area-of-interest, obtains final defeated
Go out result.
After having built cancer cell detection model and cancer cell identification model, the reference standard data that are utilized respectively in step S1
Collection is respectively trained, to improve cancer cell detection model and cancer cell identification model.
Step S3：Build lung cancer pathology image sample data collection
Knowledge is detected to remaining pathologic image using the cancer cell assay model after step S2 training
Not, lung cancer pathology figure sample data set is built.
(N-n) pathologic image data is detected first with cancer cell detection model, and to lung tissue
Pathological image data are labeled, to obtain the pathologic image of cancer-free cell and have the pathologic figure of cancer cell
Picture.
Then utilize cancer cell identification model to the lung tissue for having cancer cell in (N-n) pathologic image data
Pathological image is identified, and be labeled to pathologic image data, to obtain mark cancer cell position and type
Pathologic image.
Finally, lung cancer pathology image pattern number is built using (N-n) after the completion of mark pathologic image datas
According to collection.
In general, passing through step S1~step S3, you can data volume is big, discrimination is high, accuracy is high for structure one
Lung cancer pathology image sample data collection.
In order to further increase the discrimination and accuracy of sample data set, to the lung built by step S1~S3
Carninomatosis reason image data set is iterated optimization.
Specifically, further including step S4~S6 after step s 3.
Step S4：Update reference standard data set
Arbitrary selection m pathologic image is concentrated in the lung cancer pathology image sample data of step S3 structures, and right
These m chosen pathologic image is assessed, and the m after assessment pathologic images are added to reference to mark
Quasi- data set is to update reference standard data set.
During selection, m is less than N, and preferably m is less than or equal to the 5% of N, and more preferably m is less than or equal to
The 1% of N.
In the present embodiment, m is equal to n.
The method assessed pathologic image is：By several pathological diagnosis experts to pathologic image into
Row diagnosis, counts the accuracy rate of the m pathologic images of arbitrary selection, and to the pathologic image of marking error into
Row marks again, is supplemented in reference standard data set after carrying out cross validation to the pathologic image marked again.
Above-mentioned appraisal procedure can refer to diagnosis, mark and verification in step S1.
Step S5：Optimize cancer cell assay model
The cancer cell assay model that step S2 is built is carried out again using step S4 updated reference standard data sets
Training, to obtain the cancer cell assay model of performance optimization.
In step s 5, detailed process is with reference to step S2.
Step S6：Update lung cancer pathology image sample data collection
The lung cancer pathology image sample data collection that step S3 is built using the cancer cell assay model after step S5 optimizations
It is detected identification, updates lung cancer pathology image sample data collection.
In step s 6, detailed process is with reference to step S3.
It, can be to lung cancer pathology figure in order to make the accuracy of lung cancer pathology image sample data collection meet practice requirement
As sample data set progress successive ignition optimization, that is, step S4~S6 is repeated several times, improves lung cancer pathology image sample data collection
Accuracy.
In order to which the present invention is furture elucidated, practice is carried out using the construction method of the present invention, as follows.
Step S1：Build the reference standard data set of pathologic image
It collects pathologic original image data 218937 to open, random sampling chooses 1000, by pathological diagnosis expert
This 1000 pathologic images are diagnosed, marks and whether there is cancer cell in pathologic image, it is thin that there are cancers
The image labeling cancer cell position of born of the same parents and type, Lung Cancer Types include squamous carcinoma, gland cancer and small cell carcinoma；Every pathologic figure
As passing through three pathological diagnosis expert diagnosis and mark, and cross validation, it is ensured that annotation results are accurate, 1000 after will confirm that
Pathologic image, which is used as, refers to standard data set；Mark spends 100 hours altogether, and average every image labeling time is 6 points
Clock, wherein expert's label time is 3 minutes, two expert's acknowledging times are 3 minutes.
Step S2：Build cancer cell assay model
Cancer cell detection model is built using Resnet-101 deep neural networks and utilizes R-FCN deep neural networks
Build cancer cell identification model；Using in step S1 containing 1000 reference standard data sets respectively to cancer cell detection model
It is trained with cancer cell identification model, realization has the detection of cancer-free cell and its identification of position and type；Cancer cell detects
The model training time be 7.6 hours, the cancer cell identification model training time be 8.4 hours,
Step S3：Build lung cancer pathology image sample data collection
Using the cancer cell detection model and cancer cell identification model built in step S2 and training is completed, collection is remained
More than 210,000 pathologic original image datas of remaininging are detected identification, generate lung cancer pathology image sample data collection；It is average
Every image detection recognition speed is 0.3 second.
Step S4：Update reference standard data set
Lung cancer pathology image pattern reference data concentrate random sampling 1000 open, equally by three pathological diagnosis experts into
Row cross-diagnosis is manually verified, the accuracy rate of statistical sampling data, and the quality assessment result of sample data set is accuracy rate
87.9%；In addition, the sample data set identified to error detection in data from the sample survey is marked again, and it is added to lung tissue disease
Image reference standard data set is managed, the performance for promoting detection identification model；At this point, pathologic image reference criterion numeral
Include 2000 data according to collection.
Step S5：Optimize cancer cell assay model
Using the cancer cell detection of the updated lung cancer pathology image reference normal data set pair step S2 structures of step S4
Model and cancer cell identification model trained again to obtain the cancer cell detection model and cancer cell identification mould of performance optimization
Type.
Step S6：Update lung cancer pathology image sample data collection
The pulmonary carcinosis that cancer cell detection model and cancer cell identification model after being optimized using step S5 build step S3
Reason image sample data collection is detected identification, quality more preferably lung cancer pathology image sample data collection is obtained, at this point, pulmonary carcinosis
The quality assessment result that reason image sample data integrates is accuracy rate 90.3%.
Step S4~S6 is repeated, optimization is iterated to lung cancer pathology image sample data collection, by 10 iteration, sample
Data set quality assessment result is accuracy rate 97.4%, meets the requirement of pathological diagnosis expert.
After above-mentioned steps, extensive lung cancer pathology image data set structure is completed.
The foregoing is merely preferred embodiments of the present invention, are not intended to limit embodiments of the present invention and protection model
It encloses, to those skilled in the art, should can appreciate that all with made by description of the invention and diagramatic content
Equivalent replacement and obviously change obtained scheme, should all be included within the scope of the present invention.
Claims (10)
1. a kind of construction method of lung cancer pathology image data set, which is characterized in that include the following steps：
Step S1：Build the reference standard data set of pathologic image
N (n < N) pathologic images are arbitrarily chosen in N pathologic images, and institutes are opened to the n that selection obtains
It states pathologic image to be labeled, uses the pathologic picture construction reference standard data set after mark；
Step S2：Build cancer cell assay model
The cancer cell assay model is built based on deep neural network, and using the reference standard of step S1 structures
Data set is trained the cancer cell assay model；
Step S3：Build lung cancer pathology image sample data collection
The remaining pathologic image is carried out using the cancer cell assay model after step S2 training
Detection identification, builds lung cancer pathology image sample data collection.
2. the construction method of lung cancer pathology image data set according to claim 1, which is characterized in that further include：
Step S4：Update reference standard data set
Arbitrary selection m (m < N) pathologic is concentrated in the lung cancer pathology image sample data of step S3 structures
Image, and the m pathologic images are assessed, and the m after the assessment pathologic images are added
To the reference standard data set to update the reference standard data set.
3. the construction method of lung cancer pathology image data set according to claim 2, which is characterized in that further include：
Step S5：Optimize cancer cell assay model
Using the cancer cell assay model that step S2 is built through the step S4 updated reference standard data sets into
Row training, obtains the cancer cell assay model of performance optimization.
4. the construction method of lung cancer pathology image data set according to claim 3, which is characterized in that further include：
Step S6：Update lung cancer pathology image sample data collection
The lung cancer pathology image pattern that step S3 is built using the cancer cell assay model after step S5 optimizations
Data set is detected identification, updates the lung cancer pathology image sample data collection.
5. the construction method of lung cancer pathology image data set according to claim 1, which is characterized in that the cancer cell point
It includes cancer cell detection model and cancer cell identification model to analyse model.
6. the construction method of lung cancer pathology image data set according to claim 5, which is characterized in that be based on Resnet-
101 deep neural networks build the cancer cell detection model.
7. the construction method of lung cancer pathology image data set according to claim 5, which is characterized in that deep based on R-FCN
It spends neural network and builds the cancer cell identification model.
8. the construction method of lung cancer pathology image data set according to claim 3, which is characterized in that the step S4 is also
Including：
To the pathologic image identified by the cancer cell assay model errors in the m pathologic images
Again it is marked.
9. the construction method of lung cancer pathology image data set according to claim 1, which is characterized in that described to be labeled as wrapping
It includes：
The pathologic image is labeled with the presence or absence of cancer cell；
To there are the pathologic images of cancer cell to carry out cancer cell position mark；
To there are the pathologic images of cancer cell to carry out cancer cell-types mark.
10. the construction method of lung cancer pathology image data set according to claim 1, which is characterized in that the step S1
Further include：
Cross validation is carried out to the pathologic image after mark.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810510872.0A CN108764329A (en) | 2018-05-24 | 2018-05-24 | A kind of construction method of lung cancer pathology image data set |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810510872.0A CN108764329A (en) | 2018-05-24 | 2018-05-24 | A kind of construction method of lung cancer pathology image data set |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108764329A true CN108764329A (en) | 2018-11-06 |
Family
ID=64005873
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810510872.0A Pending CN108764329A (en) | 2018-05-24 | 2018-05-24 | A kind of construction method of lung cancer pathology image data set |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108764329A (en) |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109893732A (en) * | 2019-02-28 | 2019-06-18 | 杭州智瑞思科技有限公司 | A kind of mechanical ventilation patient-ventilator asynchrony detection method based on Recognition with Recurrent Neural Network |
| CN109979546A (en) * | 2019-04-04 | 2019-07-05 | 成都大学 | Network model analysis platform and construction method based on artificial intelligence number pathology |
| CN110335668A (en) * | 2019-05-22 | 2019-10-15 | 台州市中心医院（台州学院附属医院） | Thyroid cancer cell pathological map auxiliary analysis method and system based on deep learning |
| CN110415791A (en) * | 2019-01-29 | 2019-11-05 | 四川大学华西医院 | System and method is established in a kind of disease library |
| CN110458087A (en) * | 2019-08-07 | 2019-11-15 | 湖南省华芯医疗器械有限公司 | A kind of patient problems position mask method based on deep learning |
| CN110472581A (en) * | 2019-08-16 | 2019-11-19 | 电子科技大学 | A kind of cell image analysis method based on deep learning |
| CN110765855A (en) * | 2019-09-12 | 2020-02-07 | 杭州迪英加科技有限公司 | Pathological image processing method and system |
| CN111312372A (en) * | 2020-02-26 | 2020-06-19 | 广州金域医学检验中心有限公司 | Method and device for establishing medical image standard test data set |
| CN111325103A (en) * | 2020-01-21 | 2020-06-23 | 华南师范大学 | Cell labeling system and method |
| CN111351831A (en) * | 2020-02-26 | 2020-06-30 | 西安交通大学医学院第二附属医院 | Detection and marking device and detection method based on mass spectrometry in histology |
| CN111461165A (en) * | 2020-02-26 | 2020-07-28 | 上海商汤智能科技有限公司 | Image recognition method, recognition model training method, related device and equipment |
| CN112614573A (en) * | 2021-01-27 | 2021-04-06 | 北京小白世纪网络科技有限公司 | Deep learning model training method and device based on pathological image labeling tool |
| CN112926612A (en) * | 2019-12-06 | 2021-06-08 | 中移(成都)信息通信科技有限公司 | Pathological image classification model training method, pathological image classification method and device |
| CN113063778A (en) * | 2021-03-10 | 2021-07-02 | 南通大学 | Pleural effusion monomeric cancer cell preparation method applied to AI recognition |
| JP2023074997A (en) * | 2021-11-18 | 2023-05-30 | 倍利科技股▲ふん▼有限公司 | Ai-assisted automatic marking system and method |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101221118A (en) * | 2007-12-07 | 2008-07-16 | 东华大学 | System and method for intelligent recognizing and counting sputum smear micro-image tubercle bacillus |
| CN102304472A (en) * | 2011-05-18 | 2012-01-04 | 东华大学 | Device and method for detecting sputum smear tubercle bacillus quickly |
| CN107132268A (en) * | 2017-06-21 | 2017-09-05 | 佛山科学技术学院 | A kind of data processing equipment and system for being used to recognize cancerous lung tissue |
| CN107862694A (en) * | 2017-12-19 | 2018-03-30 | 济南大象信息技术有限公司 | A kind of hand-foot-and-mouth disease detecting system based on deep learning |
-
2018
- 2018-05-24 CN CN201810510872.0A patent/CN108764329A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101221118A (en) * | 2007-12-07 | 2008-07-16 | 东华大学 | System and method for intelligent recognizing and counting sputum smear micro-image tubercle bacillus |
| CN102304472A (en) * | 2011-05-18 | 2012-01-04 | 东华大学 | Device and method for detecting sputum smear tubercle bacillus quickly |
| CN107132268A (en) * | 2017-06-21 | 2017-09-05 | 佛山科学技术学院 | A kind of data processing equipment and system for being used to recognize cancerous lung tissue |
| CN107862694A (en) * | 2017-12-19 | 2018-03-30 | 济南大象信息技术有限公司 | A kind of hand-foot-and-mouth disease detecting system based on deep learning |
Non-Patent Citations (2)
| Title |
|---|
| JIFENG DAI等: "R-FCN: Object Detection via Region-based Fully Convolutional Networks", 《ARXIV:1605.06409V2》 * |
| 陈英豪: "半监督学习分类算法的研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110415791A (en) * | 2019-01-29 | 2019-11-05 | 四川大学华西医院 | System and method is established in a kind of disease library |
| CN109893732A (en) * | 2019-02-28 | 2019-06-18 | 杭州智瑞思科技有限公司 | A kind of mechanical ventilation patient-ventilator asynchrony detection method based on Recognition with Recurrent Neural Network |
| CN109979546A (en) * | 2019-04-04 | 2019-07-05 | 成都大学 | Network model analysis platform and construction method based on artificial intelligence number pathology |
| CN110335668A (en) * | 2019-05-22 | 2019-10-15 | 台州市中心医院（台州学院附属医院） | Thyroid cancer cell pathological map auxiliary analysis method and system based on deep learning |
| CN110458087A (en) * | 2019-08-07 | 2019-11-15 | 湖南省华芯医疗器械有限公司 | A kind of patient problems position mask method based on deep learning |
| CN110458087B (en) * | 2019-08-07 | 2022-03-11 | 湖南省华芯医疗器械有限公司 | Patient problem part labeling method based on deep learning |
| CN110472581A (en) * | 2019-08-16 | 2019-11-19 | 电子科技大学 | A kind of cell image analysis method based on deep learning |
| CN110765855A (en) * | 2019-09-12 | 2020-02-07 | 杭州迪英加科技有限公司 | Pathological image processing method and system |
| CN110765855B (en) * | 2019-09-12 | 2023-04-18 | 杭州迪英加科技有限公司 | Pathological image processing method and system |
| CN112926612A (en) * | 2019-12-06 | 2021-06-08 | 中移(成都)信息通信科技有限公司 | Pathological image classification model training method, pathological image classification method and device |
| CN111325103A (en) * | 2020-01-21 | 2020-06-23 | 华南师范大学 | Cell labeling system and method |
| CN111325103B (en) * | 2020-01-21 | 2020-11-03 | 华南师范大学 | Cell labeling system and method |
| CN111461165A (en) * | 2020-02-26 | 2020-07-28 | 上海商汤智能科技有限公司 | Image recognition method, recognition model training method, related device and equipment |
| CN111351831A (en) * | 2020-02-26 | 2020-06-30 | 西安交通大学医学院第二附属医院 | Detection and marking device and detection method based on mass spectrometry in histology |
| CN111312372A (en) * | 2020-02-26 | 2020-06-19 | 广州金域医学检验中心有限公司 | Method and device for establishing medical image standard test data set |
| CN112614573A (en) * | 2021-01-27 | 2021-04-06 | 北京小白世纪网络科技有限公司 | Deep learning model training method and device based on pathological image labeling tool |
| CN113063778A (en) * | 2021-03-10 | 2021-07-02 | 南通大学 | Pleural effusion monomeric cancer cell preparation method applied to AI recognition |
| JP2023074997A (en) * | 2021-11-18 | 2023-05-30 | 倍利科技股▲ふん▼有限公司 | Ai-assisted automatic marking system and method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108764329A (en) | A kind of construction method of lung cancer pathology image data set | |
| CN111178197B (en) | Mass R-CNN and Soft-NMS fusion based group-fed adherent pig example segmentation method | |
| CN111488921B (en) | Intelligent analysis system and method for panoramic digital pathological image | |
| CN107247971B (en) | Intelligent analysis method and system for ultrasonic thyroid nodule risk index | |
| CN108564026A (en) | Network establishing method and system for Thyroid Neoplasms smear image classification | |
| CN107203692A (en) | The implementation method of atrial fibrillation detection based on depth convolutional neural networks | |
| CN106248559A (en) | A kind of leukocyte five sorting technique based on degree of depth study | |
| CN113781397B (en) | Medical image focus detection modeling method, device and system based on federal learning | |
| CN109410238B (en) | Wolfberry identification and counting method based on PointNet + + network | |
| CN106446942A (en) | Crop disease identification method based on incremental learning | |
| CN110222701A (en) | A kind of bridge defect automatic identifying method | |
| CN108564123A (en) | A kind of Thyroid Neoplasms smear image classification method and its device | |
| CN114663431A (en) | Pancreatic tumor image segmentation method and system based on reinforcement learning and attention | |
| CN106611423B (en) | SAR image segmentation method based on ridge ripple filter and deconvolution structural model | |
| CN109544511B (en) | Method for identifying lung nodule by convolutional neural network based on particle swarm optimization | |
| CN109102515A (en) | A kind of method for cell count based on multiple row depth convolutional neural networks | |
| CN109994151A (en) | Predictive genes system is driven based on the tumour of complex network and machine learning method | |
| CN110111895A (en) | A kind of method for building up of nasopharyngeal carcinoma far-end transfer prediction model | |
| CN107578402A (en) | A kind of detecting system and its application method of the pulmonary nodule based on deep learning | |
| CN102831614B (en) | Sequential medical image quick segmentation method based on interactive dictionary migration | |
| CN108921814A (en) | A kind of Citrus Huanglongbing pathogen on-line quick detection system and method based on deep learning | |
| CN111814563B (en) | Method and device for classifying planting structures | |
| CN109871901A (en) | A kind of unbalanced data classification method based on mixing sampling and machine learning | |
| CN103593674A (en) | Cervical lymph node ultrasonoscopy feature selection method | |
| CN109558904A (en) | Image local feature classification method and device and storage medium |
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: 20181106 |