CN106096654A - A kind of cell atypia automatic grading method tactful based on degree of depth study and combination - Google Patents
A kind of cell atypia automatic grading method tactful based on degree of depth study and combination Download PDFInfo
- Publication number
- CN106096654A CN106096654A CN201610414194.9A CN201610414194A CN106096654A CN 106096654 A CN106096654 A CN 106096654A CN 201610414194 A CN201610414194 A CN 201610414194A CN 106096654 A CN106096654 A CN 106096654A
- Authority
- CN
- China
- Prior art keywords
- image
- resolution
- degree
- grade
- under
- 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
- G06F18/217—Validation; Performance evaluation; Active pattern learning techniques
- G06F18/2193—Validation; Performance evaluation; Active pattern learning techniques based on specific statistical tests
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/285—Selection of pattern recognition techniques, e.g. of classifiers in a multi-classifier system
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/03—Recognition of patterns in medical or anatomical images
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Computational Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Probability & Statistics with Applications (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of cell atypia automatic grading method tactful based on degree of depth study and combination, first degree of depth learning method is used to identify the grade of pathological tissue image block under different resolution, then use the depth model trained to combine sliding window method under each resolution and process the significantly image under current resolution, the absolute majority ballot method re-using one of combination strategy determines the grade of significantly image under current resolution, this can be obtained by the grade of significantly image under each resolution, relative majority ballot method decision-making from the grade of multiple resolution is finally used to publish picture the final grade of picture.The present invention, with significantly slice map as object of study, uses degree of depth study to add the method for sliding window and combine the mode of decision-making, can the cell atypia grade of evaluation image exactly.The method of the cell atypia automatic classification that the present invention proposes can assist doctor to pathological tissue image cancer ranking, carries out clinical diagnosis quickly and accurately.
Description
Technical field
The present invention relates to technical field of image information processing, a kind of cell tactful based on degree of depth study and combination
Atypia automatic grading method.
Background technology
Along with significantly the generation of sectioning image digital scan technology and the efficiency of scanning improve, the number of tissue pathological slice
Word shows and storage becomes practicable.Utilize digitizing technique pathological image can be carried out higher-quality analysis.Cause
For the feature of various cancerous tissue almost can be found out from tissue slice pathology image, it is possible to be used for assisting doctor to examine
Disconnected, but the existing technical research for Medical Image Processing is still little, so studying a set of dividing for pathological image
Analysis instrument is particularly significant.
The diagnosis and treatment guide issued according to WHO, the diagnosis of the grade malignancy of breast carcinoma is most widely used is promise
Fourth Chinese hierarchy system (NGS).This cover system is mainly according to three indexs of mammary gland tissue pathology image: 1) breast duct (gland
Pipe) formation degree;2) mitosis number of times;3) nucleus is heterogeneous.The formation degree of glandular tube is that assessment tumor forms glandular tube
The percentage ratio of structure, proportion is the biggest, and breast carcinoma rank is the lowest.Mitotic phase numeration assessment is in microscope 400 times amplification
Karyokinesis picture counting under multiple, quantity is the most, and breast carcinoma rank is the highest.Nucleus atypia assessment tumor cell and normal breast
The diversity of epithelial cell, difference is the biggest, and breast carcinoma rank is the highest.Each index grade malignancy score range from low to high
Being 0~3 point, the score range that three index comprehensives of the most each patient get up is 0~9.The scoring score value of patient is closer to 0 table
The grade malignancy of bright cancer is the lowest, and therefore the treatment of patient and the effect of prognosis are the best, and contrary grade malignancy is the highest, and treatment is with pre-
After effect the poorest.Therefore, the score value of three indexs being accurately determined NGS system is most important in clinic.
Owing to manual analysis method has the strongest subjectivity, different pathologist under identical objective condition people
In work scoring, there is bigger discordance.Manual analysis in addition to easily being affected by subjective and environmental factors, this mistake
Journey is also quite time-consuming laborious, and manpower cost is the highest.Computer-aided diagnosis technology can make up the defect of manual analysis.Closely
Nian Lai, the fast development of digital scan technology and computer vision technique makes computer-aided diagnosis technology be possibly realized.Meter
The appearance of calculation machine aided diagnosis technique can not only provide the most objectively analysis result for doctor, and can also reduce doctor's
Workload thus be greatly enhanced the work efficiency of doctor.The target of research computer aided system (CAD) is not configured to completely
Replace doctor, but improve the work efficiency of doctor to provide physicians with the most objective suggestion, obtain more
Diagnostic result accurately.
Summary of the invention
The technical problem to be solved is to overcome the deficiencies in the prior art, and provide a kind of based on degree of depth study and
In conjunction with the cell atypia automatic grading method of strategy, convolutional neural networks and sliding window technique is used to process significantly pathology figure
Picture;Utilizing absolute majority ballot method to determine the grade of the single resolution of image, relative majority ballot method decision diagram is as multiple resolutions
The grade of rate, obtains the level results that image is final.
The present invention solves above-mentioned technical problem by the following technical solutions:
A kind of cell atypia automatic grading method tactful based on degree of depth study and combination proposed according to the present invention, bag
Include following steps:
Step 1, the image under the different resolution of each case is marked respectively, is i.e. according to image inner cell core
Morphological characteristic and matter density to score value;According to labelling, in the pathological image of each different resolution, choose different points respectively
The image block of value grade is as training sample set corresponding under each resolution;
The convolutional neural networks that under the different resolution that step 2, employing step 1 obtain, training sample set training is different, its
In, the quantity of the resolution that the quantity of the convolutional neural networks model of training is had with image is identical;
Step 3, using the convolutional neural networks that trains as grader, and utilize sliding window technique, successively to window
Interior image block is marked;
The image block utilizing sliding window technique to generate is voted by step 4, use absolute majority ballot method, is worked as
The classification results of front resolution artwork;
Step 5, the detection image under each resolution is carried out the operation of step 3 and step 4, thus obtain every width
Classification results under image different resolution;
Step 6, each image difference is differentiated lower classification results carry out relative majority and vote method, obtain this width image
Whole classification results.
One is entered as a kind of cell atypia automatic grading method based on degree of depth study and combination strategy of the present invention
Step prioritization scheme, described score value is 1 or 2 or 3 point.
One is entered as a kind of cell atypia automatic grading method based on degree of depth study and combination strategy of the present invention
Step prioritization scheme, each case comprises the image under three kinds of resolution.
One is entered as a kind of cell atypia automatic grading method based on degree of depth study and combination strategy of the present invention
Step prioritization scheme, sliding window technique refers to utilize selected slider bar, from the beginning of the upper left side of image, from left to right, from upper
Sliding successively under to, the image block in window is all judged by the convolutional network model that often slides, it is judged that each window
Affiliated grade.
One is entered as a kind of cell atypia automatic grading method based on degree of depth study and combination strategy of the present invention
Step prioritization scheme, the image block in described step 1 is square image blocks.
The present invention uses above technical scheme compared with prior art, has following technical effect that
(1) under identical condition, the inventive method is than more standard based on manual analysis method, and elapsed time is few;
(2) the inventive method each region under resolution each to image in cell atypia classification is swept
Retouching, classification results is more comprehensive;
(3) the inventive method is for the problem of artificial single cell analysis, takes the analysis that big region and zonule combine
Method analyzes the atypia of cell.
Accompanying drawing explanation
Fig. 1 is cell atypia degree schematic diagram;Wherein (a), (b) and (c) corresponding nucleus atypia grade respectively is
The histopathology image block of 1,2,3.
Fig. 2 is image multiresolution schematic diagram;Wherein, (a) is to amplify the image under 100 resolutions, and (b) is to amplify
Image under 200 resolutions, (c) is to amplify the image under 400 resolutions.
Fig. 3 is the training structure schematic diagram of convolutional neural networks.
Detailed description of the invention
In order to make the object, technical solutions and advantages of the present invention clearer, below in conjunction with the accompanying drawings and the specific embodiments
Describe the present invention.
A kind of cell atypia automatic grading method tactful based on degree of depth study and combination of the present invention, including following step
Rapid:
Step 1, the choosing of training sample:
Training sample is to build from original data, and original data are all by the clinic of specialty pathology knowledge
Doctor was marked, and program can be marked in pathological image according to these experts and randomly select foursquare image fritter,
Wherein the length of side of these fritters is 256 pixels.Program can build corresponding to each point respectively according to the quantity of image resolution ratio
The training sample set of resolution.
Fig. 1 is cell atypia degree schematic diagram;Wherein (a) in Fig. 1 be corresponding nucleus atypia grade be the group of 1
Knit pathological image block, (b) in Fig. 1 be corresponding nucleus atypia grade be the histopathology image block of 2, (c) be corresponding carefully
Karyon atypia grade is the histopathology image block of 3.
Step 2, the training of degree of depth convolutional neural networks:
Use the sample training convolutional neural networks that step 1 is chosen as shown in Figure 3.Degree of depth convolutional neural networks is as initially
The one of artificial nerve network model feedforward neural network, obtained significant progress in recent years, particularly at image
Reason and speech recognition aspect.As trainable multitiered network structure, the convolution stage includes convolutional layer, nonlinear transformation layer, under
Sample level i.e. pond layer.
Convolutional layer extracts the feature of image by convolution kernel, and this is concept based on local receptor field.Each convolution kernel
Extract the special characteristic of all positions on input feature vector figure, it is achieved the weights on same input feature vector figure are shared.In order to extract
Features different on input feature vector figure just uses different convolution kernels to carry out convolution operation.Convolutional layer is extracted by nonlinear transformation layer
Feature carry out nonlinear mapping.Traditional convolutional neural networks uses saturation nonlinearity function to carry out nonlinear mapping
Sigmoid, tanh or softsign.Commonly used unsaturation nonlinear function in the convolutional neural networks being recently proposed
ReLU.Model carries out back-propagation gradient decline when, ReLU is than traditional saturation nonlinearity function convergence speed more
Hurry up, during such training network, iteration time can go up many soon.Pond layer is that each characteristic pattern is carried out independent operation, makes at present
With more be to use average pond or maximum two kinds of pondization operation.Average pond is calculating pixel in the window of selected field
Average, and maximum pond is the maximum calculating selected field window, and the step-length that general window translates is 1.Characteristic pattern leads to
After crossing pondization operation, the resolution reduction of characteristic pattern but remains effective feature.Meanwhile, pondization has also reached a dimensionality reduction
Purpose.After feature extraction completes, connect full articulamentum and grader by after the output characteristic figure of last layer.
Convolutional neural networks use original image pixel directly as the input of network, with traditional image recognition algorithm
Different, it is to avoid the extraction of complicated manual features;Weights share the quantity that can reduce weights, thus reduce the calculating of network
Complexity;The feature that the use of local receptor field makes whole network be observed has translation and rotational invariance.Network
Output corresponds directly to classification, and such point-to-point, end-to-end network model is greatly enhanced the precision of identification.
Step 3, use the convolutional neural networks that trained to combine sliding window to process image:
Convolutional neural networks is applied in the middle of cell atypia classification, it is contemplated that the complexity of organizational structure in tissue slice
Property, multiformity, spreading all over property and irregularities, use the method for sliding window to travel through whole image.Slider bar selected by utilization, from
The upper left side of sectioning image starts, and from left to right, slides the most successively, and the convolutional network model that often slides is all to window
Interior image block judges, it is judged that the grade belonging to each window.
Step 4, absolute majority ballot method judges the grade of single resolution hypograph:
Using sliding window technique to process significantly image in step 3, every image can produce a lot of subset, wherein subset
Quantity relevant with the step-length of the size of institute altimetric image and slip, volume machine model all can carry out classification to these subsets so that
One big width image produces a lot of subset classification results.Absolute majority ballot method is used to determine from the classification results of these subsets
Plan goes out the classification results of significantly image.Being mathematically represented as of absolute majority ballot method:
As shown in formula (1),For representing from class mark set C1,C2,C3Prediction in (corresponding respectively to scoring correspondence)
The class mark C gone outj,Represent in input picture x, be predicted to be class mark CjThe quantity of window;N is class mark set
Middle removing class mark CjOther class target quantity;If class mark CjNumber of votes obtained more than half, then be predicted as such mark, H (x) for input figure
The final result drawn by absolute majority ballot method as x.
Step 5, under each resolution detection image carry out above operation, obtain each image difference differentiate
Classification results under rate.The image of each case in general data comprises the image under three kinds of resolution, and these three is differentiated
Rate is 100 times respectively, 200 times and 400 times.So through aforesaid operations, the image of each case can obtain three cell abnormal shapes
Property classification results.
The final result of step 6, relative majority ballot method joint decision multiresolution:
The result of multiple resolution of the image of each case, uses relative majority ballot method to obtain the final of each case
Result.Relative majority ballot method be prediction who gets the most votes's labelling, if having simultaneously multiple labelling obtain the highest ticket, the most therefrom with
Machine chooses one.The voting strategy of relative majority ballot method is:
As shown in formula (2),InFor the result of above-mentioned absolute majority ballot method, F (x) is three
The most final grade of classification that under yardstick, occurrence number is most.
For the ease of public understanding technical solution of the present invention, a specific embodiment is given below.
Technical scheme provided by the present invention is applied the breast cancer tissue's figure in h and E dyeing by the present embodiment
On image set.The inventive method is tested in data base.Data are the Pathology Deparments of saab spy philanthropic hospital from Paris, FRA
The mammary gland pathological organization charts picture picked out, and these data are all jointly to be marked by two to three veteran pathologist
Note.The image of each case in data comprises the image under three kinds of resolution as in figure 2 it is shown, (a) in Fig. 2 is to amplify
Image under 100 resolutions, (b) in Fig. 2 is to amplify the image under 200 resolutions, and (c) in Fig. 2 is to amplify 400 times
Image under resolution.Image sample under three kinds of resolution of same visual angle undertissue pathological image, (a) in Fig. 2 is
Amplify the image under 100 resolutions, (a) in Fig. 2 is divided into 16 image blocks obtain (b) in Fig. 2, (b) in Fig. 2
In the size of each fritter be the same with the size of (a) in Fig. 2.Equally, every piece of image block in (c) in Fig. 2 is
By 1/4th of image block each in (b) in Fig. 2, size is the same with the size of (a) in Fig. 2.Each case
The size of the dimension of picture of three resolution is 769 × 688,1539 × 1376 and 3078 × 2752.In addition, training god
Through network when, available data is carried out the operation of data dilatation, mainly included the rotation to every image and mirror image behaviour
Make.Except training picture, also having the test picture of 124 cases in data, each case includes the breast of three different resolutions
Adenopathy example organization charts picture, is used for checking the generalization ability of automatic scoring model.
In the present embodiment, the structure of convolutional neural networks part is as shown in table 1.
Table 1
The number of plies | Operation | Port number | Size | Step-length | Activation primitive |
1 | Input | 3 | - | - | - |
2 | Convolution | 96 | 11 | 4 | ReLU |
3 | Chi Hua | 96 | 3 | 2 | - |
4 | Convolution | 256 | 5 | 1 | ReLU |
5 | Chi Hua | 256 | 3 | 2 | - |
6 | Convolution | 384 | 3 | 1 | ReLU |
7 | Convolution | 384 | 3 | 1 | ReLU |
8 | Convolution | 256 | 3 | 2 | ReLU |
9 | Chi Hua | 256 | 3 | 2 | - |
10 | Full connection | 256 | - | - | ReLU |
11 | Full connection | 128 | - | - | ReLU |
12 | Output | 3 | - | - | - |
The detection process of the present embodiment is specific as follows:
1, the generation of cell training set:
Training set intercepts in experimentation from three kinds of different resolutions the image block of 256 × 256, structure the most respectively
Build the data set comprising three kinds of different resolutions used by experiment.Training set intermediate-resolution is 100 times, 200 times, the number of 400 times
There are about 20,000,40,000 and 80,000 respectively according to collection.
2, training convolutional neural networks:
Off-the-shelf three data sets are sent into three convolutional neural networks training, the wherein structure of three convolution model
Being the same, structure is as shown in table 1.Convolutional neural networks use original image pixel directly as the input of network, with biography
The image recognition algorithm of system is different, it is to avoid the extraction of complicated manual features;Weights share the quantity that can reduce weights, from
And reduce the computation complexity of network;The feature that the use of local receptor field makes whole network be observed has translation and rotation
Turn invariance.The output of network corresponds directly to classification, such point-to-point, and end-to-end network model is greatly enhanced identification
Precision.
3, slip scan:
Using the convolutional neural networks that trains as grader, utilize sliding window technique to section from top to bottom, from a left side
To the right side, successively the little image in window is carried out classification with a pixel for step-length.
4, absolute majority ballot method:
The result being combined image block by ballot method predicts the result of whole image under this resolution.On these rank
Section, selects task of using absolute majority ballot method to complete the whole pictures classification under single resolution.
5, relative majority ballot method:
Three of three different resolution hypographs for each case predict the outcome, and the present invention uses relative majority to throw
Ticket method carrys out the result of the final prediction of this case of decision-making.
6, model evaluation:
Make in aforementioned manners test picture to be carried out cell atypia scoring, obtain the scoring of every test picture.Assessment
Standard take grading scheme, note different model prediction input histopathology image must be divided into p, and pathologist gives this pictures
Give the assessment score being divided into g, S to be this pictures, shown in evaluation criteria such as formula (3):
Finally, calculate the summation of the mark of 124 assessments, obtain the total score of the prediction test data of this model.
7, test result:
The scoring accuracy of the method that the present invention uses must be divided into 67 points, and computational efficiency is the highest, averagely at every 100
Times, 200 times, the calculating time of 400 resolution hypographs respectively may be about 1.2 seconds, 5.5 seconds and 30 seconds.
In summary, the cell atypia automatic grading method that the present invention proposes has possessed the ability of practical clinical.
Above in conjunction with accompanying drawing, embodiments of the present invention are explained in detail, but the present invention is not limited to above-mentioned enforcement
Mode, in the ken that those of ordinary skill in the art are possessed, it is also possible on the premise of without departing from present inventive concept
Make a variety of changes.
Claims (5)
1. one kind based on degree of depth study and combines tactful cell atypia automatic grading method, it is characterised in that include following
Step:
Step 1, the image under the different resolution of each case is marked respectively, is i.e. the shape according to image inner cell core
State feature and matter density are to score value;According to labelling, in the pathological image of each different resolution, choose different score values etc. respectively
The image block of level is as training sample set corresponding under each resolution;
The convolutional neural networks that under the different resolution that step 2, employing step 1 obtain, training sample set training is different, wherein, instruction
The quantity of the resolution that the quantity of the convolutional neural networks model practiced is had with image is identical;
Step 3, using the convolutional neural networks that trains as grader, and utilize sliding window technique, successively in window
Image block is marked;
The image block utilizing sliding window technique to generate is voted by step 4, use absolute majority ballot method, is currently divided
The classification results of resolution artwork;
Step 5, the detection image under each resolution is carried out the operation of step 3 and step 4, thus obtain each image
Classification results under different resolution;
Step 6, each image difference is differentiated lower classification results carry out relative majority and vote method, obtain this width image final
Classification results.
A kind of cell atypia automatic grading method tactful based on degree of depth study and combination, it is special
Levying and be, described score value is 1 or 2 or 3 point.
A kind of cell atypia automatic grading method tactful based on degree of depth study and combination, it is special
Levying and be, each case comprises the image under three kinds of resolution.
A kind of cell atypia automatic grading method tactful based on degree of depth study and combination, it is special
Levying and be, sliding window technique refers to utilize selected slider bar, from the beginning of the upper left side of image, from left to right, from top to bottom
Sliding successively, the image block in window is all judged by the convolutional network model that often slides, it is judged that belonging to each window
Grade.
A kind of cell atypia automatic grading method tactful based on degree of depth study and combination, it is special
Levying and be, the image block in described step 1 is square image blocks.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610414194.9A CN106096654A (en) | 2016-06-13 | 2016-06-13 | A kind of cell atypia automatic grading method tactful based on degree of depth study and combination |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610414194.9A CN106096654A (en) | 2016-06-13 | 2016-06-13 | A kind of cell atypia automatic grading method tactful based on degree of depth study and combination |
Publications (1)
Publication Number | Publication Date |
---|---|
CN106096654A true CN106096654A (en) | 2016-11-09 |
Family
ID=57845460
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610414194.9A Pending CN106096654A (en) | 2016-06-13 | 2016-06-13 | A kind of cell atypia automatic grading method tactful based on degree of depth study and combination |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106096654A (en) |
Cited By (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106920227A (en) * | 2016-12-27 | 2017-07-04 | 北京工业大学 | Based on the Segmentation Method of Retinal Blood Vessels that deep learning is combined with conventional method |
CN106991673A (en) * | 2017-05-18 | 2017-07-28 | 深思考人工智能机器人科技(北京)有限公司 | A kind of cervical cell image rapid classification recognition methods of interpretation and system |
CN107103601A (en) * | 2017-04-14 | 2017-08-29 | 成都知识视觉科技有限公司 | A kind of cell mitogen detection method in breast cancer points-scoring system |
CN107526946A (en) * | 2016-12-23 | 2017-12-29 | 南京理工大学 | Merge the gene expression data cancer classification method of self study and low-rank representation |
CN107886049A (en) * | 2017-10-16 | 2018-04-06 | 江苏省气象服务中心 | A kind of visibility identification method for early warning based on camera probe |
CN108288506A (en) * | 2018-01-23 | 2018-07-17 | 雨声智能科技(上海)有限公司 | A kind of cancer pathology aided diagnosis method based on artificial intelligence technology |
CN108305249A (en) * | 2018-01-24 | 2018-07-20 | 福建师范大学 | The quick diagnosis and methods of marking of full size pathological section based on deep learning |
CN108898160A (en) * | 2018-06-01 | 2018-11-27 | 中国人民解放军战略支援部队信息工程大学 | Breast cancer tissue's Pathologic Grading method based on CNN and image group Fusion Features |
CN109002886A (en) * | 2018-09-21 | 2018-12-14 | 杭州谛瞳科技有限公司 | Image processing method and electronic equipment |
CN109034208A (en) * | 2018-07-03 | 2018-12-18 | 怀光智能科技(武汉)有限公司 | A kind of cervical cell pathological section classification method of high-low resolution combination |
CN109034221A (en) * | 2018-07-13 | 2018-12-18 | 马丁 | A kind of processing method and its device of cervical cytology characteristics of image |
CN109145941A (en) * | 2018-07-03 | 2019-01-04 | 怀光智能科技(武汉)有限公司 | A kind of irregular cervical cell group's image classification method and system |
CN109635141A (en) * | 2019-01-29 | 2019-04-16 | 京东方科技集团股份有限公司 | For retrieving method, electronic equipment and the computer readable storage medium of image |
CN110188820A (en) * | 2019-05-30 | 2019-08-30 | 中山大学 | The retina OCT image classification method extracted based on deep learning sub-network characteristics |
CN110189293A (en) * | 2019-04-15 | 2019-08-30 | 广州锟元方青医疗科技有限公司 | Cell image processing method, device, storage medium and computer equipment |
CN111656393A (en) * | 2017-11-16 | 2020-09-11 | 奥斯陆大学医院 | Histological image analysis |
CN112561852A (en) * | 2019-09-26 | 2021-03-26 | 佳能株式会社 | Image determination device and image determination method |
-
2016
- 2016-06-13 CN CN201610414194.9A patent/CN106096654A/en active Pending
Cited By (28)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107526946A (en) * | 2016-12-23 | 2017-12-29 | 南京理工大学 | Merge the gene expression data cancer classification method of self study and low-rank representation |
CN106920227B (en) * | 2016-12-27 | 2019-06-07 | 北京工业大学 | The Segmentation Method of Retinal Blood Vessels combined based on deep learning with conventional method |
CN106920227A (en) * | 2016-12-27 | 2017-07-04 | 北京工业大学 | Based on the Segmentation Method of Retinal Blood Vessels that deep learning is combined with conventional method |
CN107103601A (en) * | 2017-04-14 | 2017-08-29 | 成都知识视觉科技有限公司 | A kind of cell mitogen detection method in breast cancer points-scoring system |
CN107103601B (en) * | 2017-04-14 | 2020-04-24 | 成都知识视觉科技有限公司 | Cell mitosis detection method in breast cancer scoring system |
CN106991673A (en) * | 2017-05-18 | 2017-07-28 | 深思考人工智能机器人科技(北京)有限公司 | A kind of cervical cell image rapid classification recognition methods of interpretation and system |
CN106991673B (en) * | 2017-05-18 | 2019-10-22 | 深思考人工智能机器人科技(北京)有限公司 | A kind of the cervical cell image rapid classification recognition methods and system of interpretation |
CN107886049A (en) * | 2017-10-16 | 2018-04-06 | 江苏省气象服务中心 | A kind of visibility identification method for early warning based on camera probe |
CN107886049B (en) * | 2017-10-16 | 2022-08-26 | 江苏省气象服务中心 | Visibility recognition early warning method based on camera probe |
CN111656393A (en) * | 2017-11-16 | 2020-09-11 | 奥斯陆大学医院 | Histological image analysis |
CN108288506A (en) * | 2018-01-23 | 2018-07-17 | 雨声智能科技(上海)有限公司 | A kind of cancer pathology aided diagnosis method based on artificial intelligence technology |
CN108305249A (en) * | 2018-01-24 | 2018-07-20 | 福建师范大学 | The quick diagnosis and methods of marking of full size pathological section based on deep learning |
CN108305249B (en) * | 2018-01-24 | 2022-05-24 | 福建师范大学 | Rapid diagnosis and scoring method of full-scale pathological section based on deep learning |
CN108898160A (en) * | 2018-06-01 | 2018-11-27 | 中国人民解放军战略支援部队信息工程大学 | Breast cancer tissue's Pathologic Grading method based on CNN and image group Fusion Features |
CN108898160B (en) * | 2018-06-01 | 2022-04-08 | 中国人民解放军战略支援部队信息工程大学 | Breast cancer histopathology grading method based on CNN and imaging omics feature fusion |
CN109034208A (en) * | 2018-07-03 | 2018-12-18 | 怀光智能科技(武汉)有限公司 | A kind of cervical cell pathological section classification method of high-low resolution combination |
CN109034208B (en) * | 2018-07-03 | 2020-10-23 | 怀光智能科技(武汉)有限公司 | High-low resolution combined cervical cell slice image classification system |
CN109145941B (en) * | 2018-07-03 | 2021-03-09 | 怀光智能科技(武汉)有限公司 | Irregular cervical cell mass image classification method and system |
CN109145941A (en) * | 2018-07-03 | 2019-01-04 | 怀光智能科技(武汉)有限公司 | A kind of irregular cervical cell group's image classification method and system |
CN109034221A (en) * | 2018-07-13 | 2018-12-18 | 马丁 | A kind of processing method and its device of cervical cytology characteristics of image |
CN109002886A (en) * | 2018-09-21 | 2018-12-14 | 杭州谛瞳科技有限公司 | Image processing method and electronic equipment |
CN109635141A (en) * | 2019-01-29 | 2019-04-16 | 京东方科技集团股份有限公司 | For retrieving method, electronic equipment and the computer readable storage medium of image |
CN109635141B (en) * | 2019-01-29 | 2021-04-27 | 京东方科技集团股份有限公司 | Method, electronic device, and computer-readable storage medium for retrieving an image |
US11113586B2 (en) | 2019-01-29 | 2021-09-07 | Boe Technology Group Co., Ltd. | Method and electronic device for retrieving an image and computer readable storage medium |
CN110189293A (en) * | 2019-04-15 | 2019-08-30 | 广州锟元方青医疗科技有限公司 | Cell image processing method, device, storage medium and computer equipment |
CN110188820A (en) * | 2019-05-30 | 2019-08-30 | 中山大学 | The retina OCT image classification method extracted based on deep learning sub-network characteristics |
CN110188820B (en) * | 2019-05-30 | 2023-04-18 | 中山大学 | Retina OCT image classification method based on deep learning subnetwork feature extraction |
CN112561852A (en) * | 2019-09-26 | 2021-03-26 | 佳能株式会社 | Image determination device and image determination method |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106096654A (en) | A kind of cell atypia automatic grading method tactful based on degree of depth study and combination | |
Xue et al. | An application of transfer learning and ensemble learning techniques for cervical histopathology image classification | |
CN108257135A (en) | The assistant diagnosis system of medical image features is understood based on deep learning method | |
CN107977671A (en) | A kind of tongue picture sorting technique based on multitask convolutional neural networks | |
Deng et al. | Classification of breast density categories based on SE-Attention neural networks | |
CN105760874B (en) | CT image processing system and its CT image processing method towards pneumoconiosis | |
CN107680678A (en) | Based on multiple dimensioned convolutional neural networks Thyroid ultrasound image tubercle auto-check system | |
CN107909566A (en) | A kind of image-recognizing method of the cutaneum carcinoma melanoma based on deep learning | |
Alzubaidi et al. | Robust application of new deep learning tools: an experimental study in medical imaging | |
CN109670510A (en) | A kind of gastroscopic biopsy pathological data screening system and method based on deep learning | |
CN107657602A (en) | Based on the breast structure disorder recognition methods for migrating convolutional neural networks twice | |
CN110097974A (en) | A kind of nasopharyngeal carcinoma far-end transfer forecasting system based on deep learning algorithm | |
CN106780498A (en) | Based on point depth convolutional network epithelium and matrix organization's automatic division method pixel-by-pixel | |
CN110111895A (en) | A kind of method for building up of nasopharyngeal carcinoma far-end transfer prediction model | |
CN111767952B (en) | Interpretable lung nodule benign and malignant classification method | |
CN109671053A (en) | A kind of gastric cancer image identification system, device and its application | |
CN108053398A (en) | A kind of melanoma automatic testing method of semi-supervised feature learning | |
CN106340000A (en) | Bone age assessment method | |
CN101714153A (en) | Visual perception based interactive mammography image searth method | |
CN112263217B (en) | Improved convolutional neural network-based non-melanoma skin cancer pathological image lesion area detection method | |
Yonekura et al. | Improving the generalization of disease stage classification with deep CNN for glioma histopathological images | |
CN108062749A (en) | Recognition methods, device and the electronic equipment of musculus levator ani ceasma | |
Cao et al. | An automatic breast cancer grading method in histopathological images based on pixel-, object-, and semantic-level features | |
CN114332572B (en) | Method for extracting breast lesion ultrasonic image multi-scale fusion characteristic parameters based on saliency map-guided hierarchical dense characteristic fusion network | |
CN106709421A (en) | Cell image recognition and classification method based on transform domain characteristics and CNN (Convolutional Neural Network) |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into 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: 20161109 |