CN109671072A - Cervical cancer tissues pathological image diagnostic method based on spotted arrays condition random field - Google Patents
Cervical cancer tissues pathological image diagnostic method based on spotted arrays condition random field Download PDFInfo
- Publication number
- CN109671072A CN109671072A CN201811552817.4A CN201811552817A CN109671072A CN 109671072 A CN109671072 A CN 109671072A CN 201811552817 A CN201811552817 A CN 201811552817A CN 109671072 A CN109671072 A CN 109671072A
- Authority
- CN
- China
- Prior art keywords
- image
- random field
- nucleus
- full
- condition random
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
- 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30096—Tumor; Lesion
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Biology (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
Abstract
The present invention provides the cervical cancer tissues pathological image diagnostic methods based on spotted arrays condition random field, comprising: obtains histopathology image to be processed and is pre-processed；Image segmentation is carried out to pretreated image, obtains the nucleus binary map of image and the nucleus coordinate of image, and nucleus binary map is divided into the image block of fixed size；Feature extraction is carried out to nucleus binary map, obtains the global characteristics of image, feature extraction is carried out to image block, obtains the local feature of image；The global characteristics of image and local feature are input in the conditional random field models for the multiple dimensioned array layout of rectangular area point type being pre-designed, it include nuclei picture block according to nucleus coordinate setting, obtain full figure nuclear characteristics, and it is handled using condition random field classifier trained in advance, output category result.The pathological state of patient more can be accurately provided, accuracy rate of diagnosis is improved.
Description
Technical field
The present invention relates to medical microscopic images processing technology fields, more particularly to one kind to be based on spotted arrays condition random field
Cervical cancer tissues pathological image diagnostic method.
Background technique
Cervical carcinoma is the most common gynecologic malignant tumor.Although gradually being paid attention to for the generaI investigation of this disease in recent years,
It is that but few people periodically go to hospital to check in vast rural area, meanwhile, HPV infection situation is increasingly heavier in recent years, so palace
The morbidity of neck cancer has the tendency that gradually increasing, and also has the tendency that increasingly rejuvenation.As a result, about the tissue of cervical carcinoma
Pathological research is imperative, cannot delay.
Cervical cancer tissues pathological image is intended only as basic fact, needs veteran virologist to pathological image
It judges.However, different pathological scholar or the same virologist of different time have the judgement of same pathological image
Difference, this may generate biggish error, and let alone, in the low developed area that medical resource lacks, experience is not abundant enough
Medico and virologist reliable judgement can not be made to histopathology image.
In order to improve the accuracy rate of doctor, some guide directions of doctor for lacking experience are given, increase substantially work effect
Rate simultaneously promotes integral level, it is existing in mostly using the classification methods pair such as decision tree, support vector machines and artificial neural network
Cervical cancer tissues pathological image is classified, but above-mentioned classification method does not take into account space structure relationship.Therefore,
Need a kind of cervical cancer tissues pathological image diagnostic method based on spotted arrays condition random field.
Summary of the invention
(1) technical problems to be solved
In order to solve the above problem of the prior art, the present invention provides a kind of uterine neck based on spotted arrays condition random field
Cancerous tissue pathological image diagnostic method more can accurately provide the pathological state of patient, improve accuracy rate of diagnosis.
(2) technical solution
In order to achieve the above object, the main technical schemes that the present invention uses include:
A kind of cervical cancer tissues pathological image diagnostic method based on spotted arrays condition random field, comprising the following steps:
Step S1, histopathology image to be processed is obtained；
Step S2, each image to be processed is pre-processed；
Step S3, image segmentation is carried out to described each pretreated image, obtains the nucleus of each image
Binary map positions nucleus according to image segmentation result, obtains the nucleus coordinate of each image, and by each nucleus
Binary map is divided into the image block of fixed size；
Step S4, feature extraction is carried out to each nucleus binary map, obtains the global characteristics of each image,
Feature extraction is carried out to each image block of each nucleus binary map, obtains the local feature of each image；
Step S5, the local feature of the nucleus coordinate of each image and each image is input to pre-
In the conditional random field models first designed, obtain the full figure nuclear characteristics of each image, using condition trained in advance with
Airport classifier handles the global characteristics of each image and the full figure nuclear characteristics of each image,
Output category result；
The conditional random field models being pre-designed are the multiple dimensioned array layout of rectangular area point type as shared by rectangle
The conditional probability model of the calculating rectangular centre pixel of pixel parameter shared by pixel parameter and spotted arrays.
One kind as the cervical cancer tissues pathological image diagnostic method the present invention is based on spotted arrays condition random field changes
Into before the step S1 further include:
S01, select high, medium and low differentiated tissue's pathological image subset of identical quantity as training set；
S02, step S1 to the S4 processing is carried out to each image in the training set, obtained each in training set
The global characteristics and local feature of the nucleus coordinate of image, each image, and,
The local feature of the nucleus coordinate of each image in the training set and each image is input in advance
In the conditional random field models of design, the full figure nuclear characteristics of each image in training set are obtained；
S03, the complete of each image in differentiated histopathology image subset is inputted into condition random field classifier
The full figure nuclear characteristics of office feature and each image, obtain differentiated sorter model；
Into condition random field classifier, the overall situation of each image is special in differentiated tissue's pathological image subset in input
It seeks peace the full figure nuclear characteristics of each image, obtains middle differentiation sorter model；
The overall situation that each image in low differentiated tissue's pathological image subset is inputted into condition random field classifier is special
It seeks peace the full figure nuclear characteristics of each image, obtains low differentiation sorter model.
One kind as the cervical cancer tissues pathological image diagnostic method the present invention is based on spotted arrays condition random field changes
Into, in step S2, each image to be processed is pre-processed, specifically: for each image to be processed,
First image is denoised using median filter, reuses histogram equalization to enhance picture contrast.
One kind as the cervical cancer tissues pathological image diagnostic method the present invention is based on spotted arrays condition random field changes
Into in step S3, to described each pretreated image progress image segmentation, the nucleus two-value of each image of acquisition
Figure, specifically:
It is pre- by described each according to the image partition method and preset cluster numbers K value clustered based on K-means
Treated, and image gathers for nucleus, cytoplasm, cytoplasm, four class of image labeling, grasps further according to preset morphology
Make, obtains the nucleus binary map of each image.
One kind as the cervical cancer tissues pathological image diagnostic method the present invention is based on spotted arrays condition random field changes
Into after feature extraction, using Principal Component Analysis to each image progress dimensionality reduction；Correspondingly, the step S5 includes: by institute
The global characteristics and the local feature after dimensionality reduction for stating each image are input in the conditional random field models being pre-designed.
One kind as the cervical cancer tissues pathological image diagnostic method the present invention is based on spotted arrays condition random field changes
Into, in step S4, DAISY feature that the global characteristics of each image include full figure textural characteristics and full figure is averaged；The full figure
Textural characteristics are full figure grey level histogram feature.
One kind as the cervical cancer tissues pathological image diagnostic method the present invention is based on spotted arrays condition random field changes
Into in step S4, obtaining the local feature of each image, specially extract each image block and be averaged DAISY feature descriptor.
One kind as the cervical cancer tissues pathological image diagnostic method the present invention is based on spotted arrays condition random field changes
Into using condition random field classifier trained in advance to the global characteristics of each image and each image
Full figure nuclear characteristics are handled, specifically:
It, will be described every according to the differentiated sorter model, middle differentiation sorter model and low differentiation sorter model
The global characteristics of one image, each image full figure nuclear characteristics as input variable, obtain each image
With the likelihood of each sorter model, select the classification of the highest sorter model of likelihood as final classification and diagnosis
As a result；
The calculation formula of likelihood is as follows: likelihood=full figure nucleus DAISY characteristic similarity * full figure gray scale that is averaged is straight
Square figure similarity * full figure is averaged DAISY characteristic similarity, wherein the calculating of similarity uses corr2 function.
(3) beneficial effect
The beneficial effects of the present invention are:
By the way that conditional random field models are designed as rectangular area spotted arrays symmetric configuration, which can there are many size
It is selective, various sizes of layout can be applied according to the actual size of image, be applicable not only to get by different approaches
Histopathology image, apply also for other medicine or non-medical images；The layout can selectively extract target area
Information data within the scope of dot matrix, the image data including nucleus, cytoplasm and cytoplasm, can comprehensively describe
Spatial information, the diagnostic result of system are highly efficient reliable；The layout can pass through distribution according to specific image and actual demand
To the weight of pixel shared by rectangle and pixel difference numerical value shared by spotted arrays, the calculation method of design condition probability.It should
Layout increases new dimension for observation, helps to estimate more parameters and improves estimation performance.
By means of condition random field classifier is applied to automatically analyzing for histopathology image, a set of tissue disease is formd
The auto-check system of image of science, can suitably slow down that virologist resource allocation is uneven, the problem of being not enough, can also be with
Medico is instructed rationally to be diagnosed with the insufficient doctor of experience；And the present invention more can also accurately provide patient's
3 classifications such as pathological state, including differentiated, middle differentiation, low differentiation, improve accuracy rate of diagnosis.
Detailed description of the invention
The present invention is described by means of the following drawings:
Fig. 1 is histopathology image analysis flow chart in the embodiment of the present invention；
Fig. 2 is histopathology image block schematic diagram in the embodiment of the present invention；
Fig. 3 is conditional of embodiment of the present invention random field models schematic diagram；
Fig. 4 is the flow diagram of histopathology image diagnosing method in the embodiment of the present invention, and wherein A is full figure cell
Core is averaged DAISY feature, and B is full figure grey level histogram feature, and C is that full figure is averaged DAISY feature.
Specific embodiment
In order to preferably explain the present invention, in order to understand, with reference to the accompanying drawing, by specific embodiment, to this hair
It is bright to be described in detail.
The present invention provides a kind of cervical cancer tissues pathological image diagnostic method based on spotted arrays condition random field, such as
Shown in Fig. 1, specifically includes the following steps:
Step S1, histopathology image to be processed is obtained.
Step S2, each image to be processed is pre-processed.
Step S3, image segmentation is carried out to each pretreated image, obtains the nucleus two-value of each image
Figure positions nucleus according to image segmentation result, obtains the nucleus coordinate of each image, and by each nucleus two-value
Figure is divided into the image block of fixed size.
Step S4, feature extraction is carried out to each nucleus binary map, the global characteristics of each image is obtained, to every
Each image block of one nucleus binary map carries out feature extraction, obtains the local feature of each image.
Step S5, the local feature of the nucleus coordinate of each image and each image is input to and is pre-designed
In conditional random field models, the full figure nuclear characteristics of each image are obtained, are classified using condition random field trained in advance
Device handles the global characteristics of each image and the full figure nuclear characteristics of each image, output category result.
In step sl, the format of histopathology image to be processed includes * .bmp, * .BMP, * .dip, * DIP, *
.jpg, * .JPG, * .jpeg, * JPEG, * .jpe, * .JPE, * .jfif, * JFIF, * .gif, * .GIF, * .GIF, * .GIF, *
.GIFf, * .GIFF, * .png, * .PNG.
In step s 2, image to be processed is pre-processed, specifically: first using median filter to figure to be processed
As being denoised, histogram equalization is reused to enhance picture contrast to be processed.It will be appreciated that using mean filter pair
Image to be processed is denoised, and similar effect also may be implemented.
In step s3, using the image partition method clustered based on K-means, it is 4 that cluster numbers K, which is arranged, tissue disease
Image of science gathers for nucleus, cytoplasm, cytoplasm, four class of image labeling；It is generated in conjunction with morphological operation appropriate thin
Karyon binary map.Certainly, above-mentioned image partition method is only preferred, it will be appreciated that carrying out image point using dividing ridge method
It cuts, same effect also may be implemented.Further, each nucleus binary map is divided into the image of 100 × 100 pixels
Block, as shown in Fig. 2 (scale bar of image is 50 μm).
In step s 4, feature extraction is carried out to each nucleus binary map, including full figure textural characteristics and full figure are put down
Equal DAISY feature, wherein full figure textural characteristics use full figure grey level histogram feature.To the every of each nucleus binary map
A image block carries out feature extraction, specially extracts each image block and is averaged DAISY feature descriptor.DAISY is towards dense
Ground local image characteristics description can be quickly calculated to feature extraction, point of gradient orientation histogram is carried out using Gaussian convolution
Block convergence, in this way using Gaussian convolution can it is quickly computational can the quick dense extraction for carrying out Feature Descriptor.
Certainly, it is only preferred that DAISY feature, full figure grey level histogram feature are extracted, it will be appreciated that using binary robust
Constant expansible key point (BRISK) replaces DAISY feature, replaces full figure grey level histogram feature using color moment characteristics,
Similar effect may be implemented.It is further preferred that carrying out dimensionality reduction using Principal Component Analysis (PCA), accordingly after feature extraction
, the step S5 includes: that the local feature after the global characteristics and dimensionality reduction of each image is input to the item being pre-designed
In part random field models；Multi objective is converted into a few overall target (i.e. principal component), wherein each principal component can
Reflect the most information of original variable, and information contained does not repeat mutually, simplifys a problem, while obtained result more section
Learn effective data information.
In step s 5, the conditional random field models being pre-designed are the multiple dimensioned array layout of rectangular area point type by square
The conditional probability model of the calculating rectangular centre pixel of pixel parameter shared by pixel parameter and spotted arrays shared by shape.Such as figure
Shown in 3, conditional random field models are designed as the multiple dimensioned array layout of rectangular area point type, by the ginseng of rectangular loop white image block
(herein, several parameters with several white image blocks in rectangular loop calculate the conditional probability of the image block positioned at rectangular centre
Image block and pixel are subjected to mutual analogy, the image block of gridding is regarded as a pixel, it is designed convenient for application
Conditional random field models), it include nucleus according to nucleus coordinate setting by taking in Fig. 32 × 2 spotted arrays layout as an example
Image block, using these image blocks comprising nucleus as rectangular centre pixel, by 4 white prints in 0.75 times of rectangular loop
As the sum of the average DAISY feature vector and the average DAISY feature vector of 0.25 times of rectangular loop white image block of block obtains
The average DAISY feature vector of the image block comprising nucleus as rectangular centre, being calculated all based on this includes nucleus
Image block average DAISY feature vector, the full figure nucleus for obtaining this image is averaged DAISY feature.
Using the conditional random field models of the multiple dimensioned array layout of rectangular area point type, with the following functions and advantage:
(1), size is available there are many models, 3 kinds of examples shown in including but not limited to Fig. 3, can be according to image reality
Size apply various sizes of layout, be applicable not only to the histopathology image got by different approaches, also be applicable in
In other medicine or non-medical images；(2), the model can selectively extract the Information Number within the scope of the dot matrix of target area
According to image data including nucleus, cytoplasm and cytoplasm can comprehensively describe spatial information, system is examined
Disconnected result is highly efficient reliable；(3), the model can be according to specific image and actual demand, by distributing to picture shared by rectangle
The weight of vegetarian refreshments and pixel difference numerical value shared by spotted arrays, the calculation method of design condition probability.The model is that observation increases
Add new dimension, help to estimate more parameters and improves estimation performance.It will be appreciated that the array layout in the model may be used also
To be designed as linear, round, plane, cylinder and the arrangement model of spherical shape, the effect that different array arrangement modes generates is not yet
Together, for example, increase a direction on array length, the signal-to-noise ratio of the direction can be made to improve, reduce the interference of noise.
The conditional random field models of the multiple dimensioned array layout of rectangular area point type can be indicated with following formula:
Wherein, c indicates that class label, x are input picture, and Z (θ, x) is the partition functions that distribution is normalized, θ=
{θψ, θφ, θλIt is model parameter, i corresponds to the index in image block, and n is the number of full graphics image block.As it can be seen that ψi(ci,x；
θψ) represent full figure nucleus and be averaged DAISY feature, φ (c, x；θφ) full figure grey level histogram feature is represented,Generation
Table full figure is averaged DAISY feature, and explaining condition random field of the present invention from mathematics level includes features described above information.
It is further preferred that as shown in figure 4, further comprising the steps of before above-mentioned steps S1:
S01, select high, medium and low differentiated tissue's pathological image subset of identical quantity as training set.
S02, step S1 to S4 processing is carried out to each image in training set, obtain in training set each image
The global characteristics and local feature of nucleus coordinate, each image, and, the nucleus of each image in training set is sat
The local feature of mark and each image is input in the conditional random field models being pre-designed, and obtains each figure in training set
The full figure nuclear characteristics of picture.
S03, the complete of each image in differentiated histopathology image subset is inputted into condition random field classifier
The full figure nuclear characteristics of office feature and each image, obtain differentiated sorter model.
Into condition random field classifier, the overall situation of each image is special in differentiated tissue's pathological image subset in input
It seeks peace the full figure nuclear characteristics of each image, obtains middle differentiation sorter model.
The overall situation that each image in low differentiated tissue's pathological image subset is inputted into condition random field classifier is special
It seeks peace the full figure nuclear characteristics of each image, obtains low differentiation sorter model.
As a result, in step s 5, it is handled using condition random field classifier trained in advance, specifically:
The likelihood of each histopathology image Yu above-mentioned 3 models is calculated, the highest model class of likelihood is selected
Not Zuo Wei final classification and diagnostic result, the calculation formula of likelihood is as follows: likelihood=full figure nucleus is averaged DAISY feature
Similarity × full figure grey level histogram similarity × full figure is averaged DAISY characteristic similarity, wherein the calculating of similarity uses
Corr2 function.
Compared with prior art, condition random field classifier is applied to automatically analyzing for histopathology image by the present invention,
Form the auto-check system of a set of histopathology image.The present invention can suitably slow down virologist's resource allocation not
, the problem of being not enough can also instruct medico rationally to be diagnosed with the insufficient doctor of experience；And the present invention is also
3 classifications such as the pathological state of patient, including differentiated, middle differentiation, low differentiation more can be accurately provided, it is quasi- to improve diagnosis
True rate.
It is to be appreciated that describing the skill simply to illustrate that of the invention to what specific embodiments of the present invention carried out above
Art route and feature, its object is to allow those skilled in the art to can understand the content of the present invention and implement it accordingly, but
The present invention is not limited to above-mentioned particular implementations.All various changes made within the scope of the claims are repaired
Decorations, should be covered by the scope of protection of the present invention.
Claims (8)
1. a kind of cervical cancer tissues pathological image diagnostic method based on spotted arrays condition random field, which is characterized in that including
Following steps:
Step S1, histopathology image to be processed is obtained；
Step S2, each image to be processed is pre-processed；
Step S3, image segmentation is carried out to described each pretreated image, obtains the nucleus two-value of each image
Figure positions nucleus according to image segmentation result, obtains the nucleus coordinate in each image, and by each nucleus two
Value figure is divided into the image block of fixed size；
Step S4, feature extraction is carried out to each nucleus binary map, the global characteristics of each image is obtained, to institute
The each image block for stating each nucleus binary map carries out feature extraction, obtains the local feature of each image；
Step S5, the local feature of the nucleus coordinate of each image and each image is input to and is set in advance
In the conditional random field models of meter, the full figure nuclear characteristics of each image are obtained, use condition random field trained in advance
Classifier handles the global characteristics of each image and the full figure nuclear characteristics of each image, output
Classification results；
The conditional random field models being pre-designed are the pixel as shared by rectangle of the multiple dimensioned array layout of rectangular area point type
Pixel parameter shared by point parameter and spotted arrays calculates the conditional probability model of rectangular centre pixel.
2. the cervical cancer tissues pathological image diagnostic method according to claim 1 based on spotted arrays condition random field,
It is characterized in that, before the step S1 further include:
S01, select high, medium and low differentiated tissue's pathological image subset of identical quantity as training set；
S02, step S1 to the S4 processing is carried out to each image in the training set, obtains each figure in training set
The global characteristics and local feature of the nucleus coordinate of picture, each image, and,
The local feature of the nucleus coordinate of each image in the training set and each image is input to and is pre-designed
Conditional random field models in, obtain training set in each image full figure nuclear characteristics；
S03, the overall situation that each image in differentiated histopathology image subset is inputted into condition random field classifier are special
It seeks peace the full figure nuclear characteristics of each image, obtains differentiated sorter model；
Into condition random field classifier in input in differentiated tissue's pathological image subset the global characteristics of each image and
The full figure nuclear characteristics of each image obtain middle differentiation sorter model；
Inputted into condition random field classifier in low differentiated tissue's pathological image subset the global characteristics of each image and
The full figure nuclear characteristics of each image obtain low differentiation sorter model.
3. the cervical cancer tissues pathological image diagnostic method according to claim 1 based on spotted arrays condition random field,
It is characterized in that, each image to be processed is pre-processed in step S2, specifically:
For each image to be processed, first image is denoised using median filter, reuses histogram equalization
To enhance picture contrast.
4. the cervical cancer tissues pathological image diagnostic method according to claim 1 based on spotted arrays condition random field,
It is characterized in that, carrying out image segmentation in step S3 to described each pretreated image, obtaining the thin of each image
Karyon binary map, specifically:
According to the image partition method and preset cluster numbers K value clustered based on K-means, described each is pre-processed
Image afterwards, which gathers, to be obtained for nucleus, cytoplasm, cytoplasm, four class of image labeling further according to preset morphological operation
Obtain the nucleus binary map of each image.
5. the cervical cancer tissues pathological image diagnostic method according to claim 1 based on spotted arrays condition random field,
It is characterized in that, step S4 further include:
Dimensionality reduction is carried out using local feature of the Principal Component Analysis to each image；
Correspondingly, the step S5 includes: to be input to the local feature after the global characteristics and dimensionality reduction of each image
In the conditional random field models being pre-designed.
6. the cervical cancer tissues pathological image diagnostic method according to claim 2 based on spotted arrays condition random field,
It is characterized in that,
In step S4, DAISY feature that the global characteristics of each image include full figure textural characteristics and full figure is averaged；It is described
Full figure textural characteristics are full figure grey level histogram feature.
7. the cervical cancer tissues pathological image diagnostic method according to claim 6 based on spotted arrays condition random field,
It is characterized in that,
In step S4, the local feature for obtaining each image specially extracts each image block DAISY feature that is averaged and retouches
State symbol.
8. the cervical cancer tissues pathological image diagnostic method according to claim 7 based on spotted arrays condition random field,
It is characterized in that, using condition random field classifier trained in advance to the global characteristics of each image and described each
The full figure nuclear characteristics for opening image are handled, specifically:
According to the differentiated sorter model, middle differentiation sorter model and low differentiation sorter model, by described each
The global characteristics of image, each image full figure nuclear characteristics as input variable, obtain each image and every
The likelihood of one sorter model selects the classification of the highest sorter model of likelihood as final classification and diagnosis knot
Fruit；
The calculation formula of likelihood is as follows: likelihood=full figure nucleus is averaged DAISY characteristic similarity * full figure grey level histogram
Similarity * full figure is averaged DAISY characteristic similarity, wherein the calculating of similarity uses corr2 function.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811552817.4A CN109671072A (en) | 2018-12-18 | 2018-12-18 | Cervical cancer tissues pathological image diagnostic method based on spotted arrays condition random field |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811552817.4A CN109671072A (en) | 2018-12-18 | 2018-12-18 | Cervical cancer tissues pathological image diagnostic method based on spotted arrays condition random field |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109671072A true CN109671072A (en) | 2019-04-23 |
Family
ID=66144053
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811552817.4A Pending CN109671072A (en) | 2018-12-18 | 2018-12-18 | Cervical cancer tissues pathological image diagnostic method based on spotted arrays condition random field |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109671072A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110264454A (en) * | 2019-06-19 | 2019-09-20 | 四川智动木牛智能科技有限公司 | Cervical cancer tissues pathological image diagnostic method based on more hidden layer condition random fields |
| CN112396616A (en) * | 2020-12-14 | 2021-02-23 | 南京信息工程大学 | Osteosarcoma recurrence risk prediction model based on tissue morphology analysis |
| CN113033287A (en) * | 2021-01-29 | 2021-06-25 | 杭州依图医疗技术有限公司 | Pathological image display method and device |
| CN113256627A (en) * | 2021-07-05 | 2021-08-13 | 深圳科亚医疗科技有限公司 | Apparatus and method for analysis management of cervical images, apparatus and storage medium |
| CN116959712A (en) * | 2023-07-28 | 2023-10-27 | 成都市第三人民医院 | Lung adenocarcinoma prognosis method, system, equipment and storage medium based on pathological image |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105321176A (en) * | 2015-09-30 | 2016-02-10 | 西安交通大学 | Image segmentation method based on hierarchical higher order conditional random field |
| 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 |
| US20170177943A1 (en) * | 2015-12-21 | 2017-06-22 | Canon Kabushiki Kaisha | Imaging system and method for classifying a concept type in video |
-
2018
- 2018-12-18 CN CN201811552817.4A patent/CN109671072A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105321176A (en) * | 2015-09-30 | 2016-02-10 | 西安交通大学 | Image segmentation method based on hierarchical higher order conditional random field |
| 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 |
| US20170177943A1 (en) * | 2015-12-21 | 2017-06-22 | Canon Kabushiki Kaisha | Imaging system and method for classifying a concept type in video |
Non-Patent Citations (2)
| Title |
|---|
| YIMING LIU 等: "Automatic Segmentation of Cervical Nuclei Based on Deep Learning and a Conditional Random Field", 《IEEE》 * |
| 阳维 等: "基于图像块分类器和条件随机场的显微图像分割", 《计算机应用》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110264454A (en) * | 2019-06-19 | 2019-09-20 | 四川智动木牛智能科技有限公司 | Cervical cancer tissues pathological image diagnostic method based on more hidden layer condition random fields |
| CN110264454B (en) * | 2019-06-19 | 2021-07-30 | 四川智动木牛智能科技有限公司 | Cervical cancer histopathological image diagnosis method based on multi-hidden-layer conditional random field |
| CN112396616A (en) * | 2020-12-14 | 2021-02-23 | 南京信息工程大学 | Osteosarcoma recurrence risk prediction model based on tissue morphology analysis |
| CN113033287A (en) * | 2021-01-29 | 2021-06-25 | 杭州依图医疗技术有限公司 | Pathological image display method and device |
| CN113256627A (en) * | 2021-07-05 | 2021-08-13 | 深圳科亚医疗科技有限公司 | Apparatus and method for analysis management of cervical images, apparatus and storage medium |
| CN116959712A (en) * | 2023-07-28 | 2023-10-27 | 成都市第三人民医院 | Lung adenocarcinoma prognosis method, system, equipment and storage medium based on pathological image |
| CN116959712B (en) * | 2023-07-28 | 2024-06-21 | 成都市第三人民医院 | Lung adenocarcinoma prognosis method, system, equipment and storage medium based on pathological image |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Chato et al. | Machine learning and deep learning techniques to predict overall survival of brain tumor patients using MRI images | |
| CN109671072A (en) | Cervical cancer tissues pathological image diagnostic method based on spotted arrays condition random field | |
| McKinley et al. | Ensembles of densely-connected CNNs with label-uncertainty for brain tumor segmentation | |
| Shahidi et al. | Breast cancer classification using deep learning approaches and histopathology image: A comparison study | |
| Gandomkar et al. | MuDeRN: Multi-category classification of breast histopathological image using deep residual networks | |
| Basavanhally et al. | Computerized image-based detection and grading of lymphocytic infiltration in HER2+ breast cancer histopathology | |
| Tareef et al. | Multi-pass fast watershed for accurate segmentation of overlapping cervical cells | |
| CN106570505B (en) | Method and system for analyzing histopathological images | |
| CN106340021B (en) | Blood vessel extraction method | |
| US11340324B2 (en) | Systems, methods and media for automatically segmenting and diagnosing prostate lesions using multi-parametric magnetic resonance imaging data | |
| Bai et al. | Liver tumor segmentation based on multi-scale candidate generation and fractal residual network | |
| WO2021030629A1 (en) | Three dimensional object segmentation of medical images localized with object detection | |
| Plissiti et al. | Overlapping cell nuclei segmentation using a spatially adaptive active physical model | |
| CN104376147A (en) | Image-based risk score-a prognostic predictor of survival and outcome from digital histopathology | |
| He et al. | Automatic magnetic resonance image prostate segmentation based on adaptive feature learning probability boosting tree initialization and CNN-ASM refinement | |
| CN109299679A (en) | Cervical cancer tissues pathological image diagnostic method based on sleeve configuration condition random field | |
| JP2023517058A (en) | Automatic detection of tumors based on image processing | |
| US20220028068A1 (en) | Multi-scale tumor cell detection and classification | |
| Mercan et al. | Deep feature representations for variable-sized regions of interest in breast histopathology | |
| CN112102230A (en) | Ultrasonic tangent plane identification method, system, computer equipment and storage medium | |
| Rampun et al. | Breast density classification using local ternary patterns in mammograms | |
| Pawar et al. | Local Binary Patterns Descriptor Based on Sparse Curvelet Coefficients for False‐Positive Reduction in Mammograms | |
| Madduri et al. | Classification of breast cancer histopathological images using convolutional neural networks | |
| CN109285176A (en) | A kind of cerebral tissue dividing method cut based on regularization figure | |
| Khan et al. | Breast cancer histological images nuclei segmentation and optimized classification with deep learning |
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 |
Application publication date: 20190423 |
|
| RJ01 | Rejection of invention patent application after publication |