CN109145838A - Clear cell carcinoma of kidney diagnostic method based on random Gaussian field neural network aiding - Google Patents
Clear cell carcinoma of kidney diagnostic method based on random Gaussian field neural network aiding Download PDFInfo
- Publication number
- CN109145838A CN109145838A CN201810992268.6A CN201810992268A CN109145838A CN 109145838 A CN109145838 A CN 109145838A CN 201810992268 A CN201810992268 A CN 201810992268A CN 109145838 A CN109145838 A CN 109145838A
- Authority
- CN
- China
- Prior art keywords
- neural network
- random gaussian
- formula
- gaussian field
- network model
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/69—Microscopic objects, e.g. biological cells or cellular parts
- G06V20/695—Preprocessing, e.g. image segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/69—Microscopic objects, e.g. biological cells or cellular parts
- G06V20/698—Matching; Classification
Abstract
The invention discloses a kind of clear cell carcinoma of kidney diagnostic methods based on random Gaussian field neural network aiding, in the training stage, construct the self study complex network structures model of data-driven, pathology block the outputting and inputting as network that original medical image and corresponding band are marked, the study for having supervision is carried out to network；In the judgement stage, unknown input original medical image is given, calculates site morbidity probabilistic geometry mean value to be tested, and provide confidence interval, so that doctor be assisted to determine canceration grade and physiology lesion position.After the present invention is by constructing a kind of random Gaussian field neural network model of data-driven and carrying out the study for having supervision to network model, unknown medical image can effectively be screened, auxiliary doctor positions diseased region, to help doctor preferably to judge the transparent cancer cell diseased region of kidney and lesion grade, alleviate doctor's pressure, accuracy rate is made a definite diagnosis in raising.
Description
Technical field
The present invention relates to medical field of auxiliary, saturating more particularly to a kind of kidney based on random Gaussian field neural network aiding
Clear cell carcinoma diagnostic method.
Background technique
Carcinoma of renal parenchyma is derived from the gland cancer of renal cells, and 85% is clear cell carcinoma, some is
Granulocyte carcinoma and cell mixing cancer.Often there are bleeding, necrosis, capsule change and calcification in cancer.It is born in kidney essence, infiltrates, presses after growing up
Compel, destroy renal plevis kidney calices, develop to outside kidney peplos, form hemangioma bolt or is transferred to lymph node and other organs.Pathologically,
Kidney is divided into 4 types: clear cell type kidney, granular cell type kidney, mixed cell type kidney, neoblast type kidney.Its
In, the overwhelming majority is clear cell carcinoma of kidney, accounts for the 70%~80% of kidney, cancer cell often arranges slabbing, streak, acinus
Shape or tubulose, like renal tubule.Though clear cell carcinoma is that grade malignancy is minimum in kidney, in clinical practice often
It is mixed, is classified under microscope in fact extremely difficult with granular cell carcinoma and three type of Fusoid cells cancer.Therefore it is badly in need of one kind to melt
The method for closing current artificial intelligence forward position algorithm carries out Classification and Identification to medicine contrastographic picture, and doctor is assisted to diagnose.
In artificial intelligence field of signal processing, artificial neural network is since it is with Nonlinear Modeling and self-adapting data
Ability, to be widely used in data classification and parameter mapping aspect.Adaptive base letter inside artificial neural network dependence
Link attribute between number, thus the strong correlation between study and characterize data, the classification and recurrence characteristic of response data.It is another
Aspect, the probability stochastic model based on Bayesian frame, such as random Gaussian field, provide the processing with neural network differentiated
Data means: random Gaussian field energy randomization prediction result not only provides point estimate, moreover it is possible to analyze forecast confidence area
Between, this is helpful for realistic problem.The present invention utilizes the two advantage, constructs a kind of completely new system frame
Frame, and doctor is assisted using this New Mathematical Model to carry out the diagnosis of medical imaging.
For needing to carry out the patient of clear cell carcinoma of kidney resection operation after diagnosing, in the prior art, it is generally required that will
The entire kidney of its lesion is cut off, although theory is directly simple, the cost paid for the patient is also huge.And
In the laparoscopic renal tumor resection case of current mainstream, need doctor it is subjective to canceration physiological tissue Reginal-block scanning figure
As carrying out two-dimentional identification, to make artificial judgment.This process usually needs to rely on a large amount of clinical experiences and knowledge of doctor
Accumulation can just accomplish effective pre-estimation.Therefore i.e. time-consuming and laborious, the uncertain ideal of effect.For this purpose, the present invention attempts to lead to
It crosses and designs a kind of artificial intelligence auxiliary mechanism, for improving doctor to the accurate positioning of diseased region and the assessment of lesion grade,
So as to targetedly cut off, rather than entire kidney excision.
Summary of the invention
In view of the deficiencies of the prior art, it is transparent thin to provide a kind of kidney based on random Gaussian field neural network aiding by the present invention
Born of the same parents' cancer diagnostic method by constructing a kind of artificial intelligence random Gaussian field neural network model of data-driven, and provides a large amount of
Original medical image and identification result are output and input as system, are implemented intelligent supervised learning, are completed it in learning training
Afterwards, unknown medical image can effectively be screened, auxiliary doctor positions diseased region, to help doctor better
Judge the transparent cancer cell diseased region of kidney and lesion grade, alleviates doctor's pressure, while improving and making a definite diagnosis accuracy rate.
To achieve the above object, the technical solution adopted by the present invention are as follows:
A kind of clear cell carcinoma of kidney diagnostic method based on random Gaussian field neural network aiding, in the training stage, construction
The self study complex network structures model of data-driven, the pathology block that original medical image and corresponding band are marked is as net
Network is output and input, and the study for having supervision is carried out to network；In the judgement stage, unknown input original medical image is given,
Site morbidity probabilistic geometry mean value to be tested is calculated, and provides confidence interval, so that doctor be assisted to determine canceration grade and life
Manage lesion position.
Preferably, specifically includes the following steps:
Image preprocessing: original medical image is converted to the gray level image of gray value expression by S01；
The building and training of neural network model: S02 passes through random Gaussian field for the gray value of gray level image as input
Intermediate parameters are mapped as, and link to output layer by connecting non-directed graph mode entirely, output layer is set as the knowledge by manually marking
Not as a result, to form the random Gaussian field neural network model with Nonlinear Classification function, the neural network model structure
After the completion of building, using a large amount of original medical images and the corresponding identification result with pathology block mark as the neural network
Model is output and input, and the study for having supervision is carried out to the neural network model；
S03 is identified using the neural network model: by the medical image at position to be tested through image preprocessing
Afterwards, the neural network model is inputted, prediction judges classification mode, and each classification results have certain probability, and obtain simultaneously
The confidence interval of predictive estimation is obtained, doctor assists examining by lateral comparison classification results probability size and confidence interval
It is disconnected.
Further, the original medical image is the medical scanning image of medicine clear cell carcinoma of kidney.
Preferably, S01 further include steps of using algorithm for image enhancement to original image carry out it is following a kind of or
A variety of operations: rotation, unified image size, highlights and contrast at alignment.
Preferably, S02 is further included steps of
(a1) setting x indicates that the vector that input image pixels data are constituted, y indicate output classification, construct following random Gaussian
Field neural network model:
Y (x)=W (x) [f (x)+σf∈]+σyz (1)
Wherein,
∈=∈ (x) and z=z (x) is two white Gaussian noise process with different covariance parameters respectively, and ∈'s is general
Rate is distributed as N (0, Iq), N (0, Iq) indicate that mean value is 0, covariance matrix IqIt is the Gaussian Profile of the unit matrix of Q × Q dimension,
The probability distribution of z is N (0, Ip), N (0, Ip) indicate that mean value is 0, covariance matrix IpIt is the Gauss of the unit matrix of P × P dimension
Distribution, σfAnd σyIt is energy coefficient to be estimated respectively,
W (x) is the matrix of a P × Q, wherein each element W (x)ijIt is all an independent random Gaussian field, i.e.,Wherein kwIt may be any type of positive semidefinite kernel function,
F (x)=f (f1(x),f2(x),…,fq(x)) be Q dimension vector, wherein any one element is all independent
Random Gaussian field, i.e.,
(a2) training dataset is set to be combined intoUnknown model parameters u=(f, W), f and W are respectively indicated
Bring data point intoF (x) and W (x) later has following prior probability distribution according to the definition of random Gaussian field:
p(u|σf,θf,θw)=N (0, CB) (2)
Wherein θf,θwRespectively indicate the hyper parameter that the kernel function being related in f (x) and W (x) is included, CBIt is a NQ (P
+ 1) the block diagonal battle array of × NQ (P+1) dimension,
Likelihood score function can be obtained according to (1) formula simultaneously are as follows:
It can be obtained with Bayes' theorem:
Wherein, (4) formula is target formula to be estimated, and the estimation for obtaining (4) formula can obtain the estimation of y (x), is sentenced
Other result；
(a3) (4) formula is optimized using variation Bayesian Method, to obtain optimal models structural parameters.Variation Bayes
Essence be by a kind of iterative manner come so that probability distribution q () approaching to reality to be estimated Posterior probability distribution p ()
(i.e. formula 4), i.e., by minimizing distortion function Dist:
Wherein,H [] indicates entropy function, gives firstInverse gamma distribution (IG) is distributed,
That is:
Secondly, design APPROXIMATE DISTRIBUTION q (v) is as follows:
WhereinIt is that inverse gamma is distributed, qfi,qwijIt is the Gaussian Profile of N-dimensional,
Optimal value is sought finally, being iterated to (7) formula, i.e., carries out segmented line in gradient direction using conjugate gradient decent
Property search, searching make the maximized θ of (7) formulaf,θw。
Preferably, S03 is further included steps of
(b1) for a new unknown medical image, calling S01 step first carries out image preprocessing；
(b2) classification judgement is carried out using the trained network model of S02 step, first according to Bayesian formula, accurately
Target prediction function should are as follows:
Secondly, the thought using variation Bayesian Method carries out approximation, if in above formula
It is obtained even normal probability distribution can be multiplied by two approximation probability distributions, according to the Gaussian Profile conditional probability of standard and side
The attribute of edge probability, integrates above formula, obtains the mean value mean (y of estimation judgement*) and variance cov (y*)ij, variance cov
(y*)ijAs confidence interval, as follows:
In formula, k indicates to differentiate type, δijIt is kronecker delta function, on the basis of obtaining judgement mean value,
It carries out that maximum one kind of across comparison select probability and is used as court verdict, meanwhile, determine that the judgement is tied according to variance function
Fruit has great variation possible, to constitute complete judgment criteria.
Preferably, random Gaussian field parameters, neural network model parameter by variation Bayesian Method marginalisation maximum seemingly
So estimate to obtain under degree criterion.
Compared with prior art, the beneficial effects of the present invention are 1) nerve of the present invention building based on random Gaussian field
Network model, the pathology block that a large amount of original medical image and corresponding band are marked as the input of neural network model with
Output, the study for having supervision is carried out to neural network model, then inputs the medical image at unknown new position to be tested
Neural network model to building and by there is the learning training of supervision obtains site morbidity probability Estimation judgement to be tested
Mean value and confidence interval, auxiliary doctor determine canceration grade and physiology lesion block, improve positioning of the doctor to diseased region
And the accuracy of lesion grade assessment；2) method of the invention is used to carry out assisting in diagnosis and treatment to medical image, has merged artificial intelligence
Energy algorithm, can help doctor preferably to judge the transparent cancer cell diseased region of kidney and lesion grade, thus alleviate doctor's pressure,
It improves simultaneously and makes a definite diagnosis accuracy rate, promote working efficiency；3) neural network model of the invention is the nerve based on random Gaussian field
Neural network model not only includes probability characteristics, but also includes the non-frequency characteristic of neural network, can be carried out to unknown image
Nonlinear probability classification, and confidence interval analysis is provided as a result, substantially increasing the confidence level of complementary diagnosis.
Detailed description of the invention
Fig. 1 is the stream according to the clear cell carcinoma of kidney diagnostic method based on random Gaussian field neural network aiding of embodiment
Cheng Tu；
Fig. 2 is the schematic diagram of random Gaussian field neural network model of the invention；
Fig. 3 is the schematic diagram according to the clear cell carcinoma of kidney diseased region of the invention of embodiment；
Fig. 4 is the schematic diagram according to the clear cell carcinoma of kidney diseased region of the invention of embodiment.
Specific embodiment
The invention will be further described with reference to the accompanying drawing and by specific embodiment, and following embodiment is descriptive
, it is not restrictive, this does not limit the scope of protection of the present invention.
The present invention provides a kind of clear cell carcinoma of kidney diagnostic method based on random Gaussian field neural network aiding, in training
Stage constructs the self study complex network structures model of data-driven, the pathology that original medical image and corresponding band are marked
Block is output and input as network, and the study for having supervision is carried out to network；In the judgement stage, it is original to give unknown input
Medical image calculates site morbidity probabilistic geometry mean value to be tested, and provides confidence interval, so that doctor be assisted to determine canceration
Grade and physiology lesion position.As shown in the picture, specifically includes the following steps:
Image preprocessing: original medical image is converted to the gray level image of gray value expression, and utilized traditional by S01
Algorithm for image enhancement rotated, is aligned, unified image size, is highlighted and contrast etc. operates, wherein the original doctor
The medical scanning image that image is medicine clear cell carcinoma of kidney is treated, the pathology block with mark includes being labeled with calculus and uronephrosis
Pathology block, be labeled with the pathology block of tumour；
The building and training of neural network model: S02 passes through random Gaussian field for the gray value of gray level image as input
Intermediate parameters are mapped as, and link to output layer by connecting non-directed graph mode entirely, output layer is set as the knowledge by manually marking
Not as a result, to form the random Gaussian field neural network model with Nonlinear Classification function, the neural network model structure
After the completion of building, using a large amount of original medical images and the corresponding identification result with pathology block mark as the neural network
Model is output and input, and the study for having supervision is carried out to the neural network model；Wherein, random Gaussian field parameters, nerve
Network model parameter is estimated to obtain by variation Bayesian Method under marginalisation maximum likelihood degree criterion；
S03 is identified using the neural network model: by the medical image at position to be tested through image preprocessing
Afterwards, the neural network model is inputted, prediction judges classification mode, and each classification results have certain probability, and obtain simultaneously
The confidence interval of predictive estimation is obtained, then judges that classification results, each classification results have by Neural Network model predictive
There is certain probability, and system provides the confidence interval of prediction judging result, doctor can pass through lateral comparison classification results
Probability size and possibility coverage area (confidence interval) carry out auxiliary diagnosis, so that it is determined that the canceration grade at position to be tested
With physiology lesion block.
Specifically, as shown in Fig. 2, S02 is further included steps of
(a1) setting x indicates that the vector that input image pixels data are constituted, y indicate output classification, construct following random Gaussian
Field neural network model:
Y (x)=W (x) [f (x)+σf∈]+σyz (1)
Wherein,
∈=∈ (x) and z=z (x) is two white Gaussian noise process with different covariance parameters respectively, and ∈'s is general
Rate is distributed as N (0, Iq), N (0, Iq) indicate that mean value is 0, covariance matrix IqIt is the Gaussian Profile of the unit matrix of Q × Q dimension,
The probability distribution of z is N (0, Ip), N (0, Ip) indicate that mean value is 0, covariance matrix IpIt is the Gauss of the unit matrix of P × P dimension
Distribution, σfAnd σyIt is energy coefficient to be estimated respectively,
W (x) is the matrix of a P × Q, wherein each element W (x)ijIt is all an independent random Gaussian field, i.e.,Wherein kwIt may be any type of positive semidefinite kernel function,
F (x)=f (f1(x),f2(x),…,fq(x)) be Q dimension vector, wherein any one element is all independent
Random Gaussian field, i.e.,
(a2) training dataset is set to be combined intoUnknown model parameters u=(f, W), f and W are respectively indicated
Bring data point intoF (x) and W (x) later has following prior probability distribution according to the definition of random Gaussian field:
p(u|σf,θf,θw)=N (0, CB) (2)
Wherein θf,θwRespectively indicate the hyper parameter that the kernel function being related in f (x) and W (x) is included, CBIt is a NQ (P
+ 1) the block diagonal battle array of × NQ (P+1) dimension,
Likelihood score function can be obtained according to (1) formula simultaneously are as follows:
It can be obtained with Bayes' theorem:
Wherein, (4) formula is target formula to be estimated, and the estimation for obtaining (4) formula can obtain the estimation of y (x), is sentenced
Other result；
(a3) (4) formula is optimized using variation Bayesian Method, to obtain optimal models structural parameters.Variation Bayes
Essence be by a kind of iterative manner come so that probability distribution q () approaching to reality to be estimated Posterior probability distribution p ()
(i.e. formula 4), i.e., by minimizing distortion function:
Wherein,H [] indicates entropy function, gives firstInverse gamma distribution (IG) is distributed,
That is:
Wherein, energy coefficient σ to be estimatedfAnd σyIt can be obtained from (6) formula, secondly, design APPROXIMATE DISTRIBUTION q (v) is as follows:
WhereinIt is that inverse gamma is distributed,It is the Gaussian Profile of N-dimensional,
Optimal value is sought finally, being iterated to (7) formula, i.e., carries out segmented line in gradient direction using conjugate gradient decent
Property search, searching make the maximized θ of (7) formulaf,θw。
Specifically, S03 is further included steps of
(b1) for a new unknown medical image, calling S01 step first carries out image preprocessing；
(b2) classification judgement is carried out using the trained network model of S02 step, first according to Bayesian formula, accurately
Target prediction function should are as follows:
Secondly, the thought using variation Bayesian Method carries out approximation, if in above formula
It is obtained even normal probability distribution can be multiplied by two approximation probability distributions, according to the Gaussian Profile conditional probability of standard and side
The attribute of edge probability, integrates above formula, obtains the mean value mean (y of estimation judgement*) and variance cov (y*)ij, variance cov
(y*)ijAs confidence interval, as follows:
In formula, k indicates to differentiate type, δijIt is kronecker delta function, on the basis of obtaining judgement mean value,
It carries out that maximum one kind of across comparison select probability and is used as court verdict, meanwhile, the judgement is determined according to confidence interval
As a result have great variation possible, to constitute complete judgment criteria.
Finally, it should be noted that the above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations；Although
Present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that: it still may be used
To modify to technical solution documented by previous embodiment, or some or all of the technical features are equal
Replacement；And these are modified or replaceed, technical solution of the embodiment of the present invention that it does not separate the essence of the corresponding technical solution.
Claims (6)
1. the clear cell carcinoma of kidney diagnostic method based on random Gaussian field neural network aiding, which is characterized in that
In the training stage, the self study complex network structures model of data-driven is constructed, by original medical image and corresponding band
Pathology block the outputting and inputting as network of mark, the study for having supervision is carried out to network；In the judgement stage, give unknown
Input original medical image, site morbidity probabilistic geometry mean value to be tested is calculated, and provide confidence interval, to assist curing
Teacher determines canceration grade and physiology lesion position.
2. the clear cell carcinoma of kidney diagnostic method according to claim 1 based on random Gaussian field neural network aiding,
It is characterized in that, specifically includes the following steps:
Image preprocessing: original medical image is converted to the gray level image of gray value expression by S01；
The building and training of neural network model: S02 is mapped using the gray value of gray level image as input by random Gaussian field
For intermediate parameters, and output layer is linked to by connecting non-directed graph mode entirely, output layer is set as the identification knot by manually marking
Fruit, to form the random Gaussian field neural network model with Nonlinear Classification function, the neural network model has been constructed
Cheng Hou, using a large amount of original medical images and the corresponding identification result with pathology block mark as the neural network model
Output and input, the study for having supervision is carried out to the neural network model；
S03 is identified using the neural network model: defeated by the medical image at position to be tested after image preprocessing
Enter the neural network model, prediction judges classification mode, and each classification results have certain probability, and are predicted simultaneously
The confidence interval of estimation, doctor is by lateral comparison classification results probability size and confidence interval come auxiliary diagnosis.
3. the clear cell carcinoma of kidney diagnostic method according to claim 1 based on random Gaussian field neural network aiding,
It is characterized in that, S01, which is further included steps of, carries out following one or more behaviour to original image using algorithm for image enhancement
Make: rotation, unified image size, highlights and contrast at alignment.
4. the clear cell carcinoma of kidney diagnostic method according to claim 1 based on random Gaussian field neural network aiding,
It is characterized in that, S02 is further included steps of
(a1) setting x indicates that the vector that input image pixels data are constituted, y indicate output classification, constructs following random Gaussian field mind
Through network model:
Y (x)=W (x) [f (x)+σf∈]+σyz (1)
Wherein,
∈=∈ (x) and z=z (x) is two white Gaussian noise process with different covariance parameters, the probability point of ∈ respectively
Cloth is N (0, Iq), N (0, Iq) indicate that mean value is 0, covariance matrix IqIt is the Gaussian Profile of the unit matrix of Q × Q dimension, z's
Probability distribution is N (0, Ip), N (0, Ip) indicate that mean value is 0, covariance matrix IpIt is the Gauss point of the unit matrix of P × P dimension
Cloth, σfAnd σyIt is energy coefficient to be estimated respectively,
W (x) is the matrix of a P × Q, wherein each element W (x)ijIt is all an independent random Gaussian field, i.e.,Wherein kwIt may be any type of positive semidefinite kernel function,
F (x)=f (f1(x),f2(x),…,fq(x)) be a Q dimension vector, wherein any one element be all it is independent with
Machine Gaussian field, i.e.,
(a2) training dataset is set to be combined intoUnknown model parameters u=(f, W), f and W are respectively indicated and are brought into
Data pointF (x) and W (x) later has following prior probability distribution according to the definition of random Gaussian field:
p(u|σf,θf,θw)=N (0, CB) (2)
Wherein θf,θwRespectively indicate the hyper parameter that the kernel function being related in f (x) and W (x) is included, CBIt is a NQ (P+1)
The block diagonal battle array of × NQ (P+1) dimension,
Likelihood score function can be obtained according to (1) formula simultaneously are as follows:
It can be obtained with Bayes' theorem:
Wherein, (4) formula is target formula to be estimated, and the estimation for obtaining (4) formula can obtain the estimation of y (x), obtains differentiating knot
Fruit；
(a3) (4) formula is optimized using variation Bayesian Method, so that optimal models structural parameters are obtained, the sheet of variation Bayes
Matter is by a kind of iterative manner come so that the Posterior probability distribution p () of probability distribution q () approaching to reality to be estimated is (i.e. public
Formula 4), i.e., by minimizing distortion function Dist:
Wherein,H [] indicates entropy function, gives firstDistribute inverse gamma distribution (IG), it may be assumed that
Secondly, design APPROXIMATE DISTRIBUTION q (v) is as follows:
WhereinIt is that inverse gamma is distributed,It is the Gaussian Profile of N-dimensional,
Optimal value is sought finally, being iterated to (7) formula, i.e., carries out piecewise linearity in gradient direction using conjugate gradient decent and searches
Rope, searching make the maximized θ of (7) formulaf,θw。
5. the clear cell carcinoma of kidney diagnostic method according to claim 1 based on random Gaussian field neural network aiding,
It is characterized in that, S03 is further included steps of
(b1) for a new unknown medical image, calling S01 step first carries out image preprocessing；
(b2) classification judgement is carried out using the trained network model of S02 step, first according to Bayesian formula, accurate target
Anticipation function should are as follows:
Secondly, the thought using variation Bayesian Method carries out approximation, if in above formula
It is obtained even normal probability distribution can be multiplied by two approximation probability distributions, according to the Gaussian Profile conditional probability of standard and side
The attribute of edge probability, integrates above formula, obtains the mean value mean (y of estimation judgement*) and variance cov (y*)ij, variance cov
(y*)ijAs confidence interval, as follows:
In formula, k indicates to differentiate type, δijIt is kronecker delta function, on the basis of obtaining judgement mean value, carries out
That maximum one kind of across comparison select probability is used as court verdict, meanwhile, it is more to determine that the result has according to variance function
Big variation is possible, to constitute complete judgment criteria.
6. the clear cell carcinoma of kidney diagnostic method according to claim 4 or 5 based on random Gaussian field neural network aiding,
It is characterized in that, random Gaussian field parameters, neural network model parameter are by variation Bayesian Method in marginalisation maximum likelihood degree
Estimation obtains under criterion.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810992268.6A CN109145838B (en) | 2018-08-29 | 2018-08-29 | Renal clear cell carcinoma diagnosis method based on stochastic Gaussian field neural network assistance |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810992268.6A CN109145838B (en) | 2018-08-29 | 2018-08-29 | Renal clear cell carcinoma diagnosis method based on stochastic Gaussian field neural network assistance |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109145838A true CN109145838A (en) | 2019-01-04 |
| CN109145838B CN109145838B (en) | 2021-08-17 |
Family
ID=64828713
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810992268.6A Active CN109145838B (en) | 2018-08-29 | 2018-08-29 | Renal clear cell carcinoma diagnosis method based on stochastic Gaussian field neural network assistance |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109145838B (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109820507A (en) * | 2019-03-25 | 2019-05-31 | 钟霁媛 | Disease aided diagnosis method and device |
| CN110444294A (en) * | 2019-09-10 | 2019-11-12 | 中南大学 | A kind of prostate cancer aided analysis method and equipment based on layered perception neural networks |
| CN111553505A (en) * | 2019-11-26 | 2020-08-18 | 国网浙江省电力有限公司 | Method for predicting warehousing quantity of electric power materials |
| CN111653356A (en) * | 2020-04-20 | 2020-09-11 | 浙江大学 | New coronary pneumonia screening method and new coronary pneumonia screening system based on deep learning |
| CN112242008A (en) * | 2020-10-17 | 2021-01-19 | 郑州西亚斯学院 | AR interaction visual perception method and system for transparent container |
| CN112464579A (en) * | 2021-02-02 | 2021-03-09 | 四川大学 | Identification modeling method for searching esophageal cancer lesion area based on evolutionary neural network structure |
| CN112951427A (en) * | 2021-03-16 | 2021-06-11 | 黑龙江机智通智能科技有限公司 | Abnormal cell grading system |
| WO2022110525A1 (en) * | 2020-11-24 | 2022-06-02 | 江苏大学 | Comprehensive detection apparatus and method for cancerous region |
| CN116269749A (en) * | 2023-03-06 | 2023-06-23 | 东莞市东部中心医院 | Laparoscopic bladder cancer surgical system with improved reserved nerves |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106874687A (en) * | 2017-03-03 | 2017-06-20 | 深圳大学 | Pathological section image intelligent sorting technique and device |
| US20170177812A1 (en) * | 2015-12-21 | 2017-06-22 | Elekta Ab (Publ) | Systems and methods for optimizing treatment planning |
| WO2017156640A1 (en) * | 2016-03-18 | 2017-09-21 | Fluent.Ai Inc. | Method and device for automatically learning relevance of words in a speech recognition system |
-
2018
- 2018-08-29 CN CN201810992268.6A patent/CN109145838B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170177812A1 (en) * | 2015-12-21 | 2017-06-22 | Elekta Ab (Publ) | Systems and methods for optimizing treatment planning |
| WO2017156640A1 (en) * | 2016-03-18 | 2017-09-21 | Fluent.Ai Inc. | Method and device for automatically learning relevance of words in a speech recognition system |
| CN106874687A (en) * | 2017-03-03 | 2017-06-20 | 深圳大学 | Pathological section image intelligent sorting technique and device |
Non-Patent Citations (3)
| Title |
|---|
| AMBRISH RAWAT ET.AL: "Adversarial Phenomenon in the Eyes of Bayesian Deep Learning", 《ARXIV:1711.08244V1 [STAT.ML]》 * |
| YINHAO ZHU ET.AL: "Bayesian Deep Convolutional Encoder-Decoder Networks for Surrogate Modeling and Uncertainty Quantification", 《ARXIV:1801.06879V1 [PHYSICS.COMP-PH]》 * |
| 李军 等: "虚幻面孔加工有效连通网络的非线性DCM分析", 《西安电子科技大学学报（自然科学版）》 * |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109820507B (en) * | 2019-03-25 | 2024-03-08 | 钟霁媛 | Disease auxiliary diagnosis method and device |
| CN109820507A (en) * | 2019-03-25 | 2019-05-31 | 钟霁媛 | Disease aided diagnosis method and device |
| CN110444294A (en) * | 2019-09-10 | 2019-11-12 | 中南大学 | A kind of prostate cancer aided analysis method and equipment based on layered perception neural networks |
| CN111553505A (en) * | 2019-11-26 | 2020-08-18 | 国网浙江省电力有限公司 | Method for predicting warehousing quantity of electric power materials |
| CN111653356A (en) * | 2020-04-20 | 2020-09-11 | 浙江大学 | New coronary pneumonia screening method and new coronary pneumonia screening system based on deep learning |
| CN112242008B (en) * | 2020-10-17 | 2021-09-07 | 郑州西亚斯学院 | AR interaction visual perception method and system for transparent container |
| CN112242008A (en) * | 2020-10-17 | 2021-01-19 | 郑州西亚斯学院 | AR interaction visual perception method and system for transparent container |
| US11587231B2 (en) | 2020-11-24 | 2023-02-21 | Jiangsu University | Comprehensive detection device and method for cancerous region |
| WO2022110525A1 (en) * | 2020-11-24 | 2022-06-02 | 江苏大学 | Comprehensive detection apparatus and method for cancerous region |
| CN112464579A (en) * | 2021-02-02 | 2021-03-09 | 四川大学 | Identification modeling method for searching esophageal cancer lesion area based on evolutionary neural network structure |
| CN112951427A (en) * | 2021-03-16 | 2021-06-11 | 黑龙江机智通智能科技有限公司 | Abnormal cell grading system |
| CN112951427B (en) * | 2021-03-16 | 2023-12-08 | 黑龙江机智通智能科技有限公司 | System for classifying abnormal cells |
| CN116269749A (en) * | 2023-03-06 | 2023-06-23 | 东莞市东部中心医院 | Laparoscopic bladder cancer surgical system with improved reserved nerves |
| CN116269749B (en) * | 2023-03-06 | 2023-10-10 | 东莞市东部中心医院 | Laparoscopic bladder cancer surgical system with improved reserved nerves |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109145838B (en) | 2021-08-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109145838A (en) | Clear cell carcinoma of kidney diagnostic method based on random Gaussian field neural network aiding | |
| Lal et al. | NucleiSegNet: Robust deep learning architecture for the nuclei segmentation of liver cancer histopathology images | |
| Jia et al. | Detection of cervical cancer cells based on strong feature CNN-SVM network | |
| Liu et al. | Deep learning framework based on integration of S-Mask R-CNN and Inception-v3 for ultrasound image-aided diagnosis of prostate cancer | |
| Li et al. | Computer-aided cervical cancer diagnosis using time-lapsed colposcopic images | |
| Rahman et al. | An approach for multiclass skin lesion classification based on ensemble learning | |
| Chen et al. | MMFNet: A multi-modality MRI fusion network for segmentation of nasopharyngeal carcinoma | |
| CN109461495A (en) | A kind of recognition methods of medical image, model training method and server | |
| CN107274386A (en) | A kind of cervical cell liquid-based smear artificial intelligence aids in diagosis system | |
| Sun et al. | Gastric histopathology image segmentation using a hierarchical conditional random field | |
| CN104299242A (en) | Fluorescence angiography fundus image extraction method based on NGC-ACM | |
| Chen et al. | Fourier domain structural relationship analysis for unsupervised multimodal change detection | |
| Zhu et al. | Multi-level colonoscopy malignant tissue detection with adversarial CAC-UNet | |
| Senthilkumaran et al. | Brain image segmentation | |
| Karri et al. | Explainable multi-module semantic guided attention based network for medical image segmentation | |
| Lin et al. | Label propagation for annotation-efficient nuclei segmentation from pathology images | |
| Zhai et al. | An improved full convolutional network combined with conditional random fields for brain MR image segmentation algorithm and its 3D visualization analysis | |
| CN104166994B (en) | A kind of bone suppressing method optimized based on training sample | |
| Song et al. | DPCTN: Dual path context-aware transformer network for medical image segmentation | |
| Lin et al. | Nuclei segmentation with point annotations from pathology images via self-supervised learning and co-training | |
| He et al. | Curv‐Net: Curvilinear structure segmentation network based on selective kernel and multi‐Bi‐ConvLSTM | |
| Zhang et al. | Dual parallel net: A novel deep learning model for rectal tumor segmentation via CNN and transformer with Gaussian Mixture prior | |
| Singh et al. | Detection and classification of brain tumor using hybrid feature extraction technique | |
| CN109767429A (en) | A kind of image screening method and device | |
| Nan et al. | Unsupervised tissue segmentation via deep constrained gaussian network |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |