CN107730497A - A kind of plaque within blood vessels property analysis method based on depth migration study - Google Patents
A kind of plaque within blood vessels property analysis method based on depth migration study Download PDFInfo
- Publication number
- CN107730497A CN107730497A CN201711024432.6A CN201711024432A CN107730497A CN 107730497 A CN107730497 A CN 107730497A CN 201711024432 A CN201711024432 A CN 201711024432A CN 107730497 A CN107730497 A CN 107730497A
- Authority
- CN
- China
- Prior art keywords
- mrow
- plaque
- blood vessels
- intravascular
- msup
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10101—Optical tomography; Optical coherence tomography [OCT]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10132—Ultrasound image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30101—Blood vessel; Artery; Vein; Vascular
-
- 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
- G06T2207/30204—Marker
Abstract
A kind of plaque within blood vessels property analysis method based on depth migration study, belongs to field of medical image processing, and in particular to a kind of plaque within blood vessels property analysis method.Present invention clinical modalities first obtain multi-modal intravascular image data, the attribute of handmarking's plaque within blood vessels, then the intravascular image after mark is pre-processed, input using intravascular image data after pretreatment as depth convolutional neural networks, depth convolutional neural networks training is carried out using the transfer learning mode for having supervision, and using the study that network parameter is carried out by the stochastic gradient descent mode based on back-propagation algorithm；Then ballot fusion forecasting is carried out to the intravascular image obtained in step 1 using the forecast model of the multiple cross-module states trained, ultimately produces patch species probability collection of illustrative plates.The present invention solves the problems, such as that existing method labor intensive is more, it is low artificial difference, speed to be present.The present invention can operate with blood-vessel image processing.
Description
Technical field
The invention belongs to field of medical image processing, and in particular to a kind of plaque within blood vessels property analysis method.
Background technology
Medical Image Processing is rapid with the development and ripe and clinical diagnosis technology progress of computer technology
The new branch of science and technology to grow up, nowadays the application clinically of Medical Image Processing is more and more extensive.
Cardiovascular and cerebrovascular disease is a kind of one of disease of fatal rate highest, therefore is increasingly taken seriously in clinic diagnosis, while heart and brain
The related technology of vascular diseases also turns into the focus and difficult point of field of medical image processing technical research.Injury of blood vessel is common
A kind of vascular diseases, it is dangerous as caused by cerebral thrombus, cerebral hemorrhage, acute myocardial infarction AMI etc. are all due to blood vessel breakage or damage
The high acute disease of property.It is characterized in that morbidity is urgent, it is predictable weak.The blood vessel of patient can be successfully predicted in premorbid
Vulnerability level, and take positive defensive measure, it will save the life of many people.
Now, clinically doctor carries out the prediction of blood vessel vulnerability level and depends on the intravascular image data spy of manual analysis
Point, in conjunction with other coherent detection data, blood vessel vulnerability level is calculated.Wherein analyze what intravascular image data obtained
Discriminant criterion includes：Vascular morphology, pipe diameter size, tube wall whether have patch and patch composition and size (specifically including is
It is no including calcified plaque, Lipid Plaque, fibroid patch, with the Lipid Plaque that is emitted compared with thin fiber and various mixing patches
Deng).Wherein, plaque within blood vessels attributive analysis belongs to one of most important data target therein, while is also the difficult point of analysis.
Nowadays plaque within blood vessels attributive analysis depends on manual analysis, is mainly had by the way of manual analysis as follows
Shortcoming：(1) a large amount of manpowers are expended；(2) artificial difference be present, i.e., same data, the results of measuring of different people is different；(3)
Speed is low.How to complete plaque within blood vessels attributive analysis for the Accurate Prediction that intravascular image data automates is cardiovascular and cerebrovascular
The problem of field urgent need to resolve.
The content of the invention
The present invention is solves the problems, such as that existing method expends a large amount of manpowers, artificial difference, speed to be present low, there is provided one kind
Plaque within blood vessels property analysis method based on depth migration study.
A kind of plaque within blood vessels property analysis method based on depth migration study of the present invention, passes through following technical side
Case is realized：
Step 1: clinical modalities obtain multi-modal intravascular image data, ivus image and blood vessel are specifically included
Interior OCT image, wherein the multi-modal blood vessel image for referring to different imaging means and obtaining, OCT be optical coherence tomography into
Picture；
Step 2: the attribute of handmarking's plaque within blood vessels：Doctor is special according to the image of the intravascular image data of acquisition
Sign, judge plaque within blood vessels attribute, plaque within blood vessels are classified, and image is marked；
Step 3: the intravascular image after mark is pre-processed；
Step 4: using the intravascular image data after step 3 is handled as the input of depth convolutional neural networks, adopt
Depth convolutional neural networks training is carried out with the transfer learning mode for having supervision, and using by based on the random of back-propagation algorithm
Gradient declines the study that mode carries out network parameter；
Step 5: the intravascular image obtained in step 1 is carried out using the forecast model of the multiple cross-module states trained
Ballot fusion forecasting, and export the probability of each plaque within blood vessels classification；
Step 6: generation patch species probability collection of illustrative plates, is analyzed by visual mode secondary vessel patch.
The present invention compared with prior art, it is the most prominent the characteristics of and significant beneficial effect be：
The present invention is to obtain intravascular image data by intravascular ultrasound imaging and intravascular OCT image, and image is entered
In row preprocessing process, denoising not only is carried out to image, multiple dimensioned and multiple types filtering operation also is carried out to image, from
And highlight more abundant picture material；Then the initialization of model is carried out by the way of transfer learning, then use has supervision
The method of deep learning the great amount of samples of handmarking is trained end to end, based on the deep learning trained
Model, the probability analysis to plaque within blood vessels classification is realized, and using multiple deep learning models of different modalities image data
Prediction result carry out ballot fusion.Obtain the final plaque within blood vessels class probability analysis result with higher robustness.From
The experiment of 1000 sample analyses of progress, which can be seen that each sample mean of handmarking's mode, to be needed to expend 1~2 point of doctor
Clock, and every doctor, due to relying on experience and handmarking, error probability mean height reaches 7%, automation of the invention
Analysis is completed, can more preferably liberate the labour of clinician, and provide the data of more accurate quick, embodiment medium vessels
Interior patch is classified and analyzes output patch species probability collection of illustrative plates can complete in 20S.Certain this technology can not only be applied
In intravascular ultrasound and intravascular OCT images, it can also be extended to based on other mode, such as CTA (CT angiographies), MRI
Angiogram, DSA etc. plaque within blood vessels attributive analysis.Compared with conventional method, plaque within blood vessels attribute point of the invention
Analysis method quick, accurate, automation can obtain the probability of patch classification, be doctor when carrying out the prediction of blood vessel vulnerability level
One important data target is provided, enables the clinician to carry out the prediction of blood vessel vulnerability level in conjunction with other clinical datas, carries out
The prevention work ahead of time of major disease.
Brief description of the drawings
Fig. 1 is the inventive method flow chart；
Fig. 2 is depth convolutional neural networks schematic diagram of the present invention；
Fig. 3 is the ivus image that quantity containing patch is few in the embodiment of the present invention；
Fig. 4 is the ivus image more than quantity containing patch in the embodiment of the present invention；
Fig. 5 is the intravascular OCT image that quantity containing patch is few in the embodiment of the present invention；
Fig. 6 is the intravascular OCT image more than quantity containing patch in the embodiment of the present invention.
Embodiment
Embodiment one：
A kind of as shown in figure 1, plaque within blood vessels attributive analysis side based on depth migration study that present embodiment provides
Method, specifically follow the steps below：
Step 1: clinical modalities obtain multi-modal intravascular image data, ivus image and blood vessel are specifically included
Interior OCT image, wherein the multi-modal blood vessel image for referring to different imaging means and obtaining, OCT be optical coherence tomography into
Picture；
Step 2: the attribute of handmarking's plaque within blood vessels：Doctor is special according to the image of the intravascular image data of acquisition
Sign, judge plaque within blood vessels attribute, plaque within blood vessels are classified, plaque within blood vessels include calcified plaque, Lipid Plaque, fibroid
Patch, adjoint Lipid Plaque emitted compared with thin fiber and various mixing patches etc.；And image is marked, in order to improve number
According to mark quality, same intravascular image data will carry out ballot determination according to the mark of multidigit expert；
Step 3: being pre-processed to the intravascular image after mark, the data of processing include the blood under cartesian coordinate
Image data in pipe, and the intravascular image data of polar coordinates deployed according to central point；
Step 4: using the intravascular image data after step 3 is handled as the input of depth convolutional neural networks, adopt
Depth convolutional neural networks training is carried out with the transfer learning mode for having supervision, ivus image and intravascular OCT are schemed
As being respectively trained, and using the study that network parameter is carried out by the stochastic gradient descent mode based on back-propagation algorithm；
Step 5: the intravascular image obtained in step 1 is carried out using the forecast model of the multiple cross-module states trained
Ballot fusion forecasting, increases the robustness of forecast model, and export the probability of each plaque within blood vessels classification；
Step 6: generation patch species probability collection of illustrative plates, is analyzed by visual mode secondary vessel patch.
Embodiment two：
Present embodiment is unlike embodiment one：Intravascular image is pre-processed described in step 3
Process specifically include following steps：
Step 3 one, the intravascular image after the mark of 8 gray scales is converted into 24 triple channels, using replicating passage
Mode realizes that this step operation is to realize that the parameter of network first tier migrates, ensure the image channel number and nature of input
Image channel number is consistent；
Step 3 two, using denoising method for acoustic to step 3 one convert after ivus image and intravascular OCT
View data carries out Denoising disposal；
Step 3 three, the ivus image after Denoising disposal and OCT image data are carried out it is multiple dimensioned
Filtering, the information included in the multiple dimensioned intravascular image data of expression, excavate the inherent feature of intravascular image data.
Embodiment three：
Present embodiment is unlike embodiment two：Denoising method for acoustic is preferably average described in step 3 two
Filtering method or Gaussian smoothing filter method.
Embodiment four：
Present embodiment is unlike embodiment two：Filtering method is preferably La Pula described in step 3 three
This filtering method or gal cypress (Gabor) filtering method.
Embodiment five：
Present embodiment is unlike embodiment one or two：There is the transfer learning side of supervision described in step 4
Formula carries out the training of depth convolutional neural networks and specifically includes following steps：
Step 4 one, in extensive natural scene image classification task carry out model pre-training (this training method
More classification problems may be directed to), obtain pre-training model parameter；
It is step 4 two, every in addition to last full articulamentum with pre-training model initialization depth convolutional neural networks
The parameter of layer, wherein, because the classification quantity of plaque within blood vessels is different from natural scene image classification quantity, cause last
The network parameter number of full articulamentum is different, therefore the network parameter initialized does not include the parameter of last full articulamentum,
The parameter of last full articulamentum is initialized by the way of Gauss equality initialization.Tuning instruction is carried out on this basis
Practice；This mode can allow depth learning technology to be used to greatest extent in the case where data set is limited；In certain journey
On degree, can solve the problems, such as due to the few caused model over-fitting of medical image training data.
Step 4 three, using the learning method for having supervision forecast model is trained, carried out by back-propagation algorithm
The solution of model；Individually trained for both modalities which (intravascular ultrasound and intravascular OCT image), obtain plaque within blood vessels classification
Probabilistic Prediction Model.
Embodiment six：
Present embodiment is unlike embodiment five：
Depth convolutional neural networks described in step 4 two include input layer, convolutional layer, Batch-Normalization
Layer, Pooling layers, dropout layers, full articulamentum and output layer.The decision design of depth convolutional neural networks as shown in Fig. 2
But being not limited to the network structure of the models such as network structure as shown in Figure 2, VGG, GoogLeNet, ResNet can apply
In the construction and training process of model.
Embodiment seven：
Present embodiment is unlike embodiment six：
The learning method for having supervision described in step 4 three is trained specially to forecast model：
Step 431, using the intravascular image data obtained in step 1 as sample set (X, Y), therein i-th
Individual training sample is (Xi,Yi)；I represents the sequence number of training sample, X representing input images arrangement sets, and Y represents preferable output, Xi
Represent i-th of sample in image sequence set, YiRepresent the prediction result of i-th of sample in image sequence set；
Step 4 three or two, the reality output O for calculating depth convolutional neural networks, in this stage, information makees X from input layer
For input, by conversion step by step, output layer is sent to；In the process, network perform calculating be：O=Fn(…(F2(F1
(X*W1)W2)…)Wn)；Wherein FnRepresent the level of network, W1Represent F1Layer weight matrix, W2Represent F2Layer weight matrix,
Wn represents FnLayer weight matrix；
Optimize every layer of weights using back-propagation algorithm, calculate loss letters of the reality output O with corresponding preferable output Y
Number loss；By the method backpropagation adjustment weight matrix Wn of minimization error；
Loss specific formula is as follows：
Wherein m represents the total number of training sample, and i ∈ { 1 ..., m }, j represent patch attribute classification sequence number, j ∈ 1 ...,
L }, l represents patch attribute classification sum；
Wherein function Fj(Xi) calculation formula it is as follows：
Wherein S is intermediate variable, fj(Xi) calculation formula it is as follows：
fj(Xi)=P (Yi=j | Xi；θ)
P(Yi=j | Xi；Network θ) is represented in the case of parameter θ, XiFor the output probability value of jth class patch.
Embodiment
Beneficial effects of the present invention are verified using following examples：
As shown in Fig. 2,3,4,5,6, from the intravascular image data of multiple modalities for clinically gathering multiple patients.In order to instruct
More accurately index forecast model is practised, chooses the total number m=1000, i of sample image, i.e. training sample that quantity is 1000
∈{1,…,1000}。
Step 1: using the intravascular image data of the multi-modal different patch degree as shown in Fig. 3,4,5,6 as input；
Step 2: the multi-modal intravascular image data that 8 experts of tissue obtain to clinic carries out plaque within blood vessels classification,
Calcified plaque, Lipid Plaque, fibroid patch, adjoint four kinds of the Lipid Plaque emitted compared with thin fiber are specifically divided into, and it is artificial respectively
Mark 1,2,3,4；Patch attribute classification sum l=4 i.e. in the present embodiment, j represent patch attribute classification sequence number, j ∈ 1 ...,
4 },；In order to improve the mark quality of data, same intravascular image data according to the mark of 8 experts will vote really
It is fixed；
Step 3: 8 intravascular image datas of gray scale of single channel are converted into 24 triple channel view data.Using denoising
Method for acoustic carries out Denoising disposal to multi-modal intravascular image data, and to carrying out the multi-modal intravascular of Denoising disposal
Image data carries out multiple dimensioned filtering, the information included in the multiple dimensioned intravascular image data of expression, digs to greatest extent
The inherent feature of multi-modal intravascular image data is dug, completes data prediction；
Step 4: as shown in figure 1, on the basis of above-mentioned steps, by the intravascular image data after step 3 is handled
As the input of depth convolutional neural networks, the transfer learning mode for carrying out supervision carries out deep learning network training.
Step 5: as shown in figure 1, after the end-to-end training of deep learning network by step 4, the spot that learns
Block's attribute forecast model has higher forecasting accuracy, and the sample obtained to step 1 carries out ballot fusion forecasting, output class
The probability of other 1 (calcified plaque) is 0.1, and the probability of classification 2 (Lipid Plaque) is 0.3, and the probability of classification 3 (fibroid patch) is
0.6, the probability of classification 4 (with the Lipid Plaque emitted compared with thin fiber) is 0.
Step 6: ultimately produce patch species probability collection of illustrative plates.
The present invention can also have other various embodiments, in the case of without departing substantially from spirit of the invention and its essence, this area
Technical staff works as can make various corresponding changes and deformation according to the present invention, but these corresponding changes and deformation should all belong to
The protection domain of appended claims of the invention.
Claims (7)
1. it is a kind of based on depth migration study plaque within blood vessels property analysis method, it is characterised in that methods described include with
Lower step：
Step 1: clinical modalities obtain multi-modal intravascular image data, ivus image and intravascular OCT are specifically included
Image, wherein the multi-modal blood vessel image for referring to different imaging means and obtaining, OCT is optical coherence tomography；
Step 2: the attribute of handmarking's plaque within blood vessels：Doctor sentences according to the image feature of the intravascular image data of acquisition
Determine plaque within blood vessels attribute, plaque within blood vessels are classified, and image is marked；
Step 3: the intravascular image after mark is pre-processed；
Step 4: using the intravascular image data after step 3 is handled as the input of depth convolutional neural networks, using having
The transfer learning mode of supervision carries out depth convolutional neural networks training, and using by the stochastic gradient based on back-propagation algorithm
Decline mode carries out the study of network parameter；
Step 5: the intravascular image obtained in step 1 is voted using the forecast model of the multiple cross-module states trained
Fusion forecasting, and export the probability of each plaque within blood vessels classification；
Step 6: generation patch species probability collection of illustrative plates, is analyzed by visual mode secondary vessel patch.
2. a kind of plaque within blood vessels property analysis method based on depth migration study according to claim 1, its feature
It is, the process pre-processed described in step 3 to intravascular image specifically includes following steps：
Step 3 one, the intravascular image after the mark of 8 gray scales is converted into 24 triple channels, by the way of passage is replicated
Realize；
Step 3 two, using denoising method for acoustic to step 3 one convert after ivus image and intravascular OCT image
Data carry out Denoising disposal；
Step 3 three, multiple dimensioned filter is carried out to the ivus image after Denoising disposal and OCT image data
Ripple, the multiple dimensioned inherent feature expressed the information included in intravascular image data, excavate intravascular image data.
3. a kind of plaque within blood vessels property analysis method based on depth migration study according to claim 2, its feature
It is, denoising method for acoustic described in step 3 two is preferably mean filter method or Gaussian smoothing filter method.
4. a kind of plaque within blood vessels property analysis method based on depth migration study according to claim 2, its feature
It is, filtering method described in step 3 three is preferably Laplce's filtering method or gabor filtering method.
5. a kind of plaque within blood vessels property analysis method based on depth migration study according to claim 1 or 2, it is special
Sign is have the transfer learning mode of supervision to carry out the training of depth convolutional neural networks described in step 4 and specifically include following step
Suddenly：
Step 4 one, the pre-training for carrying out in extensive natural scene image classification task model, obtain pre-training model ginseng
Number；
Step 4 two, every layer with pre-training model initialization depth convolutional neural networks in addition to last full articulamentum
Parameter, tuning training is carried out on this basis；
Step 4 three, using the learning method for having supervision forecast model is trained；Model is carried out by back-propagation algorithm
Solution, obtain the Probabilistic Prediction Model of plaque within blood vessels classification.
6. a kind of plaque within blood vessels property analysis method based on depth migration study according to claim 5, its feature
Be, depth convolutional neural networks described in step 4 two include input layer, convolutional layer, Batch-Normalization layers,
Pooling layers, dropout layers, full articulamentum and output layer.
7. a kind of plaque within blood vessels property analysis method based on depth migration study according to claim 6, its feature
It is, the learning method for having supervision described in step 4 three is trained specially to forecast model：
Step 431, using the intravascular image data obtained in step 1 as sample set (X, Y), i-th of instruction therein
It is (X to practice samplei,Yi)；I represents the sequence number of training sample, X representing input images arrangement sets, and Y represents preferable output, XiRepresent
I-th of sample in image sequence set, YiRepresent the prediction result of i-th of sample in image sequence set；
Step 4 three or two, the reality output O for calculating depth convolutional neural networks, in this stage, information is from input layer using X as defeated
Enter, by conversion step by step, be sent to output layer；In the process, network perform calculating be：O=Fn(…(F2(F1(X*
W1)W2)…)Wn)；Wherein FnRepresent the level of network, W1Represent F1Layer weight matrix, W2Represent F2Layer weight matrix, Wn
Represent FnLayer weight matrix；
Optimize every layer of weights using back-propagation algorithm, calculate loss functions of the reality output O with corresponding preferable output Y
loss；By the method backpropagation adjustment weight matrix Wn of minimization error；
Loss specific formula is as follows：
<mrow>
<mi>l</mi>
<mi>o</mi>
<mi>s</mi>
<mi>s</mi>
<mo>=</mo>
<mo>-</mo>
<mfrac>
<mn>1</mn>
<mi>m</mi>
</mfrac>
<mo>&lsqb;</mo>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>m</mi>
</munderover>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>j</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>l</mi>
</munderover>
<mo>{</mo>
<msup>
<mi>Y</mi>
<mi>i</mi>
</msup>
<mo>=</mo>
<mi>j</mi>
<mo>}</mo>
<mi>log</mi>
<mi> </mi>
<msub>
<mi>F</mi>
<mi>j</mi>
</msub>
<mrow>
<mo>(</mo>
<msup>
<mi>X</mi>
<mi>i</mi>
</msup>
<mo>)</mo>
</mrow>
<mo>&rsqb;</mo>
</mrow>
Wherein m represents the total number of training sample, and i ∈ { 1 ..., m }, j represent patch attribute classification sequence number, j ∈ { 1 ..., l }, l
Represent patch attribute classification sum；
Wherein function Fj(Xi) calculation formula it is as follows：
<mrow>
<msub>
<mi>F</mi>
<mi>j</mi>
</msub>
<mrow>
<mo>(</mo>
<msup>
<mi>X</mi>
<mi>i</mi>
</msup>
<mo>)</mo>
</mrow>
<mo>=</mo>
<msup>
<mi>e</mi>
<mrow>
<msub>
<mi>f</mi>
<mi>j</mi>
</msub>
<mrow>
<mo>(</mo>
<msup>
<mi>X</mi>
<mi>i</mi>
</msup>
<mo>)</mo>
</mrow>
</mrow>
</msup>
<mo>/</mo>
<mi>S</mi>
</mrow>
<mrow>
<mi>S</mi>
<mo>=</mo>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>j</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>l</mi>
</munderover>
<msup>
<mi>e</mi>
<mrow>
<msub>
<mi>f</mi>
<mi>l</mi>
</msub>
<mrow>
<mo>(</mo>
<msup>
<mi>X</mi>
<mi>i</mi>
</msup>
<mo>)</mo>
</mrow>
</mrow>
</msup>
</mrow>
Wherein S is intermediate variable, fj(Xi) calculation formula it is as follows：
fj(Xi)=P (Yi=j | Xi；θ)
P(Yi=j | Xi；Network θ) is represented in the case of parameter θ, XiFor the output probability value of jth class patch.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711024432.6A CN107730497B (en) | 2017-10-27 | 2017-10-27 | Intravascular plaque attribute analysis method based on deep migration learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711024432.6A CN107730497B (en) | 2017-10-27 | 2017-10-27 | Intravascular plaque attribute analysis method based on deep migration learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107730497A true CN107730497A (en) | 2018-02-23 |
| CN107730497B CN107730497B (en) | 2021-09-10 |
Family
ID=61202734
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711024432.6A Active CN107730497B (en) | 2017-10-27 | 2017-10-27 | Intravascular plaque attribute analysis method based on deep migration learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107730497B (en) |
Cited By (22)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108416769A (en) * | 2018-03-02 | 2018-08-17 | 成都斯斐德科技有限公司 | Based on pretreated IVOCT images vulnerable plaque automatic testing method |
| CN108416755A (en) * | 2018-03-20 | 2018-08-17 | 南昌航空大学 | A kind of image de-noising method and system based on deep learning |
| CN108542390A (en) * | 2018-03-07 | 2018-09-18 | 清华大学 | Vascular plaque ingredient recognition methods based on more contrast nuclear magnetic resonance images |
| CN108734208A (en) * | 2018-05-15 | 2018-11-02 | 重庆大学 | Multi-source heterogeneous data fusion system based on multi-modal depth migration study mechanism |
| CN108899075A (en) * | 2018-06-28 | 2018-11-27 | 众安信息技术服务有限公司 | A kind of DSA image detecting method, device and equipment based on deep learning |
| CN108932720A (en) * | 2018-07-24 | 2018-12-04 | 深圳先进技术研究院 | A kind of the encephalic patch and carotid plaques intelligent scissor method of magnetic resonance image |
| CN108985173A (en) * | 2018-06-19 | 2018-12-11 | 奕通信息科技（上海）股份有限公司 | Towards the depth network migration learning method for having the label apparent age data library of noise |
| CN109064442A (en) * | 2018-06-21 | 2018-12-21 | 上海遥思企业管理中心 | Vascular pressure difference modification method, device and equipment |
| CN109124764A (en) * | 2018-09-29 | 2019-01-04 | 上海联影医疗科技有限公司 | Guide device of performing the operation and surgery systems |
| CN109447940A (en) * | 2018-08-28 | 2019-03-08 | 天津医科大学肿瘤医院 | Convolutional neural networks training method, ultrasound image recognition positioning method and system |
| CN109671065A (en) * | 2018-12-14 | 2019-04-23 | 首都医科大学附属北京天坛医院 | A kind of method and system for taking bolt to judge based on head CT images |
| CN110223781A (en) * | 2019-06-03 | 2019-09-10 | 中国医科大学附属第一医院 | A kind of various dimensions plaque rupture Warning System |
| CN110215232A (en) * | 2019-04-30 | 2019-09-10 | 南方医科大学南方医院 | Ultrasonic patch analysis method in coronary artery based on algorithm of target detection |
| CN110310271A (en) * | 2019-07-01 | 2019-10-08 | 无锡祥生医疗科技股份有限公司 | Property method of discrimination, storage medium and the Vltrasonic device of carotid plaques |
| CN110428417A (en) * | 2019-08-13 | 2019-11-08 | 无锡祥生医疗科技股份有限公司 | Property method of discrimination, storage medium and the Vltrasonic device of carotid plaques |
| WO2020007277A1 (en) * | 2018-07-05 | 2020-01-09 | 北京推想科技有限公司 | Cerebral hemorrhage amount calculation method based on deep learning |
| CN110827255A (en) * | 2019-10-31 | 2020-02-21 | 杨本强 | Plaque stability prediction method and system based on coronary artery CT image |
| CN111768403A (en) * | 2020-07-09 | 2020-10-13 | 成都全景恒升科技有限公司 | Calcified plaque detection decision-making system and device based on artificial intelligence algorithm |
| CN111862009A (en) * | 2020-07-02 | 2020-10-30 | 清华大学深圳国际研究生院 | Classification method of fundus OCT images and computer-readable storage medium |
| US10970604B2 (en) | 2018-09-27 | 2021-04-06 | Industrial Technology Research Institute | Fusion-based classifier, classification method, and classification system |
| CN116740041A (en) * | 2023-06-27 | 2023-09-12 | 新疆生产建设兵团医院 | CTA scanning image analysis system and method based on machine vision |
| RU2805645C1 (en) * | 2022-12-07 | 2023-10-23 | Федеральное государственное бюджетное научное учреждение "Томский национальный исследовательский медицинский центр Российской академии наук" (Томский НИМЦ) | Method for performing intravascular optical coherence tomography of renal arteries |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1929781A (en) * | 2003-08-21 | 2007-03-14 | 依斯克姆公司 | Automated methods and systems for vascular plaque detection and analysis |
| CN101799864A (en) * | 2010-01-15 | 2010-08-11 | 北京工业大学 | Automatic identifying method of artery plaque type based on ultrasonic image in blood vessel |
| CN104376549A (en) * | 2014-11-20 | 2015-02-25 | 华北电力大学（保定） | Intravascular ultrasound image and intravascular-OCT image fusing method |
| CN104794708A (en) * | 2015-04-10 | 2015-07-22 | 浙江工业大学 | Atherosclerosis plaque composition dividing method based on multi-feature learning |
| CN105877910A (en) * | 2015-01-16 | 2016-08-24 | 上海交通大学 | Integrated system for accurate diagnosis and treatment of hardened blood vessels or tumors |
| CN106157312A (en) * | 2016-07-05 | 2016-11-23 | 董超超 | A kind of image display device registrated by Patch properties |
| CN106295584A (en) * | 2016-08-16 | 2017-01-04 | 深圳云天励飞技术有限公司 | Depth migration study is in the recognition methods of crowd's attribute |
| CN106991439A (en) * | 2017-03-28 | 2017-07-28 | 南京天数信息科技有限公司 | Image-recognizing method based on deep learning and transfer learning |
| US9767557B1 (en) * | 2016-06-23 | 2017-09-19 | Siemens Healthcare Gmbh | Method and system for vascular disease detection using recurrent neural networks |
-
2017
- 2017-10-27 CN CN201711024432.6A patent/CN107730497B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1929781A (en) * | 2003-08-21 | 2007-03-14 | 依斯克姆公司 | Automated methods and systems for vascular plaque detection and analysis |
| CN101799864A (en) * | 2010-01-15 | 2010-08-11 | 北京工业大学 | Automatic identifying method of artery plaque type based on ultrasonic image in blood vessel |
| CN104376549A (en) * | 2014-11-20 | 2015-02-25 | 华北电力大学（保定） | Intravascular ultrasound image and intravascular-OCT image fusing method |
| CN105877910A (en) * | 2015-01-16 | 2016-08-24 | 上海交通大学 | Integrated system for accurate diagnosis and treatment of hardened blood vessels or tumors |
| CN104794708A (en) * | 2015-04-10 | 2015-07-22 | 浙江工业大学 | Atherosclerosis plaque composition dividing method based on multi-feature learning |
| US9767557B1 (en) * | 2016-06-23 | 2017-09-19 | Siemens Healthcare Gmbh | Method and system for vascular disease detection using recurrent neural networks |
| CN106157312A (en) * | 2016-07-05 | 2016-11-23 | 董超超 | A kind of image display device registrated by Patch properties |
| CN106295584A (en) * | 2016-08-16 | 2017-01-04 | 深圳云天励飞技术有限公司 | Depth migration study is in the recognition methods of crowd's attribute |
| CN106991439A (en) * | 2017-03-28 | 2017-07-28 | 南京天数信息科技有限公司 | Image-recognizing method based on deep learning and transfer learning |
Non-Patent Citations (7)
| Title |
|---|
| KARIM LEKADIR 等: "A Convolutional Neural Network for Automatic Characterization of Plaque Composition in Carotid Ultrasound", 《IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATION》 * |
| XIANG LI 等: "Integrated IVUS-OCT Imaging for Atherosclerotic Plaque Characterization", 《IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS》 * |
| 周俊宇 等: "卷积神经网络在图像分类和目标检测应用综述", 《计算机工程与应用》 * |
| 宋光慧: "基于迁移学习与深度卷积特征的图像标注方法研究", 《中国博士学位论文全文数据库 信息科技辑》 * |
| 王晓斌 等: "基于优化卷积神经网络结构的交通标志识别", 《计算机应用》 * |
| 赵媛 等: "一种基于深度学习的颈动脉斑块超声图像识别方法", 《中国医疗器械信息》 * |
| 陶攀 等: "基于深度学习的超声心动图切面识别方法", 《计算机应用》 * |
Cited By (28)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108416769B (en) * | 2018-03-02 | 2021-06-04 | 成都斯斐德科技有限公司 | IVOCT image vulnerable plaque automatic detection method based on preprocessing |
| CN108416769A (en) * | 2018-03-02 | 2018-08-17 | 成都斯斐德科技有限公司 | Based on pretreated IVOCT images vulnerable plaque automatic testing method |
| CN108542390A (en) * | 2018-03-07 | 2018-09-18 | 清华大学 | Vascular plaque ingredient recognition methods based on more contrast nuclear magnetic resonance images |
| CN108416755A (en) * | 2018-03-20 | 2018-08-17 | 南昌航空大学 | A kind of image de-noising method and system based on deep learning |
| CN108416755B (en) * | 2018-03-20 | 2020-06-30 | 南昌航空大学 | Image denoising method and system based on deep learning |
| CN108734208A (en) * | 2018-05-15 | 2018-11-02 | 重庆大学 | Multi-source heterogeneous data fusion system based on multi-modal depth migration study mechanism |
| CN108985173A (en) * | 2018-06-19 | 2018-12-11 | 奕通信息科技（上海）股份有限公司 | Towards the depth network migration learning method for having the label apparent age data library of noise |
| CN109064442A (en) * | 2018-06-21 | 2018-12-21 | 上海遥思企业管理中心 | Vascular pressure difference modification method, device and equipment |
| CN108899075A (en) * | 2018-06-28 | 2018-11-27 | 众安信息技术服务有限公司 | A kind of DSA image detecting method, device and equipment based on deep learning |
| WO2020007277A1 (en) * | 2018-07-05 | 2020-01-09 | 北京推想科技有限公司 | Cerebral hemorrhage amount calculation method based on deep learning |
| CN108932720A (en) * | 2018-07-24 | 2018-12-04 | 深圳先进技术研究院 | A kind of the encephalic patch and carotid plaques intelligent scissor method of magnetic resonance image |
| CN109447940B (en) * | 2018-08-28 | 2021-09-28 | 天津医科大学肿瘤医院 | Convolutional neural network training method, ultrasonic image identification and positioning method and system |
| CN109447940A (en) * | 2018-08-28 | 2019-03-08 | 天津医科大学肿瘤医院 | Convolutional neural networks training method, ultrasound image recognition positioning method and system |
| US10970604B2 (en) | 2018-09-27 | 2021-04-06 | Industrial Technology Research Institute | Fusion-based classifier, classification method, and classification system |
| CN109124764B (en) * | 2018-09-29 | 2020-07-14 | 上海联影医疗科技有限公司 | Surgical guide device and surgical system |
| CN109124764A (en) * | 2018-09-29 | 2019-01-04 | 上海联影医疗科技有限公司 | Guide device of performing the operation and surgery systems |
| CN109671065A (en) * | 2018-12-14 | 2019-04-23 | 首都医科大学附属北京天坛医院 | A kind of method and system for taking bolt to judge based on head CT images |
| CN110215232A (en) * | 2019-04-30 | 2019-09-10 | 南方医科大学南方医院 | Ultrasonic patch analysis method in coronary artery based on algorithm of target detection |
| CN110223781A (en) * | 2019-06-03 | 2019-09-10 | 中国医科大学附属第一医院 | A kind of various dimensions plaque rupture Warning System |
| CN110310271A (en) * | 2019-07-01 | 2019-10-08 | 无锡祥生医疗科技股份有限公司 | Property method of discrimination, storage medium and the Vltrasonic device of carotid plaques |
| CN110310271B (en) * | 2019-07-01 | 2023-11-24 | 无锡祥生医疗科技股份有限公司 | Carotid plaque property discriminating method, storage medium and ultrasonic device |
| CN110428417A (en) * | 2019-08-13 | 2019-11-08 | 无锡祥生医疗科技股份有限公司 | Property method of discrimination, storage medium and the Vltrasonic device of carotid plaques |
| CN110827255A (en) * | 2019-10-31 | 2020-02-21 | 杨本强 | Plaque stability prediction method and system based on coronary artery CT image |
| CN111862009A (en) * | 2020-07-02 | 2020-10-30 | 清华大学深圳国际研究生院 | Classification method of fundus OCT images and computer-readable storage medium |
| CN111862009B (en) * | 2020-07-02 | 2024-01-16 | 清华大学深圳国际研究生院 | Classifying method of fundus OCT (optical coherence tomography) images and computer readable storage medium |
| CN111768403A (en) * | 2020-07-09 | 2020-10-13 | 成都全景恒升科技有限公司 | Calcified plaque detection decision-making system and device based on artificial intelligence algorithm |
| RU2805645C1 (en) * | 2022-12-07 | 2023-10-23 | Федеральное государственное бюджетное научное учреждение "Томский национальный исследовательский медицинский центр Российской академии наук" (Томский НИМЦ) | Method for performing intravascular optical coherence tomography of renal arteries |
| CN116740041A (en) * | 2023-06-27 | 2023-09-12 | 新疆生产建设兵团医院 | CTA scanning image analysis system and method based on machine vision |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107730497B (en) | 2021-09-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107730497A (en) | A kind of plaque within blood vessels property analysis method based on depth migration study | |
| CN109447940B (en) | Convolutional neural network training method, ultrasonic image identification and positioning method and system | |
| CN110464380A (en) | A kind of method that the ultrasound cross-section image of the late pregnancy period fetus of centering carries out quality control | |
| JP6522161B2 (en) | Medical data analysis method based on deep learning and intelligent analyzer thereof | |
| CN108257135A (en) | The assistant diagnosis system of medical image features is understood based on deep learning method | |
| CN106980899A (en) | The deep learning model and system of flow characteristic on prediction vascular tree blood flow paths | |
| CN111223553A (en) | Two-stage deep migration learning traditional Chinese medicine tongue diagnosis model | |
| CN106096632A (en) | Based on degree of depth study and the ventricular function index prediction method of MRI image | |
| CN106056126A (en) | Plaque vulnerability assessment in medical imaging | |
| CN107832737A (en) | Electrocardiogram interference identification method based on artificial intelligence | |
| CN106296699A (en) | Cerebral tumor dividing method based on deep neural network and multi-modal MRI image | |
| CN109925002A (en) | Artificial intelligence echocardiogram data collection system and its collecting method | |
| CN103116707A (en) | Heart disease intelligent diagnostic method based on case reasoning | |
| CN109215040B (en) | Breast tumor segmentation method based on multi-scale weighted learning | |
| CN110503635B (en) | Hand bone X-ray film bone age assessment method based on heterogeneous data fusion network | |
| CN109192305A (en) | A kind of heart function automatic analysis method based on deep-cycle neural network | |
| CN110148466A (en) | A kind of heart impact signal atrial fibrillation computer aided diagnosing method based on transfer learning | |
| CN105868572A (en) | Method for predicting myocardial ischemia position on basis of self-encoder | |
| CN108665449A (en) | Predict the deep learning model and device of the flow characteristic in blood flow vector path | |
| Cheng et al. | Atrial fibrillation identification with PPG signals using a combination of time-frequency analysis and deep learning | |
| CN107665737A (en) | Vascular wall stress-strain state acquisition methods, computer-readable medium and system | |
| CN111462082A (en) | Focus picture recognition device, method and equipment and readable storage medium | |
| CN104887241A (en) | Shimmer based cerebral thrombosis patient movement function rehabilitation assessment method | |
| Sehly et al. | Artificial Intelligence in echocardiography: the time is now | |
| Abbasi et al. | Automatic brain ischemic stroke segmentation with deep learning: A review |
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 | ||
| CB03 | Change of inventor or designer information | ||
| CB03 | Change of inventor or designer information |
Inventor after: Luo Gongning Inventor after: Dong Suyu Inventor after: Wang Wei Inventor after: Shu Lei Inventor after: Wang Kuanquan Inventor after: Zhang Henggui Inventor before: Wang Kuanquan Inventor before: Luo Gongning Inventor before: Dong Suyu Inventor before: Shu Lei Inventor before: Zhang Henggui |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20211215 Address after: 150000 room A18, Chuang a district, 3rd floor, building co7, Jiangbei District C, Harbin University of Commerce, Songbei District, Harbin City, Heilongjiang Province Patentee after: Harbin Hongtu Yuanchi Technology Co.,Ltd. Address before: 150001 No. 92 West straight street, Nangang District, Heilongjiang, Harbin Patentee before: HARBIN INSTITUTE OF TECHNOLOGY |