CN108492297A - The MRI brain tumors positioning for cascading convolutional network based on depth and dividing method in tumor - Google Patents
The MRI brain tumors positioning for cascading convolutional network based on depth and dividing method in tumor Download PDFInfo
- Publication number
- CN108492297A CN108492297A CN201810300057.1A CN201810300057A CN108492297A CN 108492297 A CN108492297 A CN 108492297A CN 201810300057 A CN201810300057 A CN 201810300057A CN 108492297 A CN108492297 A CN 108492297A
- Authority
- CN
- China
- Prior art keywords
- tumor
- convolutional
- layer
- classification
- depth
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- 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
- G06T2207/10088—Magnetic resonance imaging [MRI]
-
- 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
- G06T2207/20021—Dividing image into blocks, subimages or windows
-
- 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
- G06T2207/30016—Brain
-
- 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
Abstract
The present invention provides a kind of MRI brain tumors positioning cascading convolutional network based on depth and dividing method in tumor, includes the following steps：Depth cascade convolutional neural networks parted pattern is built；Model training and parameter optimization；The quick positioning of multi-modal MRI brain tumors and segmentation in tumor.The MRI brain tumors positioning provided by the invention that convolutional network is cascaded based on depth and dividing method in tumor, the depth that structure is made of full convolutional neural networks and classification convolutional neural networks cascades hybrid neural networks, cutting procedure is divided into sub-district in the positioning of complete tumors area and tumor and divides two stages, realize hierarchical MRI brain tumors be quickly accurately positioned and tumor in sub-district segmentation, complete tumors region is oriented from MRI image using full convolutional network method first, then complete tumors are further divided by edema area using image block classification method, non-reinforcing tumor area, enhance tumor area and necrotic area, realize the quick Accurate Segmentation of multi-modal MRI brain tumors being accurately positioned with sub-district in tumor.
Description
Technical field
The present invention relates to medical image analysis technical fields, and in particular to a kind of MRI cascading convolutional network based on depth
Brain tumor positions and dividing method in tumor.
Background technology
Brain tumor is to seriously endanger the major disease of human health.Wherein, glioma is the main class of malignant brain tumor
Type, although uncommon, lethality is very high.According to Document system, high-grade glioma mean survival time is 14 months.
Magnetic resonance imaging (Magnetic Resonance Imaging, MRI) is clinically most common brain tumor inspection and diagnosis hand
Section, goes out structure in brain tumor and tumor from Accurate Segmentation in MRI image, has important valence to europathology analysis and precisely diagnosis
Value, can provide important support for operation plan, radiotherapy chemotherapy program type and prognosis evaluation.
The segmentation of MRI brain tumors is needed with reference to tetra- kinds of modality images of T1, T1c, T2 and FLAIR, and each mode is wrapped again
If the three-dimensional data containing dry chip.Although manually segmentation is feasible but takes very much, and is influenced by doctor's experience, have certain main
The property seen, repeatability are poor.Therefore, it explores intelligent algorithm and carries out MRI brain tumor full-automatic dividings as necessity.
Conventional machines study uses artificial extraction feature, then trains a grader using extraction feature, then makes
Classified to image pixel with trained grader, to generate segmentation figure.But this method is by Feature Extraction Algorithm
It influences, what the feature of extraction might not be suitable for a certain specific classification task can diagnostic characteristics.However, being based on convolutional Neural
The depth learning technology of network can learn the hierarchical feature for being suitable for particular task automatically from data set, can greatly promote
Pixel classifications precision.
The present inventor passes through the study found that being currently based on the MRI brain tumor automatic division method masters of deep learning
There are two classes：Image block classification method and full convolutional network Pixel-level classification.Wherein, image block classification method uses sliding window mode
It takes the surrounding neighbors block centered on each pixel to classify, has the following disadvantages：(1) computing redundancy degree is high, splitting speed
Slowly；(2) classification is only with image block local feature, and the global characteristics of synthetic image, do not easy to produce misclassified gene；(3) mould
Type effect is directly related with training image blocks abstracting method.Entire image is directly inputted net by full convolutional network Pixel-level classification
Network, a forward calculation can complete all types of tumor regions segmentation of whole sub-picture, but have the following disadvantages：(1) medicine figure
Sub-district usually only accounts for the very little part of image in focal zone especially tumor as in, and pixel of all categories has serious uneven, whole picture
Image input training cannot solve label imbalance problem；(2) since zonule sample deficiency leads to train insufficient, figure
Picture partitioning boundary is coarse, cannot achieve the fine granularity segmentation of zonule.
Invention content
In dividing for existing MRI brain tumors, image block classification method does not utilize global context feature and splitting speed
Slowly, the complete serious imbalance of convolutional network Pixel-level classification training sample leads to the inaccurate problem of zonule partitioning boundary, this
Invention provides a kind of MRI brain tumors positioning cascading convolutional network based on depth and dividing method in tumor, and this method is by building
Cutting procedure is divided into sub-district in the positioning of complete tumors area and tumor and divides two stages by depth cascade mixing convolutional neural networks,
It realizes that multi-modal MRI brain tumors are quickly accurately positioned with sub-district in tumor to divide.
In order to solve the above-mentioned technical problem, present invention employs the following technical solutions：
A kind of MRI brain tumors positioning cascading convolutional network based on depth and dividing method in tumor, include the following steps：
S1, depth cascade convolutional neural networks parted pattern are built：
S11, depth cascade convolutional neural networks are made of net of classifying in tumor-localizing net and tumor, and the tumor-localizing net is suitable
In input tetra- mode MRI image of FLAIR, T1, T1c and T2, the binary map in tumor candidate area and normal structure is exported, in the tumor
Net of classifying is suitable for inputting the tumor candidate area of tumor-localizing net output, exports sub-district segmentation result in tumor；
S12, the tumor-localizing net are made of full convolutional network, including first to the 5th totally five convolutional layer groups, first
To the 5th totally five pond layers, convolutional layer six and convolutional layer seven, after first pond layer is located at the first convolutional layer group, described
After two pond layers are located at the second convolutional layer group, and so on, after the 5th pond layer is located at the 5th convolutional layer group, the convolution
After layer six and seven sequence of convolutional layer are set to the 5th pond layer；
Using jump connection in S13, the tumor-localizing net, the high-level semantics feature that convolutional layer seven is exported carries out 2 times
It successively merges, pixel class is carried out with final fusion feature accurate pre- with the low-level details feature behind each pond after up-sampling
It surveys；
Classification net is by two convolutional layer groups, two pond layers, three full articulamentums and a Softmax in S14, the tumor
Layer of classifying forms, wherein each convolutional layer group is divided followed by a pond layer, described three full articulamentums and a Softmax
After class layer sequence is set to the last one pond layer；
S2, model training and parameter optimization：Depth cascade convolutional neural networks are divided using the labeled data after expansion
Model carries out Training, and design object function optimization network parameter generates optimum segmentation model, specifically includes：
S21, the entire image data set after standardization and expansion is pressed 8：1：1 ratio be divided into training set, verification collection and
Test set, the tetra- mode entire image of FLAIR, T1, T1c and T2 of same brain section is defeated as the four-way of tumor-localizing net
Enter；
S22, using classification cross entropy loss function, target, object function are defined as follows as an optimization：
Wherein, Y' is segmentation tag, and Y is the probability of prediction, and C is pixel class number, and S is the number of image pixel；
S23, using stochastic gradient descent algorithm optimization object function, update tumor-localizing with error backpropagation algorithm
Pessimistic concurrency control parameter；
S24, the MRI image block data set of extraction is pressed 8：1：1 ratio is divided into training set, verification collection and test set, will
Four-way input of the tetra- modality images block of FLAIR, T1, T1c and T2 of same brain section as classification net in tumor；
S25, using the target as an optimization of the classification cross entropy loss function in step S22, but C indicates staging classification
Number, S indicate image block sample number in batch；
S26, using the stochastic gradient descent algorithm optimization object function in step S23, with error backpropagation algorithm
Pessimistic concurrency control parameter of classifying in tumor is updated, and this step, when carrying out classification net training in tumor, first in three full articulamentums is complete
Articulamentum and the second full articulamentum have used Dropout regularization methods, Dropout rates to be set as 0.50；
Quick positioning and the segmentation in tumor of S3, multi-modal MRI brain tumors comprising：
S31, after four mode MRI images after preparation standards have been trained and have been optimized as four-way input step S2
Tumor-localizing net, be automatically positioned and export the binary segmentation figure including tumor area and non-tumor area；
S32, take the four modality images block input step S2 centered on the pixel of tumor area to train and optimized after tumor in
Classification net predicts the classification of the pixel, and uses sliding window mode, is predicted one by one tumour pixel, finally obtains sub-district in tumor
Segmentation figure；
S33, sub-district segmentation figure in tumor is added on original MRI image, obtains final MRI brain tumors positioning and segmentation
Figure.
Further, the method is further comprising the steps of：
The multi-modal MRI image pretreatment of S4, brain tumor, pretreated image data are suitable for step S21, S24 and S31
Input comprising：
S41, biased field correct operation is carried out to multi-modal MRI image；
S42, the MRI image slice for extracting tetra- mode of FLAIR, T1, T1c and T2, in each MRI image slice,
Remove the gray value of highest 1% and minimum 1%；
S43, data normalization operation is carried out using following formula to the gray value of each MRI image：
Wherein, the gray value of the i-th row j row of the corresponding slice X of x (i, j),And XsIt is the mean value and standard for being sliced X respectively
Difference, x ' (i, j) are the standardized gray scales of x (i, j)；
S44, data extending technology is used to the gray level image after normalizing operation, it is initial to make the increase of training data sample
10 times；
S45, the image block that size is 33 × 33 is randomly selected centered on tumour pixel to the data set after expansion, each
Tumour classification extraction quantity is identical, and image block data ensemble average is divided into each classification in 10 groups, every group using stratified sampling method
Image block proportion it is identical.
Further, in the step S12, there are 2 convolutional layers, the convolution of convolutional layer in the first and second convolutional layer groups respectively
Core number is respectively 64 and 128, there is 3 convolutional layers, the convolution kernel of convolutional layer in the convolutional layer group of third, the 4th and the 5th respectively
Number is respectively 256,512 and 512, and the convolution kernel size of all convolutional layers is 3 × 3, step-length 1, the core size of each pond layer
For 2 × 2, step-length 2, it is 1 × 1, step-length 1 that the convolution kernel number of convolutional layer six and convolutional layer seven, which is 4096, size,.
Further, the output characteristic pattern Z in the method corresponding to any one convolution kerneliIt is calculated using following formula：
Wherein, f indicates nonlinear activation function, biIndicate that the bias term corresponding to i-th of convolution kernel, r indicate that input is logical
Road call number, k indicate input channel number, WirIndicate r-th of channel weight matrix of i-th of convolution kernel,It is convolution operation, Xr
Indicate r-th of input channel image.
Further, the tumor-localizing net further includes the linear unit R eLU of rectification, and the linear unit R eLU of rectification is used for
By output characteristic pattern Z caused by convolution kerneliEach of value carry out non-linear transfer, the linear unit R eLU definition of the rectification
It is as follows：
F (x)=max (0, x)
Wherein, f (x) indicates the linear unit function of rectification, and x is an input value.
Further, the step S13 is specifically included：By the result of convolutional layer seven carry out after 2 times of up-samplings with the 4th pond
Layer, which be added, merges to obtain fused layer 1, then will merge to obtain fusion with the addition of third pond layer after 2 times of up-samplings of the progress of fused layer 1
Layer 2, then fused layer 3 will be merged to obtain with the second pond layer addition after 2 times of up-samplings of the progress of fused layer 2, then fused layer 3 is carried out 2
Merge to obtain fused layer 4 with the first pond layer addition after times up-sampling, finally will fused layer 4 carry out 2 times of up-samplings after obtain and original
The identical characteristic pattern of beginning image size；Tumour is carried out to pixel using this characteristic pattern and normal structure 2 is classified, generates 2 pixels
Class prediction score value figure, the classification for taking predicted value high is as the final classification of pixel.
Further, in the step S14,3 convolutional layers, the convolution kernel number of convolutional layer are equipped in two convolutional layer groups
Respectively 64 and 128, the convolution kernel sizes of all convolutional layers is 3 × 3, step-length 1, the core size of each pond layer is 3 × 3,
Step-length is 2, and the size of described three full articulamentums is respectively 256,128 and 4, wherein 4 edema areas, non-reinforcing swollen for representing tumour
Tumor area, enhancing tumor area and the classification of necrotic area four.
Further, classification net further includes non-linear excitation unit Leaky ReLU, the non-linear excitation list in the tumor
Member by each of output characteristic pattern value caused by convolution kernel for carrying out non-linear transfer, the non-linear excitation unit
Leaky ReLU are defined as follows：
F (z)=max (0, z)+α min (0, z)
Wherein, f (z) indicates non-linear excitation unit function, and z is an input value, and α is Leaky parameters.
Further, in the step S14, Softmax functions are defined as follows：
Wherein, OkIt is the value of k-th of neuron of classification net output in tumor, YkIt is that input picture block belongs to k-th of classification
Probability, C are classification number.
Further, in the step S22 and S25, it has been additionally added L2 regularization terms on object function, has obtained final goal letter
Number is as follows：
Wherein, λ is regularization factors, and Q is model parameter number, and θ is network model parameter.
Further, in the step S23 and S26, specific optimization process is as follows：
mt=μ * mt-1-ηtgt
θt=θt-1+mt
Wherein, subscript t indicates that iterations, θ are network model parameter, L (θt-1) it is to work as to use θt-1For network parameter when
Loss function, gt、mtIndicate gradient, momentum and momentum coefficient, η respectively with μtIt is learning rate.Compared with prior art, this hair
The MRI brain tumors that convolutional network is cascaded based on depth of bright offer are positioned to be had the following advantages with dividing method in tumor：
1, the cascade composite nerve that structure is made of full convolutional neural networks (FCN) and classification convolutional neural networks (CNN)
Network realizes structure Accurate Segmentation in the hierarchical MRI brain tumors automatic positioning of two benches and tumor, uses full convolutional network method first
Complete tumors region is quickly oriented from MRI image, is then further divided into complete tumors using image block classification method
Edema area, non-reinforcing tumor area, enhancing tumor area and necrotic area；
2, at the tumor-localizing stage, each sub-district in tumour is incorporated as an entirety and is split, alleviated each
Between tumour sub-district and the sample imbalance problem between normal structure；
When 3, being split to sub-district in tumor, it is only necessary to be split, be subtracted using image block method to the pixel in tumour
Lack block sort quantity, improves splitting speed；
4, in tumor when the training of classification net, the image block that identical quantity can be extracted to every classification is trained, and can solve tumour
Interior sub-district sample imbalance problem, allows different classes of pixel to obtain the training of equal extent, to can get more accurate tumor
Interior partitioning boundary keeps sub-district segmentation in tumor more accurate；
5, sub-district segmentation net using small convolution kernel and has a deeper network structure in tumor, is ensuring that model parameter amount do not increase
In the case of improve the non-linear conversion ability of network, generate that level is more rich, the stronger image block classification feature of distinctive, from
And improve image block classification accuracy.
Description of the drawings
Fig. 1 is the MRI brain tumors positioning provided by the invention for cascading convolutional network based on depth and dividing method stream in tumor
Journey schematic diagram.
Fig. 2 is depth cascade convolutional neural networks parted pattern schematic diagram provided by the invention.
Fig. 3 is that sub-district divides network model schematic diagram in the tumor provided by the invention based on image block classification.
Specific implementation mode
In order to make the technical means, the creative features, the aims and the efficiencies achieved by the present invention be easy to understand, tie below
Conjunction is specifically illustrating, and the present invention is further explained.
In the description of the present invention, it is to be understood that, term " longitudinal direction ", " radial direction ", " length ", " width ", " thickness ",
The orientation of the instructions such as "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outside" or
Position relationship is to be based on the orientation or positional relationship shown in the drawings, and is merely for convenience of description of the present invention and simplification of the description, without
It is instruction or implies that signified device or element must have a particular orientation, with specific azimuth configuration and operation, therefore not
It can be interpreted as limitation of the present invention.In the description of the present invention, unless otherwise indicated, the meaning of " plurality " is two or two
More than.
In the description of the present invention, it should be noted that unless otherwise clearly defined and limited, term " installation ", " phase
Even ", " connection " shall be understood in a broad sense, for example, it may be being fixedly connected, may be a detachable connection, or be integrally connected；It can
Can also be electrical connection to be mechanical connection；It can be directly connected, can also indirectly connected through an intermediary, Ke Yishi
Connection inside two elements.For the ordinary skill in the art, above-mentioned term can be understood at this with concrete condition
Concrete meaning in invention.
It please refers to Fig.1 to shown in Fig. 3, the present invention provides a kind of MRI brain tumors positioning cascading convolutional network based on depth
With dividing method in tumor, include the following steps：
S1, depth cascade convolutional neural networks parted pattern are built：
S11, depth cascade convolutional neural networks are made of net of classifying in tumor-localizing net and tumor, and the tumor-localizing net is suitable
In input tetra- mode MRI image of FLAIR, T1, T1c and T2, the binary map in tumor candidate area and normal structure is exported, in the tumor
Net of classifying is suitable for inputting the tumor candidate area of tumor-localizing net output, exports sub-district segmentation result in tumor.
S12, the tumor-localizing net are made of full convolutional network comprising the first to the 5th totally five convolutional layer groups,
One to the 5th totally five pond layers, convolutional layer six and convolutional layer seven, it is described after first pond layer is located at the first convolutional layer group
After second pond layer is located at the second convolutional layer group, and so on, after the 5th pond layer is located at the 5th convolutional layer group, the volume
After lamination six and seven sequence of convolutional layer are set to the 5th pond layer；The input channel number of the tumor-localizing net is 4, represents MRI figures
Four mode of picture.
1 tumor-localizing network architecture hyper parameter table of table
As specific embodiment, upper table 1 is please referred to, in the step S12, have 2 respectively in the first and second convolutional layer groups
The convolution kernel number of a convolutional layer, convolutional layer is respectively 64 and 128, has 3 volumes respectively in the convolutional layer group of third, the 4th and the 5th
The convolution kernel number of lamination, convolutional layer is respectively 256,512 and 512, and the convolution kernel size of all convolutional layers is that 3 × 3, step-length is
1, the core size of each pond layer is 2 × 2, step-length 2, and the convolution kernel number of convolutional layer six and convolutional layer seven is 4096, size
It is 1 × 1, step-length 1.
As specific embodiment, the output characteristic pattern Z in the method corresponding to any one convolution kerneliUsing following formula into
Row calculates：
Wherein, f indicates nonlinear activation function, biIndicate that the bias term corresponding to i-th of convolution kernel, r indicate that input is logical
Road call number, k indicate input channel number, WirIndicate r-th of channel weight matrix of i-th of convolution kernel,It is convolution operation, Xr
Indicate r-th of input channel image.
As a preferred embodiment, in order to improve the non-linear expression ability of network, f uses rectification linear unit in formula (1)
ReLU (Rectifier Linear Units) function, activation primitives of the ReLU as full convolutional network, the rectification are linearly single
Member is used for output characteristic pattern Z caused by convolution kerneliEach of value carry out non-linear transfer, the rectification linear unit
ReLU is defined as follows：
F (x)=max (0, x) (2)
Wherein, f (x) indicates the linear unit function of rectification, and x is an input value.
Simultaneously as the output characteristic pattern after convolutional layer convolution may include a large amount of redundancy, therefore, in convolution
Redundancy feature is eliminated in the maximum pondization operation for using size to be 2 for 2 × 2, step-length after output layer, i.e., by pond layer to convolution
The output characteristic pattern of layer carries out dimensionality reduction, to make output characteristic pattern size become smaller, increases receptive field, improves the translation invariant of image
Property.
Using jump connection in S13, the tumor-localizing net, the high-level semantics feature that convolutional layer seven is exported carries out 2 times
It successively merges, pixel class is carried out with final fusion feature accurate pre- with the low-level details feature behind each pond after up-sampling
It surveys.It is specifically included：Fused layer is merged to obtain by with the 4th pond layer be added after 2 times of up-samplings of result progress of convolutional layer seven
1, then fused layer 2 will be merged to obtain with the addition of third pond layer after 2 times of up-samplings of the progress of fused layer 1, then fused layer 2 is carried out 2 times
Merge to obtain fused layer 3 after up-sampling with the second pond layer addition, then by fused layer 3 carry out after 2 times of up-samplings with the first pond layer
Fused layer 4 is merged to obtain in addition, and characteristic pattern identical with original image size is obtained after fused layer 4 is finally carried out 2 times of up-samplings；
Tumour is carried out to pixel using this characteristic pattern and normal structure 2 is classified, 2 (representing 2 classes) pixel class is generated and predicts score value figure,
The classification for taking predicted value high is as the final classification of pixel.
Classification net is by two convolutional layer groups, two pond layers, three full articulamentums and a Softmax in S14, the tumor
Layer of classifying forms, and specifically please refers to Fig.3 shown, wherein each convolutional layer group is followed by a pond layer, described three full connections
Layer and a Softmax classify layer sequence after the last one pond layer.
It as specific embodiment, please refers to shown in the following table 2, in the step S14,3 is equipped in two convolutional layer groups
The convolution kernel number of convolutional layer, convolutional layer is respectively 64 and 128, and the convolution kernel size of all convolutional layers is 3 × 3, step-length 1,
The core size of each pond layer is 3 × 3, step-length 2, and the size of described three full articulamentums is respectively 256,128 and 4, wherein 4
Represent edema area, non-reinforcing tumor area, enhancing tumor area and the classification of necrotic area four of tumour；Classification net input in the tumor
Be size centered on a certain tumour pixel it is 33 × 33 4 modality images blocks, output is the pixel in four classifications
Probability vector.
Sorter network model structure hyper parameter table in 2 tumor of table
As specific embodiment, the output characteristic pattern Z in the tumor in classification net corresponding to any one convolution kerneliAlso it adopts
It is calculated with formula (1), but nonlinear activation function uses Leaky ReLU (Rectifier Linear Units) function,
Therefore, classification net further includes non-linear excitation unit in the tumor, and the non-linear excitation unit Leaky ReLU will be for that will roll up
Output characteristic pattern Z caused by product coreiEach of value carry out non-linear transfer, the non-linear excitation unit Leaky ReLU
It is defined as follows：
F (z)=max (0, z)+α min (0, z) (3)
Wherein, f (z) indicates non-linear excitation unit function, and z is an input value, and α is Leaky parameters；As a kind of reality
Mode is applied, α is set as 0.33.
As specific embodiment, in the step S14, in order to solve more classification problems, by the defeated of the full articulamentum FC3 of third
Go out and image block prediction score value is converted into probability distribution using Softmax functions, Softmax functions are defined as follows：
Wherein, OkIt is the value of k-th of neuron of classification net output in tumor, YkIt is that input picture block belongs to k-th of classification
Probability, C are classification number, according to description above-mentioned it is found that in the tumor output of classification net have the edema area, non-reinforcing for representing tumour
Tumor area enhances tumor area and necrotic area totally four classifications.
S2, model training and parameter optimization：Depth cascade convolutional neural networks are divided using the labeled data after expansion
Model carries out Training, and design object function optimization network parameter generates optimum segmentation model, specifically includes：
S21, the entire image data set after standardization and expansion is pressed 8：1：1 ratio be divided into training set, verification collection and
Test set, the tetra- mode entire image of FLAIR, T1, T1c and T2 of same brain section is defeated as the four-way of tumor-localizing net
Enter, as an implementation, present inventor obtains 274 patient datas with segmentation tag, each mode altogether
Including 155 width sequence images, thus 274 × 155=42470 four modality images data samples are shared, extended rear training set,
Verification collection and test set are respectively 339760,42470,42470 samples.
S22, using classification cross entropy loss function, target, object function are defined as follows as an optimization：
Wherein, Y' is segmentation tag, and Y is the probability of prediction, and C is pixel class number, and S is the number of image pixel；As
A kind of embodiment, C=2, S=240 × 240=57600.
As a preferred embodiment, over-fitting in order to prevent in the step S22, is being additionally added L2 just on object function
Then change item, it is as follows to obtain final goal function：
Wherein, λ is regularization factors, and Q is model parameter number, and θ is network model parameter.
S23, using stochastic gradient descent algorithm optimization object function, update tumor-localizing with error backpropagation algorithm
Pessimistic concurrency control parameter, specific optimization process are as follows：
mt=μ * mt-1-ηtgt (8)
θt=θt-1+mt (9)
Wherein, subscript t indicates that iterations, θ are network model parameter, corresponds to θ, L (θ in formula (6)t-1) it is to work as
Use θt-1For network parameter when loss function, gt、mtIndicate gradient, momentum and momentum coefficient, η respectively with μtIt is learning rate；
As an implementation, in step S23, momentum coefficient μ=0.99, initial learning rate is set as ηt=1e-8, every 1000 times repeatedly
In generation, reduces 1/10, until 1e-10Until.
S24, the MRI image block data set of extraction is pressed 8：1：1 ratio is divided into training set, verification collection and test set, will
Four-way input of the tetra- modality images block of FLAIR, T1, T1c and T2 of same brain section as classification net in tumor；As one kind
Embodiment, the size for extracting identical quantity to each classification from the data set after expansion are 33 × 33 image blocks totally 40,
000,000, being equivalent to each classification has 10,000,000 sample, and data set is equally divided into 10 groups using stratified sampling,
Then 8 are pressed：1：1 ratio cut partition is training set, verification collection and test set, i.e. training set, verification collection and test set has 32 respectively,
000,000,4,000,000,4,000,000 samples.
S25, using classification cross entropy loss function target as an optimization, object function is with the formula (5) in step S22, but C
Indicate that staging classification number, S indicate image block sample number in batch；As an implementation, the C=4 in this step, S
=256.For the purposes of preventing over-fitting, this step is added on object function L2 regularization terms, obtains such as step S22 Chinese styles
(6) final goal function shown in.
S26, using stochastic gradient descent algorithm optimization object function, with classifying in error backpropagation algorithm update tumor
Pessimistic concurrency control parameter uses formula (7), formula (8) and the formula (9) in step S23 in specific optimization process；As an implementation,
Momentum coefficient μ=0.9 in this step, initial learning rate are set as ηt=1e-6, every 1000 iteration reduce 1/10, until 1e-8For
Only.Meanwhile using this step S26 when carrying out classification net training in tumor, network over-fitting in order to prevent, in three full articulamentums
The first complete full articulamentum FC2 of articulamentum FC1 and second used Dropout regularization methods, Dropout rates to be set as
0.50。
Quick positioning and the segmentation in tumor of S3, multi-modal MRI brain tumors comprising：
S31, after four mode MRI images after preparation standards have been trained and have been optimized as four-way input step S2
Tumor-localizing net, be automatically positioned and export the binary segmentation figure including tumor area and non-tumor area；
S32, the four modality images block input steps 3 that the Size of Neighborhood centered on the pixel of tumor area is 33 × 33 are taken to instruct
Classification net in tumor after practicing and optimizing, predicts the classification of the pixel, and use sliding window mode, is predicted one by one tumour pixel,
Finally obtain sub-district segmentation figure in tumor；Wherein, sliding window mode has been technology well known to those skilled in the art, no longer superfluous herein
It states；
S33, sub-district segmentation figure in tumor is added on original MRI image, obtains final MRI brain tumors positioning and segmentation
Figure.
As a preferred embodiment, support that the present invention carries to preferably provide image data input for aforementioned correlation step
The MRI brain tumors positioning that convolutional network is cascaded based on depth supplied and dividing method in tumor, it is further comprising the steps of：
The multi-modal MRI image pretreatment of S4, brain tumor, pretreated image data are suitable for step S21, S24 and S31
Input comprising：
S41, biased field correct operation is carried out to multi-modal MRI image, N4ITK methods specifically may be used and carry out biased field
Correct operation；
S42, the MRI image slice for extracting tetra- mode of FLAIR, T1, T1c and T2, in each MRI image slice,
Remove the gray value of highest 1% and minimum 1%；
S43, data normalization operation is carried out using following formula to the gray value of each MRI image：
Wherein, the gray value of the i-th row j row of the corresponding slice X of x (i, j),And XsIt is the mean value and standard for being sliced X respectively
Difference, x ' (i, j) are the standardized gray scales of x (i, j)；
S44, to after normalizing operation gray level image and corresponding label using flip horizontal, flip vertical, amplification 1/8 after
It reduces, rotation 45 °, 90 °, 135 °, 180 °, 225 °, 270 °, 315 ° of data extending technologies, it is first to make the increase of training data sample
10 times to begin, and aforementioned specific data extending technology has been well known to those skilled in the art, details are not described herein；
S45, the image block that size is 33 × 33 is randomly selected centered on tumour pixel to the data set after expansion, each
Tumour classification extraction quantity is identical, and image block data ensemble average is divided into each classification in 10 groups, every group using stratified sampling method
Image block proportion it is identical.
Compared with prior art, provided by the invention to be cascaded in the positioning of MRI brain tumors and tumor of convolutional network based on depth
Dividing method has the following advantages：
1, the cascade composite nerve that structure is made of full convolutional neural networks (FCN) and classification convolutional neural networks (CNN)
Network realizes structure Accurate Segmentation in the hierarchical MRI brain tumors automatic positioning of two benches and tumor, uses full convolutional network method first
Complete tumors region is quickly oriented from MRI image, is then further divided into complete tumors using image block classification method
Edema area, non-reinforcing tumor area, enhancing tumor area and necrotic area；
2, at the tumor-localizing stage, each sub-district in tumour is incorporated as an entirety and is split, alleviated each
Between tumour sub-district and the sample imbalance problem between normal structure；
When 3, being split to sub-district in tumor, it is only necessary to be split, be subtracted using image block method to the pixel in tumour
Lack block sort quantity, improves splitting speed；
4, in tumor when the training of classification net, the image block that identical quantity can be extracted to every classification is trained, and can solve tumour
Interior sub-district sample imbalance problem, allows different classes of pixel to obtain the training of equal extent, to can get more accurate tumor
Interior partitioning boundary keeps sub-district segmentation in tumor more accurate；
5, sub-district segmentation net using small convolution kernel and has a deeper network structure in tumor, is ensuring that model parameter amount do not increase
In the case of improve the non-linear conversion ability of network, generate that level is more rich, the stronger image block classification feature of distinctive, from
And improve image block classification accuracy.
Finally illustrate, the above examples are only used to illustrate the technical scheme of the present invention and are not limiting, although with reference to compared with
Good embodiment describes the invention in detail, it will be understood by those of ordinary skill in the art that, it can be to the skill of the present invention
Art scheme is modified or replaced equivalently, and without departing from the objective and range of technical solution of the present invention, should all be covered at this
In the right of invention.
Claims (11)
1. the MRI brain tumors positioning for cascading convolutional network based on depth and dividing method in tumor, which is characterized in that including following step
Suddenly：
S1, depth cascade convolutional neural networks parted pattern are built：
S11, depth cascade convolutional neural networks are made of net of classifying in tumor-localizing net and tumor, and the tumor-localizing net is suitable for defeated
Enter tetra- mode MRI image of FLAIR, T1, T1c and T2, exports the binary map in tumor candidate area and normal structure, classify in the tumor
Net is suitable for inputting the tumor candidate area of tumor-localizing net output, exports sub-district segmentation result in tumor；
S12, the tumor-localizing net are made of full convolutional network, including the first to the 5th totally five convolutional layer groups, first to
Five totally five pond layers, convolutional layer six and convolutional layers seven, after first pond layer is located at the first convolutional layer group, second pond
After change layer is located at the second convolutional layer group, and so on, after the 5th pond layer is located at the 5th convolutional layer group, the convolutional layer six
After the 5th pond layer being set to seven sequence of convolutional layer；
Using jump connection in S13, the tumor-localizing net, the high-level semantics feature that convolutional layer seven exports adopt on 2 times
It successively merges with the low-level details feature behind each pond after sample, pixel class is accurately predicted with final fusion feature；
Classification net is by two convolutional layer groups, two pond layers, three full articulamentums and a Softmax classification in S14, the tumor
Layer composition, wherein each convolutional layer group is followed by a pond layer, described three full articulamentums and a Softmax classification layer
After sequence is set to the last one pond layer；
S2, model training and parameter optimization：Convolutional neural networks parted pattern is cascaded to depth using the labeled data after expansion
Training is carried out, design object function optimization network parameter generates optimum segmentation model, specifically includes：
S21, the entire image data set after standardization and expansion is pressed 8：1：1 ratio is divided into training set, verification collection and test
Collection is inputted the tetra- mode entire image of FLAIR, T1, T1c and T2 of same brain section as the four-way of tumor-localizing net；
S22, using classification cross entropy loss function, target, object function are defined as follows as an optimization：
Wherein, Y' is segmentation tag, and Y is the probability of prediction, and C is pixel class number, and S is the number of image pixel；
S23, using stochastic gradient descent algorithm optimization object function, update tumor-localizing net mould with error backpropagation algorithm
Shape parameter；
S24, the MRI image block data set of extraction is pressed 8：1：1 ratio is divided into training set, verification collection and test set, will be same
Four-way input of the tetra- modality images block of FLAIR, T1, T1c and T2 of brain section as classification net in tumor；
S25, using the target as an optimization of the classification cross entropy loss function in step S22, but C indicates staging classification number, S
Indicate image block sample number in batch；
S26, using the stochastic gradient descent algorithm optimization object function in step S23, updated with error backpropagation algorithm
The interior pessimistic concurrency control parameter of classifying of tumor, and the classification in progress tumor of this step is netted when training, first in three full articulamentums is complete to be connected
Layer and the second full articulamentum have used Dropout regularization methods, Dropout rates to be set as 0.50；
Quick positioning and the segmentation in tumor of S3, multi-modal MRI brain tumors comprising：
S31, four mode MRI images after preparation standards have been trained and have been optimized as four-way input step S2 after it is swollen
Tumor positions net, is automatically positioned and exports the binary segmentation figure including tumor area and non-tumor area；
S32, take the four modality images block input step S2 centered on the pixel of tumor area to train and optimized after tumor in classify
Net predicts the classification of the pixel, and uses sliding window mode, is predicted one by one tumour pixel, finally obtains sub-district in tumor and divides
Figure；
S33, sub-district segmentation figure in tumor is added on original MRI image, obtains final MRI brain tumors positioning and segmentation figure.
2. the MRI brain tumors positioning according to claim 1 that convolutional network is cascaded based on depth and dividing method in tumor,
It is characterized in that, the method is further comprising the steps of：
The multi-modal MRI image pretreatment of S4, brain tumor, it is defeated that pretreated image data is suitable for step S21, S24 and S31
Enter comprising：
S41, biased field correct operation is carried out to multi-modal MRI image；
S42, the MRI image slice for extracting tetra- mode of FLAIR, T1, T1c and T2 are removed in each MRI image slice
The gray value of highest 1% and minimum 1%；
S43, data normalization operation is carried out using following formula to the gray value of each MRI image：
Wherein, the gray value of the i-th row j row of the corresponding slice X of x (i, j),And XsIt is the mean value and standard deviation for being sliced X, x ' respectively
(i, j) is the standardized gray scales of x (i, j)；
S44, data extending technology is used to the gray level image after normalizing operation, it is initial 10 so that training data sample is increased
Times；
S45, the image block that size is 33 × 33, each tumour are randomly selected centered on tumour pixel to the data set after expansion
Classification extraction quantity is identical, and image block data ensemble average is divided into the figure of each classification in 10 groups, every group using stratified sampling method
As block proportion is identical.
3. the MRI brain tumors positioning according to claim 1 or 2 for cascading convolutional network based on depth and segmentation side in tumor
Method, which is characterized in that in the step S12, there is 2 convolutional layers, the convolution of convolutional layer in the first and second convolutional layer groups respectively
Core number is respectively 64 and 128, there is 3 convolutional layers, the convolution kernel of convolutional layer in the convolutional layer group of third, the 4th and the 5th respectively
Number is respectively 256,512 and 512, and the convolution kernel size of all convolutional layers is 3 × 3, step-length 1, the core size of each pond layer
For 2 × 2, step-length 2, it is 1 × 1, step-length 1 that the convolution kernel number of convolutional layer six and convolutional layer seven, which is 4096, size,.
4. the MRI brain tumors positioning according to claim 3 that convolutional network is cascaded based on depth and dividing method in tumor,
It is characterized in that, the output characteristic pattern Z in the method corresponding to any one convolution kerneliIt is calculated using following formula：
Wherein, f indicates nonlinear activation function, biIndicate that the bias term corresponding to i-th of convolution kernel, r indicate input channel index
Number, k indicates input channel number, WirIndicate r-th of channel weight matrix of i-th of convolution kernel,It is convolution operation, XrIndicate the
R input channel image.
5. the MRI brain tumors positioning according to claim 4 that convolutional network is cascaded based on depth and dividing method in tumor,
It is characterized in that, the tumor-localizing net further includes the linear unit R eLU of rectification, and the linear unit R eLU of rectification is used for convolution
Output characteristic pattern Z caused by coreiEach of value carry out non-linear transfer, the linear unit R eLU of rectification is defined as follows：
F (x)=max (0, x)
Wherein, f (x) indicates the linear unit function of rectification, and x is an input value.
6. the MRI brain tumors positioning according to claim 1 or 2 for cascading convolutional network based on depth and segmentation side in tumor
Method, which is characterized in that the step S13 is specifically included：By the result of convolutional layer seven carry out after 2 times of up-samplings with the 4th pond layer
Be added and merge to obtain fused layer 1, then fused layer will be merged to obtain with the addition of third pond layer after 2 times of up-samplings of the progress of fused layer 1
2, then fused layer 3 will be merged to obtain with the second pond layer addition after 2 times of up-samplings of the progress of fused layer 2, then fused layer 3 is carried out 2 times
Merge to obtain fused layer 4 with the first pond layer addition after up-sampling, finally fused layer 4 is carried out obtaining after 2 times of up-samplings with it is original
The identical characteristic pattern of image size；Tumour is carried out to pixel using this characteristic pattern and normal structure 2 is classified, generates 2 pixel classes
Not Yu Ce score value figure, the classification for taking predicted value high is as the final classification of pixel.
7. the MRI brain tumors positioning according to claim 1 or 2 for cascading convolutional network based on depth and segmentation side in tumor
Method, which is characterized in that in the step S14,3 convolutional layers, the convolution kernel number of convolutional layer are equipped in two convolutional layer groups
Respectively 64 and 128, the convolution kernel sizes of all convolutional layers is 3 × 3, step-length 1, the core size of each pond layer is 3 × 3,
Step-length is 2, and the size of described three full articulamentums is respectively 256,128 and 4, wherein 4 edema areas, non-reinforcing swollen for representing tumour
Tumor area, enhancing tumor area and the classification of necrotic area four.
8. the MRI brain tumors positioning according to claim 7 that convolutional network is cascaded based on depth and dividing method in tumor,
It is characterized in that, classification net further includes non-linear excitation unit LeakyReLU in the tumor, and the non-linear excitation unit is used for will
Each of output characteristic pattern value caused by convolution kernel carries out non-linear transfer, the non-linear excitation unit Leaky ReLU
It is defined as follows：
F (z)=max (0, z)+α min (0, z)
Wherein, f (z) indicates non-linear excitation unit function, and z is an input value, and α is Leaky parameters.
9. the MRI brain tumors positioning according to claim 7 that convolutional network is cascaded based on depth and dividing method in tumor,
It is characterized in that, in the step S14, Softmax functions are defined as follows：
Wherein, OkIt is the value of k-th of neuron of classification net output in tumor, YkIt is that input picture block belongs to the general of k-th classification
Rate, C are classification number.
10. the MRI brain tumors positioning according to claim 1 or 2 for cascading convolutional network based on depth and segmentation side in tumor
Method, which is characterized in that in the step S22 and S25, L2 regularization terms have been additionally added on object function, have obtained final goal letter
Number is as follows：
Wherein, λ is regularization factors, and Q is model parameter number, and θ is network model parameter.
11. the MRI brain tumors positioning according to claim 1 or 2 for cascading convolutional network based on depth and segmentation side in tumor
Method, which is characterized in that in the step S23 and S26, specific optimization process is as follows：
mt=μ * mt-1-ηtgt
θt=θt-1+mt
Wherein, subscript t indicates that iterations, θ are network model parameter, L (θt-1) it is to work as to use θt-1For network parameter when damage
Lose function, gt、mtIndicate gradient, momentum and momentum coefficient, η respectively with μtIt is learning rate.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2017114199457 | 2017-12-25 | ||
| CN201711419945 | 2017-12-25 |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108492297A true CN108492297A (en) | 2018-09-04 |
| CN108492297B CN108492297B (en) | 2021-11-19 |
Family
ID=63314778
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810300057.1A Active CN108492297B (en) | 2017-12-25 | 2018-04-04 | MRI brain tumor positioning and intratumoral segmentation method based on deep cascade convolution network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108492297B (en) |
Cited By (49)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109360210A (en) * | 2018-10-16 | 2019-02-19 | 腾讯科技（深圳）有限公司 | Image partition method, device, computer equipment and storage medium |
| CN109410289A (en) * | 2018-11-09 | 2019-03-01 | 中国科学院武汉物理与数学研究所 | A kind of high lack sampling hyperpolarized gas lung MRI method for reconstructing of deep learning |
| CN109472784A (en) * | 2018-10-31 | 2019-03-15 | 安徽医学高等专科学校 | Based on the recognition methods for cascading full convolutional network pathological image mitotic cell |
| CN109493317A (en) * | 2018-09-25 | 2019-03-19 | 哈尔滨理工大学 | The more vertebra dividing methods of 3D based on concatenated convolutional neural network |
| CN109886929A (en) * | 2019-01-24 | 2019-06-14 | 江苏大学 | A kind of MRI tumour voxel detection method based on convolutional neural networks |
| CN109902748A (en) * | 2019-03-04 | 2019-06-18 | 中国计量大学 | A kind of image, semantic dividing method based on the full convolutional neural networks of fusion of multi-layer information |
| CN110047068A (en) * | 2019-04-19 | 2019-07-23 | 山东大学 | MRI brain tumor dividing method and system based on pyramid scene analysis network |
| CN110084823A (en) * | 2019-04-18 | 2019-08-02 | 天津大学 | Three-dimensional brain tumor image partition method based on cascade anisotropy FCNN |
| CN110097550A (en) * | 2019-05-05 | 2019-08-06 | 电子科技大学 | A kind of medical image cutting method and system based on deep learning |
| CN110097567A (en) * | 2019-04-18 | 2019-08-06 | 天津大学 | In conjunction with the three-dimensional brain tumor image partition method of improved FCNN and level set |
| CN110120033A (en) * | 2019-04-12 | 2019-08-13 | 天津大学 | Based on improved U-Net neural network three-dimensional brain tumor image partition method |
| CN110136133A (en) * | 2019-03-11 | 2019-08-16 | 嘉兴深拓科技有限公司 | A kind of brain tumor dividing method based on convolutional neural networks |
| CN110349170A (en) * | 2019-07-13 | 2019-10-18 | 长春工业大学 | A kind of full connection CRF cascade FCN and K mean value brain tumor partitioning algorithm |
| CN110427954A (en) * | 2019-07-26 | 2019-11-08 | 中国科学院自动化研究所 | The image group feature extracting method of multizone based on tumor imaging |
| CN110504032A (en) * | 2019-08-23 | 2019-11-26 | 元码基因科技(无锡)有限公司 | The method for predicting Tumor mutations load based on the image procossing of hematoxylin-eosin dye piece |
| CN110507288A (en) * | 2019-08-29 | 2019-11-29 | 重庆大学 | Vision based on one-dimensional convolutional neural networks induces motion sickness detection method |
| CN110533683A (en) * | 2019-08-30 | 2019-12-03 | 东南大学 | A kind of image group analysis method merging traditional characteristic and depth characteristic |
| CN110706214A (en) * | 2019-09-23 | 2020-01-17 | 东南大学 | Three-dimensional U-Net brain tumor segmentation method fusing condition randomness and residual error |
| CN110910335A (en) * | 2018-09-15 | 2020-03-24 | 北京市商汤科技开发有限公司 | Image processing method, image processing device and computer readable storage medium |
| CN111008974A (en) * | 2019-11-22 | 2020-04-14 | 浙江飞图影像科技有限公司 | Multi-model fusion femoral neck fracture region positioning and segmentation method and system |
| CN111126333A (en) * | 2019-12-30 | 2020-05-08 | 齐齐哈尔大学 | Garbage classification method based on light convolutional neural network |
| CN111210909A (en) * | 2020-01-13 | 2020-05-29 | 青岛大学附属医院 | Deep neural network-based rectal cancer T stage automatic diagnosis system and construction method thereof |
| WO2020108525A1 (en) * | 2018-11-30 | 2020-06-04 | 腾讯科技（深圳）有限公司 | Image segmentation method and apparatus, diagnosis system, storage medium, and computer device |
| CN111292289A (en) * | 2018-12-07 | 2020-06-16 | 中国科学院深圳先进技术研究院 | CT lung tumor segmentation method, device, equipment and medium based on segmentation network |
| CN111289251A (en) * | 2020-02-27 | 2020-06-16 | 湖北工业大学 | Rolling bearing fine-grained fault identification method |
| CN111340767A (en) * | 2020-02-21 | 2020-06-26 | 四川大学华西医院 | Method and system for processing scalp positioning image of brain tumor |
| CN111476802A (en) * | 2020-04-09 | 2020-07-31 | 山东财经大学 | Medical image segmentation and tumor detection method and device based on dense convolution model and readable storage medium |
| CN111477298A (en) * | 2020-04-03 | 2020-07-31 | 北京易康医疗科技有限公司 | Method for tracking tumor position change in radiotherapy process |
| WO2020162834A1 (en) * | 2019-02-08 | 2020-08-13 | Singapore Health Services Pte Ltd | Method and system for classification and visualisation of 3d images |
| CN111612722A (en) * | 2020-05-26 | 2020-09-01 | 星际（重庆）智能装备技术研究院有限公司 | Low-illumination image processing method based on simplified Unet full-convolution neural network |
| US20200275857A1 (en) * | 2019-03-01 | 2020-09-03 | Siemens Healthcare Gmbh | Tumor Tissue Characterization using Multi-Parametric Magnetic Resonance Imaging |
| CN111754530A (en) * | 2020-07-02 | 2020-10-09 | 广东技术师范大学 | Prostate ultrasonic image segmentation and classification method |
| CN111932486A (en) * | 2019-05-13 | 2020-11-13 | 四川大学 | Brain glioma segmentation method based on 3D convolutional neural network |
| CN111973154A (en) * | 2020-08-20 | 2020-11-24 | 山东大学齐鲁医院 | Multi-point accurate material taking system, method and device for brain tumor |
| WO2020238902A1 (en) * | 2019-05-29 | 2020-12-03 | 腾讯科技（深圳）有限公司 | Image segmentation method, model training method, apparatuses, device and storage medium |
| CN112037171A (en) * | 2020-07-30 | 2020-12-04 | 西安电子科技大学 | Multi-modal feature fusion based multi-task MRI brain tumor image segmentation method |
| CN112200810A (en) * | 2020-09-30 | 2021-01-08 | 深圳市第二人民医院(深圳市转化医学研究院) | Multi-modal automated ventricular segmentation system and method of use thereof |
| WO2021026962A1 (en) * | 2019-08-13 | 2021-02-18 | Hong Kong Applied Science and Technology Research Institute Company Limited | Medical image segmentation based on mixed context cnn model |
| CN112529915A (en) * | 2020-12-17 | 2021-03-19 | 山东大学 | Brain tumor image segmentation method and system |
| CN112634192A (en) * | 2020-09-22 | 2021-04-09 | 广东工业大学 | Cascaded U-N Net brain tumor segmentation method combined with wavelet transformation |
| CN112767417A (en) * | 2021-01-20 | 2021-05-07 | 合肥工业大学 | Multi-modal image segmentation method based on cascaded U-Net network |
| CN112837276A (en) * | 2021-01-20 | 2021-05-25 | 重庆邮电大学 | Brain glioma segmentation method based on cascaded deep neural network model |
| CN112862761A (en) * | 2021-01-20 | 2021-05-28 | 清华大学深圳国际研究生院 | Brain tumor MRI image segmentation method and system based on deep neural network |
| CN112927203A (en) * | 2021-02-25 | 2021-06-08 | 西北工业大学深圳研究院 | Glioma patient postoperative life prediction method based on multi-sequence MRI global information |
| CN109377505B (en) * | 2018-10-29 | 2021-07-06 | 哈尔滨理工大学 | MRI brain tumor image segmentation method based on multi-feature discrimination |
| CN113159171A (en) * | 2021-04-20 | 2021-07-23 | 复旦大学 | Plant leaf image fine classification method based on counterstudy |
| CN113361654A (en) * | 2021-07-12 | 2021-09-07 | 广州天鹏计算机科技有限公司 | Image identification method and system based on machine learning |
| CN114332547A (en) * | 2022-03-17 | 2022-04-12 | 浙江太美医疗科技股份有限公司 | Medical object classification method and apparatus, electronic device, and storage medium |
| CN116258671A (en) * | 2022-12-26 | 2023-06-13 | 中山大学肿瘤防治中心（中山大学附属肿瘤医院、中山大学肿瘤研究所） | MR image-based intelligent sketching method, system, equipment and storage medium |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105447872A (en) * | 2015-12-03 | 2016-03-30 | 中山大学 | Method for automatically identifying liver tumor type in ultrasonic image |
| CN106296699A (en) * | 2016-08-16 | 2017-01-04 | 电子科技大学 | Cerebral tumor dividing method based on deep neural network and multi-modal MRI image |
| CN106326937A (en) * | 2016-08-31 | 2017-01-11 | 郑州金惠计算机系统工程有限公司 | Convolutional neural network based crowd density distribution estimation method |
| CN106780498A (en) * | 2016-11-30 | 2017-05-31 | 南京信息工程大学 | Based on point depth convolutional network epithelium and matrix organization's automatic division method pixel-by-pixel |
| CN106780507A (en) * | 2016-11-24 | 2017-05-31 | 西北工业大学 | A kind of sliding window fast target detection method based on super-pixel segmentation |
| CN107016681A (en) * | 2017-03-29 | 2017-08-04 | 浙江师范大学 | Brain MRI lesion segmentation approach based on full convolutional network |
| CN107220980A (en) * | 2017-05-25 | 2017-09-29 | 重庆理工大学 | A kind of MRI image brain tumor automatic division method based on full convolutional network |
| CN107274451A (en) * | 2017-05-17 | 2017-10-20 | 北京工业大学 | Isolator detecting method and device based on shared convolutional neural networks |
| CN107463906A (en) * | 2017-08-08 | 2017-12-12 | 深图（厦门）科技有限公司 | The method and device of Face datection |
| CN107464250A (en) * | 2017-07-03 | 2017-12-12 | 深圳市第二人民医院 | Tumor of breast automatic division method based on three-dimensional MRI image |
| CN108062756A (en) * | 2018-01-29 | 2018-05-22 | 重庆理工大学 | Image, semantic dividing method based on the full convolutional network of depth and condition random field |
-
2018
- 2018-04-04 CN CN201810300057.1A patent/CN108492297B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105447872A (en) * | 2015-12-03 | 2016-03-30 | 中山大学 | Method for automatically identifying liver tumor type in ultrasonic image |
| CN106296699A (en) * | 2016-08-16 | 2017-01-04 | 电子科技大学 | Cerebral tumor dividing method based on deep neural network and multi-modal MRI image |
| CN106326937A (en) * | 2016-08-31 | 2017-01-11 | 郑州金惠计算机系统工程有限公司 | Convolutional neural network based crowd density distribution estimation method |
| CN106780507A (en) * | 2016-11-24 | 2017-05-31 | 西北工业大学 | A kind of sliding window fast target detection method based on super-pixel segmentation |
| CN106780498A (en) * | 2016-11-30 | 2017-05-31 | 南京信息工程大学 | Based on point depth convolutional network epithelium and matrix organization's automatic division method pixel-by-pixel |
| CN107016681A (en) * | 2017-03-29 | 2017-08-04 | 浙江师范大学 | Brain MRI lesion segmentation approach based on full convolutional network |
| CN107274451A (en) * | 2017-05-17 | 2017-10-20 | 北京工业大学 | Isolator detecting method and device based on shared convolutional neural networks |
| CN107220980A (en) * | 2017-05-25 | 2017-09-29 | 重庆理工大学 | A kind of MRI image brain tumor automatic division method based on full convolutional network |
| CN107464250A (en) * | 2017-07-03 | 2017-12-12 | 深圳市第二人民医院 | Tumor of breast automatic division method based on three-dimensional MRI image |
| CN107463906A (en) * | 2017-08-08 | 2017-12-12 | 深图（厦门）科技有限公司 | The method and device of Face datection |
| CN108062756A (en) * | 2018-01-29 | 2018-05-22 | 重庆理工大学 | Image, semantic dividing method based on the full convolutional network of depth and condition random field |
Non-Patent Citations (4)
| Title |
|---|
| DARKO ZIKIC ET AL: "Segmentation of Brain Tumor Tissues with Convolutional Neural Networks", 《MICCAI-BRATS CHALLENGE》 * |
| SÉRGIO PEREIRA ET AL: "Brain Tumor Segmentation Using Convolutional Neural Networks in MRI Images", 《IEEE TRANSACTIONS ON MEDICAL IMAGING》 * |
| SHAOGUO CUI ET AL: "Automatic Semantic Segmentation of Brain Gliomas from MRI Images Using a Deep Cascaded Neural Network", 《JOURNAL OF HEALTHCARE ENGINEERING》 * |
| SHAOGUO CUI ET AL: "Brain Tumor Automatic Segmentation Using Fully Convolutional Networks", 《JOURNAL OF MEDICAL IMAGING AND HEALTH INFORMATICS》 * |
Cited By (76)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110910335A (en) * | 2018-09-15 | 2020-03-24 | 北京市商汤科技开发有限公司 | Image processing method, image processing device and computer readable storage medium |
| CN110910335B (en) * | 2018-09-15 | 2023-02-24 | 北京市商汤科技开发有限公司 | Image processing method, image processing device and computer readable storage medium |
| CN109493317B (en) * | 2018-09-25 | 2020-07-07 | 哈尔滨理工大学 | 3D multi-vertebra segmentation method based on cascade convolution neural network |
| CN109493317A (en) * | 2018-09-25 | 2019-03-19 | 哈尔滨理工大学 | The more vertebra dividing methods of 3D based on concatenated convolutional neural network |
| US11403763B2 (en) | 2018-10-16 | 2022-08-02 | Tencent Technology (Shenzhen) Company Limited | Image segmentation method and apparatus, computer device, and storage medium |
| CN109360210B (en) * | 2018-10-16 | 2019-10-25 | 腾讯科技（深圳）有限公司 | Image partition method, device, computer equipment and storage medium |
| CN109360210A (en) * | 2018-10-16 | 2019-02-19 | 腾讯科技（深圳）有限公司 | Image partition method, device, computer equipment and storage medium |
| CN109377505B (en) * | 2018-10-29 | 2021-07-06 | 哈尔滨理工大学 | MRI brain tumor image segmentation method based on multi-feature discrimination |
| CN109472784A (en) * | 2018-10-31 | 2019-03-15 | 安徽医学高等专科学校 | Based on the recognition methods for cascading full convolutional network pathological image mitotic cell |
| CN109410289B (en) * | 2018-11-09 | 2021-11-12 | 中国科学院精密测量科学与技术创新研究院 | Deep learning high undersampling hyperpolarized gas lung MRI reconstruction method |
| CN109410289A (en) * | 2018-11-09 | 2019-03-01 | 中国科学院武汉物理与数学研究所 | A kind of high lack sampling hyperpolarized gas lung MRI method for reconstructing of deep learning |
| US11954863B2 (en) | 2018-11-30 | 2024-04-09 | Tencent Technology (Shenzhen) Company Limited | Image segmentation method and apparatus, diagnosis system, storage medium, and computer device |
| WO2020108525A1 (en) * | 2018-11-30 | 2020-06-04 | 腾讯科技（深圳）有限公司 | Image segmentation method and apparatus, diagnosis system, storage medium, and computer device |
| CN111292289A (en) * | 2018-12-07 | 2020-06-16 | 中国科学院深圳先进技术研究院 | CT lung tumor segmentation method, device, equipment and medium based on segmentation network |
| CN111292289B (en) * | 2018-12-07 | 2023-09-26 | 中国科学院深圳先进技术研究院 | CT lung tumor segmentation method, device, equipment and medium based on segmentation network |
| CN109886929B (en) * | 2019-01-24 | 2023-07-18 | 江苏大学 | MRI tumor voxel detection method based on convolutional neural network |
| CN109886929A (en) * | 2019-01-24 | 2019-06-14 | 江苏大学 | A kind of MRI tumour voxel detection method based on convolutional neural networks |
| WO2020162834A1 (en) * | 2019-02-08 | 2020-08-13 | Singapore Health Services Pte Ltd | Method and system for classification and visualisation of 3d images |
| CN111640118B (en) * | 2019-03-01 | 2024-03-01 | 西门子医疗有限公司 | Tumor tissue characterization using multiparameter magnetic resonance imaging |
| CN111640118A (en) * | 2019-03-01 | 2020-09-08 | 西门子医疗有限公司 | Tumor tissue characterization using multi-parameter magnetic resonance imaging |
| US20200275857A1 (en) * | 2019-03-01 | 2020-09-03 | Siemens Healthcare Gmbh | Tumor Tissue Characterization using Multi-Parametric Magnetic Resonance Imaging |
| CN109902748A (en) * | 2019-03-04 | 2019-06-18 | 中国计量大学 | A kind of image, semantic dividing method based on the full convolutional neural networks of fusion of multi-layer information |
| CN110136133A (en) * | 2019-03-11 | 2019-08-16 | 嘉兴深拓科技有限公司 | A kind of brain tumor dividing method based on convolutional neural networks |
| CN110120033A (en) * | 2019-04-12 | 2019-08-13 | 天津大学 | Based on improved U-Net neural network three-dimensional brain tumor image partition method |
| CN110097567A (en) * | 2019-04-18 | 2019-08-06 | 天津大学 | In conjunction with the three-dimensional brain tumor image partition method of improved FCNN and level set |
| CN110084823A (en) * | 2019-04-18 | 2019-08-02 | 天津大学 | Three-dimensional brain tumor image partition method based on cascade anisotropy FCNN |
| CN110047068A (en) * | 2019-04-19 | 2019-07-23 | 山东大学 | MRI brain tumor dividing method and system based on pyramid scene analysis network |
| CN110097550A (en) * | 2019-05-05 | 2019-08-06 | 电子科技大学 | A kind of medical image cutting method and system based on deep learning |
| CN111932486A (en) * | 2019-05-13 | 2020-11-13 | 四川大学 | Brain glioma segmentation method based on 3D convolutional neural network |
| US11900613B2 (en) | 2019-05-29 | 2024-02-13 | Tencent Technology (Shenzhen) Company Limited | Image segmentation method and apparatus, model training method and apparatus, device, and storage medium |
| WO2020238902A1 (en) * | 2019-05-29 | 2020-12-03 | 腾讯科技（深圳）有限公司 | Image segmentation method, model training method, apparatuses, device and storage medium |
| CN110349170A (en) * | 2019-07-13 | 2019-10-18 | 长春工业大学 | A kind of full connection CRF cascade FCN and K mean value brain tumor partitioning algorithm |
| CN110427954A (en) * | 2019-07-26 | 2019-11-08 | 中国科学院自动化研究所 | The image group feature extracting method of multizone based on tumor imaging |
| WO2021026962A1 (en) * | 2019-08-13 | 2021-02-18 | Hong Kong Applied Science and Technology Research Institute Company Limited | Medical image segmentation based on mixed context cnn model |
| US10937158B1 (en) | 2019-08-13 | 2021-03-02 | Hong Kong Applied Science and Technology Research Institute Company Limited | Medical image segmentation based on mixed context CNN model |
| CN110504032B (en) * | 2019-08-23 | 2022-09-09 | 元码基因科技(无锡)有限公司 | Method for predicting tumor mutation load based on image processing of hematoxylin-eosin staining tablet |
| CN110504032A (en) * | 2019-08-23 | 2019-11-26 | 元码基因科技(无锡)有限公司 | The method for predicting Tumor mutations load based on the image procossing of hematoxylin-eosin dye piece |
| CN110507288A (en) * | 2019-08-29 | 2019-11-29 | 重庆大学 | Vision based on one-dimensional convolutional neural networks induces motion sickness detection method |
| CN110533683A (en) * | 2019-08-30 | 2019-12-03 | 东南大学 | A kind of image group analysis method merging traditional characteristic and depth characteristic |
| CN110533683B (en) * | 2019-08-30 | 2022-04-29 | 东南大学 | Image omics analysis method fusing traditional features and depth features |
| CN110706214A (en) * | 2019-09-23 | 2020-01-17 | 东南大学 | Three-dimensional U-Net brain tumor segmentation method fusing condition randomness and residual error |
| CN110706214B (en) * | 2019-09-23 | 2022-06-17 | 东南大学 | Three-dimensional U-Net brain tumor segmentation method fusing condition randomness and residual error |
| CN111008974A (en) * | 2019-11-22 | 2020-04-14 | 浙江飞图影像科技有限公司 | Multi-model fusion femoral neck fracture region positioning and segmentation method and system |
| CN111126333A (en) * | 2019-12-30 | 2020-05-08 | 齐齐哈尔大学 | Garbage classification method based on light convolutional neural network |
| CN111126333B (en) * | 2019-12-30 | 2022-07-26 | 齐齐哈尔大学 | Garbage classification method based on light convolutional neural network |
| CN111210909A (en) * | 2020-01-13 | 2020-05-29 | 青岛大学附属医院 | Deep neural network-based rectal cancer T stage automatic diagnosis system and construction method thereof |
| CN111340767A (en) * | 2020-02-21 | 2020-06-26 | 四川大学华西医院 | Method and system for processing scalp positioning image of brain tumor |
| CN111340767B (en) * | 2020-02-21 | 2023-12-12 | 四川大学华西医院 | Brain tumor scalp positioning image processing method and system |
| CN111289251A (en) * | 2020-02-27 | 2020-06-16 | 湖北工业大学 | Rolling bearing fine-grained fault identification method |
| CN111477298A (en) * | 2020-04-03 | 2020-07-31 | 北京易康医疗科技有限公司 | Method for tracking tumor position change in radiotherapy process |
| CN111476802A (en) * | 2020-04-09 | 2020-07-31 | 山东财经大学 | Medical image segmentation and tumor detection method and device based on dense convolution model and readable storage medium |
| CN111476802B (en) * | 2020-04-09 | 2022-10-11 | 山东财经大学 | Medical image segmentation and tumor detection method, equipment and readable storage medium |
| CN111612722B (en) * | 2020-05-26 | 2023-04-18 | 星际（重庆）智能装备技术研究院有限公司 | Low-illumination image processing method based on simplified Unet full-convolution neural network |
| CN111612722A (en) * | 2020-05-26 | 2020-09-01 | 星际（重庆）智能装备技术研究院有限公司 | Low-illumination image processing method based on simplified Unet full-convolution neural network |
| CN111754530A (en) * | 2020-07-02 | 2020-10-09 | 广东技术师范大学 | Prostate ultrasonic image segmentation and classification method |
| CN111754530B (en) * | 2020-07-02 | 2023-11-28 | 广东技术师范大学 | Prostate ultrasonic image segmentation classification method |
| CN112037171B (en) * | 2020-07-30 | 2023-08-15 | 西安电子科技大学 | Multi-mode feature fusion-based multi-task MRI brain tumor image segmentation method |
| CN112037171A (en) * | 2020-07-30 | 2020-12-04 | 西安电子科技大学 | Multi-modal feature fusion based multi-task MRI brain tumor image segmentation method |
| CN111973154A (en) * | 2020-08-20 | 2020-11-24 | 山东大学齐鲁医院 | Multi-point accurate material taking system, method and device for brain tumor |
| CN111973154B (en) * | 2020-08-20 | 2022-05-24 | 山东大学齐鲁医院 | Multi-point accurate material taking system, method and device for brain tumor |
| CN112634192A (en) * | 2020-09-22 | 2021-04-09 | 广东工业大学 | Cascaded U-N Net brain tumor segmentation method combined with wavelet transformation |
| CN112634192B (en) * | 2020-09-22 | 2023-10-13 | 广东工业大学 | Cascaded U-N Net brain tumor segmentation method combining wavelet transformation |
| CN112200810A (en) * | 2020-09-30 | 2021-01-08 | 深圳市第二人民医院(深圳市转化医学研究院) | Multi-modal automated ventricular segmentation system and method of use thereof |
| CN112200810B (en) * | 2020-09-30 | 2023-11-14 | 深圳市第二人民医院(深圳市转化医学研究院) | Multi-modal automated ventricle segmentation system and method of use thereof |
| CN112529915A (en) * | 2020-12-17 | 2021-03-19 | 山东大学 | Brain tumor image segmentation method and system |
| CN112837276A (en) * | 2021-01-20 | 2021-05-25 | 重庆邮电大学 | Brain glioma segmentation method based on cascaded deep neural network model |
| CN112837276B (en) * | 2021-01-20 | 2023-09-29 | 重庆邮电大学 | Brain glioma segmentation method based on cascade deep neural network model |
| CN112767417B (en) * | 2021-01-20 | 2022-09-13 | 合肥工业大学 | Multi-modal image segmentation method based on cascaded U-Net network |
| CN112862761A (en) * | 2021-01-20 | 2021-05-28 | 清华大学深圳国际研究生院 | Brain tumor MRI image segmentation method and system based on deep neural network |
| CN112767417A (en) * | 2021-01-20 | 2021-05-07 | 合肥工业大学 | Multi-modal image segmentation method based on cascaded U-Net network |
| CN112927203A (en) * | 2021-02-25 | 2021-06-08 | 西北工业大学深圳研究院 | Glioma patient postoperative life prediction method based on multi-sequence MRI global information |
| CN113159171A (en) * | 2021-04-20 | 2021-07-23 | 复旦大学 | Plant leaf image fine classification method based on counterstudy |
| CN113361654A (en) * | 2021-07-12 | 2021-09-07 | 广州天鹏计算机科技有限公司 | Image identification method and system based on machine learning |
| CN114332547A (en) * | 2022-03-17 | 2022-04-12 | 浙江太美医疗科技股份有限公司 | Medical object classification method and apparatus, electronic device, and storage medium |
| CN116258671B (en) * | 2022-12-26 | 2023-08-29 | 中山大学肿瘤防治中心（中山大学附属肿瘤医院、中山大学肿瘤研究所） | MR image-based intelligent sketching method, system, equipment and storage medium |
| CN116258671A (en) * | 2022-12-26 | 2023-06-13 | 中山大学肿瘤防治中心（中山大学附属肿瘤医院、中山大学肿瘤研究所） | MR image-based intelligent sketching method, system, equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108492297B (en) | 2021-11-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108492297A (en) | The MRI brain tumors positioning for cascading convolutional network based on depth and dividing method in tumor | |
| Greenwald et al. | Whole-cell segmentation of tissue images with human-level performance using large-scale data annotation and deep learning | |
| Yan et al. | Breast cancer histopathological image classification using a hybrid deep neural network | |
| Marini et al. | Semi-supervised training of deep convolutional neural networks with heterogeneous data and few local annotations: An experiment on prostate histopathology image classification | |
| Duggento et al. | Deep computational pathology in breast cancer | |
| CN106408001B (en) | Area-of-interest rapid detection method based on depth core Hash | |
| CN108764241A (en) | Divide method, apparatus, computer equipment and the storage medium of near end of thighbone | |
| CN108062753A (en) | The adaptive brain tumor semantic segmentation method in unsupervised domain based on depth confrontation study | |
| CN109493346A (en) | It is a kind of based on the gastric cancer pathology sectioning image dividing method more lost and device | |
| CN109671102A (en) | A kind of composite type method for tracking target based on depth characteristic fusion convolutional neural networks | |
| Guo et al. | Domain knowledge based brain tumor segmentation and overall survival prediction | |
| Gehlot et al. | Ednfc-net: Convolutional neural network with nested feature concatenation for nuclei-instance segmentation | |
| CN114600155A (en) | Weakly supervised multitask learning for cell detection and segmentation | |
| Lokhande et al. | Carcino-Net: A deep learning framework for automated Gleason grading of prostate biopsies | |
| CN101699515A (en) | Multi-elite immune quantum clustering-based medical image segmenting system and multi-elite immune quantum clustering-based medical image segmenting method | |
| Van Buren et al. | Artificial intelligence and deep learning to map immune cell types in inflamed human tissue | |
| Khalid et al. | Deepcens: An end-to-end pipeline for cell and nucleus segmentation in microscopic images | |
| Tyagi et al. | LCSCNet: A multi-level approach for lung cancer stage classification using 3D dense convolutional neural networks with concurrent squeeze-and-excitation module | |
| Han et al. | A deep learning quantification algorithm for HER2 scoring of gastric cancer | |
| Cao et al. | 3D convolutional neural networks fusion model for lung nodule detection onclinical CT scans | |
| Li et al. | A deeply supervised convolutional neural network for brain tumor segmentation | |
| Xiang et al. | Segmentation method of multiple sclerosis lesions based on 3D‐CNN networks | |
| Sun et al. | Detection of breast tumour tissue regions in histopathological images using convolutional neural networks | |
| Wang et al. | Fast cancer metastasis location based on dual magnification hard example mining network in whole-slide images | |
| Khaliliboroujeni et al. | End-to-end metastasis detection of breast cancer from histopathology whole slide images |
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 | ||
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20200518 Address after: 400000 No. 12 Chen Cheng Road, Shapingba District, Chongqing Applicant after: CHONGQING NORMAL University Address before: No. 69 lijiatuo Chongqing District of Banan City Road 400054 red Applicant before: Chongqing University of Technology |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |