CN109754403A - Tumour automatic division method and system in a kind of CT image - Google Patents
Tumour automatic division method and system in a kind of CT image Download PDFInfo
- Publication number
- CN109754403A CN109754403A CN201811440970.8A CN201811440970A CN109754403A CN 109754403 A CN109754403 A CN 109754403A CN 201811440970 A CN201811440970 A CN 201811440970A CN 109754403 A CN109754403 A CN 109754403A
- Authority
- CN
- China
- Prior art keywords
- data
- layer
- image
- enhancing
- constructed
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 43
- 206010028980 Neoplasm Diseases 0.000 title claims abstract description 29
- 230000002708 enhancing effect Effects 0.000 claims abstract description 50
- 238000010606 normalization Methods 0.000 claims abstract description 31
- 238000012545 processing Methods 0.000 claims abstract description 27
- 230000011218 segmentation Effects 0.000 claims abstract description 16
- 238000011946 reduction process Methods 0.000 claims abstract description 9
- 238000005070 sampling Methods 0.000 claims description 24
- 230000006870 function Effects 0.000 claims description 18
- 230000009467 reduction Effects 0.000 claims description 13
- 206010019695 Hepatic neoplasm Diseases 0.000 claims description 12
- 208000014018 liver neoplasm Diseases 0.000 claims description 12
- 238000003860 storage Methods 0.000 claims description 12
- 238000012549 training Methods 0.000 claims description 11
- 238000004590 computer program Methods 0.000 claims description 10
- 210000004185 liver Anatomy 0.000 claims description 10
- 238000009826 distribution Methods 0.000 claims description 8
- 238000003709 image segmentation Methods 0.000 claims description 3
- 238000013519 translation Methods 0.000 claims description 3
- 230000008569 process Effects 0.000 abstract description 3
- 230000008859 change Effects 0.000 description 5
- 238000013527 convolutional neural network Methods 0.000 description 5
- 239000000047 product Substances 0.000 description 5
- 238000000605 extraction Methods 0.000 description 4
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000003475 lamination Methods 0.000 description 3
- 238000004891 communication Methods 0.000 description 2
- 238000002591 computed tomography Methods 0.000 description 2
- 230000008878 coupling Effects 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 230000002440 hepatic effect Effects 0.000 description 2
- 238000005192 partition Methods 0.000 description 2
- 241000208340 Araliaceae Species 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 230000005489 elastic deformation Effects 0.000 description 1
- 230000009123 feedback regulation Effects 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- PCHJSUWPFVWCPO-UHFFFAOYSA-N gold Chemical compound [Au] PCHJSUWPFVWCPO-UHFFFAOYSA-N 0.000 description 1
- 239000010931 gold Substances 0.000 description 1
- 229910052737 gold Inorganic materials 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 230000014759 maintenance of location Effects 0.000 description 1
- 230000000877 morphologic effect Effects 0.000 description 1
- 230000002980 postoperative effect Effects 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 230000008439 repair process Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 239000013589 supplement Substances 0.000 description 1
Classifications
-
- 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
- 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
-
- 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
-
- 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
- G06T2207/30096—Tumor; Lesion
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses the tumour automatic division methods and system in a kind of CT image, belong to field of medical image processing, for dividing the tumor focus region in CT image, solve the problems, such as that the segmentation precision of CT image is lower, expand comprising: carry out data enhancing to raw image data, obtains enhancing expanding data；Enhancing expanding data is normalized, normalization data is obtained；Normalization data is entered to the processing network trained, obtains segmented image；Noise reduction process is done to segmented image；To reduce the difference of different CT machine scanning bring difference original images, the scope of application and precision of processing network processes result are improved.
Description
Technical field
The present invention relates to the tumour automatic division methods in technical field of medical image processing more particularly to a kind of CT image
And system.
Background technique
Liver neoplasm is to threaten the major disease of human health, and the early detection of liver neoplasm is accurately measured and examined clinic
It controls significant with treatment.CT (Computed Tomography, Chinese are CT scan) is used as one
The detection mode of kind Cheap highly effective, has been increasingly becoming the conventional means of clinically Diagnosis of Hepatic Tumors.Fast accurate from liver
Tumor focus region is partitioned into CT image, it is not only helpful to operation program type, but also to tumor region position in art
Precise positioning has great practical value with excision and the assessment of postoperative chemicotherapy effect.
How tumor focus region is partitioned into fast accurate from CT image for liver, be doctor and scholars study one
Big project has the research of the liver neoplasm automatic division method based on deep learning framework, such as convolutional neural networks at present
(Convolutional Neural Network, CNN), full convolutional neural networks (Fully Convolutional
Network, FCN) etc..
But the network structure of CNN is more single, and the addition of full articulamentum is so that network entirety training parameter is more huge
Greatly, calculating is complex, contains much information, net training time is longer, and segmentation precision is poor.Afterwards based on the whole of this improved FCN
Body segmentation precision is still lower, and classification pixel-based does not account for the relationship between pixel, lacks Space Consistency.Hereafter
Some segmentation network architectures are proposed on the basis of these classic network frameworks again, but segmentation precision is still to be improved.
Summary of the invention
The main purpose of the present invention is to provide the tumour automatic division methods and system in a kind of CT image, it is intended to solve
The technical problem lower to the segmentation precision of CT image in the prior art.
To achieve the above object, first aspect present invention provides the tumour automatic division method in a kind of CT image, comprising:
Data enhancing is carried out to raw image data to expand, and obtains enhancing expanding data；Enhancing expanding data is normalized,
Obtain normalization data；Normalization data is entered to the processing network trained, obtains segmented image；Segmented image is done at noise reduction
Reason.
Further, described to carry out data enhancing to raw image data to expand including: based on translation rotation principle to original
Beginning image data carries out enhancing expansion, or carries out enhancing expansion to raw image data based on Stochastic Elasticity deformation principle.
Further, it includes: according to linear normalization principle to increasing that described pair of enhancing expanding data, which is normalized,
Strong expanding data and liver neoplasm goldstandard are normalized, and obtain linear normalization image data；Linearly return to described
The normalized that one change image data carries out data distribution obtains normalization data.
Further, the training method of the processing network trained includes: building the first convolutional layer, the first modified line
Property elementary layer, pond layer, dropout layers and down-sampling layer, formed constricted path；According to first convolutional layer, described first
Correct linear elementary layer, the pond layer, dropout layers and down-sampling layer extract and encode the noise reduction data, generate coding
Data；The second convolutional layer, the linear elementary layer of the second amendment and up-sampling layer are constructed, expansion path is formed；According to the volume Two
Lamination, described second correct linear elementary layer and up-sample layer decoder and divide encoding samples data, generate decoding data；Building
Probability output layer exports the decoding data.
Further, the first convolutional layer of the building, the first linear elementary layer of amendment, pond layer, dropout layers and under adopt
Sample layer includes: successively to construct three the first 3*3 convolutional layers, and construct two the first 2*2 convolutional layers, and the first 3* successively constructed
The feature port number of 3 convolutional layers and the first 2*2 convolutional layer successively increases one times since 64；After each first 3*3 convolutional layer
Construct the linear elementary layer of amendment；Between adjacent first 3*3 convolutional layer, it is between adjacent first 2*2 convolutional layer and adjacent
Pond layer is constructed between first 3*3 convolutional layer and the first 2*2 convolutional layer；First constructed after the convolutional layer of the first 3*3
First dropout layers are constructed after the convolutional layer of first 2*2, and the convolution of second the first 2*2 is constructed after dropout layers of building
Layer；Second dropout layers are constructed after second the first 2*2 convolutional layer；The second convolutional layer of the building, the second amendment are linear
Elementary layer and up-sampling layer include: successively four volume Two bases of building after second dropout layer of building, by above adopting
Sample layer and two the 2nd 3*3 convolutional layers are constituted, and the volume Two base feature port number successively constructed is with second the first 2*2 volumes
Lamination is to begin, and feature port number successively reduces one times；The building up-sampling layer before each 2nd 3*3 convolutional layer.
Further, it is distributed, and is cascaded in mirror image between the constricted path and the expansion path.
Further, the training method of the processing network trained further include: building confrontation network, the building pair
Anti- network includes: the first data of building, first data by segmentation goldstandard as standard reference, and by liver neoplasm two-value
Goldstandard and Original Liver gray level image dot product obtain；The second data are constructed, second data are by the segmented image and original
Beginning liver intensity picture point is multiplied to be arrived；Loss function is constructed, and by the first data and the second data entrance loss function, captures and
From the length space characteristics of the Pixel-level of different levels.
Second aspect of the present invention provides the automatic segmenting system of tumour in a kind of CT image, comprising: enhancing enlargement module is used
In carrying out enhancing expansion to raw image data, enhancing expanding data is obtained；Normalized module, for expanding number to enhancing
According to being normalized, normalization data is obtained；Image segmentation module, for noise reduction data to be inputted the processing net trained
Network obtains segmented image；Noise reduction module, for doing noise reduction process to segmented image.
Third aspect present invention provides a kind of electronic device, comprising: memory, processor and is stored on the memory
And the computer program that can be run on the processor, which is characterized in that when the processor executes the computer program,
Realize any one the method among the above.
Fourth aspect present invention provides a kind of computer readable storage medium, a kind of computer readable storage medium, thereon
It is stored with computer program, which is characterized in that when the computer program is executed by processor, realize any one among the above
The method.
The present invention provides the tumour automatic division method in a kind of CT image, and beneficial effect is: by original image
Data carry out enhancing expansion, do not influence on the basis that the robust for meeting variation grayscale information requires and to handle network to real information
Processing on, can reach abundant information amount to improve the extensive effect of parted pattern as a result, enable parted pattern to fit
For wider data set, the scope of application of processing network is improved；Due to the gray value for the original image that different patients obtain
It differs greatly, therefore by normalized, processing of the dividing processing network to image data can be facilitated, to reduce different CT
Machine scans the difference of bring difference original image, improves the scope of application and precision of processing network processes result.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
Some embodiments of invention for those skilled in the art without creative efforts, can also basis
These attached drawings obtain other attached drawings.
Fig. 1 is the structural schematic block diagram of the tumour automatic division method in CT of embodiment of the present invention image；
The structure of processing network of the Fig. 2 to have trained in the tumour automatic division method in CT of embodiment of the present invention image is shown
It is intended to；
Fig. 3 be CT of embodiment of the present invention image in tumour automatic division method in fight network structural schematic diagram；
Fig. 4 is the structural schematic block diagram of electronic device of the embodiment of the present invention.
Specific embodiment
In order to make the invention's purpose, features and advantages of the invention more obvious and easy to understand, below in conjunction with the present invention
Attached drawing in embodiment, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described reality
Applying example is only a part of the embodiment of the present invention, and not all embodiments.Based on the embodiments of the present invention, those skilled in the art
Member's every other embodiment obtained without making creative work, shall fall within the protection scope of the present invention.
Referring to Fig. 1, for the tumour automatic division method in a kind of CT image, comprising: S1, carried out to raw image data
Data enhancing is expanded, and enhancing expanding data is obtained；S2, enhancing expanding data is normalized, obtains normalization data；
S3, normalization data is entered to the processing network trained, obtains segmented image；S4, noise reduction process is done to segmented image.
Raw image data is carried out data enhancing to expand including: to carry out raw image data based on translation rotation principle
Enhancing is expanded, or carries out enhancing expansion to raw image data based on Stochastic Elasticity deformation principle.
Since raw image data is more dull, the data information of only one vertical positive direction, for the place trained
It being more short of for reason network information richness, dull data information causes the Generalization Capability for handling e-learning weaker, because
This needs to expand raw image data and enhanced, and obtains enhancing expanding data, to enhance the general of processing e-learning
Change performance, in the present embodiment, the process of data enhancing abides by invariance principle, and it is flat for meeting the concrete operations of invariance principle
It moves, rotation and elastic deformation not only believe the training data gray scale of original image after handling initial data by invariance principle
The robustness of breath, moreover it is possible to achieve the purpose that informative, and not influence to handle study of the network to real information, enhance place
The Generalization Capability for managing network enables to handle network suitable for wider data set；In the present embodiment, by original graph
As data are translated, rotate and elastically-deformable mode carries out enhancing expansion to raw image data, obtains enhancing and expand number
According to.
It includes: according to linear normalization principle to enhancing expanding data and liver that enhancing expanding data, which is normalized,
Dirty tumour goldstandard is normalized, and obtains linear normalization image data；Linear normalization image data is counted
Normalization data is obtained according to the normalized of distribution.
Due to the difference of CT equipment, the grey value difference for the CT image being scanned to patient is larger, in this implementation
In example, by the way that enhancing expanding data and liver neoplasm goldstandard are normalized, thus facilitate the training of processing network,
Reduce the grey value difference of different CT equipment scanning bring CT images；Specifically, gradation data is returned using linear normalization
One changes to [0,225] section, linear normalization formula are as follows:
In linear normalization formula, XnormFor normalization data, X is enhancing expanding data, Xmax、XminRespectively enhance
The maximum value and minimum value that expanding data is concentrated；It in the present embodiment, is 0/255 binaryzation to gray scale before input processing network
Liver neoplasm data carry out 0-1 standardization and are used as tumour goldstandard, and data are higher than 0.5 setting divided by with 0.5 being threshold value after 255
It is 1, is set as 0 lower than 0.5.
After it will enhance expanding data and carry out linear normalization, the normalized of data distribution is carried out, data distribution
Normalized is as follows: the hepatic data to gray scale before input processing network in the section 0-255 carries out 0 mean value standardization, will increase
Strong EDS extended data set is normalized to the data set for the normal distribution that mean value is 0, variance is 1, and the normalization formula of data distribution is such as
Under:
In the normalization formula of data distribution, μ and σ are respectively the mean value and standard deviation of raw data set.
There is noise jamming to a certain extent during handling raw image data, and the figure of these noises is special
Sign is more obvious, and accounting example is smaller on the original image, it is therefore desirable to noise reduction process is done to segmented image, in the present embodiment
In, interference noise is removed by filter, caused by filtering using the library specialized medical image processing software SimpleITK carry out
Operation, noise is shown when Fei Telei (Feret) diameter is less than 7 based on the data statistics of morphologic information, corresponding at this time
Perimeter and occupied pixel quantity are all that minimum, the identification that can be filtered without influencing other tumours is influenced within the scope of noise,
Therefore using FeretDiameter=7 as threshold value, the part less than the value will be filtered, to realize the mesh of noise reduction filtering
, enable finally obtained segmented image more accurate.
The training method for the processing network trained includes: the first convolutional layer of building, the linear elementary layer of the first amendment, Chi Hua
Layer, dropout layers and down-sampling layer form constricted path；According to the first convolutional layer, the first linear elementary layer of amendment, pond layer,
Dropout layers and down-sampling layer extract and coded samples noise reduction data, generates coded data；It constructs the second convolutional layer, second repair
Linear positive elementary layer and up-sampling layer, form expansion path；According to the second convolutional layer, the linear elementary layer of the second amendment and up-sampling
Layer decoder and partition encoding data generate decoding data；Probability output layer is constructed, decoding data is exported.
Wherein, the method for making sample noise reduction data includes: to carry out data enhancing to sample image data to expand, and obtains sample
This enhancing expanding data；Sample enhancing expanding data is normalized, samples normalization data are obtained；To sample normalizing
Change data and do noise reduction process, obtains sample noise reduction data, and the increasing with raw image data is expanded in the enhancing of sample image data
Strong extending method is consistent, and sample enhancing expanding data is consistent with the enhancing normalization processing method of expanding data, sample normalizing
It is consistent with the noise reduction process method of normalization data to change data.
Construct the first convolutional layer, the first linear elementary layer of amendment, pond layer, dropout layers and down-sampling layer include: successively
Three the first 3*3 convolutional layers are constructed, and construct two the first 2*2 convolutional layers, and the first 3*3 convolutional layer successively constructed and first
The feature port number of 2*2 convolutional layer successively increases one times since 64；An amendment is constructed after each first 3*3 convolutional layer
Linear unit layer；Between adjacent first 3*3 convolutional layer, between adjacent first 2*2 convolutional layer and the first adjacent 3*3 convolution
Pond layer is constructed between layer and the first 2*2 convolutional layer；The volume of first the first 2*2 constructed after the convolutional layer of the first 3*3
First dropout layers are constructed after lamination, and the convolutional layer of second the first 2*2 is constructed after dropout layers of building；At second
Second dropout layers are constructed after first 2*2 convolutional layer；Construct the second convolutional layer, the linear elementary layer of the second amendment and up-sampling
Layer includes: successively to construct four volume Two bases after second dropout layers of building, by up-sampling layer and two second
3*3 convolutional layer is constituted, and the volume Two base feature port number successively constructed is to begin with second the first 2*2 convolutional layer, feature
Port number successively reduces one times；The building up-sampling layer before each 2nd 3*3 convolutional layer.
Referring to Fig. 2, specifically, handling the overall U-shaped symmetrical structure of network, being divided into two stages of coding-decoding, encode
Stage is characteristic extraction part, is completed by constricted path, predominantly U-shape network left part, constricted path and classical convolution mind
It is essentially identical through network operation, 5 block are broadly divided into, wherein three block mentioning for feature by two continuous 3x3
The maximum pondization operation for down-sampling of the convolution operation (RELU is met behind each convolution), a 2x2 that take forms, and
With the beginning of start image feature port number 64, after down-sampling operation of every progress, the feature port number of image is all doubled；
Latter two block, which then introduces dropout layers on the basis of block in front, prevents network training from over-fitting occur.Meanwhile
In constricted path, the thinking of residual error network is quoted simultaneously in the convolution feature extraction operation part of each block, by using one
A shortcut based on pixel superposition, the information after raw information to be passed through to the convolution operation and convolution operation of 1x1 directly connect
It connects, keeps the two port number consistent, can supplement and increase network information transmission capacity, improve the ability of e-learning feature.Decoding
Part is the part that feature is restored, and is mainly completed by expansion path, predominantly the right part of the U-shaped network, can be divided mainly into 4
A block+1 last probability output layers (sigmoid), this four block all by a up-sampling layer, (grasp by 2x2 deconvolution
Realize) and two 3x3 convolution (RELU is met behind each convolution), and when deconvolution of every progress up-sampling, image
Feature port number halves, and exports finally by the sigmoid layer of 1x1 to probability graph belonging to each pixel.
It is distributed, and is cascaded in mirror image between constricted path and expansion path, can be supplemented by mirror image
Loss of learning part, further enriches the network information.
The training method for the processing network trained further include: building confrontation network, building confrontation network includes: building the
One data, the first data by segmentation goldstandard as standard reference, and by liver neoplasm two-value goldstandard and Original Liver gray scale
Picture point is multiplied to be arrived；The second data are constructed, the second data are obtained by segmented image and Original Liver gray level image dot product；Building damage
Function is lost, and by the first data and the second data entrance loss function, captures the length space of the Pixel-level from different levels
Feature.
Referring to Fig. 3, specifically, the input for fighting the network architecture can strictly be divided into two parts.A part is segmentation gold mark
Standard is inputted as standard referring to part, by the liver neoplasm two-value goldstandard (ground truth) and Original Liver gray scale provided
Image multiplication (dot product) obtains, and is denoted as label_mask；A part is the input of segmentation neural network forecast part, most according to segmentation network
Whole two-value predicts segmentation result figure, and be multiplied (dot product) with Original Liver grayscale image, is denoted as output_mask.Fight network
Network structure is similar to the coded portion of segmentation network, and the network is using label_mask and output_mask as input, setting
Loss function is MAE (Mean Absolute Error, mean absolute error), which can capture well and come from
The length space characteristics of the Pixel-level of different levels (information including high, medium and low layer), so as to realize image feature information
Multi-level comparison correction.Network losses function is fought by calculating the gap between standard masks and prediction mask, and is combined
The loss function for dividing network, collectively as the adjustment function of final segmentation network, to realize confrontation network to generated
The feedback regulation of parted pattern weight updates, and reaches and advanced optimizes effect.
The building of loss function uses Dice coefficient as assessment, formula are as follows:
Wherein, s1, s2 are respectively actual value and predicted value, and smotth is one for increasing the ginseng of matched curve smoothness
Number, is similar to an infinitesimal variable, and the introducing of smotth keeps function more smooth.
The loss function for fighting network is MAE (Mean Absolute Error, mean absolute error):
Wherein L is the confrontation total number of plies of network,Come from the feature extraction of input the i-th layer network of goldstandard exposure mask
Image,Come from the feature-extraction images of input segmentation the i-th layer network of prediction mask.The loss function can be fine
Ground captures the length space characteristics of the Pixel-level from different levels (information including high, medium and low layer), so as to realize figure
As the multi-level comparison of characteristic information is corrected.
Whole loss function are as follows: loss=lmae-ldice。
The embodiment of the present application provides the automatic segmenting system of tumour in a kind of CT image, comprising: enhancing enlargement module is used for
Enhancing expansion is carried out to raw image data, obtains enhancing expanding data；Normalized module, for enhancing expanding data
It is normalized, obtains normalization data；Image segmentation module, for noise reduction data to be inputted the processing net trained
Network obtains segmented image；Noise reduction module, for doing noise reduction process to segmented image.
The embodiment of the present application provides a kind of electronic device, referring to Fig. 4, the electronic device includes: memory 601, processing
Device 602 and it is stored in the computer program that can be run on memory 601 and on processor 602, processor 602 executes the calculating
When machine program, the generation method of increment Density Estimator device described in embodiment of the aforementioned figures 1 to attached drawing 4 is realized.
Further, electronic device further include: at least one input equipment 603 and at least one output equipment 604.
Above-mentioned memory 601, processor 602, input equipment 603 and output equipment 604, are connected by bus 605.
Wherein, input equipment 603 concretely camera, touch panel, physical button or mouse etc..Output equipment
604 concretely display screens.
Memory 601 can be high random access memory body (RAM, Random Access Memory) memory,
It can be non-labile memory (non-volatile memory), such as magnetic disk storage.Memory 601 is for storing one
Group executable program code, processor 602 are coupled with memory 601.
Further, the embodiment of the present application also provides a kind of computer readable storage medium, the computer-readable storages
Medium can be in the electronic device being set in the various embodiments described above, which can be earlier figures 4
Memory 601 in illustrated embodiment.It is stored with computer program on the computer readable storage medium, the program is by processor
602 realize the generation method of increment Density Estimator device described in preceding method embodiment when executing.
Further, the computer can storage medium can also be USB flash disk, mobile hard disk, read-only memory 601 (ROM,
Read-Only Memory), RAM, the various media that can store program code such as magnetic or disk.
In several embodiments provided herein, it should be understood that disclosed method, it can be by others side
Formula is realized.For example, the division of the module, only a kind of logical function partition, can there is other division in actual implementation
Mode, such as multiple module or components can be combined or can be integrated into another system, or some features can be ignored, or
It does not execute.Another point, shown or discussed mutual coupling, direct-coupling or communication connection can be by some
The indirect coupling or communication connection of interface, device or module can be electrical property, mechanical or other forms.
The module as illustrated by the separation member may or may not be physically separated, aobvious as module
The component shown may or may not be physical module, it can and it is in one place, or may be distributed over multiple
On network module.Some or all of the modules therein can be selected to realize the mesh of this embodiment scheme according to the actual needs
's.
It, can also be in addition, each functional module in each embodiment of the present invention can integrate in a processing module
It is that modules physically exist alone, can also be integrated in two or more modules in a module.Above-mentioned integrated mould
Block both can take the form of hardware realization, can also be realized in the form of software function module.
If the integrated module is realized in the form of software function module and sells or use as independent product
When, it can store in a computer readable storage medium.Based on this understanding, technical solution of the present invention is substantially
The all or part of the part that contributes to existing technology or the technical solution can be in the form of software products in other words
It embodies, which is stored in a storage medium, including some instructions are used so that a computer
Equipment (can be personal computer, server or the network equipment etc.) executes the complete of each embodiment the method for the present invention
Portion or part steps.And storage medium above-mentioned includes: USB flash disk, mobile hard disk, read-only memory (ROM, Read-Only
Memory), random access memory (RAM, Random Access Memory), magnetic or disk etc. are various can store journey
The medium of sequence code.
It should be noted that for the various method embodiments described above, describing for simplicity, therefore, it is stated as a series of
Combination of actions, but those skilled in the art should understand that, the present invention is not limited by the sequence of acts described because
According to the present invention, certain steps can use other sequences or carry out simultaneously.Secondly, those skilled in the art should also know
It knows, the embodiments described in the specification are all preferred embodiments, and related actions and modules might not all be this hair
Necessary to bright.
In the above-described embodiments, it all emphasizes particularly on different fields to the description of each embodiment, there is no the portion being described in detail in some embodiment
Point, it may refer to the associated description of other embodiments.
The above are the description to tumour automatic division method and system in a kind of CT image provided by the present invention, for
Those skilled in the art, thought according to an embodiment of the present invention have change in specific embodiments and applications
Place, to sum up, the contents of this specification are not to be construed as limiting the invention.
Claims (10)
1. the tumour automatic division method in a kind of CT image characterized by comprising
Data enhancing is carried out to raw image data to expand, and obtains enhancing expanding data；
Enhancing expanding data is normalized, normalization data is obtained；
Normalization data is entered to the processing network trained, obtains segmented image；
Noise reduction process is done to segmented image.
2. the tumour automatic division method in CT image according to claim 1, which is characterized in that
It is described to include: to raw image data progress data enhancing expansion
Enhancing expansion is carried out to raw image data based on translation rotation principle, or based on Stochastic Elasticity deformation principle to original graph
As data carry out enhancing expansion.
3. the tumour automatic division method in CT image according to claim 1, which is characterized in that
Described pair of enhancing expanding data is normalized and includes:
Enhancing expanding data and liver neoplasm goldstandard are normalized according to linear normalization principle, linearly returned
One changes image data；
The normalized for carrying out data distribution to the linear normalization image data obtains normalization data.
4. the tumour automatic division method in CT image according to claim 1, which is characterized in that
The training method of the processing network trained includes:
The first convolutional layer, the linear elementary layer of the first amendment, pond layer, dropout layers and down-sampling layer are constructed, is formed and shrinks road
Diameter；
According to first convolutional layer, the linear elementary layer of first amendment, the pond layer, dropout layers and down-sampling layer
Simultaneously coded samples noise reduction data are extracted, coded data is generated；
The second convolutional layer, the linear elementary layer of the second amendment and up-sampling layer are constructed, expansion path is formed；
According to second convolutional layer, the linear elementary layer of second amendment and up-samples layer decoder and divide the coded number
According to generation decoding data；
Probability output layer is constructed, the decoding data is exported.
5. the tumour automatic division method in CT image according to claim 4, which is characterized in that
The first convolutional layer of the building, the linear elementary layer of the first amendment, pond layer, dropout layers and down-sampling layer include:
Three the first 3*3 convolutional layers are successively constructed, and construct two the first 2*2 convolutional layers, and the first 3*3 convolution successively constructed
The feature port number of layer and the first 2*2 convolutional layer successively increases one times since 64；
The linear elementary layer of amendment is constructed after each first 3*3 convolutional layer；
Between adjacent first 3*3 convolutional layer, between adjacent first 2*2 convolutional layer and adjacent the first 3*3 convolutional layer and first
Pond layer is constructed between 2*2 convolutional layer；
First dropout layers are constructed after the convolutional layer of first the first 2*2 constructed after the convolutional layer of the first 3*3,
The convolutional layer of second the first 2*2 is constructed after dropout layers of building；
Second dropout layers are constructed after second the first 2*2 convolutional layer；
The second convolutional layer of the building, the linear elementary layer of the second amendment and up-sampling layer include:
After second dropout layers of building, four volume Two bases are successively constructed, by up-sampling layer and two the 2nd 3*3
Convolutional layer is constituted, and the volume Two base feature port number successively constructed is to begin with second the first 2*2 convolutional layer, feature channel
Number successively reduces one times；
The building up-sampling layer before each 2nd 3*3 convolutional layer.
6. the tumour automatic division method in CT image according to claim 5, which is characterized in that
It is distributed, and is cascaded in mirror image between the constricted path and the expansion path.
7. the tumour automatic division method in CT image according to claim 4, which is characterized in that
The training method of the processing network trained further include:
Building confrontation network, the building confrontation network include:
Construct the first data, first data by segmentation goldstandard as standard reference, and by liver neoplasm two-value goldstandard
And Original Liver gray level image dot product obtains；
The second data are constructed, second data are obtained by the segmented image and Original Liver gray level image dot product；
Loss function is constructed, and by the first data and the second data entrance loss function, captures the Pixel-level from different levels
Length space characteristics.
8. the automatic segmenting system of tumour in a kind of CT image characterized by comprising
Enhance enlargement module, for carrying out enhancing expansion to raw image data, obtains enhancing expanding data；
Normalized module obtains normalization data for enhancing expanding data to be normalized；
Image segmentation module obtains segmented image for noise reduction data to be inputted the processing network trained；
Noise reduction module, for doing noise reduction process to segmented image.
9. a kind of electronic device, comprising: memory, processor and be stored on the memory and can transport on the processor
Capable computer program, which is characterized in that when the processor executes the computer program, realize in claim 1 to 7
Any one the method.
10. a kind of computer readable storage medium, a kind of computer readable storage medium, are stored thereon with computer program,
It is characterized in that, when the computer program is executed by processor, realizes any one the method in claim 1 to 7.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811440970.8A CN109754403A (en) | 2018-11-29 | 2018-11-29 | Tumour automatic division method and system in a kind of CT image |
| PCT/CN2019/121594 WO2020108562A1 (en) | 2018-11-29 | 2019-11-28 | Automatic tumor segmentation method and system in ct image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811440970.8A CN109754403A (en) | 2018-11-29 | 2018-11-29 | Tumour automatic division method and system in a kind of CT image |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109754403A true CN109754403A (en) | 2019-05-14 |
Family
ID=66402563
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811440970.8A Pending CN109754403A (en) | 2018-11-29 | 2018-11-29 | Tumour automatic division method and system in a kind of CT image |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN109754403A (en) |
| WO (1) | WO2020108562A1 (en) |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110197716A (en) * | 2019-05-20 | 2019-09-03 | 广东技术师范大学 | Processing method, device and the computer readable storage medium of medical image |
| CN110717060A (en) * | 2019-09-04 | 2020-01-21 | 平安科技（深圳）有限公司 | Image mask filtering method and device and storage medium |
| CN110751627A (en) * | 2019-09-19 | 2020-02-04 | 上海联影智能医疗科技有限公司 | Image processing method, image processing device, computer equipment and storage medium |
| CN111028242A (en) * | 2019-11-27 | 2020-04-17 | 中国科学院深圳先进技术研究院 | Automatic tumor segmentation system and method and electronic equipment |
| WO2020108562A1 (en) * | 2018-11-29 | 2020-06-04 | 中国科学院深圳先进技术研究院 | Automatic tumor segmentation method and system in ct image |
| CN111652886A (en) * | 2020-05-06 | 2020-09-11 | 哈尔滨工业大学 | Liver tumor segmentation method based on improved U-net network |
| CN111739008A (en) * | 2020-06-23 | 2020-10-02 | 北京百度网讯科技有限公司 | Image processing method, device, equipment and readable storage medium |
| CN111754530A (en) * | 2020-07-02 | 2020-10-09 | 广东技术师范大学 | Prostate ultrasonic image segmentation and classification method |
| CN112529909A (en) * | 2020-12-08 | 2021-03-19 | 北京安德医智科技有限公司 | Tumor image brain region segmentation method and system based on image completion |
| CN113111684A (en) * | 2020-01-10 | 2021-07-13 | 字节跳动有限公司 | Training method and device of neural network model and image processing system |
| WO2021151275A1 (en) * | 2020-05-20 | 2021-08-05 | 平安科技（深圳）有限公司 | Image segmentation method and apparatus, device, and storage medium |
| CN113705320A (en) * | 2021-05-24 | 2021-11-26 | 中国科学院深圳先进技术研究院 | Training method, medium, and apparatus for surgical motion recognition model |
| CN114066871A (en) * | 2021-11-19 | 2022-02-18 | 江苏科技大学 | Method for training new coronary pneumonia focus region segmentation model |
| US11367181B2 (en) | 2018-12-29 | 2022-06-21 | Shanghai United Imaging Intelligence Co., Ltd. | Systems and methods for ossification center detection and bone age assessment |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114494266B (en) * | 2020-10-26 | 2024-05-28 | 中国人民解放军空军军医大学 | Cervical and peripheral multi-organ segmentation method adopting hierarchical cavity pyramid convolution |
| CN114119448A (en) * | 2021-02-05 | 2022-03-01 | 苏州大学 | Pancreas segmentation network in CT image based on improve U-shaped network |
| CN116898455B (en) * | 2023-07-06 | 2024-04-16 | 湖北大学 | Sleep electroencephalogram signal detection method and system based on deep learning model |
| CN117765532B (en) * | 2024-02-22 | 2024-05-31 | 中国科学院宁波材料技术与工程研究所 | Cornea Langerhans cell segmentation method and device based on confocal microscopic image |
| CN118229981B (en) * | 2024-05-23 | 2024-07-23 | 山东未来网络研究院(紫金山实验室工业互联网创新应用基地) | CT image tumor segmentation method, device and medium combining convolutional network and transducer |
Citations (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106408562A (en) * | 2016-09-22 | 2017-02-15 | 华南理工大学 | Fundus image retinal vessel segmentation method and system based on deep learning |
| CN106683104A (en) * | 2017-01-06 | 2017-05-17 | 西北工业大学 | Prostate magnetic resonance image segmentation method based on integrated depth convolution neural network |
| CN107680678A (en) * | 2017-10-18 | 2018-02-09 | 北京航空航天大学 | Based on multiple dimensioned convolutional neural networks Thyroid ultrasound image tubercle auto-check system |
| CN107945204A (en) * | 2017-10-27 | 2018-04-20 | 西安电子科技大学 | A kind of Pixel-level portrait based on generation confrontation network scratches drawing method |
| CN108346145A (en) * | 2018-01-31 | 2018-07-31 | 浙江大学 | The recognition methods of unconventional cell in a kind of pathological section |
| CN108492286A (en) * | 2018-03-13 | 2018-09-04 | 成都大学 | A kind of medical image cutting method based on the U-shaped convolutional neural networks of binary channel |
| CN108537793A (en) * | 2018-04-17 | 2018-09-14 | 电子科技大学 | A kind of pulmonary nodule detection method based on improved u-net networks |
| CN108596915A (en) * | 2018-04-13 | 2018-09-28 | 深圳市未来媒体技术研究院 | A kind of medical image segmentation method based on no labeled data |
| CN108596884A (en) * | 2018-04-15 | 2018-09-28 | 桂林电子科技大学 | A kind of cancer of the esophagus dividing method in chest CT image |
| CN108629784A (en) * | 2018-05-08 | 2018-10-09 | 上海嘉奥信息科技发展有限公司 | A kind of CT image intracranial vessel dividing methods and system based on deep learning |
| CN108765422A (en) * | 2018-06-13 | 2018-11-06 | 云南大学 | A kind of retinal images blood vessel automatic division method |
| CN108776969A (en) * | 2018-05-24 | 2018-11-09 | 复旦大学 | Breast ultrasound image lesion segmentation approach based on full convolutional network |
| CN108806793A (en) * | 2018-04-17 | 2018-11-13 | 平安科技（深圳）有限公司 | Lesion monitoring method, device, computer equipment and storage medium |
| CN108830912A (en) * | 2018-05-04 | 2018-11-16 | 北京航空航天大学 | A kind of interactive grayscale image color method of depth characteristic confrontation type study |
| CN108876793A (en) * | 2018-04-13 | 2018-11-23 | 北京迈格威科技有限公司 | Semantic segmentation methods, devices and systems and storage medium |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10245000B2 (en) * | 2014-12-12 | 2019-04-02 | General Electric Company | Method and system for defining a volume of interest in a physiological image |
| CN107862695A (en) * | 2017-12-06 | 2018-03-30 | 电子科技大学 | A kind of modified image segmentation training method based on full convolutional neural networks |
| CN108109152A (en) * | 2018-01-03 | 2018-06-01 | 深圳北航新兴产业技术研究院 | Medical Images Classification and dividing method and device |
| CN108171711A (en) * | 2018-01-17 | 2018-06-15 | 深圳市唯特视科技有限公司 | A kind of infant's brain Magnetic Resonance Image Segmentation method based on complete convolutional network |
| CN109754403A (en) * | 2018-11-29 | 2019-05-14 | 中国科学院深圳先进技术研究院 | Tumour automatic division method and system in a kind of CT image |
-
2018
- 2018-11-29 CN CN201811440970.8A patent/CN109754403A/en active Pending
-
2019
- 2019-11-28 WO PCT/CN2019/121594 patent/WO2020108562A1/en active Application Filing
Patent Citations (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106408562A (en) * | 2016-09-22 | 2017-02-15 | 华南理工大学 | Fundus image retinal vessel segmentation method and system based on deep learning |
| CN106683104A (en) * | 2017-01-06 | 2017-05-17 | 西北工业大学 | Prostate magnetic resonance image segmentation method based on integrated depth convolution neural network |
| CN107680678A (en) * | 2017-10-18 | 2018-02-09 | 北京航空航天大学 | Based on multiple dimensioned convolutional neural networks Thyroid ultrasound image tubercle auto-check system |
| CN107945204A (en) * | 2017-10-27 | 2018-04-20 | 西安电子科技大学 | A kind of Pixel-level portrait based on generation confrontation network scratches drawing method |
| CN108346145A (en) * | 2018-01-31 | 2018-07-31 | 浙江大学 | The recognition methods of unconventional cell in a kind of pathological section |
| CN108492286A (en) * | 2018-03-13 | 2018-09-04 | 成都大学 | A kind of medical image cutting method based on the U-shaped convolutional neural networks of binary channel |
| CN108876793A (en) * | 2018-04-13 | 2018-11-23 | 北京迈格威科技有限公司 | Semantic segmentation methods, devices and systems and storage medium |
| CN108596915A (en) * | 2018-04-13 | 2018-09-28 | 深圳市未来媒体技术研究院 | A kind of medical image segmentation method based on no labeled data |
| CN108596884A (en) * | 2018-04-15 | 2018-09-28 | 桂林电子科技大学 | A kind of cancer of the esophagus dividing method in chest CT image |
| CN108537793A (en) * | 2018-04-17 | 2018-09-14 | 电子科技大学 | A kind of pulmonary nodule detection method based on improved u-net networks |
| CN108806793A (en) * | 2018-04-17 | 2018-11-13 | 平安科技（深圳）有限公司 | Lesion monitoring method, device, computer equipment and storage medium |
| CN108830912A (en) * | 2018-05-04 | 2018-11-16 | 北京航空航天大学 | A kind of interactive grayscale image color method of depth characteristic confrontation type study |
| CN108629784A (en) * | 2018-05-08 | 2018-10-09 | 上海嘉奥信息科技发展有限公司 | A kind of CT image intracranial vessel dividing methods and system based on deep learning |
| CN108776969A (en) * | 2018-05-24 | 2018-11-09 | 复旦大学 | Breast ultrasound image lesion segmentation approach based on full convolutional network |
| CN108765422A (en) * | 2018-06-13 | 2018-11-06 | 云南大学 | A kind of retinal images blood vessel automatic division method |
Non-Patent Citations (4)
| Title |
|---|
| FAUSTO MILLETARI ETC: "V-Net: Fully Convolutional Neural Networks for Volumetric Medical Image Segmentation", 《2016 FOURTH INTERNATIONAL CONFERENCE ON 3D VISION (3DV)》 * |
| MICHAL DROZDZAL 等: "The Importance of Skip Connections in Biomedical Image Segmentation", 《DLMIA 2016, LABELS 2016: DEEP LEARNING AND DATA LABELING FOR MEDICAL APPLICATIONS》 * |
| OLAF RONNEBERGER ETC: "U-net: Convolutional networks for biomedical image segmentation", 《LECTURE NOTES IN COMPUTER SCIENCE (INCLUDING SUBSERIES LECTURE NOTES IN ARTIFICIAL INTELLIGENCE AND LECTURE NOTES IN BIOINFORMATICS)》 * |
| YUAN XUE ETC: "SegAN: Adversarial Network with Multi-scale L1 Loss for Medical Image Segmentation", 《NEUROINFORMATICS(2018)》 * |
Cited By (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020108562A1 (en) * | 2018-11-29 | 2020-06-04 | 中国科学院深圳先进技术研究院 | Automatic tumor segmentation method and system in ct image |
| US11367181B2 (en) | 2018-12-29 | 2022-06-21 | Shanghai United Imaging Intelligence Co., Ltd. | Systems and methods for ossification center detection and bone age assessment |
| US11735322B2 (en) | 2018-12-29 | 2023-08-22 | Shanghai United Imaging Intelligence Co., Ltd. | Systems and methods for ossification center detection and bone age assessment |
| CN110197716A (en) * | 2019-05-20 | 2019-09-03 | 广东技术师范大学 | Processing method, device and the computer readable storage medium of medical image |
| CN110717060A (en) * | 2019-09-04 | 2020-01-21 | 平安科技（深圳）有限公司 | Image mask filtering method and device and storage medium |
| CN110717060B (en) * | 2019-09-04 | 2023-08-18 | 平安科技（深圳）有限公司 | Image mask filtering method, device and storage medium |
| CN110751627A (en) * | 2019-09-19 | 2020-02-04 | 上海联影智能医疗科技有限公司 | Image processing method, image processing device, computer equipment and storage medium |
| CN110751627B (en) * | 2019-09-19 | 2024-01-26 | 上海联影智能医疗科技有限公司 | Image processing method, device, computer equipment and storage medium |
| CN111028242A (en) * | 2019-11-27 | 2020-04-17 | 中国科学院深圳先进技术研究院 | Automatic tumor segmentation system and method and electronic equipment |
| CN113111684B (en) * | 2020-01-10 | 2024-05-21 | 字节跳动有限公司 | Training method and device for neural network model and image processing system |
| CN113111684A (en) * | 2020-01-10 | 2021-07-13 | 字节跳动有限公司 | Training method and device of neural network model and image processing system |
| CN111652886A (en) * | 2020-05-06 | 2020-09-11 | 哈尔滨工业大学 | Liver tumor segmentation method based on improved U-net network |
| WO2021151275A1 (en) * | 2020-05-20 | 2021-08-05 | 平安科技（深圳）有限公司 | Image segmentation method and apparatus, device, and storage medium |
| CN111739008B (en) * | 2020-06-23 | 2024-04-12 | 北京百度网讯科技有限公司 | Image processing method, device, equipment and readable storage medium |
| CN111739008A (en) * | 2020-06-23 | 2020-10-02 | 北京百度网讯科技有限公司 | Image processing method, device, equipment and readable storage medium |
| CN111754530B (en) * | 2020-07-02 | 2023-11-28 | 广东技术师范大学 | Prostate ultrasonic image segmentation classification method |
| CN111754530A (en) * | 2020-07-02 | 2020-10-09 | 广东技术师范大学 | Prostate ultrasonic image segmentation and classification method |
| CN112529909A (en) * | 2020-12-08 | 2021-03-19 | 北京安德医智科技有限公司 | Tumor image brain region segmentation method and system based on image completion |
| CN113705320A (en) * | 2021-05-24 | 2021-11-26 | 中国科学院深圳先进技术研究院 | Training method, medium, and apparatus for surgical motion recognition model |
| CN114066871A (en) * | 2021-11-19 | 2022-02-18 | 江苏科技大学 | Method for training new coronary pneumonia focus region segmentation model |
Also Published As
| Publication number | Publication date |
|---|---|
| WO2020108562A1 (en) | 2020-06-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109754403A (en) | Tumour automatic division method and system in a kind of CT image | |
| Zhou et al. | Deep learning of the sectional appearances of 3D CT images for anatomical structure segmentation based on an FCN voting method | |
| Altaf et al. | Going deep in medical image analysis: concepts, methods, challenges, and future directions | |
| US11308587B2 (en) | Learning method of generative adversarial network with multiple generators for image denoising | |
| Liang et al. | MCFNet: Multi-layer concatenation fusion network for medical images fusion | |
| CN109166130B (en) | Image processing method and image processing device | |
| CN110475505A (en) | Utilize the automatic segmentation of full convolutional network | |
| DE102018111407A1 (en) | METHOD FOR MACHINE LEARNING FOR AUTOMATICALLY MODELING OF EXCESSIVE EXPENDITURE | |
| CN112949838B (en) | Convolutional neural network based on four-branch attention mechanism and image segmentation method | |
| CN110276741B (en) | Method and device for nodule detection and model training thereof and electronic equipment | |
| CN110110723B (en) | Method and device for automatically extracting target area in image | |
| Singhal et al. | Study of deep learning techniques for medical image analysis: A review | |
| CN114897780A (en) | MIP sequence-based mesenteric artery blood vessel reconstruction method | |
| CN110415253A (en) | A kind of point Interactive medical image dividing method based on deep neural network | |
| CN112949654A (en) | Image detection method and related device and equipment | |
| Yang et al. | Deep learning in medical image super resolution: a review | |
| Zhao et al. | MPSHT: multiple progressive sampling hybrid model multi-organ segmentation | |
| Sander et al. | Autoencoding low-resolution MRI for semantically smooth interpolation of anisotropic MRI | |
| Poonkodi et al. | 3D-MedTranCSGAN: 3D medical image transformation using CSGAN | |
| Lu et al. | Fine-grained calibrated double-attention convolutional network for left ventricular segmentation | |
| Li et al. | Hrinet: Alternative supervision network for high-resolution ct image interpolation | |
| CN112419283B (en) | Neural network for estimating thickness and method thereof | |
| Yuan et al. | A multiple gated boosting network for multi‐organ medical image segmentation | |
| CN110009604A (en) | The breath signal extracting method and device of angiographic image series | |
| Yang et al. | Review of Deep Learning-Based Image Inpainting Techniques |
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 |