CN107749061A - Based on improved full convolutional neural networks brain tumor image partition method and device - Google Patents
Based on improved full convolutional neural networks brain tumor image partition method and device Download PDFInfo
- Publication number
- CN107749061A CN107749061A CN201710812236.9A CN201710812236A CN107749061A CN 107749061 A CN107749061 A CN 107749061A CN 201710812236 A CN201710812236 A CN 201710812236A CN 107749061 A CN107749061 A CN 107749061A
- Authority
- CN
- China
- Prior art keywords
- msub
- mrow
- mfrac
- brain tumor
- image
- 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
- 208000003174 Brain Neoplasms Diseases 0.000 title claims abstract description 60
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 36
- 238000000034 method Methods 0.000 title claims abstract description 21
- 238000005192 partition Methods 0.000 title claims abstract description 9
- 230000011218 segmentation Effects 0.000 claims abstract description 52
- 238000012549 training Methods 0.000 claims abstract description 27
- 238000011156 evaluation Methods 0.000 claims abstract description 7
- 230000004927 fusion Effects 0.000 claims description 11
- JXSJBGJIGXNWCI-UHFFFAOYSA-N diethyl 2-[(dimethoxyphosphorothioyl)thio]succinate Chemical compound CCOC(=O)CC(SP(=S)(OC)OC)C(=O)OCC JXSJBGJIGXNWCI-UHFFFAOYSA-N 0.000 claims description 8
- 238000001914 filtration Methods 0.000 claims description 7
- 238000012360 testing method Methods 0.000 claims description 7
- 230000000295 complement effect Effects 0.000 claims description 4
- 239000012530 fluid Substances 0.000 claims description 4
- 238000003475 lamination Methods 0.000 claims description 4
- 238000005070 sampling Methods 0.000 claims description 4
- 230000008901 benefit Effects 0.000 claims description 3
- 235000013399 edible fruits Nutrition 0.000 claims description 3
- 230000002708 enhancing effect Effects 0.000 claims description 3
- 238000002075 inversion recovery Methods 0.000 claims description 3
- 230000008569 process Effects 0.000 claims description 3
- 238000009434 installation Methods 0.000 claims description 2
- 238000002844 melting Methods 0.000 claims 1
- 230000008018 melting Effects 0.000 claims 1
- 238000005481 NMR spectroscopy Methods 0.000 abstract description 5
- 230000007547 defect Effects 0.000 abstract description 4
- 238000003745 diagnosis Methods 0.000 abstract description 4
- 239000003814 drug Substances 0.000 abstract description 2
- 230000005311 nuclear magnetism Effects 0.000 abstract 1
- 238000002595 magnetic resonance imaging Methods 0.000 description 10
- 210000004556 brain Anatomy 0.000 description 9
- 206010028980 Neoplasm Diseases 0.000 description 7
- 238000003384 imaging method Methods 0.000 description 6
- 238000011176 pooling Methods 0.000 description 5
- 238000003709 image segmentation Methods 0.000 description 4
- 238000013135 deep learning Methods 0.000 description 3
- 206010030113 Oedema Diseases 0.000 description 2
- 201000011510 cancer Diseases 0.000 description 2
- 230000003902 lesion Effects 0.000 description 2
- 230000007935 neutral effect Effects 0.000 description 2
- 238000000638 solvent extraction Methods 0.000 description 2
- 238000003860 storage Methods 0.000 description 2
- 102000008186 Collagen Human genes 0.000 description 1
- 108010035532 Collagen Proteins 0.000 description 1
- 208000032612 Glial tumor Diseases 0.000 description 1
- 206010018338 Glioma Diseases 0.000 description 1
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 230000004888 barrier function Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000002775 capsule Substances 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 230000002490 cerebral effect Effects 0.000 description 1
- 210000001175 cerebrospinal fluid Anatomy 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000000701 chemical imaging Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 229920001436 collagen Polymers 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000005520 cutting process Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000002059 diagnostic imaging Methods 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 230000001969 hypertrophic effect Effects 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 206010027191 meningioma Diseases 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 230000005855 radiation Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000000630 rising effect Effects 0.000 description 1
- 210000004872 soft tissue Anatomy 0.000 description 1
- 238000012916 structural analysis Methods 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 239000013589 supplement Substances 0.000 description 1
- 210000001519 tissue Anatomy 0.000 description 1
- 210000004885 white matter Anatomy 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/40—Scaling of whole images or parts thereof, e.g. expanding or contracting
- G06T3/4007—Scaling of whole images or parts thereof, e.g. expanding or contracting based on interpolation, e.g. bilinear interpolation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/50—Image enhancement or restoration using two or more images, e.g. averaging or subtraction
-
- 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
- G06T2207/20024—Filtering details
- G06T2207/20032—Median filtering
-
- 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
- G06T2207/20212—Image combination
- G06T2207/20221—Image fusion; Image merging
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Magnetic Resonance Imaging Apparatus (AREA)
Abstract
The present invention relates to medicine equipment, to propose a kind of improved full convolutional neural networks, realize the machine segmentation of brain tumor nuclear magnetic resonance image, avoid the defects of inefficient and unstable existing for artificial segmentation, quick, reliable brain tumor segmentation result is provided, so as to provide accurate foundation for the diagnosis of brain tumor, treatment and operation guiding.Therefore, the technical solution adopted by the present invention is, it is as follows based on improved full convolutional neural networks brain tumor image partition method, step:1) image is chosen;2) full convolutional neural networks FCN models are built;3) segmentation result is tested, after FCN model trainings are good, the prediction of knub position and boundary sizes is carried out to any one brain tumor image using the model of training, and segmentation result is evaluated using corresponding evaluation index, preferably to improve FCN models.Present invention is mainly applied to nuclear-magnetism image procossing occasion.
Description
Technical field
The present invention relates to medicine equipment, specifically, is related to based on improved full convolutional neural networks brain tumor picture segmentation
Method and device.
Background technology
Brain is most important part in human body, but the brain tumor incidence of disease is in rising trend in recent years, shows according to investigations,
The brain tumor number being diagnosed in the U.S. has just increased 23,000 people newly within only 2015.The World Health Organization is according to lesion degree brain
Tumour is divided into five grades, and brain tumor is broadly divided into the benign tumours such as benign tumour and the class of malignant tumour two, meningioma and passed through
It is generally possible to get well after operative treatment, and the malignant tumour such as glioma and collagen knurl is because its intractable is difficult to cure, again
It is referred to as the cancer of the brain.Magnetic resonance imaging (Magnetic Resonance Imaging, MRI) has Noninvasive, can be in patient
Do not receive to provide the information such as shape, size and location in the case of high ionization radiation, and there is good soft tissue contrast,
Had been to be concerned by more and more people in diagnosis, treatment and the operation guiding of brain tumor.But because brain tumor is complex-shaped, size
There is randomness with position, the factors such as type difference is big cause that there is presently no the need that a kind of partitioning algorithm disclosure satisfy that clinic
Will, real-time is also unable to reach requirement, and the result of different expert's manual segmentation brain tumor images also has very big difference, and manually
Cost is higher.Therefore, the brain tumor segmenting system for studying a set of full-automatic, accuracy height and strong robustness is extremely to be necessary
's.
Have many algorithms at present to be successfully applied in the segmentation of brain tumor image, mainly include the segmentation side based on region
Method, the dividing method based on fuzzy clustering and convolutional neural networks method etc..Dividing method based on region growing is first in mesh
A seed point is chosen in mark region, then the pixel that surrounding has same nature is incorporated into seed region, until not similar
Untill the pixel aggregation of property, for region-growing method, generally existing is undesirable to the segmentation of image shadow region to ask
Topic;The dividing method of fuzzy clustering is that image pixel is classified by its similarity so that in small distance between similar individual
And it is in larger distance between inhomogeneity individual, for clustering algorithm, generally only consider the independent information of each pixel and neglect
Somewhat like the spatial information between vegetarian refreshments, it is affected by noise larger and the segmentation knot of continuum can not be obtained to finally result in algorithm
Fruit;Image partition method based on convolutional neural networks (Convolutional Neural Network, CNN) is to a picture
During element classification, it is used to train and predict using a block of pixels around the pixel as CNN input, then constantly slides
Window, the window slided every time gives CNN and carries out identification and classification, so as to realize the segmentation of whole image.But the brain based on CNN swells
Knurl dividing method can make computer storage overhead become big along with the increase of sliding window, and adjacent pixel blocks substantially weight
Multiple, convolution is calculated one by one for each block of pixels, computationally there is very big repeatability, it is most important that CNN convolution filter
Ripple device can make it that the partitioning boundary of brain tumor is unintelligible along with very big receptive field and maximum pond layer.
2015, Jonathan Long et al. by full convolutional neural networks (Fully Convolution Network,
FCN image, semantic segmentation field) is successfully applied to, and achieves significant effect.The network can be to corresponding to position in image
The each pixel put is classified, so as to realize the task of image segmentation.Full articulamentum in traditional CNN is changed chemical conversion by FCN
Convolutional layer one by one.In lower Fig. 1,5 layers are convolutional layers before traditional CNN structures, and the 6th layer and the 7th layer is one long respectively
The one-dimensional vector for 4096 is spent, the 8th layer is probability of the length for 1000 one-dimensional vector, respectively corresponding 1000 classifications;And FCN
Be expressed as convolutional layer by this 3 layers, the size of convolution kernel (port number, wide, high) be respectively (4096,1,1), (4096,1,1) and
(1000,1,1), by such convolutional layer, finally just generate one " feature vector chart ".All layers are all volumes in FCN
Lamination, without full articulamentum, so referred to as full convolutional neural networks.
The characteristic pattern that Network Capture profound FCN arrives is abstract and coarse, and these characteristic patterns have passed through substantial amounts of maximum pond
Change layer, the information such as edge, the position of target object are have lost significantly in characteristic pattern;And shallow-layer network only have passed through minority
Characteristic pattern caused by maximum pond layer Pooling is specific and fine, and what is learnt is the feature of some regional areas, these features
Figure can be to reflect the position of target object and marginal information., can be with so in order to further be optimized to segmentation result
The characteristic pattern of network shallow-layer is combined with the characteristic pattern of deep layer, this is the jump structure of FCN networks, whole to improve
The segmentation precision of image.
The content of the invention
For overcome the deficiencies in the prior art, the present invention is directed to propose a kind of improved full convolutional neural networks, realize that brain swells
The machine segmentation of knurl nuclear magnetic resonance image, avoids the defects of inefficient and unstable existing for artificial segmentation, there is provided quick, reliable
Brain tumor segmentation result, so as to for the diagnosis of brain tumor, treatment and operation guide accurate foundation is provided.It is therefore, of the invention
The technical scheme of use is, as follows based on improved full convolutional neural networks brain tumor image partition method, step:
1) image is chosen:First to magnetic resonance MRI brain tumor fluid attented inversion recoveries FLAIR, spin-spin relaxation T2
And
T1 enhancing T1C images use medium filtering denoising, then linear scale fusion are carried out to three kinds of mode, so as to more preferable
Ground is split by brain tumor;
2) full convolutional neural networks FCN models are built:Using the brain tumor image after merging well as training sample, Zhuan Jiafen
True value label of the pixel as training sample of result figure is cut, then carries out the training of FCN models;
3) segmentation result is tested, after FCN model trainings are good, any one brain tumor image is entered using the model of training
The prediction of row knub position and boundary sizes, and segmentation result is evaluated using corresponding evaluation index, so as to preferably
Improve FCN models.
Test in segmentation result step, input data is by Pooling1, Pooling2, Pooling 3, Pooling
After 4 and Pooling 5 maximum pond layer, can respectively obtain the size of artwork 1/2,1/4 size, 1/8 size, 1/16 size and
The feature vector chart of 1/32 size.The feature of 4th convolution kernel is carried out deconvolution to the figure after last Upsamping1
Upsamping2 subsidiary details, equivalent to one Interpolation Process, then the feature of the 3rd convolution kernel to Upsamping2 just now it
Rear image carries out deconvolution Upsamping3 subsidiary details again, finally again the feature of the 2nd convolution kernel to just now
Image after Upsamping3 carries out deconvolution Upsamping4 subsidiary details again, consequently, it is possible to the feature of deep layer network
Scheme to be complementary to one another with the characteristic pattern of shallow-layer network, be finally added and obtain an of a relatively high segmentation figure of precision.
FCN carries out up-sampling Upsamping, interpolation point using warp lamination to the feature vector chart of last convolutional layer
P 4 adjoint point coordinates are respectively Q11=(x1,y1), Q21=(x2,y1), Q12=(x1,y2), Q22=(x2,y2), first in X side
First time linear interpolation calculating is carried out upwards, is obtained:
Then linear interpolation is carried out in the Y direction, is obtained:
Formula (1) and formula (2) are updated into formula can obtain interpolation point P coordinate value f (x, y):
By so operation, input picture identical size finally will be returned to by the feature vector chart in pond,
So as to produce a prediction to each pixel, while the spatial information in original input picture is remained, finally up-sampled
Characteristic pattern in classified pixel-by-pixel, so as to realize image split.
Based on improved full convolutional neural networks brain tumor image segmenting device, gone by magnetic resonance MRI machine, medium filtering
Make an uproar device, linearity ratio Fusion Module, capture card, computer;
Linearity ratio Fusion Module is used to the image for choosing tri- kinds of mode of FLAIR, T2 and T1C carrying out linear scale
Fusion;
Image caused by magnetic resonance MRI machine successively through medium filtering denoising device, linearity ratio Fusion Module, adopt
Truck enters computer;
Computer installation has structure FCN model modules, test segmentation result module;
Structure FCN model modules are used for, using the brain tumor image after merging well as training sample, expert's segmentation result figure
True value label of the pixel as training sample, carry out the training of FCN models, construct FCN models;
Test segmentation result module is used for, after FCN model trainings are good, using the model of training to any one brain tumor
Image carries out the prediction of knub position and boundary sizes, and segmentation result is evaluated using corresponding evaluation index, so as to
Preferably improve FCN models.
The features of the present invention and beneficial effect are:
The present invention combines newest full convolutional neural networks image partition method, and network structure is improved, and proposes
FCN-4s network structures, and the brain tumor image of various modes is merged, so as to realize the Accurate Segmentation of brain tumor image.
Compared with the method for the traditional classical such as convolutional neural networks, its advantage is mainly reflected in:
1) it is practical:What CNN was inputted every time is a block of pixels, the probability of simply each pixel of output, and FCN
It is a kind of end-to-end end-to-end neutral net, it can input the picture of arbitrary size, and can export correspondingly sized
Split picture, it is more practical;
2) real-time is good:For FCN without going to judge the classification of each pixel as CNN sliding windows, it is a kind of segmentation of Pixel-level
Network, the problem of avoiding the repetition storage due to being brought using block of pixels and calculate convolution, calculator memory expense is small, from
And image can be made to split in real time;
3) accuracy is high:The conventional methods such as CNN simply carry out judging to divide to a pixel every time when splitting image
Class, and the feature vector chart that FCN is finally exported remains the spatial information in original input picture, to considering during pixel classifications
Relation between pixel, so it can preferably obtain the border of brain tumor image, improves the degree of accuracy of segmentation.
Brief description of the drawings:
The patient with brain tumors of the types of Fig. 1 tetra-;
The segmentation result figure of Fig. 2 expert;
The improved full convolutional neural networks segmentation MRI brain tumor algorithm flow charts of Fig. 3;
Fig. 4 FCN figures compared with CNN difference;
The bilinear interpolation up-sampled in Fig. 5 FCN;
The full convolutional neural networks structure chart of Fig. 6 MRI images;
Fig. 7 brain tumors segmentation result and evaluation figure.
Embodiment
The present invention combines medical image and deep learning algorithm, completes the segmentation of brain tumor nuclear magnetic resonance image.
This full automatic brain tumor image segmentation will produce important influence in Medical Imaging.
For traditional convolutional neural networks in the defects of image segmentation, the present invention proposes a kind of improved full convolution
Neutral net, and be successfully applied in the segmentation of brain tumor nuclear magnetic resonance image, avoid inefficient existing for artificial segmentation
And the defects of unstable.Quick, reliable brain tumor segmentation result is provided using brand-new deep learning algorithm, so as to be swollen for brain
Diagnosis, treatment and the operation guiding of knurl provide accurate foundation.
To achieve these goals, the present invention adopts the following technical scheme that:
1) image is chosen.MRI brain tumors image itself is of low quality, certain noise, the image of three kinds of mode be present
The uncorrelated and complementary information in part is provided for the segmentation of brain tumor, so in being used first to FLAIR, T2 and T1C image
Value filtering denoising, linear scale fusion then is carried out to three kinds of mode, preferably to split to brain tumor.
2) FCN models are built.Using the brain tumor image after merging well as training sample, the pixel of expert's segmentation result figure
True value label of the point as training sample, then carries out the training of FCN models.
3) segmentation result is tested.After FCN model trainings are good, any one brain tumor image is entered using the model of training
The prediction of row knub position and boundary sizes, and segmentation result is evaluated using corresponding evaluation index, so as to preferably
Improve FCN models.
The segmentation of the invention for effectively supporting brain tumor nuclear magnetic resonance image is deep learning in medical image segmentation side
One important application in face, the mode to expert's manual measurement are to supplement well.
The invention will be further described with example below in conjunction with the accompanying drawings.
1) image is chosen
Patient with brain tumors MRI imaging be three-dimensional multi-spectral imaging, generally include Flair (fluid attented inversion recovery),
Four kinds of modality images such as T1 (SPIN-LATTICE RELAXATION), T2 (spin-spin relaxation) and T1C (T1 enhancings), Fig. 1 (a) be T1 into
Picture, it is simple to operate, the structural analysis of brain tumor is frequently utilized for, but image quality is poor, can not provide more detailed image information.
Fig. 1 (b) is T1C imagings, is T1 Enhanced Imaging, and it, which can compare, clearly displays cerebral tissue structure, and due to Hypertrophic
Contrast build-up caused by the barrier breakdown of brain tumor region so that the brain tumor border in T1 Enhanced Imagings becomes brighter, very
Easily distinguish tumour and capsule becomes region.Fig. 1 (c) is T2 imagings, and area of edema is more bright compared to other imaging mode, but brain
The feature of spinal fluid and tumor region is difficult to differentiate between, and is unfavorable for lesion segmentation.Fig. 1 (d) is that FLAIR is imaged, white matter of brain in this mode
Contrast is low, but tumor region is remarkably reinforced with normal structure contrast, and area of edema border is obvious, eliminates the shadow of cerebrospinal fluid
Ring, the tumour that the ventricles of the brain are taken up with convex surface or brain ditch show apparent, is advantageous to the display of tumor region, is to split brain at present to swell
The maximally effective mode of knurl image.The characteristics of being imaged thus according to different modalities, the present invention is in order to farthest obtain brain tumor
The position in region and size boundary information, the image for choosing tri- kinds of mode of FLAIR, T2 and T1C is carried out to the fusion of linear scale.
2) FCN models are built
FCN of the present invention convolutional coding structure can be shown in See Figure 6.Input data is in the maximum pond (Pooling) by 5 times
Afterwards, the feature vector chart of the size of artwork 1/2,1/4 size, 1/8 size, 1/16 size and 1/32 size can be respectively obtained.Such as
Fruit is directly up-sampled the operation of (Upsamping) to the feature vector chart of 1/32 size, because such operation reduction
Picture is only the feature in the 5th convolution kernel, is limited to the feature that precision problem cannot be gone back among original image well, because
This iteration forward herein.The feature of 4th convolution kernel is carried out convolution benefit again to the image after last Upsamping
Details (equivalent to one Interpolation Process) is filled, that is, obtains the FCN-16s network architecture.If again the feature of the 3rd convolution kernel
Image after being up-sampled to second carries out deconvolution subsidiary details again, i.e., can obtain the FCN-8s network architecture.Finally
Image after the feature of the 2nd convolution kernel is up-sampled to third time again carries out deconvolution subsidiary details again, that is, obtains
The FCN-4s network architecture.Consequently, it is possible to the characteristic pattern of the characteristic pattern of deep layer network and shallow-layer network can be complementary to one another, finally
Addition can obtain the of a relatively high segmentation figure picture of a precision.
Specifically, FCN is up-sampled using warp lamination to the feature vector chart of last convolutional layer
(Upsamping), the rough image so as to make resolution ratio low returns to the resolution ratio of artwork.Upsamping is mainly pair
View data carry out bilinear interpolation operation, by the use of interpolation point around 4 adjoint points gray value Weighted Interpolation as the point gray scale
Value, it can be decomposed into quadratic one-dimensional linear interpolation.Can be shown in See Figure 4, interpolation point P 4 adjoint point coordinates are respectively Q11=
(x1,y1), Q21=(x2,y1), Q12=(x1,y2), Q22=(x2,y2).Carry out first time linear interpolation meter in the X direction first
Calculate, obtain:
Then linear interpolation is carried out in the Y direction, is obtained:
Formula (1) and formula (2) are updated into formula can obtain interpolation point P coordinate value f (x, y):
By so operation, input picture identical size finally will be returned to by the feature vector chart in pond,
So as to produce a prediction to each pixel, while the spatial information in original input picture is remained, finally upper
Classified pixel-by-pixel in the characteristic pattern of sampling, so as to realize that image is split.
3) segmentation result is tested
The present invention by FCN-16s, FCN-8s with it is proposed that the FCN-4s network architectures carried out contrast test, specific point
Cutting result and test index can be shown in See Figure 7.
Claims (4)
1. one kind is based on improved full convolutional neural networks brain tumor image partition method, it is characterized in that, step is as follows:
1) image is chosen:First to magnetic resonance MRI brain tumor fluid attented inversion recoveries FLAIR, spin-spin relaxation T2 and T1
Enhancing T1C images use medium filtering denoising, then linear scale fusion are carried out to three kinds of mode, so as to preferably to brain tumor
Split;
2) full convolutional neural networks FCN models are built:Split knot using the brain tumor image after merging well as training sample, expert
True value label of the pixel of fruit figure as training sample, then carry out the training of FCN models;
3) segmentation result is tested, after FCN model trainings are good, any one brain tumor image is swollen using the model of training
Knurl position and the prediction of boundary sizes, and segmentation result is evaluated using corresponding evaluation index, preferably to improve
FCN models.
2. as claimed in claim 1 based on improved full convolutional neural networks brain tumor image partition method, it is characterized in that, survey
Try in segmentation result step, input data is passing through Pooling1, Pooling2, Pooling3, Pooling4 and Pooling5
Maximum pond layer after, the spy of the size of artwork 1/2,1/4 size, 1/8 size, 1/16 size and 1/32 size can be respectively obtained
Levy vectogram.The feature of 4th convolution kernel is carried out deconvolution Upsamping2 benefits to the figure after last Upsamping1
Details is filled, equivalent to one Interpolation Process, then the feature of the 3rd convolution kernel is carried out to the image after Upsamping2 just now
Deconvolution Upsamping3 subsidiary details again, finally again the feature of the 2nd convolution kernel to Upsamping3 just now after
Image carries out deconvolution Upsamping4 subsidiary details again, consequently, it is possible to the spy of the characteristic pattern of deep layer network and shallow-layer network
Sign figure can be complementary to one another, and be finally added and obtained an of a relatively high segmentation figure of precision.
3. as claimed in claim 1 based on improved full convolutional neural networks brain tumor image partition method, it is characterized in that,
FCN carries out up-sampling Upsamping, interpolation point P 4 neighbours to the feature vector chart of last convolutional layer using warp lamination
Point coordinates is respectively Q11=(x1,y1), Q21=(x2,y1), Q12=(x1,y2), Q22=(x2,y2), carry out in the X direction first
First time linear interpolation calculates, and obtains:
<mrow>
<mi>f</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>R</mi>
<mn>1</mn>
</msub>
<mo>)</mo>
</mrow>
<mo>&ap;</mo>
<mfrac>
<mrow>
<msub>
<mi>x</mi>
<mn>2</mn>
</msub>
<mo>-</mo>
<mi>x</mi>
</mrow>
<mrow>
<msub>
<mi>x</mi>
<mn>2</mn>
</msub>
<mo>-</mo>
<msub>
<mi>x</mi>
<mn>1</mn>
</msub>
</mrow>
</mfrac>
<mi>f</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>Q</mi>
<mn>11</mn>
</msub>
<mo>)</mo>
</mrow>
<mo>+</mo>
<mfrac>
<mrow>
<mi>x</mi>
<mo>-</mo>
<msub>
<mi>x</mi>
<mn>1</mn>
</msub>
</mrow>
<mrow>
<msub>
<mi>x</mi>
<mn>2</mn>
</msub>
<mo>-</mo>
<msub>
<mi>x</mi>
<mn>1</mn>
</msub>
</mrow>
</mfrac>
<mi>f</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>Q</mi>
<mn>21</mn>
</msub>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mi>f</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>R</mi>
<mn>2</mn>
</msub>
<mo>)</mo>
</mrow>
<mo>&ap;</mo>
<mfrac>
<mrow>
<msub>
<mi>x</mi>
<mn>2</mn>
</msub>
<mo>-</mo>
<mi>x</mi>
</mrow>
<mrow>
<msub>
<mi>x</mi>
<mn>2</mn>
</msub>
<mo>-</mo>
<msub>
<mi>x</mi>
<mn>1</mn>
</msub>
</mrow>
</mfrac>
<mi>f</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>Q</mi>
<mn>12</mn>
</msub>
<mo>)</mo>
</mrow>
<mo>+</mo>
<mfrac>
<mrow>
<mi>x</mi>
<mo>-</mo>
<msub>
<mi>x</mi>
<mn>1</mn>
</msub>
</mrow>
<mrow>
<msub>
<mi>x</mi>
<mn>2</mn>
</msub>
<mo>-</mo>
<msub>
<mi>x</mi>
<mn>1</mn>
</msub>
</mrow>
</mfrac>
<mi>f</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>Q</mi>
<mn>22</mn>
</msub>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>2</mn>
<mo>)</mo>
</mrow>
</mrow>
Then linear interpolation is carried out in the Y direction, is obtained:
<mrow>
<mi>f</mi>
<mrow>
<mo>(</mo>
<mi>P</mi>
<mo>)</mo>
</mrow>
<mo>&ap;</mo>
<mfrac>
<mrow>
<msub>
<mi>y</mi>
<mn>2</mn>
</msub>
<mo>-</mo>
<mi>y</mi>
</mrow>
<mrow>
<msub>
<mi>y</mi>
<mn>2</mn>
</msub>
<mo>-</mo>
<msub>
<mi>y</mi>
<mn>1</mn>
</msub>
</mrow>
</mfrac>
<mi>f</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>R</mi>
<mn>1</mn>
</msub>
<mo>)</mo>
</mrow>
<mo>+</mo>
<mfrac>
<mrow>
<mi>y</mi>
<mo>-</mo>
<msub>
<mi>y</mi>
<mn>1</mn>
</msub>
</mrow>
<mrow>
<msub>
<mi>y</mi>
<mn>2</mn>
</msub>
<mo>-</mo>
<msub>
<mi>y</mi>
<mn>1</mn>
</msub>
</mrow>
</mfrac>
<mi>f</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>R</mi>
<mn>2</mn>
</msub>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>3</mn>
<mo>)</mo>
</mrow>
</mrow>
Formula (1) and formula (2) are updated into formula can obtain interpolation point P coordinate value f (x, y):
<mrow>
<mtable>
<mtr>
<mtd>
<mrow>
<mi>f</mi>
<mrow>
<mo>(</mo>
<mi>x</mi>
<mo>,</mo>
<mi>y</mi>
<mo>)</mo>
</mrow>
<mo>&ap;</mo>
<mfrac>
<mrow>
<mi>f</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>Q</mi>
<mn>11</mn>
</msub>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<mn>2</mn>
</msub>
<mo>-</mo>
<msub>
<mi>x</mi>
<mn>1</mn>
</msub>
<mo>)</mo>
<mo>(</mo>
<msub>
<mi>y</mi>
<mn>2</mn>
</msub>
<mo>-</mo>
<msub>
<mi>y</mi>
<mn>1</mn>
</msub>
<mo>)</mo>
</mrow>
</mfrac>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<mn>2</mn>
</msub>
<mo>-</mo>
<mi>x</mi>
<mo>)</mo>
</mrow>
<mrow>
<mo>(</mo>
<msub>
<mi>y</mi>
<mn>2</mn>
</msub>
<mo>-</mo>
<mi>y</mi>
<mo>)</mo>
</mrow>
<mo>+</mo>
<mfrac>
<mrow>
<mi>f</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>Q</mi>
<mn>21</mn>
</msub>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<mn>2</mn>
</msub>
<mo>-</mo>
<msub>
<mi>x</mi>
<mn>1</mn>
</msub>
<mo>)</mo>
<mo>(</mo>
<msub>
<mi>y</mi>
<mn>2</mn>
</msub>
<mo>-</mo>
<msub>
<mi>y</mi>
<mn>1</mn>
</msub>
<mo>)</mo>
</mrow>
</mfrac>
<mrow>
<mo>(</mo>
<mi>x</mi>
<mo>-</mo>
<msub>
<mi>x</mi>
<mn>1</mn>
</msub>
<mo>)</mo>
</mrow>
<mrow>
<mo>(</mo>
<msub>
<mi>y</mi>
<mn>2</mn>
</msub>
<mo>-</mo>
<mi>y</mi>
<mo>)</mo>
</mrow>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mo>+</mo>
<mfrac>
<mrow>
<mi>f</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>Q</mi>
<mn>12</mn>
</msub>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<mn>2</mn>
</msub>
<mo>-</mo>
<msub>
<mi>x</mi>
<mn>1</mn>
</msub>
<mo>)</mo>
<mo>(</mo>
<msub>
<mi>y</mi>
<mn>2</mn>
</msub>
<mo>-</mo>
<msub>
<mi>y</mi>
<mn>1</mn>
</msub>
<mo>)</mo>
</mrow>
</mfrac>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<mn>2</mn>
</msub>
<mo>-</mo>
<mi>x</mi>
<mo>)</mo>
</mrow>
<mrow>
<mo>(</mo>
<mi>y</mi>
<mo>-</mo>
<msub>
<mi>y</mi>
<mn>1</mn>
</msub>
<mo>)</mo>
</mrow>
<mo>+</mo>
<mfrac>
<mrow>
<mi>f</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>Q</mi>
<mn>22</mn>
</msub>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<mn>2</mn>
</msub>
<mo>-</mo>
<msub>
<mi>x</mi>
<mn>1</mn>
</msub>
<mo>)</mo>
<mo>(</mo>
<msub>
<mi>y</mi>
<mn>2</mn>
</msub>
<mo>-</mo>
<msub>
<mi>y</mi>
<mn>1</mn>
</msub>
<mo>)</mo>
</mrow>
</mfrac>
<mrow>
<mo>(</mo>
<mi>x</mi>
<mo>-</mo>
<msub>
<mi>x</mi>
<mn>1</mn>
</msub>
<mo>)</mo>
</mrow>
<mrow>
<mo>(</mo>
<mi>y</mi>
<mo>-</mo>
<msub>
<mi>y</mi>
<mn>1</mn>
</msub>
<mo>)</mo>
</mrow>
</mrow>
</mtd>
</mtr>
</mtable>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>4</mn>
<mo>)</mo>
</mrow>
</mrow>
By so operation, input picture identical size finally will be returned to by the feature vector chart in pond, so as to
A prediction is produced to each pixel, while remains the spatial information in original input picture, finally in the spy of up-sampling
Classified pixel-by-pixel in sign figure, so as to realize that image is split.
4. one kind is based on improved full convolutional neural networks brain tumor image segmenting device, it is characterized in that, set by magnetic resonance MRI
Standby, medium filtering denoising device, linearity ratio Fusion Module, capture card, computer;
Linearity ratio Fusion Module is used to the image for choosing tri- kinds of mode of FLAIR, T2 and T1C carrying out melting for linear scale
Close;
Image caused by magnetic resonance MRI machine is successively through medium filtering denoising device, linearity ratio Fusion Module, capture card
Into computer;
Computer installation has structure FCN model modules, test segmentation result module;
Structure FCN model modules are used for, using the brain tumor image after merging well as training sample, the picture of expert's segmentation result figure
True value label of the vegetarian refreshments as training sample, the training of FCN models is carried out, constructs FCN models;
Test segmentation result module is used for, after FCN model trainings are good, using the model of training to any one brain tumor image
The prediction of knub position and boundary sizes is carried out, and segmentation result is evaluated using corresponding evaluation index, so as to more preferable
Improve FCN models in ground.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710812236.9A CN107749061A (en) | 2017-09-11 | 2017-09-11 | Based on improved full convolutional neural networks brain tumor image partition method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710812236.9A CN107749061A (en) | 2017-09-11 | 2017-09-11 | Based on improved full convolutional neural networks brain tumor image partition method and device |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107749061A true CN107749061A (en) | 2018-03-02 |
Family
ID=61254935
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710812236.9A Pending CN107749061A (en) | 2017-09-11 | 2017-09-11 | Based on improved full convolutional neural networks brain tumor image partition method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107749061A (en) |
Cited By (30)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108537192A (en) * | 2018-04-17 | 2018-09-14 | 福州大学 | A kind of remote sensing image ground mulching sorting technique based on full convolutional network |
CN108564582A (en) * | 2018-04-23 | 2018-09-21 | 重庆大学 | A kind of MRI brain tumor automatic identifying methods based on deep neural network |
CN108765447A (en) * | 2018-04-26 | 2018-11-06 | 深圳博脑医疗科技有限公司 | A kind of image partition method, image segmentation device and electronic equipment |
CN108898140A (en) * | 2018-06-08 | 2018-11-27 | 天津大学 | Brain tumor image segmentation algorithm based on improved full convolutional neural networks |
CN108961214A (en) * | 2018-05-31 | 2018-12-07 | 天津大学 | Brain tumor MRI three-dimensional dividing method based on improved continuous type maximum-flow algorithm |
CN109146886A (en) * | 2018-08-19 | 2019-01-04 | 沈阳农业大学 | A kind of RGBD image, semantic segmentation optimization method based on depth density |
CN109214388A (en) * | 2018-11-21 | 2019-01-15 | 济南浪潮高新科技投资发展有限公司 | A kind of lesion segmentation approach and device based on personalized converged network |
CN109389585A (en) * | 2018-09-20 | 2019-02-26 | 东南大学 | A kind of brain tissue extraction method based on full convolutional neural networks |
CN109584254A (en) * | 2019-01-07 | 2019-04-05 | 浙江大学 | A kind of heart left ventricle's dividing method based on the full convolutional neural networks of deep layer |
CN109800873A (en) * | 2019-01-29 | 2019-05-24 | 北京旷视科技有限公司 | Image processing method and device |
CN110009619A (en) * | 2019-04-02 | 2019-07-12 | 清华大学深圳研究生院 | A kind of image analysis method based on fluorescence-encoded liquid phase biochip |
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 |
CN110120048A (en) * | 2019-04-12 | 2019-08-13 | 天津大学 | In conjunction with the three-dimensional brain tumor image partition method for improving U-Net and CMF |
CN110163852A (en) * | 2019-05-13 | 2019-08-23 | 北京科技大学 | The real-time sideslip detection method of conveyer belt based on lightweight convolutional neural networks |
CN110210483A (en) * | 2019-06-13 | 2019-09-06 | 上海鹰瞳医疗科技有限公司 | Medical image lesion region dividing method, model training method and equipment |
CN110297483A (en) * | 2018-03-21 | 2019-10-01 | 广州极飞科技有限公司 | To operating area boundary acquisition methods, device, operation flight course planning method |
CN110533676A (en) * | 2019-09-06 | 2019-12-03 | 青岛海信医疗设备股份有限公司 | Tumor image dividing method, device and terminal device |
CN110619633A (en) * | 2019-09-10 | 2019-12-27 | 武汉科技大学 | Liver image segmentation method based on multi-path filtering strategy |
WO2020019671A1 (en) * | 2018-07-23 | 2020-01-30 | 哈尔滨工业大学(深圳) | Breast lump detection and classification system and computer-readable storage medium |
CN110853738A (en) * | 2019-10-12 | 2020-02-28 | 北京量健智能科技有限公司 | Imaging method and device under action of contrast agent |
CN110910405A (en) * | 2019-11-20 | 2020-03-24 | 湖南师范大学 | Brain tumor segmentation method and system based on multi-scale cavity convolutional neural network |
WO2020078269A1 (en) * | 2018-10-16 | 2020-04-23 | 腾讯科技(深圳)有限公司 | Method and device for three-dimensional image semantic segmentation, terminal and storage medium |
WO2020151355A1 (en) * | 2019-01-25 | 2020-07-30 | 厦门大学 | Deep learning-based magnetic resonance spectroscopy reconstruction method |
CN111493828A (en) * | 2020-04-23 | 2020-08-07 | 四川大学 | Sequence-to-sequence sleep disorder detection method based on full convolution network |
WO2020164493A1 (en) * | 2019-02-14 | 2020-08-20 | 腾讯科技(深圳)有限公司 | Method and apparatus for filtering medical image area, and storage medium |
CN112034456A (en) * | 2020-08-27 | 2020-12-04 | 五邑大学 | Smoke inspection system, method, control device and storage medium |
CN112289447A (en) * | 2020-10-30 | 2021-01-29 | 四川大学华西医院 | Surgical incision healing grade discrimination system |
CN112912924A (en) * | 2018-10-22 | 2021-06-04 | 皇家飞利浦有限公司 | Correctness of prediction algorithm segmentation |
CN114092494A (en) * | 2021-11-29 | 2022-02-25 | 长春工业大学 | Brain MR image segmentation method based on superpixel and full convolution neural network |
CN115239688A (en) * | 2022-08-09 | 2022-10-25 | 四川大学华西医院 | Brain metastasis tumor identification method and system based on magnetic resonance contrast enhanced 3D-T1WI image |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106600571A (en) * | 2016-11-07 | 2017-04-26 | 中国科学院自动化研究所 | Brain tumor automatic segmentation method through fusion of full convolutional neural network and conditional random field |
CN106780499A (en) * | 2016-12-07 | 2017-05-31 | 电子科技大学 | A kind of multi-modal brain tumor image partition method based on stacking autocoding network |
CN107016681A (en) * | 2017-03-29 | 2017-08-04 | 浙江师范大学 | Brain MRI lesion segmentation approach based on full convolutional network |
-
2017
- 2017-09-11 CN CN201710812236.9A patent/CN107749061A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106600571A (en) * | 2016-11-07 | 2017-04-26 | 中国科学院自动化研究所 | Brain tumor automatic segmentation method through fusion of full convolutional neural network and conditional random field |
CN106780499A (en) * | 2016-12-07 | 2017-05-31 | 电子科技大学 | A kind of multi-modal brain tumor image partition method based on stacking autocoding network |
CN107016681A (en) * | 2017-03-29 | 2017-08-04 | 浙江师范大学 | Brain MRI lesion segmentation approach based on full convolutional network |
Non-Patent Citations (1)
Title |
---|
AVI BEN-COHEN等: "Fully Convolutional Network for Liver Segmentation and Lesions Detection", 《INTERNATIONAL WORKSHOP ON DEEP LEARNING IN MEDICAL IMAGE ANALYSIS》 * |
Cited By (50)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110297483A (en) * | 2018-03-21 | 2019-10-01 | 广州极飞科技有限公司 | To operating area boundary acquisition methods, device, operation flight course planning method |
CN110297483B (en) * | 2018-03-21 | 2020-12-18 | 广州极飞科技有限公司 | Method and device for obtaining boundary of area to be operated and operation route planning method |
CN108537192B (en) * | 2018-04-17 | 2021-11-30 | 福州大学 | Remote sensing image earth surface coverage classification method based on full convolution network |
CN108537192A (en) * | 2018-04-17 | 2018-09-14 | 福州大学 | A kind of remote sensing image ground mulching sorting technique based on full convolutional network |
CN108564582A (en) * | 2018-04-23 | 2018-09-21 | 重庆大学 | A kind of MRI brain tumor automatic identifying methods based on deep neural network |
CN108564582B (en) * | 2018-04-23 | 2020-08-11 | 重庆大学 | MRI brain tumor image automatic optimization method based on deep neural network |
CN108765447B (en) * | 2018-04-26 | 2021-02-12 | 深圳博脑医疗科技有限公司 | Image segmentation method, image segmentation device and electronic equipment |
CN108765447A (en) * | 2018-04-26 | 2018-11-06 | 深圳博脑医疗科技有限公司 | A kind of image partition method, image segmentation device and electronic equipment |
CN108961214A (en) * | 2018-05-31 | 2018-12-07 | 天津大学 | Brain tumor MRI three-dimensional dividing method based on improved continuous type maximum-flow algorithm |
CN108898140A (en) * | 2018-06-08 | 2018-11-27 | 天津大学 | Brain tumor image segmentation algorithm based on improved full convolutional neural networks |
WO2020019671A1 (en) * | 2018-07-23 | 2020-01-30 | 哈尔滨工业大学(深圳) | Breast lump detection and classification system and computer-readable storage medium |
CN109146886A (en) * | 2018-08-19 | 2019-01-04 | 沈阳农业大学 | A kind of RGBD image, semantic segmentation optimization method based on depth density |
CN109389585B (en) * | 2018-09-20 | 2021-11-02 | 东南大学 | Brain tissue extraction method based on full convolution neural network |
CN109389585A (en) * | 2018-09-20 | 2019-02-26 | 东南大学 | A kind of brain tissue extraction method based on full convolutional neural networks |
WO2020078269A1 (en) * | 2018-10-16 | 2020-04-23 | 腾讯科技(深圳)有限公司 | Method and device for three-dimensional image semantic segmentation, terminal and storage medium |
US11861501B2 (en) | 2018-10-16 | 2024-01-02 | Tencent Technology (Shenzhen) Company Limited | Semantic segmentation method and apparatus for three-dimensional image, terminal, and storage medium |
CN112912924B (en) * | 2018-10-22 | 2024-05-14 | 皇家飞利浦有限公司 | Accuracy of predictive algorithm segmentation |
CN112912924A (en) * | 2018-10-22 | 2021-06-04 | 皇家飞利浦有限公司 | Correctness of prediction algorithm segmentation |
CN109214388A (en) * | 2018-11-21 | 2019-01-15 | 济南浪潮高新科技投资发展有限公司 | A kind of lesion segmentation approach and device based on personalized converged network |
CN109584254A (en) * | 2019-01-07 | 2019-04-05 | 浙江大学 | A kind of heart left ventricle's dividing method based on the full convolutional neural networks of deep layer |
US11782111B2 (en) | 2019-01-25 | 2023-10-10 | Xiamen University | Method for reconstructing magnetic resonance spectrum based on deep learning |
WO2020151355A1 (en) * | 2019-01-25 | 2020-07-30 | 厦门大学 | Deep learning-based magnetic resonance spectroscopy reconstruction method |
CN109800873B (en) * | 2019-01-29 | 2021-03-23 | 北京旷视科技有限公司 | Image processing method and device |
CN109800873A (en) * | 2019-01-29 | 2019-05-24 | 北京旷视科技有限公司 | Image processing method and device |
US11995821B2 (en) | 2019-02-14 | 2024-05-28 | Tencent Technology (Shenzhen) Company Limited | Medical image region screening method and apparatus and storage medium |
WO2020164493A1 (en) * | 2019-02-14 | 2020-08-20 | 腾讯科技(深圳)有限公司 | Method and apparatus for filtering medical image area, and storage medium |
CN110009619A (en) * | 2019-04-02 | 2019-07-12 | 清华大学深圳研究生院 | A kind of image analysis method based on fluorescence-encoded liquid phase biochip |
CN110120048A (en) * | 2019-04-12 | 2019-08-13 | 天津大学 | In conjunction with the three-dimensional brain tumor image partition method for improving U-Net and CMF |
CN110120048B (en) * | 2019-04-12 | 2023-06-06 | 天津大学 | Three-dimensional brain tumor image segmentation method combining improved U-Net and CMF |
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 |
CN110163852A (en) * | 2019-05-13 | 2019-08-23 | 北京科技大学 | The real-time sideslip detection method of conveyer belt based on lightweight convolutional neural networks |
CN110163852B (en) * | 2019-05-13 | 2021-10-15 | 北京科技大学 | Conveying belt real-time deviation detection method based on lightweight convolutional neural network |
CN110210483A (en) * | 2019-06-13 | 2019-09-06 | 上海鹰瞳医疗科技有限公司 | Medical image lesion region dividing method, model training method and equipment |
CN110533676A (en) * | 2019-09-06 | 2019-12-03 | 青岛海信医疗设备股份有限公司 | Tumor image dividing method, device and terminal device |
CN110533676B (en) * | 2019-09-06 | 2022-08-16 | 青岛海信医疗设备股份有限公司 | Tumor image segmentation method and device and terminal equipment |
CN110619633A (en) * | 2019-09-10 | 2019-12-27 | 武汉科技大学 | Liver image segmentation method based on multi-path filtering strategy |
CN110619633B (en) * | 2019-09-10 | 2023-06-23 | 武汉科技大学 | Liver image segmentation method based on multipath filtering strategy |
CN110853738A (en) * | 2019-10-12 | 2020-02-28 | 北京量健智能科技有限公司 | Imaging method and device under action of contrast agent |
CN110853738B (en) * | 2019-10-12 | 2023-08-18 | 拜耳股份有限公司 | Imaging method and device under action of contrast agent |
CN110910405B (en) * | 2019-11-20 | 2023-04-18 | 湖南师范大学 | Brain tumor segmentation method and system based on multi-scale cavity convolutional neural network |
CN110910405A (en) * | 2019-11-20 | 2020-03-24 | 湖南师范大学 | Brain tumor segmentation method and system based on multi-scale cavity convolutional neural network |
CN111493828B (en) * | 2020-04-23 | 2021-11-16 | 四川大学 | Sequence-to-sequence sleep disorder detection method based on full convolution network |
CN111493828A (en) * | 2020-04-23 | 2020-08-07 | 四川大学 | Sequence-to-sequence sleep disorder detection method based on full convolution network |
CN112034456B (en) * | 2020-08-27 | 2023-10-17 | 五邑大学 | Smoke inspection system, method, control device and storage medium |
CN112034456A (en) * | 2020-08-27 | 2020-12-04 | 五邑大学 | Smoke inspection system, method, control device and storage medium |
CN112289447A (en) * | 2020-10-30 | 2021-01-29 | 四川大学华西医院 | Surgical incision healing grade discrimination system |
CN112289447B (en) * | 2020-10-30 | 2022-03-08 | 四川大学华西医院 | Surgical incision healing grade discrimination system |
CN114092494A (en) * | 2021-11-29 | 2022-02-25 | 长春工业大学 | Brain MR image segmentation method based on superpixel and full convolution neural network |
CN115239688A (en) * | 2022-08-09 | 2022-10-25 | 四川大学华西医院 | Brain metastasis tumor identification method and system based on magnetic resonance contrast enhanced 3D-T1WI image |
CN115239688B (en) * | 2022-08-09 | 2024-03-12 | 四川大学华西医院 | Brain metastasis recognition method and system based on magnetic resonance contrast enhancement 3D-T1WI image |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107749061A (en) | Based on improved full convolutional neural networks brain tumor image partition method and device | |
US11967072B2 (en) | Three-dimensional object segmentation of medical images localized with object detection | |
US11783451B2 (en) | Systems and methods for reducing colored noise in medical images using deep neural network | |
CN102737250B (en) | To the method and system of spinal injury automatic detection in 3D medical images | |
CN103339653B (en) | The method delineating tissue injury | |
CN107644420A (en) | Blood-vessel image dividing method, MRI system based on central line pick-up | |
CN110910405B (en) | Brain tumor segmentation method and system based on multi-scale cavity convolutional neural network | |
CN110188792A (en) | The characteristics of image acquisition methods of prostate MRI 3-D image | |
CN110084823A (en) | Three-dimensional brain tumor image partition method based on cascade anisotropy FCNN | |
WO2007079207A2 (en) | An integrated segmentation and classification approach applied to medical applications analysis | |
CN115496771A (en) | Brain tumor segmentation method based on brain three-dimensional MRI image design | |
CN112767417B (en) | Multi-modal image segmentation method based on cascaded U-Net network | |
CN112529909A (en) | Tumor image brain region segmentation method and system based on image completion | |
CN106127263A (en) | The human brain magnetic resonance image (MRI) classifying identification method extracted based on three-dimensional feature and system | |
Tian et al. | Improving in vivo human cerebral cortical surface reconstruction using data-driven super-resolution | |
CN108053401A (en) | A kind of B ultrasound image processing method and device | |
Meera et al. | A review on automatic detection of brain tumor using computer aided diagnosis system through MRI | |
Mooij et al. | Automatic segmentation of prostate zones | |
CN114565572A (en) | Cerebral hemorrhage CT image classification method based on image sequence analysis | |
CN113313728B (en) | Intracranial artery segmentation method and system | |
Zhang et al. | 3D cross-scale feature transformer network for brain MR image super-resolution | |
CN111429404B (en) | Imaging system and method for detecting cardiovascular and cerebrovascular vessels | |
CN117392389A (en) | MT-SASS network-based kidney cancer MRI image segmentation classification method | |
AlZubi | 3D multiresolution statistical approaches for accelerated medical image and volume segmentation | |
CN116664590A (en) | Automatic segmentation method and device based on dynamic contrast enhancement magnetic resonance image |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20180302 |