CN108986124A - In conjunction with Analysis On Multi-scale Features convolutional neural networks retinal vascular images dividing method - Google Patents
In conjunction with Analysis On Multi-scale Features convolutional neural networks retinal vascular images dividing method Download PDFInfo
- Publication number
- CN108986124A CN108986124A CN201810635753.8A CN201810635753A CN108986124A CN 108986124 A CN108986124 A CN 108986124A CN 201810635753 A CN201810635753 A CN 201810635753A CN 108986124 A CN108986124 A CN 108986124A
- Authority
- CN
- China
- Prior art keywords
- image
- layer
- retinal
- neural networks
- convolutional
- 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
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
- G06T7/12—Edge-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/40—Image enhancement or restoration by the use of histogram techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/03—Recognition of patterns in medical or anatomical images
Abstract
The invention belongs to technical field of image processing, to realize automatically extracting and dividing for retinal vessel, the anti-interference ability to the factors such as blood vessel shade and tissue deformation is improved, keeps the Average Accuracy of vessel segmentation higher.The present invention, in conjunction with Analysis On Multi-scale Features convolutional neural networks retinal vascular images dividing method.Firstly, suitably being pre-processed to retinal images, including carry out restricted contrast self-adapting histogram equilibrium processing and gamma brightness adjustment；Simultaneously, data amplification is carried out for the less problem of retinal image data, cutting piecemeal is carried out to experimental image, secondly, combining the retinal vessel of Analysis On Multi-scale Features to divide network by building, spatial pyramid cavity pondization is introduced into coding-decoder architecture convolutional neural networks, pass through successive ignition, the autonomous optimization for carrying out model parameter, realizes the automatic cutting procedure of Pixel-level retinal vessel, obtains retinal vessel segmentation figure.Present invention is mainly applied to the design and manufacture occasions of medical instrument.
Description
Technical field
The invention belongs to artificial intelligence fields to combine with field of medical image processing, is related to a kind of multiple dimensioned based on combining
The Segmentation Method of Retinal Blood Vessels of the convolutional neural networks of feature can be realized and automatically extract and divide blood vessel tree Image.
Background technique
In numerous fundus oculi diseases, cataract, age-related macular degeneration and diabetic retinopathy are four substantially
Blind inducement, disease incidence is high and harm is serious.Since eyeground is the position that human body can uniquely observe directly blood vessel, to eye
The analysis of base map picture and the major way for being treated as prevention and diagnosis fundus oculi disease.Wherein, eye fundus image blood vessel segmentation is weight
The disease quantitative analysis means wanted, many researchs are unfolded for the blood vessel segmentation of eye fundus image.However current retinal vessel
Segmentation relies primarily on expert and carries out manual markings, and diagnosis process takes considerable time, therefore studies a kind of automation extraction blood vessel
Method it is particularly important.
Currently, the blood vessel segmentation method of eye fundus image is broadly divided into the unsupervised method with supervision.Unsupervised method benefit
Extract target blood with certain rule of the relational design between characteristics of image, mainly include matching matrix, morphological analysis,
Blood vessel tracking, means clustering algorithm etc..There is the method for supervision to construct pixel classifier by extracting feature, introducing supervision message,
Better segmentation effect can be obtained.Feature extraction generally includes the methods of Gabor filtering, Gauss matched filtering.Pixel classifications
Method generally comprises the classifiers such as naive Bayesian, SVM, random forest.However, traditional measure of supervision needs hand-designed phase
Feature is closed, needs to expend great effort and time in parameter optimization, while this mode is affected by designer's subjective experience.
In recent years, the fast development of deep learning especially convolutional neural networks theory, starts in industry and academia
One research boom, image segmentation can be regarded as a kind of pixel classifications problem, the pixel-level image based on deep learning point
The main thought of segmentation method is characteristics of image to be automatically extracted using convolutional neural networks, and believe from feedback network study to supervision
Breath recycles deconvolution to operate enlarged image resolution ratio, restores original image size, the result after being divided.
Summary of the invention
In order to overcome the deficiencies of the prior art, it the present invention is directed to realize automatically extracting and dividing for retinal vessel, improves pair
The anti-interference ability of the factors such as blood vessel shade and tissue deformation, keeps the Average Accuracy of vessel segmentation higher.For this purpose, this hair
It is bright the technical solution adopted is that, in conjunction with Analysis On Multi-scale Features convolutional neural networks retinal vascular images dividing method.Firstly, to view
Nethike embrane image is suitably pre-processed, including carries out restricted contrast self-adapting histogram equilibrium processing and gamma brightness tune
It is whole；Meanwhile data amplification is carried out for the less problem of retinal image data, cutting piecemeal is carried out to experimental image, secondly,
It combines the retinal vessel of Analysis On Multi-scale Features to divide network by constructing, spatial pyramid cavity pondization is introduced into coding-decoding
Device structure convolutional neural networks, by successive ignition, the autonomous optimization for carrying out model parameter realizes Pixel-level retinal vessel certainly
Dynamic cutting procedure, obtains retinal vessel segmentation figure.
Specific refinement step is as follows:
Step 1: retinal vascular images are pre-processed
1) gray level image is converted by colored eye fundus image；
2) image normalization is handled；
3) the limited adaptive histogram equalization processing of degree of comparing；
4) gamma adjustment is carried out, brightness of image range is adjusted；
5) image pixel value is normalized, [0,1] is converted by the range of [0,255]；
Step 2: image block and building training sample, the size of original image are W × H, and the size of image block is p × p,
An extracted region images block is chosen in original image, wherein abscissa range isOrdinate range isCenter of the n point as image block is randomly selected in above-mentioned zone, then cuts out n having a size of p × p's
Image block is as training sample, wherein and selected image block center need to ensure in blood-vessel image effective coverage, meanwhile, to view
The standard drawing of retinal vasculature segmentation is also similarly pre-processed, and is corresponded with training data, and training image piecemeal is chosen
1/10 as verifying collection, remaining sample is as training set；
Step 3: convolutional neural networks building and training process in conjunction with Analysis On Multi-scale Features；
Firstly, one U-shaped network of coding-decoder architecture of construction, feature extraction and resolution ratio for blood-vessel image
Restore, which includes five parts: input layer, convolutional layer, pond layer, up-sampling layer and output layer, the network of building are as follows:
Coded portion: input layer, two convolutional layers, the first pond layer, two convolutional layers, the second pond layer, decoded portion:
First up-sampling layer, two convolutional layers, the second up-sampling layer, two convolutional layers, output layers, coding and decoding part constitutes one
A U-shaped segmentation network；
Step 4: building test sample；
Step 5: test process is similarly pre-processed test image block, then is input to trained network
It is tested, obtains retinal vessel segmentation probability graph；Image block after segmentation is spliced, composite artwork size, is passed through
Output obtains retinal vessel segmentation figure after binaryzation.
Space gold tower cavity pond structure: the second pond layer of U-shaped network is below in conjunction with a spatial pyramid type
Pond structure, which is cascaded by four various sizes of convolutional layers: one 1 × 1 convolutional layer and three cavity volumes
The void ratio rate of lamination, three empty convolution is 1,2,4 respectively, and convolution kernel all uses 3 × 3 size, with voidage
Increase, convolution operation can further expansion receptive field, by the way that various sizes of empty convolutional layer is merged, can will part it is thin
Information is saved in conjunction with high-level semantics information, so that network model be made to learn multilayer information representation abundant；Then parallel by four
The output characteristic layer of convolutional layer merges, behind addition one 1 × 1 convolutional layer carry out batch normalization operation and one 3
× 3 convolutional layer reconnects the decoded portion of above-mentioned U-shaped network；By gradient descent method to the parameter in back-propagation process
It optimizes, when error, which is gradually reduced, to tend towards stability, network, which can consider, has restrained, the corresponding label of the image that runs off vascular
Probability graph obtains the optimal weights and offset of network, saves as the training pattern of the database.
Input picture is I (x, y), and normalized process is as follows:
I (x, y) indicates that the pixel value in (x, y), μ (x, y) indicate the mean value of image-region pixel in formula, and σ (x, y) is indicated
The standard deviation of image-region, the image-region of Ω expression calculating mean value and variance, i, the mobile step-length of the region j internal coordinate point,Pixel value after indicating normalization, m × n representative image area pixel sum；By normalized, original image turns
Turning to mean value is 0, the image that variance is 1,It indicates to carry out normalized image into the pixel value after gray scale stretching.
When constructing test sample, in order to divide the image into the image block of several identical sizes completely, in image peripheral
Zero padding is carried out, the size of test image is W × H, and image block size is p × p, wherein moving step length s × s of image block,
Specific fill rule is as follows:
Mod=(H-p) %s, mod=(W-p) %s (6)
In formulaWithFilled height and width are respectively indicated, the quantity m for the image block that above-mentioned piecemeal obtains is indicated
Are as follows:
The features of the present invention and beneficial effect are:
The present invention, can be complete by the local detail information of bottom and top layer by U-shaped network of the building based on deep learning
Office's feature combines, and learns the relationship of each pixel and surrounding pixel in retinal images, utilizes the inside of blood vessel and non-vascular
Feature realizes the blood vessel segmentation of Pixel-level.
Present invention incorporates spatial pyramid type pond structures, expand receptive field, multi-scale image feature is melted
It closes, while empty convolution greatly reduces training parameter, dramatically improves the robustness and accuracy of blood vessel segmentation.
Compared to advanced calculations such as split plot design, support vector machines, extreme learning machine and full convolutional neural networks based on edge
Method, the convolutional neural networks method of combination Analysis On Multi-scale Features proposed by the present invention, which can be realized, automatically extracts characteristics of image,
More preferable to retinal vessel segmentation effect, identification accuracy is higher.In Relational database, every evaluation index of blood vessel segmentation
Advanced international standard is all reached.
Detailed description of the invention:
Fig. 1 is overall flow figure of the invention；
Fig. 2 is the sample effect picture of training image piecemeal: being (a) block diagram of original image；It (b) is corresponding expert point
Cut standard drawing；
Fig. 3 is the network structure of the convolutional neural networks in conjunction with Analysis On Multi-scale Features；
Fig. 4 is retinal vessel segmentation result: (a) original image；(b) pretreated image；(c) retinal vessel point
Cut figure；(d) the manual segmentation figure of the first expert；
Fig. 5 is the assessment of segmentation effect: (a) the method for the present invention and second expert's segmentation standard figure comparative evaluation index；(b)
ROC curve diagram.
Specific embodiment
The invention proposes a kind of Segmentation Method of Retinal Blood Vessels based on the convolutional neural networks for combining Analysis On Multi-scale Features.
Firstly, suitably pre-processed to retinal images, including carry out restricted contrast self-adapting histogram equilibrium processing and
Gamma brightness adjustment.Meanwhile we have carried out data amplification for the less problem of retinal image data, to experimental image into
Row cuts piecemeal, expands broad applicability of the invention.Secondly, the present invention combines the retina of Analysis On Multi-scale Features by building
Spatial pyramid cavity pondization is introduced coding-decoder architecture convolutional neural networks, by repeatedly changing by blood vessel segmentation network
The generation autonomous optimization for carrying out model parameter, realizes the automatic cutting procedure of Pixel-level retinal vessel, obtains retinal vessel segmentation
Figure.On the one hand coding combines the local detail information of bottom with top layer global characteristics with decoding process；On the other hand it introduces
Spatial pyramid cavity convolution pond module can expand receptive field, multi-scale image feature is merged, at the same cavity
Convolution greatly reduces training parameter, dramatically improves the robustness and accuracy of blood vessel segmentation.
Specific refinement step is as follows:
Step 1: retinal vascular images are pre-processed.
1) gray level image is converted by colored eye fundus image.
2) image normalization is handled.In order to generate the image of similarity number magnitude pixel value, need to return original image
One change processing.
3) the limited adaptive histogram equalization processing of degree of comparing.It can be improved retinal fundus images after processing
Quality, the brightness of balanced eye fundus image make it be more suitable for Subsequent vessel extraction.
4) gamma adjustment is carried out, brightness of image is adjusted to suitable range.
5) image pixel value is normalized, [0,1] is converted by the range of [0,255].
Step 2: image block and building training sample.Since training samples number is less, it would be desirable to image data
Carry out data amplification.Assuming that the size of original image is W × H, the size of image block is p × p, and one is chosen in original image
Extracted region images block, wherein abscissa range beOrdinate range isIn above-mentioned zone
Center of the n point as image block is randomly selected, then cuts out the n image blocks having a size of p × p as training sample.Wherein,
Selected image block center need to ensure in blood-vessel image effective coverage.Meanwhile we to retinal vessel segmentation standard drawing
Also same pretreatment has been carried out, has been corresponded with training data.We choose 1/10 conduct verifying collection of training image piecemeal,
Remaining sample is as training set.
Step 3: convolutional neural networks building and training process in conjunction with Analysis On Multi-scale Features.
Firstly, one U-shaped network of coding-decoder architecture of construction, feature extraction and resolution ratio for blood-vessel image
Restore.The network mainly includes five parts: input layer, convolutional layer, pond layer, up-sampling layer (warp lamination) and output layer.
The network constructed in the present invention are as follows:
Coded portion: input layer, two convolutional layers (conv1, conv2), the first pond layer (pool1), two convolutional layers
(conv3, conv4), the second pond layer (pool2).Decoded portion: the first up-sampling layer (up1), two convolutional layers (conv7,
Conv8), the second up-sampling layer (up2), two convolutional layers (conv9, conv10), output layer.Coding and decoding part constitutes
One U-shaped segmentation network.
Space gold tower cavity pond structure: because blood-vessel image includes various sizes of thickness vascular tree, in order to more
Well combine multi-scale image feature, the present invention U-shaped network the second pond layer (pool2) below in conjunction with a space gold
The pond structure of word tower.The structure is mainly cascaded by four various sizes of convolutional layers: one 1 × 1 convolutional layer
(conv-s1) distinguish with three empty convolutional layers (conv-s2, conv-s3, conv-s4), the void ratio of three empty convolution
It is (rate=1,2,4), convolution kernel all uses 3 × 3 size.As voidage increases, convolution operation being capable of further expansion
Receptive field.By the way that various sizes of empty convolutional layer is merged, can by local detail information in conjunction with high-level semantics information, from
And network model is made to learn multilayer information representation abundant.Meanwhile compared to common convolution operation, empty convolution does not increase additionally again
Add parameter, substantially reduces the training time.Then we merge the output characteristic layer of four parallel-convolution layers, behind add
Add one 1 × 1 convolutional layer (conv5) to carry out batch normalization operation and one 3 × 3 convolutional layer (conv6), reconnects
State the decoded portion of U-shaped network.This model optimizes parameter in back-propagation process by gradient descent method, when error by
When gradually decline tends towards stability, network, which can consider, has restrained, and the corresponding label probability figure of the image that runs off vascular obtains network
Optimal weights and offset save as the training pattern of the database.
Step 4: building test sample.In order to divide the image into the image block of several identical sizes, Wo Men completely
Image peripheral carries out zero padding.
Step 5: test process.Test image block is similarly pre-processed, then is input to and has trained to obtain network
It is tested, obtains retinal vessel segmentation probability graph.Image block after segmentation is spliced, composite artwork size, is passed through
Output obtains retinal vessel segmentation figure after binaryzation, shows and divides consistent binary map with expert.By to segmentation result
Parameter evaluation is carried out, the present invention reaches 95% or more accuracy rate.
It is specifically described with reference to the accompanying drawing:
Technology frame chart of the invention is as shown in Figure 1.Specific refinement step is as follows:
Step 1: retinal vascular images are pre-processed.
1) gray level image is converted by colored eye fundus image.
2) image normalization is handled.In order to generate the image of similarity number magnitude pixel value, need to return original image
One change processing.Assuming that input picture is I (x, y), treatment process is as follows:
I (x, y) indicates that the pixel value in (x, y), μ (x, y) indicate the mean value of image-region pixel in formula, and σ (x, y) is indicated
The standard deviation of image-region, the image-region of Ω expression calculating mean value and variance, i, the mobile step-length of the region j internal coordinate point,Pixel value after indicating normalization, m × n representative image area pixel sum.By normalized, original image turns
Turning to mean value is 0, the image that variance is 1.It indicates to carry out normalized image into the pixel value after gray scale stretching.
3) the limited adaptive histogram equalization processing of degree of comparing.It can be improved retinal fundus images after processing
Quality, the brightness of balanced eye fundus image make it be more suitable for Subsequent vessel extraction.
4) gamma adjustment is carried out, brightness of image is adjusted to suitable range.
5) image pixel value is normalized, [0,1] is converted by the range of [0,255].
Step 2: image block and building training sample.Since training samples number is less, it would be desirable to image data
Carry out data amplification.Assuming that the size of original image is W × H, the size of image block is p × p, and one is chosen in original image
Extracted region images block, wherein abscissa range beOrdinate range isIn above-mentioned zone
Center of the n point as image block is randomly selected, then cuts out the n image blocks having a size of p × p as training sample.Wherein,
Selected image block center need to ensure in blood-vessel image effective coverage.The image block size used in the present invention is 32 × 32.
Meanwhile we have also carried out same pretreatment to the standard drawing of retinal vessel segmentation, correspond with training data.We
Choose training image piecemeal 1/10 collects as verifying, remaining sample is as training set.The training of image block in the present invention
Sample and corresponding segmentation standard pattern sheet are as shown in Figure 2.
Step 3: in conjunction with convolutional neural networks building and the training test process of Analysis On Multi-scale Features.
Firstly, one U-shaped network of coding-decoder architecture of construction, feature extraction and resolution ratio for blood-vessel image
Restore.The network mainly includes five parts: input layer, convolutional layer, pond layer, up-sampling layer (warp lamination) and output layer.
The network constructed in the present invention are as follows:
Coded portion: input layer, two convolutional layers (conv1, conv2), the first pond layer (pool1), two convolutional layers
(conv3, conv4), the second pond layer (pool2).Decoded portion: the first up-sampling layer (up1), two convolutional layers (conv7,
Conv8), the second up-sampling layer (up2), two convolutional layers (conv9, conv10), output layer.Coding and decoding part constitutes
One U-shaped segmentation network.
The process that wherein coded portion convolution is realized is as follows:
f(X；W, b)=W*X+b (5)
F (X in formula；W, b) it is represented as the characteristic pattern of output, X is the characteristic pattern of preceding layer input, and W and b represent convolution kernel
And offset, wherein * represents convolution operation.Since the convolution sum pondization operation of cataloged procedure can make characteristic pattern become smaller, in order to extensive
The size of restored map needs to expand resolution ratio using up-sampling or deconvolution operation.
Coded portion convolutional layer obtains the constant characteristic pattern of size using zero padding mode in the present invention, and convolution kernel all uses 3
× 3 size, step-length 1 connect the activation primitive of a ReLU behind convolutional layer.The Chi Huahe that pond layer uses having a size of 2 ×
2, on the one hand the maximum pond layer (max-pooling) that step-length is 2, pondization operation is reduced the effect of parameter, dimensionality reduction degree, another
Aspect reduces convolution process parameter error, more retains texture information, has rotational invariance.Each pondization operates can will be special
The number for levying layer is double, and upper sampling process can halve feature number of layers.It up-samples layer and carries out transposition convolution, convolution kernel uses 3
× 3 size, step-length 2, equally by the way of zero padding.It is exported in the last layer using 1 × 1 32 characteristic patterns of convolution kernel
It is trained for target mapping.Model is to intersect entropy function using softmax classification layer, loss function, can be by uninterrupted
Ground is updated and optimizes to weight parameter.Over-fitting in order to prevent, the present invention are enhanced by data and after each convolutional layers
The mode for adding random deactivating layer (dropout) optimizes.
Space gold tower cavity pond structure: because blood-vessel image includes various sizes of thickness vascular tree, in order to more
Well combine multi-scale image feature, the present invention U-shaped network the second pond layer (pool2) below in conjunction with a space gold
The pond structure of word tower.The structure is mainly cascaded by four various sizes of convolutional layers: one 1 × 1 convolutional layer
(conv-s1) distinguish with three empty convolutional layers (conv-s2, conv-s3, conv-s4), the void ratio of three empty convolution
It is (rate=1,2,4), convolution kernel all uses 3 × 3 size.As voidage increases, convolution operation being capable of further expansion
Receptive field.By the way that various sizes of empty convolutional layer is merged, can by local detail information in conjunction with high-level semantics information, from
And network model is made to learn multilayer information representation abundant.Meanwhile compared to common convolution operation, empty convolution does not increase additionally again
Add parameter, substantially reduces the training time.Then we merge the output characteristic layer of four parallel-convolution layers, behind add
Add one 1 × 1 convolutional layer (conv5) to carry out batch normalization operation and one 3 × 3 convolutional layer (conv6), reconnects
State the decoded portion of U-shaped network.Wherein, the feature number of layers of four parallel-convolution layers is all 128, and feature number of layers subtracts after merging
It is less 256.This model optimizes parameter in back-propagation process by gradient descent method, tends to be steady when error is gradually reduced
Periodically, network, which can consider, has restrained, the corresponding label probability figure of the image that runs off vascular, obtain network optimal weights and partially
Shifting amount saves as the training pattern of the database.The convolutional neural networks model knot for the more size characteristics of combination that this patent proposes
Structure is as shown in Figure 3.
Step 4: piecemeal is carried out to test image and corresponding standard drawing.Assuming that the size of test image is W × H, image block
Having a size of p × p, wherein moving step length s × s of image block, in order to divide the image into the figure of several identical sizes completely
As block, we carry out zero padding in image peripheral, and the rule of filling is as follows:
Mod=(H-p) %s, mod=(W-p) %s (6)
In formulaWithRespectively indicate filled height and width.The quantity m for the image block that above-mentioned piecemeal obtains can be with
It indicates are as follows:
Step 5: test image block is similarly pre-processed, then be input to trained network is tested,
Obtain retinal vessel segmentation probability graph.Image block after segmentation is spliced, composite artwork size is defeated after binaryzation
Retinal vessel segmentation figure is obtained out, is showed and is divided consistent binary map with expert.It is commented by carrying out parameter to segmentation result
Estimate, this patent obtains 95% or more accuracy rate, and specific segmentation result is as shown in Figure 4.The multiple dimensioned spy of combination that this patent proposes
The network of the convolutional neural networks of sign has good segmentation effect, better than existing in the indexs such as sensitivity, specificity, accuracy
Some advanced algorithms.Specific index evaluation is as shown in Figure 5.
Although above in conjunction with diagram, invention has been described, and the invention is not limited to above-mentioned specific implementations
Mode, the above mentioned embodiment is only schematical, rather than restrictive, and those skilled in the art are at this
Under the enlightenment of invention, without deviating from the spirit of the invention, many variations can also be made, these belong to of the invention
Within protection.
It will be appreciated by those skilled in the art that attached drawing is the schematic diagram of a preferred embodiment, the embodiments of the present invention
Serial number is for illustration only, does not represent the advantages or disadvantages of the embodiments.
The foregoing is merely presently preferred embodiments of the present invention, is not intended to limit the invention, it is all in spirit of the invention and
Within principle, any modification, equivalent replacement, improvement and so on be should all be included in the protection scope of the present invention.
Claims (5)
1. a kind of combination Analysis On Multi-scale Features convolutional neural networks retinal vascular images dividing method, characterized in that firstly, to view
Nethike embrane image is suitably pre-processed, including carries out restricted contrast self-adapting histogram equilibrium processing and gamma brightness tune
It is whole；Meanwhile data amplification is carried out for the less problem of retinal image data, cutting piecemeal is carried out to experimental image, secondly,
It combines the retinal vessel of Analysis On Multi-scale Features to divide network by constructing, spatial pyramid cavity pondization is introduced into coding-decoding
Device structure convolutional neural networks, by successive ignition, the autonomous optimization for carrying out model parameter realizes Pixel-level retinal vessel certainly
Dynamic cutting procedure, obtains retinal vessel segmentation figure.
2. Analysis On Multi-scale Features convolutional neural networks retinal vascular images dividing method is combined as described in claim 1, it is special
Sign is that specific refinement step is as follows:
Step 1: retinal vascular images are pre-processed
1) gray level image is converted by colored eye fundus image；
2) image normalization is handled；
3) the limited adaptive histogram equalization processing of degree of comparing；
4) gamma adjustment is carried out, brightness of image range is adjusted；
5) image pixel value is normalized, [0,1] is converted by the range of [0,255]；
Step 2: image block and building training sample, the size of original image are W × H, and the size of image block is p × p, in original
An extracted region images block is chosen in beginning image, wherein abscissa range isOrdinate range isCenter of the n point as image block is randomly selected in above-mentioned zone, then cuts out n having a size of p × p's
Image block is as training sample, wherein and selected image block center need to ensure in blood-vessel image effective coverage, meanwhile, to view
The standard drawing of retinal vasculature segmentation is also similarly pre-processed, and is corresponded with training data, and training image piecemeal is chosen
1/10 as verifying collection, remaining sample is as training set；
Step 3: convolutional neural networks building and training process in conjunction with Analysis On Multi-scale Features；
Firstly, one U-shaped network of coding-decoder architecture of construction, the feature extraction and resolution ratio for blood-vessel image are extensive
Multiple, which includes five parts: input layer, convolutional layer, pond layer, up-sampling layer and output layer, the network of building are as follows:
Coded portion: input layer, two convolutional layers, the first pond layer, two convolutional layers, the second pond layer, decoded portion: first
Layer, two convolutional layers, the second up-sampling layer, two convolutional layers, output layers are up-sampled, coding and decoding part constitutes a U
The segmentation network of type；
Step 4: building test sample；
Step 5: test process is similarly pre-processed test image block, then is input to trained network and is carried out
Test obtains retinal vessel segmentation probability graph；Image block after segmentation is spliced, composite artwork size, by two-value
Output obtains retinal vessel segmentation figure after change.
3. Analysis On Multi-scale Features convolutional neural networks retinal vascular images dividing method is combined as claimed in claim 2, it is special
Sign is space gold tower cavity pond structure: pond of the second pond layer of U-shaped network below in conjunction with a spatial pyramid type
Change structure, which is cascaded by four various sizes of convolutional layers: one 1 × 1 convolutional layer and three empty convolutional layers,
The void ratio rate of three empty convolution is 1,2,4 respectively, and convolution kernel all uses 3 × 3 size, as voidage increases,
Convolution operation can further expansion receptive field local detail can be believed by merging various sizes of empty convolutional layer
Breath is in conjunction with high-level semantics information, so that network model be made to learn multilayer information representation abundant；Then by four parallel-convolutions
The output characteristic layer of layer merges, behind addition one 1 × 1 convolutional layer carry out batch normalization operation and one 3 × 3
Convolutional layer reconnects the decoded portion of above-mentioned U-shaped network；The parameter in back-propagation process is carried out by gradient descent method excellent
Change, when error, which is gradually reduced, to tend towards stability, network, which can consider, has restrained, the corresponding label probability of the image that runs off vascular
Figure, obtains the optimal weights and offset of network, saves as the training pattern of the database.
4. Analysis On Multi-scale Features convolutional neural networks retinal vascular images dividing method is combined as claimed in claim 2, it is special
Sign is that input picture is I (x, y), and normalized process is as follows:
I (x, y) indicates that the pixel value in (x, y), μ (x, y) indicate the mean value of image-region pixel in formula, and σ (x, y) indicates image
The standard deviation in region, the image-region of Ω expression calculating mean value and variance, i, the mobile step-length of the region j internal coordinate point,
Pixel value after indicating normalization, m × n representative image area pixel sum；By normalized, original image is converted into
Value is 0, the image that variance is 1,It indicates to carry out normalized image into the pixel value after gray scale stretching.
5. Analysis On Multi-scale Features convolutional neural networks retinal vascular images dividing method is combined as claimed in claim 2, it is special
Sign is, when constructing test sample, in order to divide the image into the image block of several identical sizes completely, carries out in image peripheral
Zero padding, the size of test image are W × H, and image block size is p × p, wherein moving step length s × s of image block, specifically
Fill rule is as follows:
Mod=(H-p) %s, mod=(W-p) %s (6)
In formulaWithFilled height and width are respectively indicated, the quantity m for the image block that above-mentioned piecemeal obtains is indicated are as follows:
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810635753.8A CN108986124A (en) | 2018-06-20 | 2018-06-20 | In conjunction with Analysis On Multi-scale Features convolutional neural networks retinal vascular images dividing method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810635753.8A CN108986124A (en) | 2018-06-20 | 2018-06-20 | In conjunction with Analysis On Multi-scale Features convolutional neural networks retinal vascular images dividing method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108986124A true CN108986124A (en) | 2018-12-11 |
Family
ID=64541475
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810635753.8A Pending CN108986124A (en) | 2018-06-20 | 2018-06-20 | In conjunction with Analysis On Multi-scale Features convolutional neural networks retinal vascular images dividing method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108986124A (en) |
Cited By (68)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109544585A (en) * | 2018-12-19 | 2019-03-29 | 中国石油大学(华东) | A kind of cholelithiasis CT medical image data Enhancement Method based on lightweight convolutional neural networks |
| CN109636802A (en) * | 2019-01-18 | 2019-04-16 | 天津工业大学 | Pulmonary parenchyma based on depth convolutional neural networks is through CT image partition method |
| CN109685813A (en) * | 2018-12-27 | 2019-04-26 | 江西理工大学 | A kind of U-shaped Segmentation Method of Retinal Blood Vessels of adaptive scale information |
| CN109727259A (en) * | 2019-01-07 | 2019-05-07 | 哈尔滨理工大学 | A kind of retinal images partitioning algorithm based on residual error U-NET network |
| CN109784402A (en) * | 2019-01-15 | 2019-05-21 | 中国第一汽车股份有限公司 | Quick unmanned vehicle Driving Scene dividing method based on multi-level features fusion |
| CN109859210A (en) * | 2018-12-25 | 2019-06-07 | 上海联影智能医疗科技有限公司 | A kind of medical data processing unit and method |
| CN109961446A (en) * | 2019-03-27 | 2019-07-02 | 深圳视见医疗科技有限公司 | CT/MR three-dimensional image segmentation processing method, device, equipment and medium |
| CN109996071A (en) * | 2019-03-27 | 2019-07-09 | 上海交通大学 | Variable bit rate image coding, decoding system and method based on deep learning |
| CN110120055A (en) * | 2019-04-12 | 2019-08-13 | 浙江大学 | Fundus fluorescein angiography image based on deep learning is without perfusion area automatic division method |
| CN110136147A (en) * | 2019-05-21 | 2019-08-16 | 湖北工业大学 | A kind of method, apparatus and storage medium of the Medical Image Segmentation based on U-Net model |
| CN110189337A (en) * | 2019-05-31 | 2019-08-30 | 广东工业大学 | A kind of automatic Pilot image, semantic dividing method |
| CN110197491A (en) * | 2019-05-17 | 2019-09-03 | 上海联影智能医疗科技有限公司 | Image partition method, device, equipment and storage medium |
| CN110222726A (en) * | 2019-05-15 | 2019-09-10 | 北京字节跳动网络技术有限公司 | Image processing method, device and electronic equipment |
| CN110363776A (en) * | 2019-06-28 | 2019-10-22 | 联想(北京)有限公司 | Image processing method and electronic equipment |
| CN110414380A (en) * | 2019-07-10 | 2019-11-05 | 上海交通大学 | A kind of students ' behavior detection method based on target detection |
| CN110472483A (en) * | 2019-07-02 | 2019-11-19 | 五邑大学 | A kind of method and device of the small sample semantic feature enhancing towards SAR image |
| CN110544261A (en) * | 2019-09-04 | 2019-12-06 | 东北大学 | Blast furnace tuyere coal injection state detection method based on image processing |
| CN110689526A (en) * | 2019-09-09 | 2020-01-14 | 北京航空航天大学 | Retinal blood vessel segmentation method and system based on retinal fundus image |
| CN110751636A (en) * | 2019-10-12 | 2020-02-04 | 天津工业大学 | Fundus image retinal arteriosclerosis detection method based on improved coding and decoding network |
| CN110781895A (en) * | 2019-10-10 | 2020-02-11 | 湖北工业大学 | Image semantic segmentation method based on convolutional neural network |
| CN110796162A (en) * | 2019-09-18 | 2020-02-14 | 平安科技（深圳）有限公司 | Image recognition method, image recognition model training method, image recognition device, image recognition training device and storage medium |
| CN110866550A (en) * | 2019-11-01 | 2020-03-06 | 云南大学 | Convolutional neural network, pyramid strip pooling method and malicious software classification method |
| 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 |
| CN110992364A (en) * | 2019-12-31 | 2020-04-10 | 重庆艾可立安医疗器械有限公司 | Retina image recognition method and device, computer equipment and storage medium |
| CN110991617A (en) * | 2019-12-02 | 2020-04-10 | 华东师范大学 | Construction method of kaleidoscope convolution network |
| CN111000563A (en) * | 2019-11-22 | 2020-04-14 | 北京理工大学 | Automatic measuring method and device for retinal artery and vein diameter ratio |
| CN111079698A (en) * | 2019-12-30 | 2020-04-28 | 杭州小马智趣科技有限公司 | Method and device for recognizing tangram toy |
| CN111080646A (en) * | 2019-11-25 | 2020-04-28 | 杭州电子科技大学 | Improved image segmentation method based on wide-activation convolutional neural network |
| CN111127487A (en) * | 2019-12-27 | 2020-05-08 | 电子科技大学 | Real-time multi-tissue medical image segmentation method |
| CN111145170A (en) * | 2019-12-31 | 2020-05-12 | 电子科技大学 | Medical image segmentation method based on deep learning |
| CN111145200A (en) * | 2019-11-13 | 2020-05-12 | 西北大学 | Blood vessel center line tracking method combining convolutional neural network and cyclic neural network |
| CN111160114A (en) * | 2019-12-10 | 2020-05-15 | 深圳数联天下智能科技有限公司 | Gesture recognition method, device, equipment and computer readable storage medium |
| CN111160276A (en) * | 2019-12-31 | 2020-05-15 | 重庆大学 | U-shaped cavity full-volume integral cutting network identification model based on remote sensing image |
| TWI694414B (en) * | 2018-12-18 | 2020-05-21 | 國立成功大學 | Analyzing method, computer program product, and electric device using convolutional neural network |
| CN111242949A (en) * | 2020-01-02 | 2020-06-05 | 浙江工业大学 | Fundus image blood vessel segmentation method based on full convolution neural network multi-scale features |
| CN111405295A (en) * | 2020-02-24 | 2020-07-10 | 核芯互联科技(青岛)有限公司 | Video coding unit segmentation method, system and hardware implementation method |
| CN111462013A (en) * | 2020-04-03 | 2020-07-28 | 西安交通大学 | Single-image rain removing method based on structured residual learning |
| CN111476301A (en) * | 2019-12-26 | 2020-07-31 | 山东中科先进技术研究院有限公司 | Medical image classification method and system based on machine learning |
| CN111488964A (en) * | 2019-01-29 | 2020-08-04 | 北京市商汤科技开发有限公司 | Image processing method and device and neural network training method and device |
| CN111524142A (en) * | 2020-03-10 | 2020-08-11 | 浙江工业大学 | Automatic segmentation method for cerebrovascular image |
| CN111598894A (en) * | 2020-04-17 | 2020-08-28 | 哈尔滨工业大学 | Retina blood vessel image segmentation system based on global information convolution neural network |
| CN111612856A (en) * | 2020-05-25 | 2020-09-01 | 中南大学 | Retina neovascularization detection method and imaging method for color fundus image |
| CN111666842A (en) * | 2020-05-25 | 2020-09-15 | 东华大学 | Shadow detection method based on double-current-cavity convolution neural network |
| CN111860233A (en) * | 2020-07-06 | 2020-10-30 | 中国科学院空天信息创新研究院 | SAR image complex building extraction method and system based on attention network selection |
| CN111862099A (en) * | 2020-06-04 | 2020-10-30 | 杭州深睿博联科技有限公司 | Blood vessel segmentation method and device based on pyramid framework and coarse-to-fine strategy |
| CN111915612A (en) * | 2020-07-24 | 2020-11-10 | 山东大学 | Image segmentation method and system based on multi-encoder convolutional neural network |
| CN112053363A (en) * | 2020-08-19 | 2020-12-08 | 苏州超云生命智能产业研究院有限公司 | Retinal vessel segmentation method and device and model construction method |
| CN112085745A (en) * | 2020-09-07 | 2020-12-15 | 福建农林大学 | Retinal vessel image segmentation method of multi-channel U-shaped full convolution neural network based on balanced sampling splicing |
| CN112102283A (en) * | 2020-09-14 | 2020-12-18 | 北京航空航天大学 | Retina fundus blood vessel segmentation method based on depth multi-scale attention convolution neural network |
| CN112183635A (en) * | 2020-09-29 | 2021-01-05 | 南京农业大学 | Method for realizing segmentation and identification of plant leaf lesions by multi-scale deconvolution network |
| CN112200773A (en) * | 2020-09-17 | 2021-01-08 | 苏州慧维智能医疗科技有限公司 | Large intestine polyp detection method based on encoder and decoder of cavity convolution |
| CN112215267A (en) * | 2020-09-25 | 2021-01-12 | 天津大学 | Hyperspectral image-oriented depth space spectrum subspace clustering method |
| CN112348825A (en) * | 2020-10-16 | 2021-02-09 | 佛山科学技术学院 | DR-U-net network method and device for retinal blood flow image segmentation |
| CN112396613A (en) * | 2020-11-17 | 2021-02-23 | 平安科技（深圳）有限公司 | Image segmentation method and device, computer equipment and storage medium |
| CN112508931A (en) * | 2020-12-18 | 2021-03-16 | 闽江学院 | Leukocyte segmentation method based on U-Net and ResNet |
| CN112562058A (en) * | 2020-11-23 | 2021-03-26 | 西安电子科技大学 | Rapid establishing method of intracranial vascular simulation three-dimensional model based on transfer learning |
| CN112712526A (en) * | 2020-12-31 | 2021-04-27 | 杭州电子科技大学 | Retinal vessel segmentation method based on asymmetric convolutional neural network dual channels |
| CN113011340A (en) * | 2021-03-22 | 2021-06-22 | 华南理工大学 | Cardiovascular surgery index risk classification method and system based on retina image |
| CN113076966A (en) * | 2020-01-06 | 2021-07-06 | 字节跳动有限公司 | Image processing method and device, neural network training method and storage medium |
| CN113240677A (en) * | 2021-05-06 | 2021-08-10 | 浙江医院 | Retina optic disc segmentation method based on deep learning |
| CN113793348A (en) * | 2021-09-24 | 2021-12-14 | 河北大学 | Retinal vessel segmentation method and device |
| CN114693698A (en) * | 2020-12-30 | 2022-07-01 | 四川大学华西医院 | Neural network-based computer-aided lung airway segmentation method |
| CN114742917A (en) * | 2022-04-25 | 2022-07-12 | 桂林电子科技大学 | CT image segmentation method based on convolutional neural network |
| CN115205300A (en) * | 2022-09-19 | 2022-10-18 | 华东交通大学 | Fundus blood vessel image segmentation method and system based on cavity convolution and semantic fusion |
| CN115294126A (en) * | 2022-10-08 | 2022-11-04 | 南京诺源医疗器械有限公司 | Intelligent cancer cell identification method for pathological image |
| CN115444378A (en) * | 2022-11-10 | 2022-12-09 | 成都成电金盘健康数据技术有限公司 | Large-scale group injury online detection system |
| CN116269378A (en) * | 2023-01-09 | 2023-06-23 | 西安电子科技大学 | Psychological health state detection device based on skin nicotinic acid response video analysis |
| CN112396613B (en) * | 2020-11-17 | 2024-05-10 | 平安科技（深圳）有限公司 | Image segmentation method, device, computer equipment and storage medium |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101526994A (en) * | 2009-04-03 | 2009-09-09 | 山东大学 | Fingerprint image segmentation method irrelevant to collecting device |
| CN103996018A (en) * | 2014-03-03 | 2014-08-20 | 天津科技大学 | Human-face identification method based on 4DLBP |
| CN106407917A (en) * | 2016-09-05 | 2017-02-15 | 山东大学 | Dynamic scale distribution-based retinal vessel extraction method and system |
| CN106920227A (en) * | 2016-12-27 | 2017-07-04 | 北京工业大学 | Based on the Segmentation Method of Retinal Blood Vessels that deep learning is combined with conventional method |
| CN106952243A (en) * | 2017-03-14 | 2017-07-14 | 哈尔滨工程大学 | UUV Layer Near The Sea Surface infrared image self adaptation merger histogram stretches Enhancement Method |
| CN107256550A (en) * | 2017-06-06 | 2017-10-17 | 电子科技大学 | A kind of retinal image segmentation method based on efficient CNN CRF networks |
| CN107392244A (en) * | 2017-07-18 | 2017-11-24 | 厦门大学 | The image aesthetic feeling Enhancement Method returned based on deep neural network with cascade |
| US20180114116A1 (en) * | 2016-10-26 | 2018-04-26 | Sentient Technologies (Barbados) Limited | Cooperative evolution of deep neural network structures |
-
2018
- 2018-06-20 CN CN201810635753.8A patent/CN108986124A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101526994A (en) * | 2009-04-03 | 2009-09-09 | 山东大学 | Fingerprint image segmentation method irrelevant to collecting device |
| CN103996018A (en) * | 2014-03-03 | 2014-08-20 | 天津科技大学 | Human-face identification method based on 4DLBP |
| CN106407917A (en) * | 2016-09-05 | 2017-02-15 | 山东大学 | Dynamic scale distribution-based retinal vessel extraction method and system |
| US20180114116A1 (en) * | 2016-10-26 | 2018-04-26 | Sentient Technologies (Barbados) Limited | Cooperative evolution of deep neural network structures |
| CN106920227A (en) * | 2016-12-27 | 2017-07-04 | 北京工业大学 | Based on the Segmentation Method of Retinal Blood Vessels that deep learning is combined with conventional method |
| CN106952243A (en) * | 2017-03-14 | 2017-07-14 | 哈尔滨工程大学 | UUV Layer Near The Sea Surface infrared image self adaptation merger histogram stretches Enhancement Method |
| CN107256550A (en) * | 2017-06-06 | 2017-10-17 | 电子科技大学 | A kind of retinal image segmentation method based on efficient CNN CRF networks |
| CN107392244A (en) * | 2017-07-18 | 2017-11-24 | 厦门大学 | The image aesthetic feeling Enhancement Method returned based on deep neural network with cascade |
Cited By (106)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TWI694414B (en) * | 2018-12-18 | 2020-05-21 | 國立成功大學 | Analyzing method, computer program product, and electric device using convolutional neural network |
| CN109544585A (en) * | 2018-12-19 | 2019-03-29 | 中国石油大学(华东) | A kind of cholelithiasis CT medical image data Enhancement Method based on lightweight convolutional neural networks |
| CN109859210A (en) * | 2018-12-25 | 2019-06-07 | 上海联影智能医疗科技有限公司 | A kind of medical data processing unit and method |
| CN109685813A (en) * | 2018-12-27 | 2019-04-26 | 江西理工大学 | A kind of U-shaped Segmentation Method of Retinal Blood Vessels of adaptive scale information |
| CN109685813B (en) * | 2018-12-27 | 2020-10-13 | 江西理工大学 | U-shaped retinal vessel segmentation method capable of adapting to scale information |
| CN109727259A (en) * | 2019-01-07 | 2019-05-07 | 哈尔滨理工大学 | A kind of retinal images partitioning algorithm based on residual error U-NET network |
| CN109784402A (en) * | 2019-01-15 | 2019-05-21 | 中国第一汽车股份有限公司 | Quick unmanned vehicle Driving Scene dividing method based on multi-level features fusion |
| CN109636802A (en) * | 2019-01-18 | 2019-04-16 | 天津工业大学 | Pulmonary parenchyma based on depth convolutional neural networks is through CT image partition method |
| CN111488964A (en) * | 2019-01-29 | 2020-08-04 | 北京市商汤科技开发有限公司 | Image processing method and device and neural network training method and device |
| CN109996071A (en) * | 2019-03-27 | 2019-07-09 | 上海交通大学 | Variable bit rate image coding, decoding system and method based on deep learning |
| CN109961446A (en) * | 2019-03-27 | 2019-07-02 | 深圳视见医疗科技有限公司 | CT/MR three-dimensional image segmentation processing method, device, equipment and medium |
| CN109996071B (en) * | 2019-03-27 | 2020-03-27 | 上海交通大学 | Variable code rate image coding and decoding system and method based on deep learning |
| CN109961446B (en) * | 2019-03-27 | 2021-06-01 | 深圳视见医疗科技有限公司 | CT/MR three-dimensional image segmentation processing method, device, equipment and medium |
| CN110120055A (en) * | 2019-04-12 | 2019-08-13 | 浙江大学 | Fundus fluorescein angiography image based on deep learning is without perfusion area automatic division method |
| CN110222726A (en) * | 2019-05-15 | 2019-09-10 | 北京字节跳动网络技术有限公司 | Image processing method, device and electronic equipment |
| CN110197491B (en) * | 2019-05-17 | 2021-08-17 | 上海联影智能医疗科技有限公司 | Image segmentation method, device, equipment and storage medium |
| CN110197491A (en) * | 2019-05-17 | 2019-09-03 | 上海联影智能医疗科技有限公司 | Image partition method, device, equipment and storage medium |
| CN110136147A (en) * | 2019-05-21 | 2019-08-16 | 湖北工业大学 | A kind of method, apparatus and storage medium of the Medical Image Segmentation based on U-Net model |
| CN110189337A (en) * | 2019-05-31 | 2019-08-30 | 广东工业大学 | A kind of automatic Pilot image, semantic dividing method |
| CN110363776A (en) * | 2019-06-28 | 2019-10-22 | 联想(北京)有限公司 | Image processing method and electronic equipment |
| CN110472483B (en) * | 2019-07-02 | 2022-11-15 | 五邑大学 | SAR image-oriented small sample semantic feature enhancement method and device |
| CN110472483A (en) * | 2019-07-02 | 2019-11-19 | 五邑大学 | A kind of method and device of the small sample semantic feature enhancing towards SAR image |
| CN110414380A (en) * | 2019-07-10 | 2019-11-05 | 上海交通大学 | A kind of students ' behavior detection method based on target detection |
| CN110544261A (en) * | 2019-09-04 | 2019-12-06 | 东北大学 | Blast furnace tuyere coal injection state detection method based on image processing |
| CN110544261B (en) * | 2019-09-04 | 2023-08-29 | 东北大学 | Method for detecting coal injection state of blast furnace tuyere based on image processing |
| CN110689526B (en) * | 2019-09-09 | 2022-06-07 | 北京航空航天大学 | Retinal blood vessel segmentation method and system based on retinal fundus image |
| CN110689526A (en) * | 2019-09-09 | 2020-01-14 | 北京航空航天大学 | Retinal blood vessel segmentation method and system based on retinal fundus image |
| CN110796162B (en) * | 2019-09-18 | 2023-08-29 | 平安科技（深圳）有限公司 | Image recognition method, image recognition training method, image recognition device, image recognition training method, image recognition training device and storage medium |
| WO2021051520A1 (en) * | 2019-09-18 | 2021-03-25 | 平安科技（深圳）有限公司 | Image identification method, identification model training method, related device, and storage medium |
| CN110796162A (en) * | 2019-09-18 | 2020-02-14 | 平安科技（深圳）有限公司 | Image recognition method, image recognition model training method, image recognition device, image recognition training device and storage medium |
| CN110781895A (en) * | 2019-10-10 | 2020-02-11 | 湖北工业大学 | Image semantic segmentation method based on convolutional neural network |
| CN110781895B (en) * | 2019-10-10 | 2023-06-20 | 湖北工业大学 | Image semantic segmentation method based on convolutional neural network |
| CN110751636A (en) * | 2019-10-12 | 2020-02-04 | 天津工业大学 | Fundus image retinal arteriosclerosis detection method based on improved coding and decoding network |
| CN110751636B (en) * | 2019-10-12 | 2023-04-21 | 天津工业大学 | Fundus image retinal arteriosclerosis detection method based on improved coding and decoding network |
| CN110866550A (en) * | 2019-11-01 | 2020-03-06 | 云南大学 | Convolutional neural network, pyramid strip pooling method and malicious software classification method |
| CN110866550B (en) * | 2019-11-01 | 2022-06-14 | 云南大学 | Convolutional neural network, pyramid strip pooling method and malicious software classification method |
| CN111145200A (en) * | 2019-11-13 | 2020-05-12 | 西北大学 | Blood vessel center line tracking method combining convolutional neural network and cyclic 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 |
| 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 |
| CN111000563A (en) * | 2019-11-22 | 2020-04-14 | 北京理工大学 | Automatic measuring method and device for retinal artery and vein diameter ratio |
| CN111080646A (en) * | 2019-11-25 | 2020-04-28 | 杭州电子科技大学 | Improved image segmentation method based on wide-activation convolutional neural network |
| CN111080646B (en) * | 2019-11-25 | 2023-09-05 | 杭州电子科技大学 | Improved image segmentation method based on wide-activation convolutional neural network |
| CN110991617B (en) * | 2019-12-02 | 2020-12-01 | 华东师范大学 | Construction method of kaleidoscope convolution network |
| CN110991617A (en) * | 2019-12-02 | 2020-04-10 | 华东师范大学 | Construction method of kaleidoscope convolution network |
| CN111160114B (en) * | 2019-12-10 | 2024-03-19 | 深圳数联天下智能科技有限公司 | Gesture recognition method, gesture recognition device, gesture recognition equipment and computer-readable storage medium |
| CN111160114A (en) * | 2019-12-10 | 2020-05-15 | 深圳数联天下智能科技有限公司 | Gesture recognition method, device, equipment and computer readable storage medium |
| CN111476301A (en) * | 2019-12-26 | 2020-07-31 | 山东中科先进技术研究院有限公司 | Medical image classification method and system based on machine learning |
| CN111127487A (en) * | 2019-12-27 | 2020-05-08 | 电子科技大学 | Real-time multi-tissue medical image segmentation method |
| CN111127487B (en) * | 2019-12-27 | 2022-04-19 | 电子科技大学 | Real-time multi-tissue medical image segmentation method |
| CN111079698A (en) * | 2019-12-30 | 2020-04-28 | 杭州小马智趣科技有限公司 | Method and device for recognizing tangram toy |
| CN111160276B (en) * | 2019-12-31 | 2023-05-12 | 重庆大学 | U-shaped cavity full convolution segmentation network identification model based on remote sensing image |
| CN110992364A (en) * | 2019-12-31 | 2020-04-10 | 重庆艾可立安医疗器械有限公司 | Retina image recognition method and device, computer equipment and storage medium |
| CN111145170A (en) * | 2019-12-31 | 2020-05-12 | 电子科技大学 | Medical image segmentation method based on deep learning |
| CN110992364B (en) * | 2019-12-31 | 2023-11-28 | 重庆艾可立安医疗器械有限公司 | Retina image recognition method, retina image recognition device, computer equipment and storage medium |
| CN111160276A (en) * | 2019-12-31 | 2020-05-15 | 重庆大学 | U-shaped cavity full-volume integral cutting network identification model based on remote sensing image |
| CN111242949B (en) * | 2020-01-02 | 2021-06-18 | 浙江工业大学 | Fundus image blood vessel segmentation method based on full convolution neural network multi-scale features |
| CN111242949A (en) * | 2020-01-02 | 2020-06-05 | 浙江工业大学 | Fundus image blood vessel segmentation method based on full convolution neural network multi-scale features |
| CN113076966A (en) * | 2020-01-06 | 2021-07-06 | 字节跳动有限公司 | Image processing method and device, neural network training method and storage medium |
| CN113076966B (en) * | 2020-01-06 | 2023-06-13 | 字节跳动有限公司 | Image processing method and device, training method of neural network and storage medium |
| CN111405295A (en) * | 2020-02-24 | 2020-07-10 | 核芯互联科技(青岛)有限公司 | Video coding unit segmentation method, system and hardware implementation method |
| CN111524142B (en) * | 2020-03-10 | 2023-06-30 | 浙江工业大学 | Automatic segmentation method for cerebrovascular images |
| CN111524142A (en) * | 2020-03-10 | 2020-08-11 | 浙江工业大学 | Automatic segmentation method for cerebrovascular image |
| CN111462013A (en) * | 2020-04-03 | 2020-07-28 | 西安交通大学 | Single-image rain removing method based on structured residual learning |
| CN111462013B (en) * | 2020-04-03 | 2022-03-22 | 西安交通大学 | Single-image rain removing method based on structured residual learning |
| CN111598894A (en) * | 2020-04-17 | 2020-08-28 | 哈尔滨工业大学 | Retina blood vessel image segmentation system based on global information convolution neural network |
| CN111666842B (en) * | 2020-05-25 | 2022-08-26 | 东华大学 | Shadow detection method based on double-current-cavity convolution neural network |
| CN111612856B (en) * | 2020-05-25 | 2023-04-18 | 中南大学 | Retina neovascularization detection method and imaging method for color fundus image |
| CN111666842A (en) * | 2020-05-25 | 2020-09-15 | 东华大学 | Shadow detection method based on double-current-cavity convolution neural network |
| CN111612856A (en) * | 2020-05-25 | 2020-09-01 | 中南大学 | Retina neovascularization detection method and imaging method for color fundus image |
| CN111862099A (en) * | 2020-06-04 | 2020-10-30 | 杭州深睿博联科技有限公司 | Blood vessel segmentation method and device based on pyramid framework and coarse-to-fine strategy |
| CN111860233A (en) * | 2020-07-06 | 2020-10-30 | 中国科学院空天信息创新研究院 | SAR image complex building extraction method and system based on attention network selection |
| CN111915612A (en) * | 2020-07-24 | 2020-11-10 | 山东大学 | Image segmentation method and system based on multi-encoder convolutional neural network |
| CN112053363B (en) * | 2020-08-19 | 2023-12-15 | 苏州超云生命智能产业研究院有限公司 | Retina blood vessel segmentation method, retina blood vessel segmentation device and model construction method |
| CN112053363A (en) * | 2020-08-19 | 2020-12-08 | 苏州超云生命智能产业研究院有限公司 | Retinal vessel segmentation method and device and model construction method |
| CN112085745A (en) * | 2020-09-07 | 2020-12-15 | 福建农林大学 | Retinal vessel image segmentation method of multi-channel U-shaped full convolution neural network based on balanced sampling splicing |
| CN112102283A (en) * | 2020-09-14 | 2020-12-18 | 北京航空航天大学 | Retina fundus blood vessel segmentation method based on depth multi-scale attention convolution neural network |
| CN112102283B (en) * | 2020-09-14 | 2022-12-16 | 北京航空航天大学 | Retina fundus blood vessel segmentation method based on depth multi-scale attention convolution neural network |
| CN112200773A (en) * | 2020-09-17 | 2021-01-08 | 苏州慧维智能医疗科技有限公司 | Large intestine polyp detection method based on encoder and decoder of cavity convolution |
| CN112215267B (en) * | 2020-09-25 | 2022-11-01 | 天津大学 | Hyperspectral image-oriented depth space spectrum subspace clustering method |
| CN112215267A (en) * | 2020-09-25 | 2021-01-12 | 天津大学 | Hyperspectral image-oriented depth space spectrum subspace clustering method |
| CN112183635A (en) * | 2020-09-29 | 2021-01-05 | 南京农业大学 | Method for realizing segmentation and identification of plant leaf lesions by multi-scale deconvolution network |
| CN112348825A (en) * | 2020-10-16 | 2021-02-09 | 佛山科学技术学院 | DR-U-net network method and device for retinal blood flow image segmentation |
| CN112396613B (en) * | 2020-11-17 | 2024-05-10 | 平安科技（深圳）有限公司 | Image segmentation method, device, computer equipment and storage medium |
| CN112396613A (en) * | 2020-11-17 | 2021-02-23 | 平安科技（深圳）有限公司 | Image segmentation method and device, computer equipment and storage medium |
| CN112562058B (en) * | 2020-11-23 | 2024-02-23 | 西安电子科技大学 | Method for quickly establishing intracranial vascular simulation three-dimensional model based on transfer learning |
| CN112562058A (en) * | 2020-11-23 | 2021-03-26 | 西安电子科技大学 | Rapid establishing method of intracranial vascular simulation three-dimensional model based on transfer learning |
| CN112508931A (en) * | 2020-12-18 | 2021-03-16 | 闽江学院 | Leukocyte segmentation method based on U-Net and ResNet |
| CN114693698B (en) * | 2020-12-30 | 2023-06-20 | 四川大学华西医院 | Computer-aided lung airway segmentation method based on neural network |
| CN114693698A (en) * | 2020-12-30 | 2022-07-01 | 四川大学华西医院 | Neural network-based computer-aided lung airway segmentation method |
| CN112712526A (en) * | 2020-12-31 | 2021-04-27 | 杭州电子科技大学 | Retinal vessel segmentation method based on asymmetric convolutional neural network dual channels |
| CN112712526B (en) * | 2020-12-31 | 2024-02-27 | 杭州电子科技大学 | Retina blood vessel segmentation method based on asymmetric convolutional neural network double channels |
| CN113011340A (en) * | 2021-03-22 | 2021-06-22 | 华南理工大学 | Cardiovascular surgery index risk classification method and system based on retina image |
| CN113011340B (en) * | 2021-03-22 | 2023-12-19 | 华南理工大学 | Cardiovascular operation index risk classification method and system based on retina image |
| CN113240677A (en) * | 2021-05-06 | 2021-08-10 | 浙江医院 | Retina optic disc segmentation method based on deep learning |
| CN113240677B (en) * | 2021-05-06 | 2022-08-02 | 浙江医院 | Retina optic disc segmentation method based on deep learning |
| CN113793348B (en) * | 2021-09-24 | 2023-08-11 | 河北大学 | Retinal blood vessel segmentation method and device |
| CN113793348A (en) * | 2021-09-24 | 2021-12-14 | 河北大学 | Retinal vessel segmentation method and device |
| CN114742917A (en) * | 2022-04-25 | 2022-07-12 | 桂林电子科技大学 | CT image segmentation method based on convolutional neural network |
| CN114742917B (en) * | 2022-04-25 | 2024-04-26 | 桂林电子科技大学 | CT image segmentation method based on convolutional neural network |
| CN115205300A (en) * | 2022-09-19 | 2022-10-18 | 华东交通大学 | Fundus blood vessel image segmentation method and system based on cavity convolution and semantic fusion |
| CN115205300B (en) * | 2022-09-19 | 2022-12-09 | 华东交通大学 | Fundus blood vessel image segmentation method and system based on cavity convolution and semantic fusion |
| CN115294126B (en) * | 2022-10-08 | 2022-12-16 | 南京诺源医疗器械有限公司 | Cancer cell intelligent identification method for pathological image |
| CN115294126A (en) * | 2022-10-08 | 2022-11-04 | 南京诺源医疗器械有限公司 | Intelligent cancer cell identification method for pathological image |
| CN115444378A (en) * | 2022-11-10 | 2022-12-09 | 成都成电金盘健康数据技术有限公司 | Large-scale group injury online detection system |
| CN116269378B (en) * | 2023-01-09 | 2023-11-17 | 西安电子科技大学 | Psychological health state detection device based on skin nicotinic acid response video analysis |
| CN116269378A (en) * | 2023-01-09 | 2023-06-23 | 西安电子科技大学 | Psychological health state detection device based on skin nicotinic acid response video analysis |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108986124A (en) | In conjunction with Analysis On Multi-scale Features convolutional neural networks retinal vascular images dividing method | |
| CN107437092B (en) | The classification method of retina OCT image based on Three dimensional convolution neural network | |
| CN106920227B (en) | The Segmentation Method of Retinal Blood Vessels combined based on deep learning with conventional method | |
| CN106408562B (en) | Eye fundus image Segmentation Method of Retinal Blood Vessels and system based on deep learning | |
| CN109376636B (en) | Capsule network-based eye fundus retina image classification method | |
| CN112132817B (en) | Retina blood vessel segmentation method for fundus image based on mixed attention mechanism | |
| CN108806792A (en) | Deep learning facial diagnosis system | |
| CN109584246A (en) | Based on the pyramidal DCM cardiac muscle diagnosis and treatment irradiation image dividing method of Analysis On Multi-scale Features | |
| CN109345538A (en) | A kind of Segmentation Method of Retinal Blood Vessels based on convolutional neural networks | |
| CN108615010A (en) | Facial expression recognizing method based on the fusion of parallel convolutional neural networks characteristic pattern | |
| CN112150476B (en) | Coronary artery sequence blood vessel segmentation method based on space-time discriminant feature learning | |
| CN108648191A (en) | Pest image-recognizing method based on Bayes's width residual error neural network | |
| Cao et al. | Gastric cancer diagnosis with mask R-CNN | |
| CN107358262A (en) | The sorting technique and sorter of a kind of high-definition picture | |
| CN109671094A (en) | A kind of eye fundus image blood vessel segmentation method based on frequency domain classification | |
| CN114881962B (en) | Retina image blood vessel segmentation method based on improved U-Net network | |
| CN111815562A (en) | Retinal vessel segmentation method combining U-Net and self-adaptive PCNN | |
| Paul et al. | Octx: Ensembled deep learning model to detect retinal disorders | |
| Shrivastava et al. | A Comprehensive Analysis of Machine Learning Techniques in Biomedical Image Processing Using Convolutional Neural Network | |
| CN107229937A (en) | A kind of retinal vessel sorting technique and device | |
| Ma et al. | Retinal vessel segmentation based on generative adversarial network and dilated convolution | |
| Revathi et al. | Particle Swarm Optimization based Detection of Diabetic Retinopathy using a Novel Deep CNN | |
| Huang et al. | CSAUNet: A cascade self-attention u-shaped network for precise fundus vessel segmentation | |
| Tu et al. | DRPAN: a novel adversarial network approach for retinal vessel segmentation | |
| CN106709448A (en) | Face convolutional neural network-based traditional Chinese medicine constitution recognition method |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181211 |