CN110176007A - Crystalline lens dividing method, device and storage medium - Google Patents
Crystalline lens dividing method, device and storage medium Download PDFInfo
- Publication number
- CN110176007A CN110176007A CN201910412464.6A CN201910412464A CN110176007A CN 110176007 A CN110176007 A CN 110176007A CN 201910412464 A CN201910412464 A CN 201910412464A CN 110176007 A CN110176007 A CN 110176007A
- Authority
- CN
- China
- Prior art keywords
- crystalline lens
- shape template
- initial
- lens structure
- lens
- 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
- 210000000695 crystalline len Anatomy 0.000 title claims abstract description 309
- 238000000034 method Methods 0.000 title claims abstract description 38
- 238000012545 processing Methods 0.000 claims abstract description 45
- 238000009499 grossing Methods 0.000 claims abstract description 36
- 238000003062 neural network model Methods 0.000 claims abstract description 28
- 230000011218 segmentation Effects 0.000 claims abstract description 24
- 238000005516 engineering process Methods 0.000 claims description 11
- 239000000284 extract Substances 0.000 claims description 11
- 238000000605 extraction Methods 0.000 claims description 7
- 238000012549 training Methods 0.000 claims description 7
- 238000004590 computer program Methods 0.000 claims description 5
- 239000000126 substance Substances 0.000 claims description 2
- 238000010586 diagram Methods 0.000 description 8
- 238000013527 convolutional neural network Methods 0.000 description 6
- 238000012014 optical coherence tomography Methods 0.000 description 6
- 230000008569 process Effects 0.000 description 6
- 238000013461 design Methods 0.000 description 5
- 238000003475 lamination Methods 0.000 description 5
- 230000004913 activation Effects 0.000 description 4
- 238000003708 edge detection Methods 0.000 description 4
- 208000002177 Cataract Diseases 0.000 description 3
- 238000001514 detection method Methods 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 230000001054 cortical effect Effects 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 201000010099 disease Diseases 0.000 description 2
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 210000001747 pupil Anatomy 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 244000291564 Allium cepa Species 0.000 description 1
- 235000002732 Allium cepa var. cepa Nutrition 0.000 description 1
- 241000208340 Araliaceae Species 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 230000004888 barrier function Effects 0.000 description 1
- 210000005252 bulbus oculi Anatomy 0.000 description 1
- 210000004087 cornea Anatomy 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 208000030533 eye disease Diseases 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 238000010191 image analysis Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- 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
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10101—Optical tomography; Optical coherence tomography [OCT]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30041—Eye; Retina; Ophthalmic
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Probability & Statistics with Applications (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The embodiment of the present invention provides a kind of crystalline lens dividing method, device and storage medium, this method comprises: extracting crystalline lens image-region from original image；By presetting neural network model, the initial lens structure in crystalline lens image-region is obtained；Edge-smoothing processing is carried out to initial lens structure using shape template, the lens structure after being divided, which is by being trained to crystalline lens sample.The automatic segmentation to lens structure can be realized by default neural network model and shape template, to improve the accuracy of lens structure segmentation while reducing cost of labor.
Description
Technical field
The present embodiments relate to art of image analysis more particularly to a kind of crystalline lens dividing method, device and storage to be situated between
Matter.
Background technique
As common blinding eye disease, cataract is to lead to phacoscotasmus for some reason, to influence view
Film imaging, causes patient not see thing.Wherein, leading portion optical coherence tomography (Anterior Segment Optical
Coherence Tomography, referred to as: AS-OCT) can be used to auxiliary diagnosis include a variety of ophthalmology diseases such as cataract.Tool
Body, AS-OCT is a kind of diagnostic mode of non-intrusion type fanout free region, is measured in white using the density of crystalline lens different structure
The severity of the ophthalmology diseases such as barrier.
Currently, the lens structure segmentation based on AS-OCT image is all performed manually by mostly.But manual segmentation crystalline lens
The problem that structure has accuracy low, and cost of labor is higher.
Summary of the invention
The embodiment of the present invention provides a kind of crystalline lens dividing method, device and storage medium, by dividing crystalline lens automatically
Structure improves the accuracy of lens structure segmentation while reducing cost of labor.
In a first aspect, the embodiment of the present invention provides a kind of crystalline lens dividing method, comprising:
Crystalline lens image-region is extracted from original image；
By presetting neural network model, the initial lens structure in the crystalline lens image-region is obtained；
Edge-smoothing processing is carried out to the initial lens structure using shape template, the crystalline lens knot after being divided
Structure, the shape template are by being trained to crystalline lens sample.
In a kind of possible embodiment, the shape template includes first shape template and the second shape template, wherein
The first shape template is used to carry out the crystalline lens pronucleus of the initial lens structure edge-smoothing processing, and described second
Shape template is used to carry out edge-smoothing processing to core after the crystalline lens of the initial lens structure.
In a kind of possible embodiment, the number of the shape template be it is multiple, it is described using shape template to described
Initial lens structure carries out edge-smoothing processing, the lens structure after being divided, comprising: calculates multiple shape moulds
The similarity of plate and the initial lens structure；The maximum shape template of similarity is chosen, to the initial lens structure
Carry out edge-smoothing processing, the lens structure after being divided.
In a kind of possible embodiment, the phase for calculating the multiple shape templates and the initial lens structure
Like degree, comprising:
For shape template described in each, the shape template and the initial crystalline lens knot are obtained by following steps
The similarity of structure:
The product for calculating the normalized parameter of the shape template and the initial lens structure is the first median, institute
State most narrow spacing of the symmetry axis in rotary course in corresponding all distances that normalized parameter is the initial lens structure
From；
Calculate first median and default bias amount and be the second median；
Boundary coding is carried out to the initial lens structure according to formula (1), obtains target value；
According to the target value and second median, the similarity is determined；
F (c, θ)=| | c-Pθ| | formula (1)
Wherein, c indicates the center point coordinate of the initial lens structure；PθIndicate pair of the initial lens structure
Claim the intersecting point coordinate on the boundary of axis and the initial lens structure, wherein θ indicates the angle of symmetry axis relative datum line, institute
Symmetry axis is stated since reference line, is rotated with predetermined angle；| | | | indicate norm sign；{ f (c, θ) } indicates the target value.
In a kind of possible embodiment, the shape template, training is obtained in the following manner:
According to the boundary point coordinate of the crystalline lens sample, the center point coordinate of the crystalline lens sample is obtained；
Obtain the intersecting point coordinate of the symmetry axis of the crystalline lens sample and the boundary of the crystalline lens sample, wherein described
Symmetry axis is rotated since reference line with predetermined angle；
According to the center point coordinate and the intersecting point coordinate, obtain the central point of the crystalline lens sample to intersection point away from
From；
The distance is normalized using the normalized parameter of the crystalline lens sample, obtains the crystalline lens
The nuclear boundary of sample, the normalized parameter are the minimum range in the symmetry axis rotary course in the corresponding distance；
Extract the intermediate region of the nuclear boundary；
The corresponding intermediate region of the described crystalline lens sample of M is clustered using default clustering algorithm, obtains N number of shape
Template, M and N are positive integer, and M is greater than N.
In a kind of possible embodiment, the default clustering algorithm includes any of following clustering algorithm: K mean value
Algorithm, fuzzy C-mean algorithm FCM clustering algorithm.
It is described that crystalline lens image-region is extracted from original image in a kind of possible embodiment, comprising: to use
Canny edge detecting technology extracts the crystalline lens image-region from the original image.
Second aspect, the embodiment of the present invention provide a kind of crystalline lens segmenting device, comprising:
Extraction module, for extracting crystalline lens image-region from original image；
Processing module, for obtaining initial crystalline in the crystalline lens image-region by presetting neural network model
Body structure；And edge-smoothing processing is carried out to the initial lens structure using preset algorithm, the crystalline lens after being divided
Structure.
The third aspect, the embodiment of the present invention provide a kind of crystalline lens segmenting device, including memory and processor, Yi Jicun
Store up the computer program executed on the memory for the processor；It is real that the processor executes the computer program
Now following operation:
Crystalline lens image-region is extracted from original image；
By presetting neural network model, the initial lens structure in the crystalline lens image-region is obtained；
Edge-smoothing processing is carried out to the initial lens structure using preset algorithm, the crystalline lens knot after being divided
Structure.
Fourth aspect, the embodiment of the present invention provide a kind of computer readable storage medium, including computer-readable instruction, when
When processor reads and executes the computer-readable instruction, so that the processor performs the following operations:
Crystalline lens image-region is extracted from original image；
By presetting neural network model, the initial lens structure in the crystalline lens image-region is obtained；
Edge-smoothing processing is carried out to the initial lens structure using preset algorithm, the crystalline lens knot after being divided
Structure.
Crystalline lens dividing method, device and storage medium provided in an embodiment of the present invention, are extracted from original image first
Crystalline lens image-region；Later, by presetting neural network model, the initial crystalline lens knot in crystalline lens image-region is obtained
Structure, and edge-smoothing processing is carried out to initial lens structure using shape template, the lens structure after being divided, the shape
Shape template is by being trained to crystalline lens sample.It can be realized pair by default neural network model and shape template
The automatic segmentation of lens structure, to improve the accuracy of lens structure segmentation while reducing cost of labor.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
Some embodiments of invention for those of ordinary skill in the art without creative efforts, can be with
It obtains other drawings based on these drawings.
Fig. 1 is the flow chart for the crystalline lens dividing method that one embodiment of the invention provides；
Fig. 2 is the application exemplary diagram for the crystalline lens dividing method that one embodiment of the invention provides；
Fig. 3 shows a kind of lenticular nuclear structure；
Fig. 4 is the exemplary diagram of the shape template before the normalized that one embodiment of the invention provides；
Fig. 5 (a) is the exemplary diagram for the first shape template that one embodiment of the invention provides；
Fig. 5 (b) is the exemplary diagram for the second shape template that one embodiment of the invention provides；
Fig. 6 is the structural schematic diagram for the crystalline lens segmenting device that one embodiment of the invention provides；
Fig. 7 be another embodiment of the present invention provides crystalline lens segmenting device structural schematic diagram.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
The specification of the embodiment of the present invention, claims and term " first " in above-mentioned attached drawing and " second " etc. are to use
In distinguishing similar object, without being used to describe a particular order or precedence order.It should be understood that the data used in this way exist
It can be interchanged in appropriate situation, so that the embodiment of the present invention described herein for example can be in addition to illustrating herein or describing
Those of other than sequence implement.In addition, term " includes " and " having " and their any deformation, it is intended that covering is not
Exclusive includes, for example, the process, method, system, product or equipment for containing a series of steps or units be not necessarily limited to it is clear
Step or unit those of is listed on ground, but is not clearly listed or for these process, methods, product or is set
Standby intrinsic other step or units.
Currently, for the grade classification of cataract, in the world using LOCSII phacoscotasmus classification standard.This point
Class standard human intervention is larger, and the different doctor of experience has a certain difference the classification of different structure, therefore, accurately
It is partitioned into lenticular structure and calculates turbidity automatically, it appears is particularly important.
Since a large amount of medical images of hand labeled are the tasks of a cumbersome and easy error, it is based on above-mentioned, this hair
Bright embodiment provides a kind of crystalline lens dividing method, device and storage medium, by default neural network model and shape template,
The automatic segmentation to lens structure is realized, to improve the accurate of lens structure segmentation while reducing cost of labor
Degree.Wherein, shape template can improve the coarse boundary of the initial lens structure after presetting Processing with Neural Network, to obtain
The higher lens structure of accuracy.
Fig. 1 is the flow chart for the crystalline lens dividing method that one embodiment of the invention provides.The embodiment provides a kind of crystalline
Body dividing method, the crystalline lens dividing method can be executed by crystalline lens segmenting device.The crystalline lens segmenting device can pass through
The mode of software and/or hardware is realized.Illustratively, which can include but is not limited to computer, server
Equal electronic equipments.Wherein, server can be a server, or the server cluster consisted of several servers, or
Person is a cloud computing service center.
As shown in Figure 1, the embodiment provide crystalline lens dividing method the following steps are included:
S101, crystalline lens image-region is extracted from original image.
Wherein, original image can be the target image that actual acquisition arrives, which not only includes crystalline lens image
Region can also include other eyeballs, for example, cornea, vitreum, etc..Here " crystalline lens image-region " is real
Region where the lens structure to be split of border.Supplementary explanation, crystalline lens image-region are less than the big of original image
It is small.
Optionally, which may include: that crystalline lens figure is extracted from original image using canny edge detecting technology
As region.Crystalline lens image-region is extracted by canny edge detecting technology, interference letter extra in original image can be reduced
Breath.
Wherein, canny edge detecting technology is a kind of multistage edge detection algorithm, and target is to find an optimal edge
Detection algorithm.Specifically, optimal edge detection is meant that:
(1) optimal detection: algorithm can identify the actual edge in image as much as possible, missing inspection true edge it is general
The probability of rate and erroneous detection non-edge is all as small as possible；
(2) oplimal Location criterion: the position of the positional distance actual edge point of the marginal point detected is nearest, or by
It is minimum in the degree for the true edge that influence of noise causes the edge detected to deviate object；
(3) test point and marginal point correspond: the marginal point and actual edge point of operator detection should be corresponded.
It should be noted that canny edge detecting technology is only a kind of example, mentioned from original image with illustrating how to realize
Crystalline lens image-region is taken, but the embodiment of the present invention is not limited system, crystalline substance can be extracted from original image with other technologies
Shape body image-region.
S102, the initial lens structure by presetting neural network model, in acquisition crystalline lens image-region.
Wherein, the full convolutional neural networks model of U-shaped that default neural network model can obtain for preparatory training.U-shaped is complete
Convolutional neural networks can make default neural network model have preferable stability and scalability, the learning characteristic of deep learning
Ability it is preferable.
Illustratively, which can be with specifically: crystalline lens image-region is inputted default neural network model, is obtained pre-
If the output of neural network model is initial lens structure.
How the application of the embodiment major embodiment to default neural network model uses default neural network model,
Its training process can refer to related description, and details are not described herein again.
S103, edge-smoothing processing is carried out to initial lens structure using shape template, the crystalline lens after being divided
Structure.
Wherein, shape template is by being trained to crystalline lens sample.
It is appreciated that through default neural network model treated initial lens structure, boundary be it is irregular, especially
It is the segmentation of lens nucleus, and therefore, the embodiment of the present invention improves the coarse boundary in lens nucleus region using shape template, is obtained
Obtain last segmentation result.
With reference to Fig. 2, the segmentation process that an original image is handled through above-mentioned S101 to S103 is shown.Wherein, 1 AS-OCT is indicated
Image；2 indicate area-of-interest (Region of Interest, referred to as: ROI), i.e., crystalline lens is divided by crystalline lens, the example
Three regions: the region cortex (cortex), the region core (nucleus), pupil region；3 indicate the full convolutional neural networks of U-shaped, use
In prediction cut zone；4 indicate shape template, for improving the coarse boundary of lenticular core region.As shown in Fig. 2, having one
Part core region is classified as cortical area by mistake, in order to solve this problem, designs a kind of shape template, Lai Gaishan crystalline lens
The coarse segmentation of nuclear periphery.
Hereinafter, illustrating default neural network model in conjunction with Fig. 2.With reference to Fig. 2, which includes:
Coded portion (left-hand component) and decoded portion (right-hand component).
Wherein, each uncoiling lamination includes a cascade in decoded portion, and input exports complete from upper uncoiling lamination
The information of convolutional layer is corresponded in office's information and coded portion.Uncoiling lamination extracts information from corresponding convolutional layer, and merges local spy
The information of sign avoids local interference to more effectively handle the information of object to be split.
Coded portion includes six convolutional layers, and each convolutional layer includes two to three sub- convolutional layers, each sub- convolutional layer
It will use an activation primitive (Relu) and the maximum pond (MaxPooling) of 2x2；In order to keep effective restored image and
Feature is extracted, decoded portion also includes six uncoiling laminations, and each uncoiling lamination includes a cascade, comes from corresponding characteristic layer
Up-sampled with space, be later with two convolution sums, one activation primitive (Relu), input from upper one layer of global information with
The information of respective layer during network code.It should also be noted that convolutional layer may include the sub- convolutional layer of predetermined number, this is pre-
If mutually cascading between the sub- convolutional layer of number, sub- convolutional layer is using Relu activation primitive and maximum pondization processing.
Default neural network model has used six layers of network structure, biggish in the crystalline lens image area size of input
In the case of, deeper network significantly more efficient can extract global information, to accurately divide lens structure, relatively be applicable in
Existing network structure.
For example, the size of crystalline lens image-region is 1024*1024, what Conv<3x3>with Relu was indicated is convolution kernel
For 3*3, activation primitive Relu, the intersection entropy loss used (Cross Entropy Loss) function is each side-
Output layers of output can achieve the effect of the feature using different levels in side-output layers of use<1x1>convolution kernel
Fruit can further promote segmentation effect.
Using the complete lenticular pupil region of convolutional neural networks model prediction of U-shaped, cortical area and core region, so as to
It is enough to reach preferable training on small data, it can be to avoid over-fitting；And the full convolutional neural networks model of U-shaped can be by making
The cut zone for obtaining clear boundary is connected with jump.
In the embodiment, crystalline lens image-region is extracted first from original image；Then, by presetting neural network mould
Type obtains the initial lens structure in crystalline lens image-region, and carries out side to initial lens structure using shape template
Edge smoothing processing, the lens structure after being divided, the shape template are by being trained to crystalline lens sample.
The automatic segmentation to lens structure can be realized by default neural network model and shape template, thus reducing cost of labor
While, improve the accuracy of lens structure segmentation.
In the above-described embodiments, in a kind of possible implementation, the shape template may include first shape template
With the second shape template.Wherein, first shape template is used to carry out edge-smoothing to the crystalline lens pronucleus of initial lens structure
Processing, the second shape template are used to carry out edge-smoothing processing to core after the crystalline lens of initial lens structure.
It is appreciated that the number of the shape template is multiple.Optionally, S103, using shape template to initial crystalline
Body structure carries out edge-smoothing processing, and the lens structure after being divided may include: the multiple shape templates of calculating and initial
The similarity of lens structure；The maximum shape template of similarity is chosen, edge-smoothing processing is carried out to initial lens structure,
Lens structure after being divided.Illustratively, the crystalline lens of multiple first shape templates and initial lens structure is calculated
The similarity of pronucleus chooses the maximum first shape template of similarity, carries out side to the crystalline lens pronucleus of initial lens structure
Edge smoothing processing；Calculate the similarity of multiple second shape templates with core after the crystalline lens of initial lens structure；It chooses similar
Maximum second shape template is spent, edge-smoothing processing is carried out to core after the crystalline lens of initial lens structure, after obtaining segmentation
Lens structure.
Further, the similarity for calculating multiple shape templates and initial lens structure may include:
For each shape template, the similarity of shape template Yu initial lens structure is obtained by following steps:
The product for calculating the normalized parameter of shape template and initial lens structure is the first median, normalization ginseng
Number is minimum range of the symmetry axis of initial lens structure in rotary course in corresponding all distances；
Calculate the first median and default bias amount and be the second median；
Boundary coding is carried out to initial lens structure according to formula (1), obtains target value；
According to target value and the second median, the similarity is determined；
F (c, θ)=| | c-Pθ| | formula (1)
Wherein, c indicates the center point coordinate of the initial lens structure；PθIndicate pair of the initial lens structure
Claim the intersecting point coordinate on the boundary of axis and the initial lens structure, wherein θ indicates the angle of symmetry axis relative datum line, right
Claim axis since reference line, is rotated with predetermined angle；| | | | indicate norm sign；{ f (c, θ) } indicates the target value.
For example, the number of shape template is N, N is positive integer, is expressed as { f (cn, θ) }, n=1,2,3 ..., N, cn
Indicate that the center point coordinate of shape template n, θ indicate the rotation angle of symmetry axis；Initial lens structure is expressed as St={ xj,
yj, the center point coordinate of initial lens structure isL indicates the edge sampling of initial lens structure
Point number, normalized parameter are expressed as zt, wherein zt=| | c-p1| |, p1Indicate the symmetry axis rotation of initial lens structure
In the process when minimum range in corresponding distance, the intersecting point coordinate of symmetry axis and boundary leads to then for each shape template
It crosses following steps and obtains the similarity of shape template Yu initial lens structure:
One, the first median T is calculatedn: Tn={ f (cn,θ)}×zt。
Two, the second median T ' is calculatedn: T 'n=Tn+ offset, offset indicate default bias amount, value be -10, -
9,…,9,10}。
Three, target value { f (c, θ) } is calculated.
Four, according to target value and the second median, similarity is determined.Specifically, target value and the second median are calculated
Difference: Dn=f (c, θ)-T 'n, DnIt is smaller, illustrate that shape template at this time is more similar to initial lens structure, to obtain each
The similarity of shape template and initial lens structure.
Above embodiments illustrate how to use shape template, it will be illustrated next and how to train to obtain shape template.Specifically
Ground, training obtains shape template in the following manner:
According to the boundary point coordinate of crystalline lens sample, the center point coordinate of crystalline lens sample is obtained；Obtain crystalline lens sample
Symmetry axis and crystalline lens sample boundary intersecting point coordinate, wherein symmetry axis is rotated since reference line with predetermined angle；
According to center point coordinate and intersecting point coordinate, the central point of crystalline lens sample is obtained to the distance of intersection point；Use crystalline lens sample
Normalized parameter, which is adjusted the distance, to be normalized, and the nuclear boundary of crystalline lens sample is obtained, and normalized parameter is symmetry axis rotation
Minimum range in corresponding distance in the process；Extract the intermediate region of nuclear boundary；Using default clustering algorithm to M crystalline lens
The corresponding intermediate region of sample is clustered, and obtains N number of shape template, M and N are positive integer, and M is greater than N.
Lenticular structure is the structure of a similar onion, and lens nucleus structure is smooth curved-surface structure, by
The inspiration of this inspiration designs lenticular nuclear structure as shown in Figure 3 herein, with central point, with the friendship of symmetry axis and boundary
Point, the distance between encoded.Wherein, different figure layers shares identical central point, and can be by away from central point
Distance is distinguished from each other.
With reference to Fig. 3, the boundary representation of crystalline lens sample m is Si={ xi,yi, n=1,2,3 ..., N, crystalline lens sample m's
Center point coordinate is expressed as(xi,yi) be choose crystalline lens sample m i-th of sample point coordinate,
The corresponding shape template of crystalline lens sample m is defined as following formula:
f(cm, θ)=| | cm-pθ||
Wherein, PθIndicate the intersecting point coordinate of the symmetry axis of crystalline lens sample m and the boundary of crystalline lens sample m, θ is from reference line
(dotted line in Fig. 3) start, with 5 degree of predetermined angle rotations, in this way, the boundary in different images can be compiled as
Shown in Fig. 4.In Fig. 4, horizontal axis indicates that θ, the longitudinal axis are indicated | | cm-pθ||.Use the normalized parameter z of crystalline lens sample mm=| |
cm-pm1| | shape template shown in Fig. 4 is normalized, the shape template as shown in Fig. 5 (a) and Fig. 5 (b) is obtained,
In, Fig. 5 (a) indicates first shape template (symmetry axis rotates counterclockwise, 0 degree -180 degree), and Fig. 5 (b) indicates the second shape template
(symmetry axis rotates counterclockwise, and 180 degree -360 is spent).
The intermediate region of nuclear boundary is thick line portion in Fig. 3；It is corresponding to M crystalline lens sample using default clustering algorithm
Intermediate region is clustered, and obtains N number of shape template, M and N are positive integer, and M is greater than N.
Optionally, the default clustering algorithm may include any of following clustering algorithm: K mean value (K-mean) is calculated
Method, fuzzy C-mean algorithm (Fuzzy C-means, referred to as: FCM) clustering algorithm, etc..Wherein, poly- for K-mean algorithm and FCM
The detailed description of class algorithm can refer to the relevant technologies, and details are not described herein again.
It is compared with current crystalline lens dividing method, the present invention at least has the advantages that
(1) present invention designs a kind of based on the full-automatic lens structure dividing method of deep learning, due to data acquisition phase
To difficulty, so existing crystalline lens partitioning scheme is all relied on using the mode manually divided, consistency and accuracy
The experience of segmentation personnel, so, an automatic splitting scheme is very significant for effectively stable segmentation.
(2) according to the matched method of Structural Feature Design shape template of lens nucleus, make segmentation result close to really
Physical structure considers the feature in lenticular internal structure, therefore, designs a shape template and comes in learning training sample
Then shape is corrected test data, can effectively be split to structure.
(3) divide lens structure using the full convolutional neural networks of U-shaped, network can be good at being trained and learning number
Feature in.
(4) strong antijamming capability has preferable generalization ability.
Fig. 6 is the structural schematic diagram for the crystalline lens segmenting device that one embodiment of the invention provides.As shown in fig. 6, crystalline lens
Segmenting device 60 includes: extraction module 61 and processing module 62.Wherein:
The extraction module 61, for extracting crystalline lens image-region from original image.
The processing module 62, connect with extraction module 61, for obtaining extraction module 61 by presetting neural network model
The obtained initial lens structure in crystalline lens image-region；And using shape template to the initial lens structure into
The processing of row edge-smoothing, the lens structure after being divided, the shape template is by being trained to crystalline lens sample
It obtains.
Optionally, shape template may include first shape template and the second shape template.Wherein, the first shape mould
Plate is used to carry out the crystalline lens pronucleus of the initial lens structure edge-smoothing processing, second shape template for pair
Core carries out edge-smoothing processing after the crystalline lens of the initial lens structure.
In the above-described embodiments, the number of the shape template is multiple, and processing module 62 is for using shape template
Edge-smoothing processing is carried out to the initial lens structure, when lens structure after divide, specifically: calculating is multiple
The similarity of the shape template and the initial lens structure；The maximum shape template of similarity is chosen, to described initial
Lens structure carries out edge-smoothing processing, the lens structure after being divided.
Optionally, processing module 62 is similar to the initial lens structure for calculating multiple shape templates
When spending, it is specifically used for:
For shape template described in each, the shape template and the initial crystalline lens knot are obtained by following steps
The similarity of structure:
The product for calculating the normalized parameter of the shape template and the initial lens structure is the first median, institute
State most narrow spacing of the symmetry axis in rotary course in corresponding all distances that normalized parameter is the initial lens structure
From；
Calculate first median and default bias amount and be the second median；
Boundary coding is carried out to the initial lens structure according to formula (1), obtains target value；
According to the target value and second median, the similarity is determined；
F (c, θ)=| | c-Pθ| | formula (1)
Wherein, c indicates the center point coordinate of the initial lens structure；PθIndicate pair of the initial lens structure
Claim the intersecting point coordinate on the boundary of axis and the initial lens structure, wherein θ indicates the angle of symmetry axis relative datum line, institute
Symmetry axis is stated since reference line, is rotated with predetermined angle；| | | | indicate norm sign；{ f (c, θ) } indicates the target value.
Further, the shape template can train in the following manner acquisition:
According to the boundary point coordinate of the crystalline lens sample, the center point coordinate of the crystalline lens sample is obtained；
Obtain the intersecting point coordinate of the symmetry axis of the crystalline lens sample and the boundary of the crystalline lens sample, wherein described
Symmetry axis is rotated since reference line with predetermined angle；
According to the center point coordinate and the intersecting point coordinate, obtain the central point of the crystalline lens sample to intersection point away from
From；
The distance is normalized using the normalized parameter of the crystalline lens sample, obtains the crystalline lens
The nuclear boundary of sample, the normalized parameter are the minimum range in the symmetry axis rotary course in the corresponding distance；
Extract the intermediate region of the nuclear boundary；
The corresponding intermediate region of the described crystalline lens sample of M is clustered using default clustering algorithm, obtains N number of shape
Template, M and N are positive integer, and M is greater than N.
Wherein, the default clustering algorithm includes any of following clustering algorithm: K mean algorithm, and FCM cluster is calculated
Method, etc..
In addition, extraction module 61 can be specifically used for: using canny edge detecting technology, extracted from original image brilliant
Shape body image-region.
The crystalline lens segmenting device that the embodiment provides extracts crystalline lens image-region first from original image；Then,
By presetting neural network model, the initial lens structure in crystalline lens image-region is obtained, and using shape template to first
Beginning lens structure carries out edge-smoothing processing, and the lens structure after being divided, which is by crystalline lens
What sample was trained.It can realize and the automatic of lens structure is divided by presetting neural network model and shape template
It cuts, to improve the accuracy of lens structure segmentation while reducing cost of labor.
Fig. 7 be another embodiment of the present invention provides crystalline lens segmenting device structural schematic diagram.As shown in fig. 7, crystalline
Body segmenting device 70 includes memory 71 and processor 72, and is stored in the calculating executed on memory 71 for processor 72
Machine program.Processor 72 executes computer program and crystalline lens segmenting device 70 is made to realize following operation:
Crystalline lens image-region is extracted from original image；
By presetting neural network model, the initial lens structure in the crystalline lens image-region is obtained；
Edge-smoothing processing is carried out to the initial lens structure using shape template, the crystalline lens knot after being divided
Structure, the shape template are by being trained to crystalline lens sample.
It should be noted that the embodiment of the present invention is not limited for the number of memory 71 and processor 72,
All can be one or more, Fig. 7 is illustrated for one；It, can be by more between memory 71 and processor 72
Kind mode is carried out wired or is wirelessly connected.
In some embodiments, the shape template includes first shape template and the second shape template, wherein described first
Shape template is used to carry out edge-smoothing processing, second shape template to the crystalline lens pronucleus of the initial lens structure
Edge-smoothing processing is carried out for core after the crystalline lens to the initial lens structure.
Optionally, the number of the shape template is multiple, described to use shape template to the initial lens structure
Edge-smoothing processing is carried out, the lens structure after being divided may include: the multiple shape templates of calculating and described first
The similarity of beginning lens structure；The maximum shape template of similarity is chosen, it is flat to carry out edge to the initial lens structure
Sliding processing, the lens structure after being divided.
Further, the similarity for calculating the multiple shape templates and the initial lens structure, can wrap
It includes: for each described shape template, obtaining the shape template and the initial lens structure by following steps
Similarity: the product for calculating the normalized parameter of the shape template and the initial lens structure is the first median, institute
State most narrow spacing of the symmetry axis in rotary course in corresponding all distances that normalized parameter is the initial lens structure
From；Calculate first median and default bias amount and be the second median；According to formula (1) to the initial crystalline lens
Structure carries out boundary coding, obtains target value；According to the target value and second median, the similarity is determined.
Optionally, the shape template can train in the following manner acquisition: according to the boundary of the crystalline lens sample
Point coordinate, obtains the center point coordinate of the crystalline lens sample；Obtain the symmetry axis and the crystalline lens of the crystalline lens sample
The intersecting point coordinate on the boundary of sample, wherein the symmetry axis is rotated since reference line with predetermined angle；According to the center
Point coordinate and the intersecting point coordinate, obtain the central point of the crystalline lens sample to the distance of intersection point；Use the lens-like
The distance is normalized in this normalized parameter, obtains the nuclear boundary of the crystalline lens sample, the normalization
Parameter is the minimum range in the symmetry axis rotary course in the corresponding distance；Extract the middle area of the nuclear boundary
Domain；The corresponding intermediate region of the described crystalline lens sample of M is clustered using default clustering algorithm, obtains N number of shape template,
M and N is positive integer, and M is greater than N.
Wherein, the default clustering algorithm may include any of following clustering algorithm: K mean algorithm, Fuzzy C are equal
Value FCM clustering algorithm.
In addition, above-mentioned extract crystalline lens image-region from original image, it may include: using canny edge detection skill
Art extracts crystalline lens image-region from original image.
On the basis of the above, further, crystalline lens segmenting device 70 can also export the lens structure after segmentation.Cause
This, crystalline lens segmenting device 70 can also include display screen 73.The display screen 73 is used to export the lens structure after segmentation.
Wherein, display screen 73 can be capacitance plate, electromagnetic screen or infrared screen.In general, display screen 73 is used for basis
The instruction of processor 72 shows data, is also used to receive the touch operation for acting on display screen 73, and corresponding signal is sent
To processor 72 or the other component of crystalline lens segmenting device 70.It optionally, further include red when display screen 73 is infrared screen
The surrounding of display screen 73 is arranged in outer touching box, the infrared touch frame, can be also used for receiving infrared signal, and this is infrared
Signal is sent to the other component of processor 72 or crystalline lens segmenting device 70.
The embodiment of the present invention also provides a kind of computer readable storage medium, including computer-readable instruction, works as processor
When reading and executing the computer-readable instruction, so that the processor is executed such as the step in above-mentioned any embodiment.
Those of ordinary skill in the art will appreciate that: realize that all or part of the steps of above method embodiment can pass through
The relevant hardware of program instruction is completed, and program above-mentioned can be stored in a computer readable storage medium, the program
When being executed, step including the steps of the foregoing method embodiments is executed；And storage medium above-mentioned includes: read-only memory (Read-Only
Memory, referred to as: ROM), random access memory (Random Access Memory, referred to as: RAM), magnetic or disk etc.
The various media that can store program code.
Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention., rather than its limitations；To the greatest extent
Pipe present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that: its according to
So be possible to modify the technical solutions described in the foregoing embodiments, or to some or all of the technical features into
Row equivalent replacement；And these are modified or replaceed, various embodiments of the present invention technology that it does not separate the essence of the corresponding technical solution
The range of scheme.
Claims (10)
1. a kind of crystalline lens dividing method characterized by comprising
Crystalline lens image-region is extracted from original image；
By presetting neural network model, the initial lens structure in the crystalline lens image-region is obtained；
Edge-smoothing processing is carried out to the initial lens structure using shape template, the lens structure after being divided,
The shape template is by being trained to crystalline lens sample.
2. the method according to claim 1, wherein the shape template includes first shape template and the second shape
Shape template, wherein the first shape template is used to carry out edge-smoothing to the crystalline lens pronucleus of the initial lens structure
Processing, second shape template are used to carry out edge-smoothing processing to core after the crystalline lens of the initial lens structure.
3. the method according to claim 1, wherein the number of the shape template be it is multiple, it is described use shape
Shape template carries out edge-smoothing processing to the initial lens structure, the lens structure after being divided, comprising:
Calculate the similarity of multiple shape templates Yu the initial lens structure；
The maximum shape template of similarity is chosen, edge-smoothing processing is carried out to the initial lens structure, after obtaining segmentation
Lens structure.
4. according to the method described in claim 3, it is characterized in that, described calculate multiple shape templates and the initial crystalline substance
The similarity of shape body structure, comprising:
For shape template described in each, the shape template and the initial lens structure are obtained by following steps
Similarity:
The product for calculating the normalized parameter of the shape template and the initial lens structure is the first median, described to return
One change parameter is minimum range of the symmetry axis of the initial lens structure in rotary course in corresponding all distances；
Calculate first median and default bias amount and be the second median；
Boundary coding is carried out to the initial lens structure according to formula (1), obtains target value；
According to the target value and second median, the similarity is determined；
F (c, θ)=| | c-Pθ| | formula (1)
Wherein, c indicates the center point coordinate of the initial lens structure；PθIndicate the symmetry axis of the initial lens structure
With the intersecting point coordinate on the boundary of the initial lens structure, wherein θ indicates the angle of symmetry axis relative datum line, described right
Claim axis since reference line, is rotated with predetermined angle；| | | | indicate norm sign；{ f (c, θ) } indicates the target value.
5. training obtains in the following manner the method according to claim 1, wherein the shape template:
According to the boundary point coordinate of the crystalline lens sample, the center point coordinate of the crystalline lens sample is obtained；
Obtain the intersecting point coordinate of the symmetry axis of the crystalline lens sample and the boundary of the crystalline lens sample, wherein described symmetrical
Axis is rotated since reference line with predetermined angle；
According to the center point coordinate and the intersecting point coordinate, the central point of the crystalline lens sample is obtained to the distance of intersection point；
The distance is normalized using the normalized parameter of the crystalline lens sample, obtains the crystalline lens sample
Nuclear boundary, the normalized parameter is the minimum range in the symmetry axis rotary course in the corresponding distance；
Extract the intermediate region of the nuclear boundary；
The corresponding intermediate region of the described crystalline lens sample of M is clustered using default clustering algorithm, obtains N number of shape mould
Plate, M and N are positive integer, and M is greater than N.
6. according to the method described in claim 5, it is characterized in that, the default clustering algorithm includes in following clustering algorithm
Any one:
K mean algorithm, fuzzy C-mean algorithm FCM clustering algorithm.
7. method according to any one of claims 1-5, which is characterized in that described to extract crystalline lens from original image
Image-region, comprising:
Using canny edge detecting technology, the crystalline lens image-region is extracted from the original image.
8. a kind of crystalline lens segmenting device characterized by comprising
Extraction module, for extracting crystalline lens image-region from original image；
Processing module, for obtaining the initial crystalline lens knot in the crystalline lens image-region by presetting neural network model
Structure；And edge-smoothing processing is carried out to the initial lens structure using shape template, the crystalline lens knot after being divided
Structure, the shape template are by being trained to crystalline lens sample.
9. a kind of crystalline lens segmenting device, which is characterized in that including memory and processor, and be stored on the memory
The computer program executed for the processor；
The processor executes the computer program and realizes following operation:
Crystalline lens image-region is extracted from original image；
By presetting neural network model, the initial lens structure in the crystalline lens image-region is obtained；
Edge-smoothing processing is carried out to the initial lens structure using shape template, the lens structure after being divided,
The shape template is by being trained to crystalline lens sample.
10. a kind of computer readable storage medium, which is characterized in that including computer-readable instruction, when processor is read and is held
When the row computer-readable instruction, so that the processor performs the following operations:
Crystalline lens image-region is extracted from original image；
By presetting neural network model, the initial lens structure in the crystalline lens image-region is obtained；
Edge-smoothing processing is carried out to the initial lens structure using shape template, the lens structure after being divided,
The shape template is by being trained to crystalline lens sample.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910412464.6A CN110176007A (en) | 2019-05-17 | 2019-05-17 | Crystalline lens dividing method, device and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910412464.6A CN110176007A (en) | 2019-05-17 | 2019-05-17 | Crystalline lens dividing method, device and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110176007A true CN110176007A (en) | 2019-08-27 |
Family
ID=67691473
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910412464.6A Pending CN110176007A (en) | 2019-05-17 | 2019-05-17 | Crystalline lens dividing method, device and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110176007A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111210404A (en) * | 2019-12-24 | 2020-05-29 | 中国科学院宁波工业技术研究院慈溪生物医学工程研究所 | Method and device for classifying lens segmentation difficulty |
| CN115712363A (en) * | 2022-11-21 | 2023-02-24 | 北京中科睿医信息科技有限公司 | Interface color display method, device, equipment and medium |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1776717A (en) * | 2005-12-01 | 2006-05-24 | 上海交通大学 | Method for identifying shoes print at criminal scene |
| CN101256631A (en) * | 2007-02-26 | 2008-09-03 | 富士通株式会社 | Method, apparatus, program and readable storage medium for character recognition |
| CN101719272A (en) * | 2009-11-26 | 2010-06-02 | 上海大学 | Three-dimensional image segmentation method based on three-dimensional improved pulse coupled neural network |
| CN102254172A (en) * | 2011-06-16 | 2011-11-23 | 电子科技大学 | Method for segmenting fingerprint image based on cellular neural network and morphology |
| CN103984416A (en) * | 2014-06-10 | 2014-08-13 | 北京邮电大学 | Gesture recognition method based on acceleration sensor |
| CN106023220A (en) * | 2016-05-26 | 2016-10-12 | 史方 | Vehicle exterior part image segmentation method based on deep learning |
| CN106780454A (en) * | 2016-12-08 | 2017-05-31 | 苏州汉特士视觉科技有限公司 | Vision positioning method and automatic feed dividing feeding device based on edge back projection |
-
2019
- 2019-05-17 CN CN201910412464.6A patent/CN110176007A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1776717A (en) * | 2005-12-01 | 2006-05-24 | 上海交通大学 | Method for identifying shoes print at criminal scene |
| CN101256631A (en) * | 2007-02-26 | 2008-09-03 | 富士通株式会社 | Method, apparatus, program and readable storage medium for character recognition |
| CN101719272A (en) * | 2009-11-26 | 2010-06-02 | 上海大学 | Three-dimensional image segmentation method based on three-dimensional improved pulse coupled neural network |
| CN102254172A (en) * | 2011-06-16 | 2011-11-23 | 电子科技大学 | Method for segmenting fingerprint image based on cellular neural network and morphology |
| CN103984416A (en) * | 2014-06-10 | 2014-08-13 | 北京邮电大学 | Gesture recognition method based on acceleration sensor |
| CN106023220A (en) * | 2016-05-26 | 2016-10-12 | 史方 | Vehicle exterior part image segmentation method based on deep learning |
| CN106780454A (en) * | 2016-12-08 | 2017-05-31 | 苏州汉特士视觉科技有限公司 | Vision positioning method and automatic feed dividing feeding device based on edge back projection |
Non-Patent Citations (1)
| Title |
|---|
| 看准资讯: "黄金比例把明星脸都变成了什么鬼", 《搜狐网》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111210404A (en) * | 2019-12-24 | 2020-05-29 | 中国科学院宁波工业技术研究院慈溪生物医学工程研究所 | Method and device for classifying lens segmentation difficulty |
| CN115712363A (en) * | 2022-11-21 | 2023-02-24 | 北京中科睿医信息科技有限公司 | Interface color display method, device, equipment and medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Li et al. | Computer‐assisted diagnosis for diabetic retinopathy based on fundus images using deep convolutional neural network | |
| TWI715117B (en) | Method, device and electronic apparatus for medical image processing and storage mdeium thereof | |
| Gao et al. | Classification of CT brain images based on deep learning networks | |
| US10223610B1 (en) | System and method for detection and classification of findings in images | |
| US20220198230A1 (en) | Auxiliary detection method and image recognition method for rib fractures based on deep learning | |
| Banar et al. | Towards fully automated third molar development staging in panoramic radiographs | |
| CN104573309B (en) | Device and method for computer-aided diagnosis | |
| Van Rikxoort et al. | Automatic lung segmentation from thoracic computed tomography scans using a hybrid approach with error detection | |
| Shao et al. | Brain ventricle parcellation using a deep neural network: Application to patients with ventriculomegaly | |
| US20210236080A1 (en) | Cta large vessel occlusion model | |
| CN109872325B (en) | Full-automatic liver tumor segmentation method based on two-way three-dimensional convolutional neural network | |
| CN106682435A (en) | System and method for automatically detecting lesions in medical image through multi-model fusion | |
| JP2021002338A (en) | Method and system for image segmentation and identification | |
| CN107977952A (en) | Medical image cutting method and device | |
| Sreelakshmy et al. | [Retracted] An Automated Deep Learning Model for the Cerebellum Segmentation from Fetal Brain Images | |
| Almotiri et al. | A multi-anatomical retinal structure segmentation system for automatic eye screening using morphological adaptive fuzzy thresholding | |
| US20240185428A1 (en) | Medical Image Analysis Using Neural Networks | |
| Yang et al. | A deep learning segmentation approach in free‐breathing real‐time cardiac magnetic resonance imaging | |
| David et al. | Retinal Blood Vessels and Optic Disc Segmentation Using U‐Net | |
| US11475568B2 (en) | Method for controlling display of abnormality in chest x-ray image, storage medium, abnormality display control apparatus, and server apparatus | |
| CN110176007A (en) | Crystalline lens dividing method, device and storage medium | |
| CN110009641A (en) | Lens segmentation method, device and storage medium | |
| Li et al. | A deep-learning method for the end-to-end prediction of intracranial aneurysm rupture risk | |
| Arzhaeva et al. | Computer‐aided detection of interstitial abnormalities in chest radiographs using a reference standard based on computed tomography | |
| Zheng et al. | Adaptive segmentation of vertebral bodies from sagittal MR images based on local spatial information and Gaussian weighted chi-square distance |
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 |
Application publication date: 20190827 |
|
| RJ01 | Rejection of invention patent application after publication |